Compare commits

..

235 Commits

Author SHA1 Message Date
Ryuichi Leo Takashige 0955966b2a Claude-generated settings, no idea if it works 2026-02-26 13:49:05 +00:00
Ryuichi Leo Takashige 20ea13f047 use memory pressure 2026-02-26 13:16:28 +00:00
Ryuichi Leo Takashige 4ff1578140 cleanup 2026-02-25 21:46:31 +00:00
Ryuichi Leo Takashige a5873bc1fd remove test script 2026-02-25 20:54:52 +00:00
Ryuichi Leo Takashige dc1ce2a2cf cleanup 2026-02-25 20:54:02 +00:00
Ryuichi Leo Takashige ff57b00dc6 cleanup 2026-02-25 20:44:56 +00:00
Ryuichi Leo Takashige d3222c498a Loosen conditions 2026-02-25 19:36:05 +00:00
Ryuichi Leo Takashige 2f719d62a7 Handle low memory better 2026-02-25 19:25:18 +00:00
Evan Quiney ba611f9cd0 Revert "report macmon failures more aggressively (#1618)" (#1625)
this pr broke macmon - revert it
2026-02-25 19:15:55 +00:00
Evan Quiney eab3e0b456 report macmon failures more aggressively (#1618)
macmon appears to be going silent. time it out and restart it after
3*interval (Default 3s) and report warning.

Co-authored-by: Alex Cheema <41707476+AlexCheema@users.noreply.github.com>
2026-02-25 17:49:10 +00:00
Evan Quiney c4e874e97d skip nan logprobs on tokens (#1622)
sometimes we generate NaN logprobs for tokens, this causes pydantic
validation errors on the receiving end. in this case we should just not
send the logprob items
2026-02-25 17:44:15 +00:00
rltakashige e23c3a3026 Address Mac Mini pipeline GPU timeouts (#1620)
## Motivation
Users were reporting GPU timeout errors on Mac Minis, which we never saw
on testing with Mac Studios. It also seems to only happen with large
models.

## Changes
Eval specific distributed operations.

## Why It Works

As I wrote in a Slack message:
Basically, prefill is too slow for pipeline communications. If there are
both communications and GPU operations as part of an mlx graph, the
communications become subject to the GPU's 5 second command buffer
timeout.

For normal generation, I added evals to the communications (only during
prefill, as it slows down decode) to do this, fixing GPU timeouts.

But we don't do this during warmup, as the prompt is absolutely tiny.
This is still too slow on an M4 Pro on some models that it causes a GPU
timeout during warmup...


----------------------
This was one of the issues. However, there is another issue:

mx.all_gather sometimes reads stale data with FAST_SYNCH enabled. I'm
still investigating the root cause, but the code as it is now works on
Mac Minis.



## Test Plan

### Manual Testing
<img width="2762" height="1808" alt="image"
src="https://github.com/user-attachments/assets/27c88542-606c-4551-8f7c-bd2c0471f54e"
/>

<img width="2820" height="1898" alt="image"
src="https://github.com/user-attachments/assets/0ba3478c-ee39-438d-902c-92893db23d05"
/>


### Automated Testing
Needs a bunch on mac minis
2026-02-25 17:37:32 +00:00
Alex Cheema 190e63e56d fix: log exceptions causing silent node shutdown (#1621)
## Motivation

Nodes silently shut down mid-inference with no exception logged. The
logs show a clean-looking shutdown cascade (unsubscribe all topics →
stop worker → runner communication closed) but no error explaining
*why*. This makes debugging cluster issues extremely difficult.

## Changes

**`src/exo/routing/router.py`** — Added `try/except Exception` around
`_networking_recv()` and `_networking_recv_connection_messages()` loops.
Logs the root cause at ERROR level via
`logger.opt(exception=...).error(...)` before re-raising. Uses
`Exception` (not `BaseException`) so clean SIGTERM cancellation is
unaffected.

**`src/exo/main.py`** — Wrapped `anyio.run(node.run)` in
`try/except/finally`:
- `except BaseException`: logs the fatal exception at CRITICAL level
through loguru
- `finally`: ensures `logger.info("EXO Shutdown complete")` and
`logger_cleanup()` always run, guaranteeing the async log queue is
flushed before process exit

## Why It Works

The Router's recv loops (`_networking_recv`,
`_networking_recv_connection_messages`) are infinite `while True` loops
calling Rust bindings with **no exception handling**. When the Rust
networking layer raises `ConnectionError` (e.g. channel closed), the
exception cascades silently through anyio task groups: Router → Node →
all components shut down. The exception was never logged because (1) no
try-except anywhere in the cascade, and (2) `logger_cleanup()` was
skipped when `anyio.run()` raised, so loguru's async queue was never
flushed.

## Test Plan

### Manual Testing
- Run `uv run exo`, send SIGTERM → should see clean shutdown with "EXO
Shutdown complete", no error/critical logs
- Next time the silent shutdown reproduces, logs should now show the
actual exception with full traceback at ERROR level from the recv loop,
plus CRITICAL level from main()

### Automated Testing
- All 222 existing tests pass (1 pre-existing failure in
`test_python.py::test_sleep_on_multiple_items` unrelated to this change)
- `uv run basedpyright` — 0 errors
- `uv run ruff check` — all checks passed
- `nix fmt` — applied

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com>
2026-02-25 08:24:18 -08:00
Alex Cheema 7660893538 fix: replace flaky internet checks with explicit offline mode (#1615)
## Motivation

The socket-based internet connectivity probes (connecting to
1.1.1.1/8.8.8.8/1.0.0.1 every 10 seconds) are flaky — they produce false
negatives on some networks/ISPs, causing downloads to silently fail or
get stuck. Instead of dynamically detecting connectivity, this replaces
it with an explicit offline mode that the user opts into.

## Changes

- **Removed internet check infrastructure**: Deleted
`_test_internet_connection()` (socket probes) and
`_check_internet_connection()` (10s polling loop) from
`DownloadCoordinator`
- **Renamed `internet_connection` → `offline`** on `ShardDownloader`
base class and all subclasses (`SingletonShardDownloader`,
`CachedShardDownloader`, `ResumableShardDownloader`)
- **Removed `on_connection_lost` callbacks** from download calls — no
longer needed without dynamic connectivity detection
- **Added `EXO_OFFLINE` env var support** in `constants.py` and `Args`
default in `main.py`
- **Added Offline Mode toggle** to macOS app Settings → General tab,
which sets `EXO_OFFLINE=true` in the process environment and restarts

## Why It Works

The download behavior (`skip_internet` conditionals in
`download_utils.py`) is unchanged — we just changed what drives it.
Instead of a flaky socket probe setting
`internet_connection=True/False`, the user explicitly sets offline mode
via:
1. `--offline` CLI flag
2. `EXO_OFFLINE=true` environment variable
3. macOS app Settings → General → Offline Mode toggle

This is more reliable and predictable. Users on flaky networks no longer
get intermittent download failures.

## Test Plan

### Manual Testing
- `uv run exo --offline` starts without internet checks, rejects
downloads for unavailable models
- `EXO_OFFLINE=true uv run exo` behaves identically
- macOS app Settings toggle persists across restarts and passes env var
to the exo process

### Automated Testing
- All existing tests pass (222 passed), including 8 offline-mode
specific tests
- `uv run basedpyright` — 0 errors
- `uv run ruff check` — all checks passed
- `nix fmt` — 0 files changed

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com>
2026-02-24 13:36:19 -08:00
Evan Quiney 9a2d2a4a7c bump (#1608)
should be release for 1.0.68 - let's synch our py version with our app
version - 0.3.68.
next minor release should be 1.4.0 and 0.4.0 respectively.

Co-authored-by: rltakashige <rl.takashige@gmail.com>
2026-02-24 19:14:15 +00:00
Alex Cheema bea64fe85e fix: prevent stale loading state and conversation loss when switching chats (#1613)
## Motivation

When navigating back to a previous conversation that used a different
model than the one currently loading, the UI incorrectly displayed the
other model's loading/download progress bar instead of the conversation
messages. Additionally, attempting to continue that old conversation by
sending a message would create an entirely new chat rather than
continuing the existing one.

## Changes

All changes in `dashboard/src/routes/+page.svelte`:

1. **Added fallthrough reset in the `chatLaunchState` restore
`$effect`**: When switching to a conversation whose model has no active
instance (not running, not downloading, not loading), the effect now
resets `chatLaunchState` to `"idle"` instead of leaving stale state from
a different model.

2. **Added `skipCreate` parameter to `launchModelForChat()`**: When
continuing an existing conversation (has messages),
`createConversation()` is skipped so the old conversation is preserved
rather than replaced with a new empty one.

3. **Reordered view conditional**: Progress views
(downloading/loading/launching) take priority, then conversation
messages (when idle with messages or model ready), then model selector
(idle with no messages). This ensures:
   - Old conversations display normally when their model isn't running
- Download progress shows when relaunching a model for any conversation
   - Model selector only appears when there are no messages

## Why It Works

- The `$effect` fallthrough prevents `chatLaunchState` from retaining a
stale value (e.g., `"downloading"`) from Model A when the user switches
to a conversation using Model B that has no active state.
- The `skipCreate` flag ensures `launchModelForChat` can be reused for
both new conversations (from model picker) and continuing existing ones
(from chat input) without always creating a new conversation.
- The reordered view logic ensures each state maps to the correct UI:
active launch → progress, existing messages → chat view, nothing → model
selector.

## Test Plan

### Manual Testing
- Load an old conversation while a different model is downloading →
should see conversation messages, not the other model's loading bar
- Send a message from that old conversation → model should
launch/download with progress shown → conversation continues with the
response appended
- Select a new model from the picker → should still create a new
conversation as before
- New conversation with model download → progress bar shows correctly

---------

Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com>
2026-02-24 19:08:31 +00:00
rltakashige 14526d281a update mlx 2 (#1611)
## Motivation

GPU locks because of prompt progress callbacks taking time. Current
solution: Don't fix it, make the symptom better

## Changes
Shortened timeout by 2x
Get event leak fixes from latest upstream
2026-02-24 18:30:48 +00:00
Evan Quiney 73e50df827 fix glm5 tool calling (#1612)
glm5 is a deepseekv32 model, so was parsing dsml style tool calls
instead of glm style tool calls. fix it!!
2026-02-24 18:21:18 +00:00
Alex Cheema 2b417f28ff fix: sync model selectors between sidebar and chat input (#1610)
## Motivation

The "Load Model" dropdown in the right sidebar and the model selector
above the chat input were operating on independent state
(`selectedPreviewModelId` vs `selectedChatModel`). This caused several
UX issues:
- Selecting a model in one selector didn't update the other
- After going home from a chat, the chat selector showed "SELECT MODEL"
while the sidebar still showed the model
- On page refresh, only the sidebar restored the saved model
- Sending a chat message with a running model briefly showed the
recommended models view instead of starting the chat

## Changes

- **`handleModelPickerSelect`**: Also sets `selectedChatModel` when a
model is picked from the sidebar dropdown
- **`handleChatPickerSelect`**: Also sets `selectedPreviewModelId` when
a model is picked from the chat selector
- **`handleGoHome`**: Restores `selectedChatModel` from the sidebar's
`selectedModelId` instead of clearing it
- **`applyLaunchDefaults`**: Syncs `selectedChatModel` when restoring
saved defaults on page load
- **`handleChatSend`**: Sets `chatLaunchState = "ready"` before calling
`sendMessage` when the model is already running, ensuring the chat view
renders correctly on view transition

## Why It Works

The two selectors were backed by different store properties that were
never kept in sync. By updating both properties in every selection path
(sidebar pick, chat pick, go-home, page restore), they always reflect
the same model. The `chatLaunchState = "ready"` fix mirrors what
`launchModelForChat` already does (line 2649) and prevents the chat
state from being "idle" when transitioning from the welcome view to the
chat view.

## Test Plan

### Manual Testing
<!-- Hardware: MacBook Pro M4 Max 48GB -->
- Select a model from sidebar "Load Model" dropdown → chat input
selector updates to match
- Select a model from chat input selector → sidebar dropdown updates to
match
- Launch a model, chat with it, click "Go Home" → both selectors show
the same model
- Refresh the page → both selectors show the previously selected model
- With a running model, type and send a message from the welcome view →
chat starts directly without flashing the recommended models view

### Automated Testing
No new automated tests — this is a UI state synchronization fix in the
Svelte dashboard with no backend changes.

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com>
2026-02-24 17:31:20 +00:00
Alex Cheema b65982ddd7 fix: improve text contrast on HOME and DOWNLOADS nav links (#1609)
## Motivation

Follow-up to #1601 (downloads page contrast fix). The HOME and DOWNLOADS
navigation links in the top-right header use `text-exo-light-gray`
(`oklch(0.6 0 0)`) which is too dim against the dark header background.

## Changes

Changed both nav links in `HeaderNav.svelte` from `text-exo-light-gray`
to `text-white/70` for better visibility. Hover state
(`text-exo-yellow`) is unchanged.

## Why It Works

`text-white/70` provides noticeably better contrast against
`bg-exo-dark-gray` while still looking subdued relative to the yellow
accent color on hover. This is consistent with the approach used in
#1601.

## Test Plan

### Manual Testing
- Verified both links are clearly readable on the home page and
downloads page
- Hover state still transitions to yellow as expected

### Automated Testing
- Dashboard builds successfully

---------

Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com>
2026-02-24 13:48:40 +00:00
rltakashige 2fe689315b download .model files in exo bench (#1607)
## Motivation

failed again for kimi on a machine that had never downloaded it.

## Test Plan

### Manual Testing
it worked this time
2026-02-24 11:13:04 +00:00
Alex Cheema 644c5573ce fix: improve text contrast on downloads page (#1601)
## Summary
- Bumps opacity on dark-grey text/icons on the downloads page that were
nearly invisible against the dark background
- Informational text (GB sizes, speeds, disk free, model IDs) → full
`text-exo-light-gray`
- Interactive icons (delete, resume, retry) → `/70` at rest
- Hover-only elements (download button) → `/60`

## Test plan
- [ ] Open http://localhost:52415/downloads with models downloaded
- [ ] Verify GB downloaded amounts are clearly visible
- [ ] Verify delete trash icons are visible (not just on hover)
- [ ] Verify download speed text is readable
- [ ] Verify "paused" labels and resume buttons are visible
- [ ] Verify "X GB free" disk labels in column headers are readable

🤖 Generated with [Claude Code](https://claude.com/claude-code)

---------

Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com>
2026-02-24 10:42:49 +00:00
rltakashige 12c3015f52 fix qwen moe tensor sharding (#1604)
## Motivation

<!-- Why is this change needed? What problem does it solve? -->
<!-- If it fixes an open issue, please link to the issue here -->

## Changes

<!-- Describe what you changed in detail -->

## Why It Works

<!-- Explain why your approach solves the problem -->

## Test Plan

### Manual Testing
<!-- Hardware: (e.g., MacBook Pro M1 Max 32GB, Mac Mini M2 16GB,
connected via Thunderbolt 4) -->
<!-- What you did: -->
<!-- - -->

### Automated Testing
<!-- Describe changes to automated tests, or how existing tests cover
this change -->
<!-- - -->
2026-02-23 21:23:27 +00:00
rltakashige 365dd68d9a Final fixes for release (#1603)
## Motivation

<!-- Why is this change needed? What problem does it solve? -->
<!-- If it fixes an open issue, please link to the issue here -->

## Changes

<!-- Describe what you changed in detail -->

## Why It Works

<!-- Explain why your approach solves the problem -->

## Test Plan

### Manual Testing
<!-- Hardware: (e.g., MacBook Pro M1 Max 32GB, Mac Mini M2 16GB,
connected via Thunderbolt 4) -->
<!-- What you did: -->
<!-- - -->

### Automated Testing
<!-- Describe changes to automated tests, or how existing tests cover
this change -->
<!-- - -->
2026-02-23 21:10:15 +00:00
Alex Cheema d3d129581e test: verify instance deletion cancels ongoing tasks (#1508)
## Summary
- The cancellation logic for issue #1215 already exists in
`get_transition_events()` (`src/exo/master/placement.py:208-227`) — when
an instance is deleted, `TaskStatusUpdated(Cancelled)` events are
emitted for all Pending/Running tasks on that instance
- Combined with PR #1276's token-boundary cancellation in runners, the
full pipeline works end-to-end
- However, the existing test
`test_get_transition_events_delete_instance` passed `{}` for tasks, so
this path was never exercised
- This PR adds 4 tests covering the cancellation behavior:
  - Running tasks are cancelled on instance deletion
  - Pending tasks are cancelled on instance deletion
  - Completed/Failed/TimedOut/Cancelled tasks are left alone
  - Only tasks matching the deleted instance are cancelled

Closes #1215

## Test plan
- [x] `uv run pytest src/exo/master/tests/test_placement.py -v` — all 15
tests pass
- [x] `uv run basedpyright` — 0 errors
- [x] `uv run ruff check` — all checks passed

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com>
2026-02-23 20:12:23 +00:00
Alex Cheema c90a0cec78 fix: suppress closure errors in runnersupervisor and force spawn start method (#1547)
some errors could be thrown during shutdown - we can dismiss these safely

co-authored by me :)
2026-02-23 18:30:41 +00:00
Alex Cheema e8c1337168 fix: add download/resume buttons to pending downloads (#1581)
## Summary
- Adds a **resume button** (download icon) to paused pending downloads
(those with partial progress)
- Adds a **download button** to not-started pending downloads
- Both buttons call the existing `startDownload()` function which
handles both new downloads and resuming partial ones
- Previously, paused downloads only showed a "paused" label with no
action, and not-started downloads showed "..." with no way to trigger
them

## Test plan
- [ ] Build dashboard (`cd dashboard && npm run build`)
- [ ] Start exo, navigate to Downloads tab
- [ ] Verify paused downloads show a clickable resume (download arrow)
icon below the progress bar
- [ ] Verify not-started pending downloads show a clickable download
icon
- [ ] Click both button types and confirm downloads start/resume

> Note: Screenshot could not be captured because the dashboard requires
the exo backend API to render, and exo has a pre-existing
`Keypair.generate()` startup bug on main.

🤖 Generated with [Claude Code](https://claude.com/claude-code)

---------

Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com>
2026-02-23 18:20:35 +00:00
Alex Cheema 7024ddcf3e fix: detect completed downloads by checking final file exists (#1582)
## Summary

Split from #1547 per review feedback.

When scanning existing download status, `get_downloaded_size()` returns
bytes from either the final file or its `.partial` counterpart — so a
`.partial` file with all bytes downloaded (e.g. process killed before
hash verification and rename) could be falsely reported as complete.

The previous approach (commit 3b54e7df) added a byte-comparison fallback
in the coordinator (`downloaded >= total > 0`), but this suffered from
the same `.partial` conflation issue.

**Fix:** Check whether the final (non-`.partial`) file actually exists
on disk before marking status as `"complete"` in `download_utils.py`.
This is the only reliable signal that hash verification passed and the
rename from `.partial` succeeded. The coordinator-level byte comparison
is removed since the source now reports correctly.

### Changes
- `download_utils.py`: Add `final_file_exists` check — only report
`"complete"` when the renamed, hash-verified file exists with matching
size
- `coordinator.py`: Revert to simple `progress.status == "complete"`
check, removing the byte-comparison fallback

**Note:** The corresponding byte-comparison workaround in #1547 should
also be removed.

## Test plan
- [x] basedpyright — 0 errors
- [x] ruff check — passes
- [x] pytest — 218 passed, 1 skipped (1 pre-existing Rust bindings
failure)

🤖 Generated with [Claude Code](https://claude.com/claude-code)

---------

Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com>
2026-02-23 18:12:07 +00:00
Alex Cheema dc89ba662a feat: add info button to model picker variant rows (#1589)
## Motivation

When a model has multiple quantization variants (e.g., 4-bit, 8-bit),
expanding the group shows each variant's quantization and size — but
there's no way to see which specific nodes have each variant downloaded.
The top-level model group has an (i) info button, but the individual
variant rows don't. This makes it hard to tell which nodes have which
quantization.

## Changes

- Added an (i) info button to each expanded variant row in
`ModelPickerGroup.svelte`
- When clicked, opens the existing info modal scoped to that single
variant (showing its quantization, size, and which nodes have it
downloaded)
- Changed the variant row element from `<button>` to `<div
role="button">` to allow nesting the info `<button>` (matching the
pattern used by the top-level model row)

## Why It Works

Reuses the existing `onShowInfo` callback and info modal by constructing
a synthetic single-variant `ModelGroup` from the clicked variant. No
changes needed to `ModelPickerModal.svelte` — the info modal already
handles single-variant groups correctly.

## Test Plan

### Manual Testing
<!-- Hardware: MacBook Pro -->
- Open dashboard, open model picker
- Expand a model with multiple variants — each variant row now has an
(i) icon
- Click a variant's (i) — the info modal shows that single variant's
quantization/size and which nodes have it downloaded

### Automated Testing
- Dashboard builds successfully with `npm run build`
- `nix fmt` passes

Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com>
2026-02-23 09:53:36 -08:00
Alex Cheema 5cd96b5060 feat: seamless chat UX with auto model selection and smart recommendations (#1590)
## Motivation

The current chat experience requires users to manually select a model
before they can start chatting. This creates unnecessary friction —
especially for new users who may not know which model to pick.
Additionally, models that are already running or downloading appear
greyed out in the picker if available memory is low, making them
impossible to select even though they're already loaded.

## Changes

### New: ChatModelSelector component
- Category-based model recommendations (Best for Coding, Writing,
Agentic, Biggest) shown on the idle chat screen
- Tiered model ranking system (`AUTO_TIERS`) that picks the best model
fitting available memory
- Fixed-position tooltips that escape overflow-hidden ancestors

### Auto model selection from chat
- Typing in an empty chat auto-picks the best model using tier rankings
- Prefers already-running models over launching smaller new ones (avoids
the "launched a big model but recommends a smaller one" bug)
- If a model is already downloading/loading, attaches to existing
progress instead of launching a duplicate

### Launch-from-chat state machine
- Full `idle → launching → downloading → loading → ready` flow with
inline progress display
- `pendingAutoMessage` queue: messages typed during model launch are
remembered and sent once model is ready
- Fixed `$effect` race condition where restore effect competed with
auto-advance effect, causing lost messages

### Model picker improvements
- Instance status badges: green dot (ready), blue pulsing (downloading),
yellow pulsing (loading) on both group and variant rows
- "Ready" availability filter in ModelFilterPopover
- **Models with active instances (ready, downloading, loading) are never
greyed out** regardless of memory — they're already loaded/in-progress
and must be selectable

### Home page model display
- Model picker button shows the best running model name instead of "—
SELECT MODEL —" when a model is already active
- Uses the same `bestRunningModelId` tier-based derived as the chat
auto-selection

### Chat sidebar & form
- ChatSidebar conversation management with rename, delete, "New Chat"
- `userForcedIdle` guard prevents reactive effects from overriding
explicit user actions (e.g., clicking "New Chat")
- ChatForm simplified: uses ModelPickerModal instead of inline dropdown,
`modelDisplayOverride` prop for showing auto-selected models

## Why It Works

The tier-based auto-selection (`getAutoTierIndex` + `pickAutoModel`)
ensures users always get the best model their cluster can run. By
checking running instances first and comparing tiers, we avoid the bug
where loading a large model reduces available memory and causes a
smaller model to be recommended. The `$effect` ordering fix (moving the
`chatLaunchState === "ready"` guard after the `hasRunningInstance`
check) ensures pending messages are always delivered. The instance
status override in ModelPickerGroup (`anyVariantHasInstance`) ensures
models with active instances bypass memory-based greying regardless of
available RAM.

## Test Plan

### Manual Testing
<!-- Hardware: (e.g., MacBook Pro M1 Max 32GB, Mac Mini M2 16GB,
connected via Thunderbolt 4) -->
<!-- What you did: -->
- Launch exo, navigate to Chat view
- Verify idle chat shows category recommendations (Coding, Writing,
Agentic, Biggest) based on available memory
- Type a message without selecting a model → verify auto-selection picks
the best tier model
- Launch a model, go to New Chat → verify the input bar shows the
running model name (not "SELECT MODEL")
- Launch a model, go to Home → verify model picker button shows the
running model
- While a model is downloading, send a message → verify download
progress appears and message is queued
- Open model picker with a model running → verify it is NOT greyed out
and is selectable
- Click "New Chat" while in a conversation → verify it resets to idle
with model selector

### Automated Testing
- Dashboard builds successfully (`cd dashboard && npm run build`)
- No TypeScript errors in the build output


🤖 Generated with [Claude Code](https://claude.com/claude-code)

---------

Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com>
2026-02-23 09:32:55 -08:00
rltakashige 05986f77aa add exo bench protobuf dependency (#1596)
kimi k2.5 requires protobuf
2026-02-23 16:45:22 +00:00
vskiwi dab7ed4821 fix: handle gossipsub MessageTooLarge error to prevent silent crash (#1583)
## Summary

Large prompts (70K+ tokens / ~500KB+ JSON) cause exo to silently crash.
The root cause is an unhandled `PublishError::MessageTooLarge` from
gossipsub when serialized `TextGeneration` commands exceed the 1MB
`max_transmit_size` limit.

The error propagates as a generic Python exception through the PyO3
bindings. Since `_networking_publish` in `router.py` only catches
`NoPeersSubscribedToTopicError` and `AllQueuesFullError`, the unhandled
exception crashes the networking async task, causing exo to shut down
silently — no error message, no API response.

## Changes

- **Rust (PyO3 bindings):** Add `MessageTooLargeError` exception class
and handle `PublishError::MessageTooLarge` explicitly in the gossipsub
publish path, matching the existing pattern for
`NoPeersSubscribedToTopicError` and `AllQueuesFullError`
- **Python (router):** Catch `MessageTooLargeError` in
`_networking_publish` and log a warning with the message size,
preventing the networking task from crashing

## Reproduction

On a multi-node cluster with a large model (e.g., GLM-5 754B tensor
parallel over JACCL RDMA):
1. Send a chat completion request with ~70K+ tokens
2. exo silently shuts down — no error logged, curl gets no response
3. With shorter prompts (< ~50K tokens): works fine

## Test plan

- Verified `cargo check` passes for `networking` and `exo_pyo3_bindings`
crates
- Verified `ruff check` passes for modified Python files
- Manual testing on 4× Mac Studio M3 Ultra cluster: 50K token requests
pass, 70K+ previously caused silent shutdown, now logs a warning and
drops the oversized message gracefully

Co-authored-by: vsm <vsm@nomail.com>
Co-authored-by: Cursor <cursoragent@cursor.com>
Co-authored-by: rltakashige <rl.takashige@gmail.com>
2026-02-23 16:21:21 +00:00
Evan Quiney 2261014715 runner process checks (#1592)
partial solve to some of the more mysterious failures

notably, runner now switches to failed after EXO RUNNER MUST OOM
2026-02-23 16:11:18 +00:00
Evan Quiney 61d2a2b6cf add lazy task group (#1569)
in a few places we instantiate task groups early, in others we have an
optional task group.

this standardizes these patterns into a lazy task group, which queues
tasks until it is entered, at which point it enters its inner task group
and starts all its queued tasks. it should be noted that no queued tasks
will be started until the group itself is started.
2026-02-23 16:05:59 +00:00
Evan Quiney 0ff99a2c40 fix isinstance for qwen3Moe (#1595)
we were checking if qwen3 had a transformers Qwen3DecoderLayer rather
than an mlx Qwen3MoeDecoderLayer causing an assertion error on loading
qwen models - this corrects it to the actual layer type
2026-02-23 15:39:13 +00:00
rltakashige fbb80e1cc9 Address ring slowdown by turning on FAST SYNCH (#1594)
## Motivation

Large models + large prompts + pipeline RING = 0.2tps generation speed
Large models + large prompts + pipeline JACCL = 15 tps generation speed

Why? Well, MLX_METAL_FAST_SYNCH is on in pipeline JACCL.

## Changes

Just turn on fast synch everywhere, especially as GPU locks are old news

Also, changed to use mx.device_info as mx.metal.device_info is going to
be deprecated.

## Why It Works

Some magic thing that happens in the mlx backend. I really tried to find
a regression but couldn't. I will probably try again at some point.

## Test Plan

### Manual Testing
Did a bunch, no longer 0.2tps

### Automated Testing
We'll do that today.
2026-02-23 15:14:58 +00:00
Jake Hillion 8d94eab6c6 bench: fix KeyError on DownloadCompleted total field
The bench harness accessed the serialized DownloadCompleted field as
"totalBytes", but the Python field is `total: Memory` which serializes
to "total" (not snake_case, so the camelCase alias generator leaves it
unchanged). This caused a KeyError when --danger-delete-downloads
needed to free disk space by deleting existing models.

Changed the key from "totalBytes" to "total" to match the actual
serialized JSON structure.

Test plan:
- CI
- Reproduced KeyError on unfixed code by filling disk on a test node
  to 14GB free and running exo-bench with a 16GB model
  (Meta-Llama-3.1-8B-Instruct-bf16) with --danger-delete-downloads
- Verified fixed code successfully deletes smaller models to free
  space and completes the benchmark run
2026-02-23 14:17:41 +00:00
Alex Cheema f370452d7e Better onboarding UX (#1533)
## Summary
- **Complete onboarding wizard**: 7-step flow guiding new users from
Welcome → Your Devices (topology) → Add More Devices (animation) →
Choose Model → Download → Load → Chat
- **Native macOS integration**: NSPopover welcome callout anchored to
menu bar icon on first launch, polished DMG installer with
drag-to-Applications arrow
- **Dashboard UX polish**: auto-download on model select, toast
notifications, connection banner, skeleton loading, download progress in
header, recommended model tags, sidebar hidden in home state for cleaner
first impression
- **Settings & menu bar overhaul**: native Settings window with Advanced
tab, onboarding reset, chat sidebar toggle

## Test plan
- [ ] Fresh install: verify onboarding wizard appears and flows Welcome
→ Topology → Animation → Model → Download → Load → Chat
- [ ] Verify topology shows real device data in onboarding step 2
- [ ] Verify selecting a model in the main dashboard picker
auto-triggers download
- [ ] Verify chat sidebar is hidden on home view, appears when chat is
active
- [ ] Verify DMG installer has white background with curved arrow
- [ ] Verify NSPopover appears anchored to menu bar icon on first launch

🤖 Generated with [Claude Code](https://claude.com/claude-code)

---------

Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com>
Co-authored-by: Ryuichi Leo Takashige <leo@exolabs.net>
2026-02-23 11:27:28 +00:00
Alex Cheema a4c2aa2b87 fix: raise error when MlxJaccl requested without RDMA cycles (#1585)
## Summary
- When MlxJaccl (RDMA) placement is requested but no RDMA-connected
cycles exist, raise a clear ValueError instead of silently falling back
to non-RDMA cycles
- Split from #1519 for independent review

## Test plan
- [x] basedpyright — 0 errors
- [x] ruff check — passes

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com>
2026-02-23 10:09:01 +00:00
Alex Cheema 7312c535b4 feat: add user context prompt and GitHub issue option to macOS bug report (#1544)
## Summary

- When clicking "Send Bug Report" in the macOS app, users are now
prompted with "What's the issue? (optional)" before the diagnostic
upload begins
- The user's description is included in the uploaded report JSON
(`user_description` field)
- After successful upload, a "Create GitHub Issue" button opens the
browser to `github.com/exo-explore/exo/issues/new` pre-filled with the
user's description, macOS version, and EXO version

## Changes

- **`ContentView.swift`**: Replaced simple button with multi-phase
inline UI (idle → prompting → sending → success/failure). Added
`openGitHubIssue()` helper using `URLComponents` for pre-filled GitHub
issue URLs.
- **`BugReportService.swift`**: Added `userDescription` parameter to
`sendReport()` and `makeReportJson()`, included in the JSON payload when
non-empty.
- Removed the previous `BugReportModal.svelte` approach (the feature
belongs in the macOS app, not the dashboard).

## Test plan

- [ ] Build the Xcode project and run the macOS app
- [ ] Click "Send Bug Report" → verify text prompt appears
- [ ] Type a description, click Send → verify upload succeeds and
"Create GitHub Issue" button appears
- [ ] Click "Create GitHub Issue" → verify browser opens with pre-filled
template
- [ ] Test Cancel returns to idle, test empty description still works
- [ ] Test failure case (e.g. with exo stopped) shows error and dismiss

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com>
2026-02-22 15:37:43 +00:00
Alex Cheema 18717023ad chore: remove deprecated MlxIbv dashboard references (#1584)
## Summary
- Remove legacy MlxIbvInstance references from ChatSidebar and ModelCard
components
- MlxIbv was replaced by MlxJaccl; these are leftover type checks
- Split from #1519 for independent review

## Test plan
- [x] Visual inspection of dashboard components

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com>
2026-02-22 06:56:12 -08:00
Alex Cheema 1780e4ade4 fix: change RDMA AVAILABLE to RDMA NOT ENABLED warning (#1580)
## Summary
- Changed blue info badge "RDMA AVAILABLE" to yellow warning badge "RDMA
NOT ENABLED" — more accurately describes the state
- Added hover tooltip with enable instructions to all views (was missing
in 2 of 4 instances)
- Warning icon instead of info icon, consistent with other cluster
warnings (TB cycle, macOS mismatch)

## Screenshots

**Badge (yellow warning):**
![RDMA warning
badge](https://raw.githubusercontent.com/exo-explore/exo/3f7bdb482c5011d60f140aa84ab21023032e4a57/rdma-warning.png)

**Hover tooltip with instructions:**
![RDMA warning
hover](https://raw.githubusercontent.com/exo-explore/exo/3f7bdb482c5011d60f140aa84ab21023032e4a57/rdma-warning-hover.png)

## Test plan
- [x] Dashboard builds successfully
- [ ] Verify badge appears when 2+ TB5 nodes have RDMA disabled
- [ ] Verify hover tooltip shows in normal layout
- [ ] Verify hover tooltip shows in topology-only mode
- [ ] Verify dismiss button works
- [ ] Verify compact badge in status bar shows yellow warning

🤖 Generated with [Claude Code](https://claude.com/claude-code)

---------

Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com>
Co-authored-by: rltakashige <rl.takashige@gmail.com>
2026-02-20 21:40:07 +00:00
Jake Hillion ab9273e723 downloads: add read_only flag to DownloadCompleted for EXO_MODELS_PATH
Models in EXO_MODELS_PATH are pre-downloaded into read-only directories
and must not be deleted. The DownloadCoordinator had no awareness of
these paths, so they never appeared as completed downloads in cluster
state, and the bench harness could attempt to delete them when freeing
disk space.

Added a `read_only: bool` field to `DownloadCompleted` (default False).
The DownloadCoordinator now checks `resolve_model_in_path` in
`_start_download`, proactively scans EXO_MODELS_PATH in
`_emit_existing_download_progress` to emit DownloadCompleted events for
all pre-downloaded models (overriding DownloadPending from the regular
scan), and refuses deletion of read-only models. The bench harness
filters out read-only models from deletion candidates.

Test plan:
- Ran with EXO_MODELS_PATH. Available models now show as downloaded in
  the UI. There isn't good UI for the fact they can't be deleted, but it
  should work with exo_bench.
2026-02-20 20:27:45 +00:00
Jake Hillion 71e48c0f62 model-cards: add missing metadata for Qwen3 Coder Next variants (#1576)
The Qwen3-Coder-Next model card TOML files were missing family,
quantization, base_model, and capabilities fields. This caused them not
to appear under the Qwen family filter in the dashboard's model picker.

Added the missing metadata to all five variants (4bit, 5bit, 6bit, 8bit,
bf16), matching the format used by the existing Qwen3-Coder-480B model
cards.

Test plan:
- Eyeballs
2026-02-20 18:25:49 +00:00
Jake Hillion 42da58c297 worker: add EXO_MODELS_PATH for pre-downloaded model directories
Users with pre-existing model files (e.g. on shared NFS mounts or from
prior downloads) had no way to point exo at those directories without
going through the download coordinator. EXO_MODELS_DIR only moves the
download target directory, it doesn't support read-only search paths.

Added EXO_MODELS_PATH environment variable as a colon-separated list of
directories to search for models. When the worker's plan loop encounters
a DownloadModel task, it checks these directories first and emits a
synthetic DownloadCompleted event if found, bypassing the download
coordinator entirely. The runner's build_model_path also checks these
directories first so the correct path is used during model loading.

This keeps the existing event sourcing state machine unchanged — the
DownloadCompleted event propagates naturally through the system, so
_load_model and all downstream logic work without modification.

Test plan:
- `s1@s1s-Mac-Studio ~ % EXO_LIBP2P_NAMESPACE=jake EXO_MODELS_PATH="/Volumes/Definitely Leo's SSD" nix --extra-experimental-features 'nix-command flakes' run github:exo-explore/exo/f2babbc2f742357d97dc177619fec062ef545be4`
- Started mlx-community/Qwen3-Coder-Next-4bit - it's present on the disk
  and it worked.
- Renamed one safetensor of mlx-community/Qwen3-Coder-Next-4bit on the
  disk. It then started the download locally, as expected.
2026-02-20 18:17:56 +00:00
Mustafa Alp Yılmaz 6b5a705959 fix: immediate cancel check after prefill completes (#1575)
## Problem

When a request is cancelled during prefill, the cancellation is not
detected until `check_for_cancel_every` additional tokens have been
generated. This is because `tokens_since_last_cancel_check` is
initialized to `0`, meaning the first cancel check only happens after
generating `check_for_cancel_every` tokens post-prefill.

For long prefills (which are the most likely to be cancelled), this adds
unnecessary latency before the cancellation is actually honoured.

## Fix

Initialize `tokens_since_last_cancel_check` to `check_for_cancel_every`
instead of `0`, so the very first token generated after prefill triggers
an immediate cancel check.

```diff
- tokens_since_last_cancel_check = 0
+ tokens_since_last_cancel_check = check_for_cancel_every
```

## Impact

- Cancellations issued during prefill are detected immediately when
generation begins
- No change in behaviour for non-cancelled requests (the counter resets
to `0` after each check as before)
- 1 line changed

Co-authored-by: rltakashige <rl.takashige@gmail.com>
2026-02-20 18:00:59 +00:00
Alex Cheema 6b54a27019 fix: add downloaded_bytes to DownloadPending event (#1564)
## Summary
- Add downloaded_bytes field to existing DownloadPending event for
accurate resume progress
- Minimal change per maintainer directive — no new download states
introduced

## Test plan
- [x] 42 tests passed, 1 skipped
- [x] Verified downloaded_bytes populates correctly for partial
downloads

🤖 Generated with [Claude Code](https://claude.com/claude-code)

---------

Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com>
Co-authored-by: rltakashige <rl.takashige@gmail.com>
2026-02-20 17:54:18 +00:00
rltakashige e01f50a5cd Update mlx fork (#1565)
## Motivation

Some fixes upstream. This sort of commit will probably be quite common
until GPU locks are resolved.
2026-02-20 17:23:52 +00:00
Evan Quiney 1093080214 cancel active downloads on coordinator shutdown (#1567)
we were seeing some crashes as lost download tasks were trying to push
data toward a deleted coordinator. this cancels download tasks with the
coordinator's shutdown on master election
2026-02-20 17:17:43 +00:00
rltakashige 1a2b8b044a Refactor runner into separate runners (#1570)
## Motivation

We're going to be refactoring the llm inference code, so we should split
the runner up into parts while we can.

## Test Plan

### Manual Testing
Works on single node, at least.

### Automated Testing
Passes CI. Will be tested by our tests today.
2026-02-20 17:11:01 +00:00
Evan Quiney dc8d42b4dc add system ids (#1536)
addresses some election edge cases where a new worker with an old master
would get stuck on the old workers buffer index - we now use new system
ids each time we instantiate a node, and each event-producing system has
a unique system id for its lifespan (until the master moves).
2026-02-20 15:41:59 +00:00
Jake Hillion d484b062e8 bench: add download timing to bench output (#1566)
The bench script downloads models during the planning phase but doesn't
record how long the download took, making it difficult to track download
performance for a given model over time.

Modified `run_planning_phase` to return download metadata: whether a
fresh download occurred, the wall-clock duration, and the model size in
bytes. These fields are included in every JSON output row alongside the
existing per-run metrics, and a summary line is logged to the console.

This allows filtering bench results by `download_occurred` and grouping
by `model_id` to compute average download times across runs.

Test plan:

```
# existing model
jake@maverick:/data/users/jake/repos/exo/ > nix run .#exo-bench -- --host s1 --model mlx-community/gpt-oss-120b-MXFP4-Q8 --pp 128 --tg 128
...
2026-02-20 15:23:49.081 | INFO     | __main__:main:340 - Planning phase: checking downloads...
2026-02-20 15:23:49.152 | INFO     | harness:run_planning_phase:402 - Started download on 12D3KooWKx41iikn188ozrxSdoG26g88jFCfie9wEA1eQR8csbPm
2026-02-20 15:23:49.184 | INFO     | __main__:main:352 - Download: model already cached
...
Wrote results JSON: bench/results.json
jake@maverick:/data/users/jake/repos/exo/ > cat bench/results.json
[
  {
    "elapsed_s": 2.9446684420108795,
    "output_text_preview": "The user just typed a long series of \"a\". Possibly they are testing. There's no explicit question. Could be they want a response? Might be a test of handling long input. We can respond politely, ask i",
    "stats": {
      "prompt_tps": 117.7872141515621,
      "generation_tps": 85.49598231498028,
      "prompt_tokens": 129,
      "generation_tokens": 128,
      "peak_memory_usage": {
        "inBytes": 68215145744
      }
    },
    "model_short_id": "gpt-oss-120b-MXFP4-Q8",
    "model_id": "mlx-community/gpt-oss-120b-MXFP4-Q8",
    "placement_sharding": "Pipeline",
    "placement_instance_meta": "MlxRing",
    "placement_nodes": 1,
    "instance_id": "68babc2a-6e94-4c70-aa07-7ec681f7c856",
    "pp_tokens": 128,
    "tg": 128,
    "repeat_index": 0
  }
]%
# no change to output
```

```
# missing model
jake@maverick:/data/users/jake/repos/exo/ > nix run .#exo-bench -- --host s1 --model mlx-community/Meta-Llama-3.1-8B-Instruct-4bit --pp 128 --tg 128
...
2026-02-20 15:24:42.553 | INFO     | __main__:main:340 - Planning phase: checking downloads...
2026-02-20 15:24:42.625 | INFO     | harness:run_planning_phase:402 - Started download on 12D3KooWKx41iikn188ozrxSdoG26g88jFCfie9wEA1eQR8csbPm
2026-02-20 15:25:37.494 | INFO     | __main__:main:350 - Download: 54.9s (freshly downloaded)
...
Wrote results JSON: bench/results.json
jake@maverick:/data/users/jake/repos/exo/ > cat bench/results.json
[
  {
    "elapsed_s": 1.500349276990164,
    "output_text_preview": "It seems like you've entered a large number of 'a's. If you'd like to discuss something or ask a question, I'm here to help. If not, is there anything else I can assist you with? \n\nIf you're intereste",
    "stats": {
      "prompt_tps": 395.43264952543666,
      "generation_tps": 128.03520443181478,
      "prompt_tokens": 129,
      "generation_tokens": 128,
      "peak_memory_usage": {
        "inBytes": 5116952079
      }
    },
    "model_short_id": "Meta-Llama-3.1-8B-Instruct-4bit",
    "model_id": "mlx-community/Meta-Llama-3.1-8B-Instruct-4bit",
    "placement_sharding": "Pipeline",
    "placement_instance_meta": "MlxRing",
    "placement_nodes": 1,
    "instance_id": "ccd9bd71-d4cc-4b75-a37f-98090544626a",
    "pp_tokens": 128,
    "tg": 128,
    "repeat_index": 0,
    "download_duration_s": 54.88322358299047
  }
]%
# one new field
```
2026-02-20 15:33:08 +00:00
Alex Cheema e32b649d2f fix: enable psutil fallback for memory monitoring when macmon is missing (#1478)
## Summary
- On macOS, memory monitoring relied exclusively on `macmon` — the
psutil fallback was explicitly disabled (`memory_poll_rate = None`)
- When `macmon` is not installed (e.g., mac-mini-2 through mac-mini-4 in
our cluster), **no memory data was reported**, causing nodes to show 0GB
memory in the cluster state
- This blocked the scheduler from placing shards on those nodes since it
had no memory data to work with
- Fix: when `macmon` is not found on Darwin, fall back to psutil-based
memory polling (`memory_poll_rate = 1`)

## Root cause
`InfoGatherer` has two memory monitoring paths:
1. `macmon` (Darwin-only): provides memory + GPU/CPU/power stats
2. `psutil` (non-Darwin fallback): provides memory via
`MemoryUsage.from_psutil()`

Line 378 disabled psutil on Darwin: `memory_poll_rate = None if
IS_DARWIN else 1`
Line 389 only starts macmon if the binary exists: `if
shutil.which("macmon") is not None`

If macmon is missing on Darwin, **neither path runs** — zero memory
reported.

## Test plan
- [ ] Verify `uv run basedpyright` passes (0 errors confirmed)
- [ ] Verify `uv run ruff check` passes (confirmed)
- [ ] Verify `uv run pytest src/exo/utils/info_gatherer/` passes (2/2
confirmed)
- [ ] Deploy to cluster nodes without macmon and verify memory appears
in `/state`

🤖 Generated with [Claude Code](https://claude.com/claude-code)

---------

Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com>
Co-authored-by: rltakashige <rl.takashige@gmail.com>
2026-02-20 13:12:22 +00:00
Alex Cheema bddad7e79c feat: show ETA on prefill progress bar (#1557)
## Summary
- Show estimated time remaining during prefill (prompt processing phase)
- Track prefill start time via performance.now() and extrapolate from
observed token throughput
- Display ~Xs remaining or ~Xm Ys remaining next to the percentage on
the progress bar
- Wait 200ms before showing ETA to ensure a stable sample window

## Changes
**PrefillProgressBar.svelte**: Add etaText derived computation that
calculates remaining time from (remainingTokens / tokensPerMs). Renders
in a new flex row below the progress bar alongside the percentage.

**app.svelte.ts**: Add startedAt: number field to PrefillProgress
interface. Set on first prefill_progress SSE event, preserved across
subsequent updates.

## Test plan
- [ ] Start inference with a long prompt (10k+ tokens) on a multi-node
cluster
- [ ] Verify the progress bar shows ~Xs remaining after ~200ms of
prefill
- [ ] Verify the ETA decreases as prefill progresses
- [ ] Verify short prefills (<200ms) dont flash a briefly-visible ETA
- [ ] Verify ETA disappears when prefill completes and token generation
begins

Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com>
Co-authored-by: rltakashige <rl.takashige@gmail.com>
2026-02-20 12:37:56 +00:00
rltakashige addf73a144 Add support for Ollama API (#1560)
## Motivation

Ollama has a bunch of integrations, such as OpenWebUI, that are very
handy. Let's support it :)

## Test Plan

### Manual Testing
<img width="3426" height="1998" alt="image"
src="https://github.com/user-attachments/assets/44b07f1e-308e-4ff1-9a11-922d8279939f"
/>
2026-02-20 12:03:27 +00:00
Mustafa Alp Yılmaz a16ff2c047 fix: correct misleading docstring in seed_models (#1561)
## Summary
- Fixed stale docstring in `seed_models()` that referenced
`.cache/huggingface/hub` when the function actually moves models to
`EXO_MODELS_DIR` (resolved via `ensure_models_dir()`)
- The old docstring was misleading for AI coding agents analyzing the
codebase, causing incorrect conclusions about model storage paths

## Changes
`src/exo/download/download_utils.py`: Updated docstring from `"Move
model in resources folder of app to .cache/huggingface/hub"` to `"Move
models from resources folder to EXO_MODELS_DIR."`

Co-authored-by: rltakashige <rl.takashige@gmail.com>
2026-02-20 11:57:55 +00:00
rltakashige 3006c8ea4e Ensure coordinator is rank 0 (#1559)
## Motivation

Coordinator can be a random rank. Let's just fix this to rank 0 as
that's what we typically assume.

## Test Plan

### Manual Testing
Works as normal on 2 nodes.


Let's wait for a little more testing to merge this.

---------

Co-authored-by: Evan <evanev7@gmail.com>
2026-02-20 11:46:24 +00:00
rltakashige f662c129dd Prioritise tb for ring instances (#1556)
## Motivation

TB has better bandwidth and latency than ethernet. We should prioritise
TB5 where possible. This drastically improves distributed image
generation performance.

## Test Plan

### Manual Testing
Saw on the dashboard that TB (169.254) addresses were prioritised.

Tested that image models scale much better.

### Automated Testing
No regression on Kimi K2.5
2026-02-19 21:32:48 +00:00
Evan Quiney c45ff9ad43 memory tidy (#1558)
add some pythonic extensions to memory, did a bunch of cleanup.
2026-02-19 21:15:33 +00:00
rltakashige 7031901ae5 Prevent common fatal crashes (#1555)
## Motivation
Occasionally, memory does not get released when we shut down. There is
no reason to delay deleting the model.

Also handles can become None during shutdown, causing TypeErrors which
are not handled and bringing down exo.

Similarly, we were closing the event sender in the wrong place.

Also let's not verify the SSL certificate for http connections to local
peers, as this is failing sometimes and crashing.

## Test Plan

### Manual Testing
No more crashes as described.
2026-02-19 20:51:17 +00:00
rltakashige cf648a53b8 Add thinking in thinking blocks, and fix DeepSeek interleaved tool calls (#1548)
## Motivation

OpenCode shows <think> tags and not thinking blocks as we aren't
following the API specs properly.

Claude was also getting horrible prefix cache hits because it sends
headers.

## Changes

Handle thinking tokens properly by placing them in think tags for each
of the API endpoints.
Also support DeepSeekV3.2 tool calling properly as a minor feature.
Strips Claude headers at the API level.

## Test Plan

### Manual Testing
Tested OpenCode manually
Needs testing with Claude.

### Automated Testing
All CI and tests passing - added a new e2e test for DeepSeekV32 tool
parsing.
2026-02-19 18:44:49 +00:00
Alex Cheema 94b2ce6922 feat: Mac Studio en2 RDMA port warning v2 (#1551)
Rebuilt from scratch (replaces PR #1543). Detects when Mac Studio uses
RDMA over en2 (TB5 port next to Ethernet) which does not support RDMA.
Shows dismissible warning banner with hover tooltip showing affected
devices, SVG rear panel illustration, and fix instructions. 205 lines in
+page.svelte.

---------

Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com>
Co-authored-by: rltakashige <rl.takashige@gmail.com>
2026-02-19 18:39:17 +00:00
rltakashige 423ed0f07f Strip Claude headers to improve prefix cache hit rates (#1552)
## Motivation
Our hits are really bad at the moment (0.2%). This PR makes it 98.5% on
average.

## Changes

Also adds an example for how to run Claude using Exo.

## Why It Works
Claude sends some billing and session headers that change with each
message.

## Test Plan

### Manual Testing
Works in manual testing.
2026-02-19 18:29:34 +00:00
Evan Quiney ed001f2409 remove prefillprogress event (#1550)
this should never have been a separate event, but i didnt quite
communicate that well when this was merged. convert PrefillProgress to a
chunk like the rest of the runner responses.

tested with Llama-3.3-70B, prefill progress events still show up in the
dashboard as usual
2026-02-19 18:23:28 +00:00
Evan Quiney 4c4c6ce99f simplify rust ident module
this is partly dead code, partly narrowing the rust-python boundary in
prep for future rewrites. no testing as this is all type safe
refactoring.
2026-02-19 17:19:31 +00:00
Jake Hillion 42e1e7322b bench: restore --danger-delete-downloads planning phase (#1542)
c2f2111b extracted shared utilities from exo_bench.py into harness.py
but accidentally dropped the run_planning_phase function and
--danger-delete-downloads CLI argument in the process.

Restored run_planning_phase in harness.py (where its dependencies now
live) and re-added the --danger-delete-downloads argument to
add_common_instance_args. Re-wired the planning phase call in
exo_bench.py's main() before the benchmark loop.
2026-02-19 15:42:02 +00:00
Alex Cheema aa3f106fb9 fix: import ResponsesStreamEvent and DRY up SSE formatting (#1499)
## Summary
- `ResponsesStreamEvent` was defined in `openai_responses.py` as a union
of all 11 streaming event types but never imported or used anywhere in
the codebase
- Import it in the responses adapter and add a `_format_sse(event:
ResponsesStreamEvent) -> str` helper
- Replace 13 hardcoded `f"event: {type}\ndata:
{event.model_dump_json()}\n\n"` strings with `_format_sse()` calls

## Test plan
- [x] `uv run basedpyright` — 0 errors
- [x] `uv run ruff check` — all checks passed
- [x] `nix fmt` — 0 files changed
- [x] `uv run pytest` — 188 passed, 1 skipped

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com>
2026-02-19 13:40:24 +00:00
Mustafa Alp Yılmaz 2e29605194 fix: finalize cancel tasks (#1498)
# Cancel task finalization (main.py)

After forwarding the cancel to the runner supervisor, emit TaskStatusUpdated(Complete) for the cancel task itself. This ensures the cancel task is properly removed from state.tasks.
2026-02-19 13:27:34 +00:00
Evan Quiney cacb456cb2 remove nightly (#1538)
we have no good need for rust nightly (nor futures, for that matter)
2026-02-19 12:55:31 +00:00
rltakashige 51021f6fc6 Add cancellation button and the ability to cancel during prefill (#1540)
## Motivation
There's no way to easily use the cancellation features we added! Also,
prefill can take ages so let's allow cancelling out of that.

## Changes

Wiring up our existing functionality to easily cancel during generation
(and adding stuff to do so during prefill)

## Test Plan

### Manual Testing
Tested it works during both prefill and decode.

### Automated testing
Needs testing to see if this causes a GPU timeout error on large prefill
on large models in pipeline parallel. However, from manually testing GLM
5 pipeline ring on 2 nodes, and from reading the code, it does not seem
like this will be the case.
2026-02-19 11:40:59 +00:00
Alex Cheema 025ed9fd82 feat: add prefill progress bar for long prompts (#1181)
## Motivation

Users processing long prompts have no visibility into when token
generation will start. This feature adds a progress bar showing prefill
progress, giving users real-time feedback during prompt processing.

## Changes

### Backend
- Added `PrefillProgress` event type with `command_id`,
`processed_tokens`, `total_tokens`
- Added `PrefillProgressResponse` type (though now using direct callback
approach)
- Wired `prompt_progress_callback` through MLX's `stream_generate()`
- Progress events sent directly from callback for real-time updates (not
batched)
- API generates SSE named events: `event: prefill_progress\ndata: {...}`
- Added `PrefillProgressData` dataclass and `StreamEvent` union type in
API

### Dashboard
- Added `PrefillProgress` interface to store
- Updated SSE parsing to handle `event:` lines (named events)
- Created `PrefillProgressBar.svelte` with animated progress bar
- Shows "Processing prompt: X/Y tokens" with percentage
- Progress bar disappears when first token arrives

## Why It Works

MLX's `stream_generate()` accepts a `prompt_progress_callback(processed,
total)` that's called after each prefill chunk. By sending events
directly from this callback (rather than yielding from the generator),
progress updates are sent in real-time during prefill.

Using SSE named events (`event: prefill_progress`) maintains full
OpenAI/Claude API compatibility - standard clients ignore named events
they don't recognize, while the exo dashboard explicitly listens for
them.

## Test Plan

### Manual Testing
- Hardware: MacBook Pro M3 Max
- Set `prefill_step_size=256` for more frequent updates
- Tested with long prompts (pasted large documents)
- Verified progress bar updates incrementally during prefill
- Confirmed progress bar disappears when generation starts
- Tested with curl - standard `data:` events still work normally

Here is it working:


https://github.com/user-attachments/assets/5cc6f075-c5b2-4a44-bb4d-9efb246bc5fe


### Automated Testing
- Type checker passes (0 errors)
- All 192 tests pass
- Dashboard builds successfully

### API Compatibility
- Named SSE events are ignored by OpenAI SDK clients
- Regular token data uses standard `data: {...}` format
- `[DONE]` sentinel works as expected

---

**Note:** `prefill_step_size` is temporarily set to 256 for testing.
Should be changed back to 2048 before merging for production
performance.

---------

Co-authored-by: Claude Opus 4.5 <noreply@anthropic.com>
Co-authored-by: Evan <evanev7@gmail.com>
Co-authored-by: Ryuichi Leo Takashige <leo@exolabs.net>
2026-02-19 03:18:25 +00:00
rltakashige 19bc09550d Add status=downloaded filter for model endpoint (#1539)
## Motivation

https://github.com/exo-explore/exo/issues/1346#issuecomment-3831427905


## Test Plan

### Manual Testing
**Without filter**
<img width="1708" height="1010" alt="Screenshot 2026-02-18 at 22 26 22"
src="https://github.com/user-attachments/assets/f4bf7142-717d-4042-ac28-d8a55a8e45e7"
/>

**With filter**
<img width="1723" height="1021" alt="Screenshot 2026-02-18 at 22 26 45"
src="https://github.com/user-attachments/assets/40a522d5-c6e6-4148-b21a-02caa1221ebe"
/>
2026-02-18 22:34:11 +00:00
Alex Cheema 7cadca4f27 Try multiple endpoints for internet connectivity check (#1516)
## Summary
- `_test_internet_connection()` previously only tried `1.1.1.1:443`,
which some ISPs/networks block, causing exo to incorrectly report no
internet and fail downloads on startup
- Now tries `1.1.1.1`, `8.8.8.8`, and `1.0.0.1` in sequence, succeeding
if any endpoint responds
- Returns early on first success for minimal latency in the common case

Fixes #1425

## Test plan
- [ ] Verify downloads work on networks that block `1.1.1.1`
- [ ] Verify existing behavior unchanged on networks where `1.1.1.1`
works
- [ ] Verify `internet_connection` is set to `False` only when all three
endpoints fail

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com>
Co-authored-by: rltakashige <rl.takashige@gmail.com>
2026-02-18 22:10:07 +00:00
rltakashige 24e99ce197 Cleanup mistakes (#1537)
Oops
2026-02-18 22:05:26 +00:00
Alex Cheema 315992549b fix: unblock MpReceiver.close() to prevent shutdown hang (#1511)
## Summary

- `MpReceiver.close()` did not unblock threads stuck on `queue.get()` in
`receive_async()`, causing abandoned threads (via
`abandon_on_cancel=True`) to keep the Python process alive indefinitely
after tests pass
- This caused the `aarch64-darwin` CI jobs in PR #1462 to hang for ~6
hours until the GitHub Actions timeout killed them
- Sends an `_MpEndOfStream` sentinel before closing the buffer,
mirroring what `MpSender.close()` already does

## Test plan

- [x] `uv run basedpyright` — 0 errors
- [x] `uv run ruff check` — clean
- [x] `nix fmt` — 0 changed
- [x] `uv run pytest` — 188 passed, 1 skipped in 12s (no hang)

🤖 Generated with [Claude Code](https://claude.com/claude-code)

---------

Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com>
Co-authored-by: rltakashige <rl.takashige@gmail.com>
Co-authored-by: Ryuichi Leo Takashige <leo@exolabs.net>
2026-02-18 21:59:02 +00:00
Alex Cheema ce5a65d3b9 Add MiniMax M2.5 model cards (#1514)
## Summary
- Adds model cards for MiniMax M2.5 in three quantizations: 4bit (~129
GB), 6bit (~186 GB), 8bit (~243 GB)
- No code changes needed — `MiniMaxM2ForCausalLM` is already in the
tensor parallel whitelist and `MiniMaxShardingStrategy` is already
implemented in `auto_parallel.py`
- Credit to @vskiwi for confirming MiniMax M2.5 works out of the box
with existing code

Closes #1480

## Test plan
- [x] `basedpyright` passes with 0 errors
- [x] `ruff check` passes
- [x] `pytest` passes (260 passed, 1 skipped)
- [ ] Verify MiniMax M2.5 models appear in model selector on dashboard

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com>
Co-authored-by: rltakashige <rl.takashige@gmail.com>
2026-02-18 21:11:13 +00:00
rltakashige c2f2111b88 Fix tool calling (#1529)
## Motivation

GPT OSS tool calling issues.

## Changes

Fixes those and adds a bunch of evals for tool calling.
Fixes GLM5 prefix caching, where CacheList wasn't getting handled
properly.
Extracts a bunch of the setup functionality of exo bench to a harness
that can be reused elsewhere, such as in the tool calling eval.

## Test Plan
### Automated Testing
Let's run the evals for all models
2026-02-18 20:29:18 +00:00
Alex Cheema 6c322ebb72 feat: only show thinking toggle for models that support it (#1497)
## Summary
- Adds `thinking_toggle` capability to 26 model cards that support
toggling thinking mode on/off
- GPT-OSS models (20b, 120b) excluded — they always think and don't
support toggling
- Dashboard UI updated to check for `thinking_toggle` capability before
showing the toggle button

## Test plan
- [x] `uv run basedpyright` — 0 errors
- [x] `uv run ruff check` — all checks passed
- [x] `nix fmt` — 0 files changed
- [x] `uv run pytest` — 188 passed, 0 failed
- [x] Security review passed (no secrets, eval/exec, innerHTML, or dep
changes)

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com>
2026-02-18 17:05:00 +00:00
vskiwi 2ebe6216b4 feat: add explicit --offline mode for air-gapped clusters (#1525)
## Motivation

Closes #1510

There is currently no reliable way to run exo on an air-gapped or offline cluster where models are pre-staged on local disks. The two existing mechanisms — `--no-downloads` and `HF_HUB_OFFLINE=1` — each cover only a subset of the problem:

1. **`--no-downloads` blocks model loading**: When passed, `DownloadCoordinator` is not created. No `NodeDownloadProgress` events are ever emitted, so `_model_needs_download()` in `plan.py` perpetually returns `DownloadModel`, short-circuiting `_load_model()` and preventing the model from ever being loaded.

2. **`HF_HUB_OFFLINE=1` doesn't cover exo's aiohttp code**: exo's download pipeline primarily uses raw `aiohttp` for HTTP operations (file list fetching, file downloads, HEAD verification), not the `huggingface_hub` library. These calls will attempt connections and time out on air-gapped networks.

3. **`skip_internet` is not propagated to `download_file_with_retry()`**: Even when `internet_connection = False`, the `_download_file()` function still makes HTTP HEAD calls via `file_meta()` to verify local files and unconditionally attempts downloads for missing files.

## Changes

### `src/exo/main.py`
- Add `--offline` flag to `Args` with env var detection (`EXO_OFFLINE=1`, `HF_HUB_OFFLINE=1`)
- Pass `offline` to `DownloadCoordinator` at creation and re-creation (election loop)

### `src/exo/download/coordinator.py`
- Add `offline: bool = False` field
- In offline mode: set `internet_connection = False` immediately in `__post_init__`, skip `_test_internet_connection()` ping (avoids 3s timeout), skip `_check_internet_connection` periodic loop
- In `_start_download()`: if model is not fully available locally, emit `DownloadFailed` with clear message instead of starting a download task

### `src/exo/download/download_utils.py`
- Add `skip_internet: bool` parameter to `download_file_with_retry()` and `_download_file()`
- When `skip_internet=True` in `_download_file()`: return local file immediately without HTTP HEAD verification; raise `FileNotFoundError` for missing files
- Propagate `skip_internet` from `download_shard()` to `download_file_with_retry()`

### `src/exo/download/tests/test_offline_mode.py` (new)
- 8 tests covering `_download_file`, `download_file_with_retry`, and `fetch_file_list_with_cache` in offline mode

## Why It Works

Unlike `--no-downloads` which disables `DownloadCoordinator` entirely, `--offline` keeps the coordinator running in a restricted mode. The existing `_emit_existing_download_progress()` disk scanner still runs every 60 seconds, emitting `DownloadCompleted` events for pre-staged models. These events flow through the event-sourcing pipeline and populate `state.downloads`, which unblocks `_model_needs_download()` in `plan.py` — no changes to the planning logic required.

```
--offline flag
  → DownloadCoordinator (offline mode)
    → Skip 1.1.1.1 ping, internet_connection = False
    → _emit_existing_download_progress scans disk
      → Emits DownloadCompleted for pre-staged models
        → _model_needs_download sees DownloadCompleted
          → _load_model proceeds normally
```

## Test Plan

### Automated Testing
- `ruff check` — passes
- 8 new tests in `test_offline_mode.py` — all pass
- 11 existing download tests in `test_download_verification.py` — all pass (no regressions)

### Manual Testing
1. Pre-stage a model on disk (e.g., `~/.exo/models/mlx-community--Qwen3-0.6B-4bit/`)
2. Start exo with `--offline` (or `EXO_OFFLINE=1`)
3. Place an instance via API or dashboard
4. Verify: model loads into memory and inference works without any network calls

### Environment
- macOS (Apple Silicon), multi-node cluster with Thunderbolt interconnect
- Models pre-staged via rsync / NFS mount
2026-02-18 16:18:09 +00:00
ciaranbor f54c80b121 Ciaran/image edit api (#1500)
## Motivation

- Image editing previously ignored input image dimensions, always
defaulting to 1024x1024
- Size dropdown was hidden in edit mode, giving users no control over
output dimensions
- Portrait/landscape presets used non-standard aspect ratios (1024x1365
/ 1365x1024)

## Changes

- Added "auto" size option that uses input image dimensions for edits,
defaults to 1024x1024 for generation
- Introduced ImageSize Literal type and normalize_image_size() validator
(replaces raw str size fields)
  - Updated portrait/landscape presets to standard 1024x1536 / 1536x1024
  - Made size selector visible in edit mode (previously hidden)
  - Default size changed from "1024x1024" to "auto"

## Why It Works

- "auto" reads actual input image dimensions via PIL at generation time,
so edits preserve the original aspect ratio
- Pydantic field_validator on both ImageGenerationTaskParams and
ImageEditsTaskParams normalizes None → "auto", keeping the API
backward-compatible

## Test Plan

### Manual Testing

- Verify image edits output at the input image's native resolution when
size is "auto"
- Verify size dropdown appears and works in both generate and edit modes
2026-02-18 16:05:39 +00:00
rltakashige 48b8f86395 Add support for GLM 5 (#1526)
## Motivation

Add GLM 5 support in favor of #1513 

## Changes

<!-- Describe what you changed in detail -->

## Why It Works

<!-- Explain why your approach solves the problem -->

## Test Plan

### Manual Testing
<!-- Hardware: (e.g., MacBook Pro M1 Max 32GB, Mac Mini M2 16GB,
connected via Thunderbolt 4) -->
<!-- What you did: -->
<!-- - -->

### Automated Testing
<!-- Describe changes to automated tests, or how existing tests cover
this change -->
<!-- - -->
2026-02-18 14:04:06 +00:00
Evan 5cbd6377a2 prioritize official model cards over custom model cards
our old model card search path would override official model cards with
custom model cards - our packaged model cards should always be the
default here
2026-02-18 13:20:05 +00:00
Evan Quiney 8f01523ddb remove dead code (#1496) 2026-02-18 11:43:27 +00:00
Alex Cheema 3addeadea8 Update mlx-lm to 0.30.7 (#1520)
## Summary
- Bumps `mlx-lm` from 0.30.6 to 0.30.7 in `pyproject.toml` and `uv.lock`

## Test plan
- [x] `uv lock` resolves successfully
- [x] `basedpyright` — no new errors (63 pre-existing in unrelated
`test_tool_call_tracker.py`)
- [x] `ruff check` — all checks passed
- [x] `nix fmt` — no formatting changes
- [x] `pytest` — 188 passed, 1 skipped

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com>
2026-02-18 11:14:23 +00:00
rltakashige f2be929211 Leo/address rdma gpu locks 2 (#1515)
Same as #1489 . Had to revert and redo thanks to Claude.

---------

Co-authored-by: Jake Hillion <jake@hillion.co.uk>
Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com>
2026-02-17 14:00:52 -08:00
rltakashige 83af8c63fa Revert "Use custom fork that resolves GPU locks" (#1502)
Reverts exo-explore/exo#1489

Goddammit Claude...
2026-02-17 18:18:54 +00:00
Evan Quiney eccc6298d1 Revert "Add MetaInstance declarative layer (#1447)"
This reverts commit a962a28afc.
2026-02-17 18:11:47 +00:00
Evan Quiney c8997217cf Revert "feat: better onboarding UX for new users (#1479)"
This reverts commit 490d2e46ba.
2026-02-17 18:02:32 +00:00
Alex Cheema 490d2e46ba feat: better onboarding UX for new users (#1479)
## Summary

- **Auto-open dashboard** in browser on first launch (uses
`~/.exo/.dashboard_opened` marker)
- **Welcome overlay** with "Choose a Model" CTA button when no model
instance is running
- **Tutorial progress messages** during model download → loading → ready
lifecycle stages
- **Fix conversation sidebar** text contrast — bumped to white text,
added active state background
- **Simplify technical jargon** — sharding/instance type/min nodes
hidden behind collapsible "Advanced Options" toggle; strategy display
hidden behind debug mode
- **Polished DMG installer** with drag-to-Applications layout, custom
branded background, and AppleScript-configured window positioning

## Test plan

- [ ] Launch exo for the first time (delete `~/.exo/.dashboard_opened`
to simulate) — browser should auto-open
- [ ] Verify welcome overlay appears on topology when no model is loaded
- [ ] Launch a model and verify download/loading/ready messages appear
in instance cards
- [ ] Check conversation sidebar text is readable (white on dark, yellow
when active)
- [ ] Verify "Advanced Options" toggle hides/shows sharding controls
- [ ] Build DMG with `packaging/dmg/create-dmg.sh` and verify
drag-to-Applications layout

🤖 Generated with [Claude Code](https://claude.com/claude-code)

---------

Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com>
2026-02-17 17:52:49 +00:00
rltakashige facf2d4d03 Use custom fork that resolves GPU locks (#1489)
## Motivation

There is an issue on Macs that means that an explicit synchronization is
necessary for memory to be updated from L1 cache. This means that GPU
locks can occur when a spin wait does not see the updated timestamp.

## Changes

Updated in my own personal fork.

## Why It Works

https://github.com/ARM-software/acle/releases

## Test Plan

### Manual Testing
Tested manually that no GPU locks occur (even with multiple simultaneous
instances running) and that the performance differential is negligible
(267 vs 269 tps on Llama 3.2 1B at an approx 10k context.)


------------------------------------------------------
I have seen a GPU lock, specifically when sending a particularly large
chat completion while the model was loading. However, I have since been
unable to reproduce and this may be something I did wrong. Please do
create an issue and tag me if any GPU locks do occur.

---------

Co-authored-by: Jake Hillion <jake@hillion.co.uk>
Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com>
2026-02-17 17:48:43 +00:00
Alex Cheema a962a28afc Add MetaInstance declarative layer (#1447)
## Motivation

Users currently manage instances directly, which means if a node
disconnects or connections break, the instance dies and nothing
recreates it. MetaInstance is a declarative primitive: "ensure an
instance matching these parameters always exists." The reconciler
watches for unhealthy or missing backing instances and re-places them
automatically.

## Changes

- **MetaInstance type** (`meta_instance.py`): declarative constraint
with `model_id`, `min_nodes`, optional `node_ids`, and `sharding`
- **Reconciler** (`reconcile.py`): `find_unsatisfied_meta_instances`
checks which MetaInstances lack a healthy backing instance,
`try_place_for_meta_instance` creates one
- **Master loop** (`main.py`): periodically reconciles unsatisfied
MetaInstances; immediate placement on `CreateMetaInstance` command
- **API** (`api.py`): `create_meta_instance` / `delete_meta_instance` /
`GET /meta_instances` endpoints; delete cascades to backing instances
with task cancellation
- **Binding via `meta_instance_id` on Instance** (`instances.py`): no
separate binding event or backing map — the instance carries its parent
MetaInstance ID directly, eliminating race conditions in the reconciler
- **Dashboard**: sidebar shows MetaInstances with their backing instance
status; orphan instances (created directly) still shown separately
- **Tests**: constraint matching, connection health, unsatisfied
detection, exclusive binding, cascade delete with task cancellation

### Recent improvements

- **fix: cancel active tasks on cascade delete** — `DeleteMetaInstance`
now emits `TaskStatusUpdated(Cancelled)` for any Pending/Running tasks
on backing instances before emitting `InstanceDeleted`. Previously,
cascade-deleting backing instances left orphaned task references in
state.
- **Lifecycle logging** — added `logger.info`/`logger.warning` for:
`CreateMetaInstance` (model, min_nodes, sharding), `DeleteMetaInstance`
(with cascade count), reconciler placement success/failure, and retry
decisions with attempt counts in `InstanceHealthReconciler`.
- **GET `/meta_instances` endpoint** — lists all meta-instances without
needing to fetch full state.
- **2 regression tests** — `test_cascade_delete_cancels_active_tasks`
and `test_cascade_delete_skips_completed_tasks` verify the
cascade-delete event sequence.

## Why It Works

Putting `meta_instance_id` on `BaseInstance` makes binding inherent to
instance creation. When the reconciler creates an instance for a
MetaInstance, it tags it via `model_copy`. When the instance is deleted,
the binding disappears with it. This avoids the two bugs that a separate
binding mechanism would introduce:
1. Stale exclusion sets — the reconciler loop can't accidentally bind
two MetaInstances to the same instance
2. Delete ordering race — no window between deleting an instance and its
binding where the reconciler could re-place

## Test Plan

### Manual Testing
<!-- Hardware: (e.g., MacBook Pro M1 Max 32GB, Mac Mini M2 16GB,
connected via Thunderbolt 4) -->
- Created MetaInstance via dashboard, verified instance placed
- Verified delete cascades (deleting MetaInstance removes backing
instance)
- Verified orphan instances still work independently

### Automated Testing
- 30 tests in `test_meta_instance_edge_cases.py`: lifecycle, retry
logic, error handling, concurrent operations, cascade delete with task
cancellation
- 24 tests in `test_reconcile.py`: constraint matching, connection
health (single/multi-node, edge removal, IP changes), unsatisfied
detection, exclusive binding, idempotency
- All 261 tests pass
- basedpyright 0 errors, ruff clean, dashboard builds

---------

Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com>
2026-02-17 09:48:19 -08:00
Alex Cheema db79c350c1 Fix graceful process shutdown in macOS app (#1372)
## Motivation

Fixes #1370

When the macOS app stops exo, GPU/system memory isn't released. This
happens because:

1. The macOS app calls `process.terminate()` (SIGTERM) but the Python
process only registers a graceful shutdown handler for SIGINT, not
SIGTERM. SIGTERM's default Python behavior raises `SystemExit` which
bypasses the cleanup cascade (runner subprocess MLX cleanup via
`mx.clear_cache()`, channel closing, etc.).
2. The app doesn't wait for the process to actually finish cleanup — it
immediately nils out the process reference.

## Changes

**`src/exo/main.py`**: Register SIGTERM handler alongside SIGINT so the
graceful shutdown cascade (`Node.shutdown()` → cancel task group →
worker/runner cleanup → `mx.clear_cache()` + `gc.collect()`) runs
regardless of which signal is received.

**`app/EXO/EXO/ExoProcessController.swift`**: Replace immediate
`process.terminate()` with escalating shutdown per @Evanev7's
suggestion:
1. Send SIGINT via `process.interrupt()` — triggers the registered
Python handler for graceful cleanup
2. Wait up to 5 seconds for the process to exit
3. If still running, escalate to SIGTERM via `process.terminate()`
4. Wait up to 3 seconds
5. If still running, force kill via SIGKILL

The escalation runs in a detached `Task` so the UI updates immediately
(status → stopped) without blocking.

## Why It Works

The root cause is that SIGTERM wasn't triggering the graceful shutdown
path. By registering a SIGTERM handler in Python and sending SIGINT
first from the macOS app, the process gets a chance to run the full
cleanup cascade: cancelling the task group, shutting down runners (which
call `del model; mx.clear_cache(); gc.collect()`), closing channels, and
flushing logs. The escalation to SIGTERM and SIGKILL ensures the process
always terminates even if graceful shutdown hangs.

## Test Plan

### Manual Testing
<!-- Hardware: Mac Studio M4 Max 128GB -->
- Start exo via macOS app, load a model, run inference
- Stop via the toggle switch, verify memory is released without
requiring a system restart
- Test rapid stop/start (restart) to ensure no race conditions

### Automated Testing
- `uv run basedpyright` — 0 errors
- `uv run ruff check` — passes
- `nix fmt` — no changes

---------

Co-authored-by: Claude Opus 4.5 <noreply@anthropic.com>
Co-authored-by: Evan Quiney <evanev7@gmail.com>
2026-02-17 09:03:54 -08:00
Alex Cheema d6301ed593 dashboard: redesign downloads page as model×node table (#1465)
## Motivation

The current downloads page uses a node-centric card grid layout that is
messy and hard to read — the same model across different nodes appears
in separate cards, and deep nesting wastes space. This makes it
difficult to quickly see which models are on which nodes.

## Changes

Rewrote the downloads page
(`dashboard/src/routes/downloads/+page.svelte`) from a card grid to a
clean table layout:

- **Rows** = models (unique across all nodes)
- **Columns** = nodes (with disk free shown in header)
- **Cells** show status at a glance:
  -  Green checkmark + size for completed downloads
  - 🟡 Yellow percentage + mini progress bar + speed for active downloads
  - `...` for pending downloads
  -  Red X for failed downloads
  - `--` for models not present on a node
- Delete/download action buttons appear on row hover
- Model name column is sticky on horizontal scroll (for many-node
clusters)
- Models sorted by number of nodes with completed downloads
- Imported shared utilities from `$lib/utils/downloads` instead of
inline re-implementations

### Backend: model directory in download events

- Added `model_directory` field to `BaseDownloadProgress` so all
download status events include the on-disk path
- Added `_model_dir()` helper to `DownloadCoordinator` to compute the
path from `EXO_MODELS_DIR`
- Dashboard uses this to show file location and enable "open in Finder"
for completed downloads

### Info modal

- Clicking a model name opens an info modal showing card details
(family, quantization, capabilities, storage size, layer count, tensor
parallelism support)

### Other fixes

- Fixed model name truncation in the table
- Excluded `tests/start_distributed_test.py` from pytest collection (CLI
script that calls `sys.exit()` at import time)

## Test Plan

- [x] `uv run basedpyright` — 0 errors
- [x] `uv run ruff check` — all passed
- [x] `nix fmt` — clean
- [x] `uv run pytest` — 188 passed, 1 skipped

🤖 Generated with [Claude Code](https://claude.com/claude-code)

---------

Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com>
2026-02-17 14:31:47 +00:00
Evan Quiney 6d1ca6689b don't time out node identities (#1493)
currently nodes leaving and rejoining the cluster can lose their identity. We have no need to delete this data on node timing out, so let's just persist it.
2026-02-17 11:48:28 +00:00
Evan c01b6fff21 eprint banner
our banner was being printed to stdout but should be printed to stderr
as its essentially a log message
2026-02-17 11:43:06 +00:00
Jake Hillion 8392e78afe bench: add spec for automatic canary benchmarks (#1483)
Adds all the models that can fit onto a single M3 Ultra for single
machine benchmarks. Fixes the macOS version, GPU spec, and chip type for
maximum reproducibility. Specifies the minimum memory accordingly for
each type of model, using the smallest machine available (the smallest
M3 Ultra is 96GiB).

Test plan:
- Running this with some code that makes machines of this spec available
and stores the results. It works.

This will become part of a larger testing/stability strategy once we've
collected more of the data.
2026-02-17 10:52:05 +00:00
Evan 86735ece78 begins
begins
2026-02-16 19:26:19 +00:00
Evan Quiney 2759e92334 api cancellation (#1276)
closing the http request to the api now
- sends a cancellation from the api
- writes that canellation in the master
- worker plans off the cancellation
- runner observes that cancellation after every generation step (+1
communication per token)
- cancellation happens synchronously to prevent gpu locks

closes #61

---------

Co-authored-by: Alex Cheema <alexcheema123@gmail.com>
Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com>
2026-02-16 05:30:07 -08:00
Jake Hillion 131fb141a6 bench: add --danger-delete-downloads flag with planning phase
exo bench previously relied on the worker's plan loop to download
models, which could fail silently or run into disk space issues during
benchmarking. This made it difficult to diagnose download failures.

Added a planning phase that runs before benchmarking to explicitly
handle downloads. It checks available disk space on each node via the
/state endpoint and starts downloads via POST /download/start. When
the --danger-delete-downloads flag is set and there's insufficient
space, it deletes existing models from smallest to largest until
there's room for the benchmark model.

Test plan:
- CI

```
jake@maverick:/data/users/jake/repos/exo/ > nix run .#exo-bench -- --pp 128,2048,4096 --tg 128 --stdout --settle-timeout 10 --host s1 --model mlx-community/gpt-oss-120b-MXFP4-Q8
PyTorch was not found. Models won't be available and only tokenizers, configuration and file/data utilities can be used.
2026-02-16 12:12:11.807 | INFO     | __main__:main:710 - pp/tg mode: combinations (product) - 3 pairs
Warning: You are sending unauthenticated requests to the HF Hub. Please set a HF_TOKEN to enable higher rate limits and faster downloads.
2026-02-16 12:12:13.455 | DEBUG    | __main__:main:725 - [exo-bench] loaded tokenizer: mlx-community/gpt-oss-120b-MXFP4-Q8 for prompt sizer
2026-02-16 12:12:13.473 | DEBUG    | __main__:main:761 - exo-bench model: short_id=gpt-oss-120b-MXFP4-Q8 full_id=mlx-community/gpt-oss-120b-MXFP4-Q8
2026-02-16 12:12:13.473 | INFO     | __main__:main:762 - placements: 1
2026-02-16 12:12:13.474 | INFO     | __main__:main:764 -   - Pipeline / MlxRing / nodes=1
2026-02-16 12:12:13.474 | INFO     | __main__:main:771 - Planning phase: checking downloads...
Traceback (most recent call last):
  File "/nix/store/q31kmbcfr5bf97290bvbnhrvpc3fv824-source/bench/exo_bench.py", line 885, in <module>
    raise SystemExit(main())
                     ~~~~^^
  File "/nix/store/q31kmbcfr5bf97290bvbnhrvpc3fv824-source/bench/exo_bench.py", line 772, in main
    run_planning_phase(
    ~~~~~~~~~~~~~~~~~~^
        client,
        ^^^^^^^
    ...<4 lines>...
        settle_deadline,
        ^^^^^^^^^^^^^^^^
    )
    ^
  File "/nix/store/q31kmbcfr5bf97290bvbnhrvpc3fv824-source/bench/exo_bench.py", line 367, in run_planning_phase
    raise RuntimeError(
    ...<2 lines>...
    )
RuntimeError: Insufficient disk on 12D3KooWE2C7dzC9d9YJMEfWK3g8og7JdZj3HHXZ8VmGrXYAEnEj: need 65GB, have 55GB. Use --danger-delete-downloads to free space.
jake@maverick:/data/users/jake/repos/exo/ > nix run .#exo-bench -- --pp 128,2048,4096 --tg 128 --stdout --settle-timeout 10 --host s1 --model mlx-community/gpt-oss-120b-MXFP4-Q8 --danger-delete-downloads
PyTorch was not found. Models won't be available and only tokenizers, configuration and file/data utilities can be used.
2026-02-16 12:12:19.626 | INFO     | __main__:main:710 - pp/tg mode: combinations (product) - 3 pairs
2026-02-16 12:12:21.262 | DEBUG    | __main__:main:725 - [exo-bench] loaded tokenizer: mlx-community/gpt-oss-120b-MXFP4-Q8 for prompt sizer
2026-02-16 12:12:21.280 | DEBUG    | __main__:main:761 - exo-bench model: short_id=gpt-oss-120b-MXFP4-Q8 full_id=mlx-community/gpt-oss-120b-MXFP4-Q8
2026-02-16 12:12:21.280 | INFO     | __main__:main:762 - placements: 1
2026-02-16 12:12:21.280 | INFO     | __main__:main:764 -   - Pipeline / MlxRing / nodes=1
2026-02-16 12:12:21.280 | INFO     | __main__:main:771 - Planning phase: checking downloads...
2026-02-16 12:12:21.336 | INFO     | __main__:run_planning_phase:386 - Deleting mlx-community/Qwen3-0.6B-4bit from 12D3KooWE2C7dzC9d9YJMEfWK3g8og7JdZj3HHXZ8VmGrXYAEnEj (335MB)
2026-02-16 12:12:21.350 | INFO     | __main__:run_planning_phase:386 - Deleting mlx-community/Llama-3.2-1B-Instruct-4bit from 12D3KooWE2C7dzC9d9YJMEfWK3g8og7JdZj3HHXZ8VmGrXYAEnEj (679MB)
2026-02-16 12:12:21.363 | INFO     | __main__:run_planning_phase:386 - Deleting mlx-community/Llama-3.2-3B-Instruct-4bit from 12D3KooWE2C7dzC9d9YJMEfWK3g8og7JdZj3HHXZ8VmGrXYAEnEj (1740MB)
2026-02-16 12:12:21.373 | INFO     | __main__:run_planning_phase:386 - Deleting mlx-community/Llama-3.2-3B-Instruct-8bit from 12D3KooWE2C7dzC9d9YJMEfWK3g8og7JdZj3HHXZ8VmGrXYAEnEj (3264MB)
2026-02-16 12:12:21.384 | INFO     | __main__:run_planning_phase:386 - Deleting mlx-community/GLM-4.7-Flash-8bit from 12D3KooWE2C7dzC9d9YJMEfWK3g8og7JdZj3HHXZ8VmGrXYAEnEj (30366MB)
2026-02-16 12:12:21.413 | INFO     | __main__:run_planning_phase:407 - Started download on 12D3KooWE2C7dzC9d9YJMEfWK3g8og7JdZj3HHXZ8VmGrXYAEnEj
```

It's not pretty but it works!
2026-02-16 13:06:38 +00:00
Evan Quiney 2d8bfc2e3c fix: PlaceInstanceParams broken field validator
our field validator for PlaceInstance was wrong - we can just rely on default behaviour here anyway!
2026-02-16 03:58:43 -08:00
ciaranbor 042999f728 Ciaran/message deletion (#1409)
## Motivation

When a user deletes a message during an active streamed generation, it
can cause unexpected behavior. The delete confirmation text was also
misleading — it said "all responses after it" only for user messages,
which didn't accurately describe the behavior (all messages after the
deleted one are removed, regardless of role)

## Changes

- Prevent deletion during streaming: Disabled the delete button and
blocked handleDeleteClick when loading is true, with a visual indication
(dimmed button, cursor-not-allowed, tooltip change)
- Clarified delete confirmation text: Replaced role-specific wording
with a simpler, accurate message:
    - Last message: "Delete this message?"
- Any other message: "Delete this message and all messages after it?"

## Why It Works

Guarding on the loading state at both the click handler and the button's
disabled attribute ensures no deletion can be triggered while a response
is being streamed

## Test Plan

### Manual Testing

- Verify the delete button is visually disabled and non-clickable while
a response is streaming
- Verify the tooltip shows "Cannot delete while generating" during
streaming
  - Verify the last message shows "Delete this message?" confirmation
- Verify non-last messages show "Delete this message and all messages
after it?" confirmation
  - Verify deletion works normally when not streaming
2026-02-16 11:46:41 +00:00
ciaranbor b61dc2eb35 Prevent image editing without image input (#1410)
## Motivation

Models that only support image editing (ImageToImage but not
TextToImage) would silently attempt text-to-image generation when a user
submitted a text prompt without an attached image

## Changes

- Added an early return guard in handleSubmit() that prevents submission
when the selected model only supports image editing and no image is
attached (isEditOnlyWithoutImage)
- Fixed the text-to-image generation branch to use the more specific
modelSupportsTextToImage() check instead of the broad isImageModel(),
ensuring only models with TextToImage capability trigger generation from
text alone
- The existing isEditOnlyWithoutImage derived state (which was already
used for UI hints like placeholder text and button disabling) now also
blocks the actual submit path

## Why It Works

The text-to-image fallback now correctly checks
modelSupportsTextToImage() directly, so edit-only models no longer fall
through to the generation path


## Test Plan

### Manual Testing

- Select an edit-only image model (e.g., one with only ImageToImage
capability)
- Verify the send button is disabled and placeholder reads "Attach an
image to edit..." when no image is attached
  - Attach an image and verify the form becomes submittable
- Select a text-to-image model and verify text-only prompts still
trigger generation normally
  - Ensure pressing `enter` doesn't bypass check
2026-02-16 11:39:59 +00:00
rltakashige 36a7115b6f Pass usage and generation stats through all adapters correctly (#1461)
## Motivation

Exo is not returning usage stats correctly at the moment.

## Changes

- Correctly pass usage stats instead of generation stats.
- Pass usage stats within tool calls.

## Test Plan

### Manual Testing
Needs manual testing.

### Automated Testing
Passes CI.
2026-02-16 11:20:04 +00:00
Jake Hillion 0b7d88b43b python: add hermetic basedpyright typecheck to nix flake check
The existing CI typecheck job used `uv run basedpyright` which depends
on a non-hermetic uv sync step. This replaces it with a fully hermetic
typecheck as a Nix flake check using the uv2nix virtual environment.

Added a `typecheckVenv` with dev dependencies, a `linuxOverlay` to
ignore native shared library deps (NVIDIA, torch, triton, mlx) that
aren't needed at type-check time, and `passthru` preservation plus
`.pyi` stub copying on the `exo-pyo3-bindings` overlay so basedpyright
can resolve the Rust bindings types. Also guarded the `mlx` Nix build
override to macOS only since it requires Metal. Removed the old
non-hermetic `typecheck` CI job since `nix flake check` now covers it.

The hermetic check ensures type checking uses exactly the locked
dependency versions and catches type errors without requiring a
working uv/pip environment.

Test plan:
- CI (`nix flake check` runs on x86_64-linux, aarch64-linux, aarch64-darwin)
- Verified `nix build ".#checks.x86_64-linux.typecheck"` passes with 0 errors
2026-02-16 11:09:23 +00:00
Alex Cheema 1c3cc699d4 fix: add missing getModelFitStatus prop to Recent tab (#1470)
## Summary
- Clicking the **Recent** tab in the Model Picker crashed with
`TypeError: e.getModelFitStatus is not a function`
- The `ModelPickerGroup` component in the Recent tab section was missing
the `{getModelFitStatus}` prop, while all other tabs (e.g., the main
model list) passed it correctly
- Added the missing `{getModelFitStatus}` prop so the Recent tab renders
without errors, matching the behavior of the other tabs

## Test plan
- [ ] Open the dashboard and click **SELECT MODEL**
- [ ] Switch to the **Recent** tab — verify it renders without crashing
- [ ] Confirm model fit status indicators display correctly on recent
models
- [ ] Verify the other tabs (All, Favorites) still work as before

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com>
2026-02-13 16:37:32 +00:00
rltakashige 5a28642790 Add support for Step 3.5 flash! (#1460)
## Motivation

Working version of #1366 

## Changes

Add Step 3.5 Flash

## Test Plan

### Manual Testing
Works!

### Automated Testing
Running two processes tensor/pipeline sharded gives same logits as
single process.
2026-02-13 12:10:18 +00:00
Alex Cheema 6950f94109 dashboard: show macOS version in debug mode (#1454)
## Motivation

When debugging cluster issues, it's useful to see which macOS version
each node is running — especially since version mismatches can cause
compatibility problems. The OS version data is already collected by the
identity gatherer but wasn't shown in the topology graph.

## Changes

- Added OS version label (e.g. "macOS 15.2") to each node in the
topology graph when debug mode is enabled
- Renders below the existing TB and RDMA debug labels using the same
styling conventions
- Sources data from the existing `nodeIdentities` store (no backend
changes needed)

## Why It Works

The `nodeIdentities` store already contains `osVersion` for each node.
We simply read it in the `TopologyGraph` component and append a text
label in the debug section, following the exact same pattern as the TB
and RDMA labels.

## Test Plan

### Manual Testing
<!-- Hardware: MacBook Pro -->
- Enable debug mode in the dashboard
- Verify OS version label appears below TB/RDMA labels on each node
- Verify label disappears when debug mode is disabled

### Automated Testing
- Dashboard build passes (`npm run build`)

---------

Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com>
Co-authored-by: rltakashige <rl.takashige@gmail.com>
Co-authored-by: Ryuichi Leo Takashige <leo@exolabs.net>
2026-02-12 17:56:55 +00:00
Alex Cheema d0c44273db feat: add enable_thinking toggle for thinking-capable models (#1457)
## Motivation

Fixes #1456. Models like DeepSeek V3.2, Qwen3, and GLM-4.7 always run in
thinking mode because their chat templates auto-inject `<think>`. Users
need a way to disable thinking for models that support both modes.

## Changes

**API**: Added `enable_thinking: bool | None` to `ChatCompletionRequest`
and `TextGenerationTaskParams`. Passed through the adapter to
`tokenizer.apply_chat_template()` as a kwarg (only when explicitly set,
so models without the template variable are unaffected).

**Dashboard**: Added a thinking toggle button in the chat input area.
Visible only when the selected model has both "text" and "thinking"
capabilities.

## Why It Works

Most thinking model chat templates (DeepSeek, Qwen3, GLM) accept
`enable_thinking` as a Jinja template variable. Passing
`enable_thinking=False` prevents the template from injecting `<think>`,
matching the vLLM convention.

## Test Plan

### Manual Testing
- `curl` with `"enable_thinking": false` against a thinking model — no
`<think>` in output
- Dashboard toggle visible for thinking models, hidden for text-only
models

### Automated Testing
- basedpyright: 0 errors
- ruff: clean
- pytest: 188 passed
- dashboard build: success

---------

Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com>
2026-02-12 17:35:24 +00:00
Jake Hillion cc33213842 bench: add --settle-timeout for cluster startup retry (#1449)
exo_bench.py fails if started too soon after a cluster starts because
the topology hasn't populated yet, resulting in no valid placements.

Extracted the preview-fetch-and-filter logic into a
`fetch_and_filter_placements` helper and added a retry loop with
exponential backoff (1s initial, 2x multiplier, 60s cap). The new
`--settle-timeout` flag controls how long to retry (default 0 = try
once, preserving existing behaviour). Each retry logs a warning
explaining the cluster may still be settling.

Test plan:
- Tested on several freshly started clusters. This used to fail a lot,
  now it succeeds.
2026-02-12 16:38:09 +00:00
Alex Cheema 62e8110e97 fix: prevent DownloadModel TaskCreated event flood (#1452)
## Motivation

When a model download fails repeatedly (e.g. `ContentLengthError` on a
large model like `zai-org/GLM-5`), the download coordinator accumulates
duplicate progress callbacks — one per retry cycle. Each callback
independently throttles at 1 event/sec, so after N retries, every
download progress tick generates N events instead of 1. After an hour of
failures (~60 retry cycles), this produces ~60 `NodeDownloadProgress`
events/sec, overwhelming the master, delaying heartbeats, and causing
the node to time itself out.

### The callback accumulation cycle
1. `_start_download_task()` calls
`shard_downloader.on_progress(callback)` which **appends** to a list
2. Download fails → `DownloadFailed` status set, but old callback stays
in the list
3. 60s later: `_emit_existing_download_progress()` scans disk → resets
status to `DownloadPending`
4. Worker sends new `StartDownload` → coordinator accepts (guard didn't
check `DownloadFailed`)
5. `_start_download_task()` appends **another** callback
6. Each callback has its own throttle → N callbacks = N events per
progress tick

## Changes

### Commit 1: `src/exo/worker/main.py`
Move the `DownloadModel` backoff check **before** `TaskCreated` emission
in `plan_step()`. Previously `TaskCreated` was emitted unconditionally
every 0.1s even when backoff blocked the download command.

### Commit 2: `src/exo/download/coordinator.py`
1. **Register progress callback once** in `__post_init__` instead of
per-download in `_start_download_task()`. Uses a per-model throttle dict
instead of per-callback closure variables.
2. **Add `DownloadFailed` to the `_start_download()` guard** so
redundant `_start_download_task()` calls don't happen. Retries still
work because `_emit_existing_download_progress` resets `DownloadFailed`
→ `DownloadPending` by scanning disk every 60s.

## Why It Works

The root cause was callbacks accumulating in
`ResumableShardDownloader.on_progress_callbacks` (a list that only
appends, never clears). By registering one callback per coordinator
lifetime and guarding against re-entry on `DownloadFailed`, we ensure
exactly one progress event per model per progress tick regardless of how
many retry cycles have occurred.

## Test Plan

### Manual Testing
- Verified the download retry flow: failed download → 60s scan resets
status → new `StartDownload` accepted → download retries with single
callback

### Automated Testing
- `uv run basedpyright` — 0 errors
- `uv run ruff check` — passes
- `uv run pytest` — 188 passed

---------

Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com>
2026-02-11 23:50:43 +00:00
Alex Cheema 98773437f3 Make info gatherer monitors resilient with retry loops and timeouts (#1448)
## Motivation

Info gatherer monitors could silently stop posting events, causing stale
node state after rejoins. The macmon monitor was especially fragile — it
had no retry loop, so a crash or parse error would kill it permanently.
Worse, the unhandled exception would propagate to the TaskGroup and take
down *all* sibling monitors. Additionally, none of the monitors had
timeouts on their subprocess calls, so a hung `system_profiler` or
`networksetup` could stall a monitor indefinitely.

## Changes

- Wrap `_monitor_macmon` in a `while True` retry loop with `except
Exception`, matching the pattern used by all other monitors
- Add `fail_after` timeouts to all monitor loop bodies:
- 10s for lightweight commands (`_monitor_misc`, `_watch_system_info`,
`_gather_iface_map` init)
- 30s for heavier commands (`_monitor_system_profiler_thunderbolt_data`,
`_monitor_thunderbolt_bridge_status`)
- Remove unused `CalledProcessError` and `cast` imports

## Why It Works

All monitors now follow the same resilient pattern: `while True` → `try`
with `fail_after` → `except Exception` (logs warning) → `sleep`. If a
subprocess hangs, the timeout fires and `TimeoutError` is caught by the
existing `except Exception` handler. If macmon crashes, it restarts
after the interval instead of dying permanently. No single monitor
failure can cascade to kill the others.

## Test Plan

### Manual Testing
<!-- Hardware: macOS with macmon installed -->
<!-- What you did: -->
- Run exo, kill macmon process (`kill $(pgrep macmon)`), verify it
restarts and metrics resume
- Verify all monitors continue posting events after simulated hangs

### Automated Testing
- All 188 existing tests pass
- basedpyright: 0 errors
- ruff: all checks passed

🤖 Generated with [Claude Code](https://claude.com/claude-code)

---------

Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com>
2026-02-11 23:07:36 +00:00
Jake Hillion a8acb3cafb dashboard: show available disk space on downloads page
The downloads page previously only showed the approximate space used by
downloaded models (summed from completed download sizes), but did not
show how much disk space was actually available. This made it difficult
to know if a download would succeed before pressing the button.

Added disk space tracking to the InfoGatherer that polls the models
directory partition every 30 seconds. The DiskUsage type captures total
and available space, which flows through the event system to State and
is exposed via the /state API. The dashboard now displays "X on disk /
Y available" for each node in the downloads view.

Test plan:
- CI
2026-02-11 21:57:28 +00:00
Alex Cheema a0721dbe57 feat: warn when cluster nodes have mismatched macOS versions (#1436)
## Motivation

When nodes in an exo cluster run different macOS versions, inference can
produce incompatible results or fail silently. Users currently have no
way to know this from the dashboard.

## Changes

- Added `get_os_version()` to `system_info.py` that returns the macOS
version (e.g. `"15.3"`) or platform name for non-Mac nodes
- Added `os_version` field to `NodeIdentity` and
`StaticNodeInformation`, gathered once at startup
- Propagated `os_version` through the event sourcing pipeline
(`apply.py`)
- Exposed `nodeIdentities` from the dashboard store with `osVersion`
- Added a derived `macosVersionMismatch` check in `+page.svelte` that
triggers when 2+ macOS nodes report different versions
- Rendered a yellow "INCOMPATIBLE macOS VERSIONS" warning badge
(matching the existing Thunderbolt Bridge cycle warning style) with a
hover tooltip listing each node's name and version, in all three
topology view sizes (large, medium, compact)

## Why It Works

The OS version is a static property gathered once at node startup via
`platform.mac_ver()`. It flows through the existing
`StaticNodeInformation` → `NodeGatheredInfo` event → `NodeIdentity`
state pipeline, so no new event types or state fields beyond
`os_version` on `NodeIdentity` are needed. The dashboard derives the
mismatch by comparing `osVersion` across all nodes whose version looks
like a macOS version string (starts with a digit).

## Test Plan

### Manual Testing
Hardware: 4x Mac Studio M2 Ultra 512GB (s18, s17 (2), james, mike),
connected via Thunderbolt
- s18 and s17 (2) on macOS 26.2, james and mike on macOS 26.3
- Verified the "INCOMPATIBLE macOS VERSIONS" warning badge appears in
the topology view
- Verified the hover tooltip lists all four nodes with their respective
versions
- Screenshots attached in comment below

### Automated Testing
- basedpyright: 0 errors
- ruff check: all checks passed
- nix fmt: no formatting changes needed
- Dashboard builds successfully

---------

Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com>
2026-02-11 21:18:59 +00:00
Alex Cheema 50e2bcf93e fix: RDMA debug labels, TB5 info box, and rdma_ctl status detection (#1437)
## Motivation

Several RDMA/Thunderbolt UX issues in the dashboard and macOS app:

1. **Debug mode showed "? ?" for RDMA connections** — the topology view
only extracted IPs from socket connections, not RDMA interface names
2. **No way to detect if RDMA is actually enabled** — the system only
knew about TB5 hardware and RDMA topology edges, not whether `rdma_ctl`
was enabled on each node
3. **False "RDMA AVAILABLE" info box** — showed on Mac Minis with idle
TB5 ports even when RDMA was already enabled, and on single nodes with
TB5
4. **macOS app only showed local RDMA status** — ran `rdma_ctl` locally
with no visibility into other nodes in the cluster

## Changes

### Dashboard: Fix RDMA debug labels (`0abc90c4`)
- Added `sourceRdmaIface` and `sinkRdmaIface` to `TopologyEdge`
interface
- Updated `TopologyGraph.svelte` and `ModelCard.svelte` to show `RDMA
en2 → en3` instead of `? ?`

### Dashboard: TB5 RDMA info box (`a3795552`, `8ce8e173`)
- Added dismissible info box when 2+ nodes have TB5 hardware but RDMA is
disabled
- Includes setup instructions (Recovery mode → `rdma_ctl enable` →
reboot, TB5 cables, macOS version match)
- Requires 2+ exo nodes with TB5 to avoid false positives from
single-node setups

### Backend: `rdma_ctl status` detection (`ae07239b`)
- Added `RdmaCtlStatus` event to `info_gatherer.py` — runs `rdma_ctl
status` with 5s timeout, `shutil.which` guard, and `OSError` handling
(polls every 10s on macOS)
- Added `NodeRdmaCtlStatus` model to `profiling.py` and `node_rdma_ctl`
field to `State`
- Handle in `apply.py` (event apply + node timeout cleanup)
- Exposed `nodeRdmaCtl` in dashboard store (`app.svelte.ts`)
- Info box detection now uses actual RDMA status instead of TB5 link
speeds

### Dashboard: Per-node RDMA debug labels (`ae07239b`)
- Debug mode shows `RDMA:ON` (green) or `RDMA:OFF` (dim) per node in
topology view, below the TB bridge label

### macOS app: Cluster-wide RDMA status from `/state` (`a1455b61`,
`d0d77b63`)
- Added `NodeRdmaCtlStatus` to `ClusterState.swift` — decoded from
`/state` endpoint
- Replaced local-only `rdma_ctl status` check with cluster-wide
`nodeRdmaCtl` from state
- Debug section shows per-node RDMA enabled/disabled for all nodes in
the cluster
- Still shows local `ibv_devices` and `ibv_devinfo` details (device
names, active ports) for richer local debugging

## Files changed

| Area | File | Change |
|------|------|--------|
| Backend | `src/exo/utils/info_gatherer/info_gatherer.py` |
`RdmaCtlStatus` event, monitor task |
| Backend | `src/exo/shared/types/profiling.py` | `NodeRdmaCtlStatus`
model |
| Backend | `src/exo/shared/types/state.py` | `node_rdma_ctl` field |
| Backend | `src/exo/shared/apply.py` | Event handler + timeout cleanup
|
| Dashboard | `dashboard/src/lib/stores/app.svelte.ts` | `nodeRdmaCtl` +
`nodeThunderbolt` in store |
| Dashboard | `dashboard/src/routes/+page.svelte` | Info box with RDMA
detection + instructions |
| Dashboard | `dashboard/src/lib/components/TopologyGraph.svelte` | RDMA
debug labels per node + fix "? ?" |
| Dashboard | `dashboard/src/lib/components/ModelCard.svelte` | RDMA
interface display fix |
| App | `app/EXO/EXO/Models/ClusterState.swift` | `NodeRdmaCtlStatus`
struct + decode |
| App | `app/EXO/EXO/ContentView.swift` | Cluster-wide RDMA view + local
device details |
| App | `app/EXO/EXO/Services/NetworkStatusService.swift` | Remove local
`rdma_ctl`, keep `ibv_*` |

## Test Plan

- [x] `uv run basedpyright` — 0 errors
- [x] `uv run ruff check` — pass
- [x] `nix fmt` — clean
- [x] `cd dashboard && npm run build` — success
- [x] `uv run pytest` — 188 passed
- [x] Xcode build — compiles (only pre-existing `dist/exo` resource
error)
- [x] Deployed to Mac Minis — `nodeRdmaCtl` shows `enabled: true`, no
false info box
- [x] Deployed to James cluster — RDMA debug labels show correctly

🤖 Generated with [Claude Code](https://claude.com/claude-code)

---------

Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com>
2026-02-11 20:43:11 +00:00
Alex Cheema 7bed91c9c2 feat: add Recent tab to model picker (#1440)
## Motivation

When frequently switching between models, it's tedious to search through
the full model list to find ones you've used before. A "Recent" tab
provides quick access to previously launched models.

## Changes

- **New store** (`dashboard/src/lib/stores/recents.svelte.ts`):
`RecentsStore` class persisting recently launched model IDs with
timestamps to localStorage (key: `exo-recent-models`). Caps at 20
entries, deduplicates on re-launch (moves to top).
- **FamilySidebar**: Added "Recent" tab between Favorites and Hub,
conditionally shown when there are recent models.
- **FamilyLogos**: Added clock/history icon for the recents tab.
- **ModelPickerModal**: Added `recentModelIds`/`hasRecents` props.
Derives single-variant `ModelGroup[]` from recent IDs and renders them
using the same `ModelPickerGroup` component as all other tabs —
consistent styling, memory grey-out, favorites, info button, download
indicators.
- **+page.svelte**: Calls `recordRecentLaunch(modelId)` after successful
instance launch. Passes reactive recent state to the modal.

## Why It Works

Follows the exact same pattern as the existing Favorites feature
(localStorage persistence, conditional tab display, reactive Svelte 5
`$state`/`$derived`). Recent models are wrapped as single-variant
`ModelGroup` objects so they reuse `ModelPickerGroup` for identical row
rendering across all tabs.

## Test Plan

### Manual Testing
<!-- Hardware: MacBook Pro -->
- Launch a model instance → reopen model picker → "Recent" tab appears
with the launched model
- Launch a second model → it appears at top of the Recent list
- Re-launch the first model → it moves back to top
- Search within the Recent tab filters the list
- Models that don't fit in memory are greyed out (same as All tab)
- Close/reopen browser → recents persist from localStorage

### Automated Testing
- Dashboard builds successfully (`npm run build`)

---------

Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com>
Co-authored-by: rltakashige <rl.takashige@gmail.com>
2026-02-11 12:34:08 -08:00
Hunter Bown 48caea4a23 feat: add intermediate model-fit state for cluster-capacity-only models (#1441)
## Motivation
In the model picker, non-runnable models were not clearly separated
between two different cases:

- model exceeds currently available RAM but can fit in total cluster
capacity
- model exceeds total cluster capacity

That made it harder to distinguish "not runnable right now" from "too
large for this cluster."

## Changes
- Added a tri-state fit status in dashboard model-picker flow:
  - `fits_now`
  - `fits_cluster_capacity`
  - `too_large`
- Updated dashboard logic to compute both available cluster RAM and
total cluster RAM.
- Passed fit status through picker components.
- Updated model size color mapping:
  - `fits_now` -> white/gray
  - `fits_cluster_capacity` -> orange
  - `too_large` -> red
- Updated group ordering for non-runnable models:
- orange groups (`fits_cluster_capacity`) are listed above red groups
(`too_large`).

## Why It Works
Launch safety is unchanged: selection is still gated by existing
placement feasibility (`canModelFit`), so models that cannot run now
remain disabled.

The new fit status is used for visual distinction and ordering only:
- runnable now
- fits cluster capacity but not free RAM now
- too large for cluster capacity

## Before
Non-runnable models were not clearly distinguished by temporary capacity
vs hard capacity limit.

## After
Model picker clearly separates states by both color and order:
- Runnable now (white/gray)
- Fits cluster capacity but not free RAM now (orange, disabled)
- Exceeds cluster capacity (red, disabled)

## Test Plan
### Manual testing
- Open model picker with live cluster memory telemetry.
- Verify white/gray models are selectable.
- Verify orange models are disabled and appear above red models.
- Verify red models are disabled and appear below orange models.

### Automated checks
- `npm --prefix dashboard run build` passes.
- `uv run basedpyright` passes.
- `uv run ruff check` passes.
- `npm --prefix dashboard run check` reports existing pre-change Svelte
diagnostics (same known SVG title/a11y items).
- `uv run pytest` in this local environment exits during collection due
existing `tests/start_distributed_test.py` `SystemExit` usage.
- No Python code was changed.

---------

Co-authored-by: Alex Cheema <41707476+AlexCheema@users.noreply.github.com>
2026-02-11 00:21:35 +00:00
Mustafa Alp Yılmaz eead50b4cd Fix setrlimit crash when hard file descriptor limit < 65535 (#1430)
## Summary

`exo` crashes on startup when the system's hard file descriptor limit is
below 65535, which occurs in macOS LaunchDaemon environments, Docker
containers, and other restricted setups.

**Root cause:** `resource.setrlimit(resource.RLIMIT_NOFILE, (max(soft,
65535), hard))` raises `ValueError` when `hard < 65535` because the soft
limit cannot exceed the hard limit.

**Fix:**
- **`main.py`**: Clamp the target soft limit to `min(max(soft, 65535),
hard)` so it never exceeds the hard limit
- **`utils_mlx.py`**: Query current limits instead of hardcoding `(2048,
4096)`, which both crashed on restricted systems and incorrectly lowered
the hard limit when it was set higher

## Test plan

- [x] `basedpyright` passes with 0 errors
- [x] `ruff check` passes
- [x] Verify startup works on system with hard limit < 65535 (tested in
macOS LaunchDaemon with hard limit 10240)
- [x] Verify startup still works on default macOS (hard limit typically
unlimited)
2026-02-10 22:46:22 +00:00
Jake Hillion 199df64cfc util: remove VecExt trait, inline at call site (#1446)
VecExt added a .map() convenience method on Vec<T> that simply called
.into_iter().map(f).collect(). This thin wrapper provided no
optimisation benefit and obscured a standard iterator pattern behind a
nightly feature gate and an extra dependency.

Replaced the single call site in exo_pyo3_bindings with the equivalent
iterator chain and removed the ext module, the extend dependency, and
the trait_alias feature gate from the util crate.

Test plan:
- CI
2026-02-10 20:45:57 +00:00
Ryuichi Leo Takashige dc7ade8052 set the mlx hash 2026-02-10 20:16:17 +00:00
Ryuichi Leo Takashige dc781497c5 update mlx to 0.30.6 2026-02-10 20:16:17 +00:00
Jake Hillion c37eb24331 util: remove dead code (#1445)
The util crate contained several unused items: NonemptyArray,
BoxedSliceExt, a blanket Sealed trait, an empty alias module, six unused
nightly feature gates, and five unused Cargo dependencies (thiserror,
once_cell, internment, derive_more, bon, recursion).

Removed all items that had no references outside their own definitions,
keeping only WakerDeque, VecExt, and the trait_alias feature gate which
are actively used by the networking and exo_pyo3_bindings crates.

Test plan:
- CI
2026-02-10 20:10:54 +00:00
Jake Hillion 8af2af6328 nix: override apple-sdk to 26.2 and enable MLX_BUILD_CPU (#1443)
The pinned nixpkgs provides apple-sdk 26.0, but building MLX requires
SDK 26.2. The upstream package reads versions.json via a relative path
at eval time, so it can't be overridden through callPackage args.

Added a thin overlay that copies the upstream apple-sdk source and
patches only metadata/versions.json to point at SDK 26.2. Also enabled
MLX_BUILD_CPU in the MLX nix build.

This avoids vendoring the entire apple-sdk package (~2200 lines) while
still getting the SDK version we need.

Test plan:
- CI
- Built and ran on two machines connected with Thunderbolt 5 - Kimi K2.5
starts in Tensor+RDMA and seems sensible.
2026-02-10 19:53:53 +00:00
rltakashige 43728b2047 Send all exo logs (#1439)
## Motivation

Log rotation adds a bunch of .zst files. Let's send them all in the bug
reports.

This PR also standardises the logs so that they all include the
timestamp.
2026-02-10 19:28:33 +00:00
rltakashige 1699fcfb9f standardise logs (#1442)
## Motivation

Standardises exo.log and event_log

## Changes

- exo.log and exo.log.zst are now in a exo_log directory.
- event_log is now timestamped and not numbered. The timestamps can be
sorted as they are YYYY-MM-DD-HH-MM

## Test Plan

### Manual Testing
Nothing crashes.
2026-02-10 19:13:25 +00:00
rltakashige 009b43c662 add log rotation for .exo/exo.log (#1438)
## Motivation

.exo.log currently contains all past history. This just makes it hard to
read and is unnecessarily expensive even on disk.

## Why It Works

Just uses loguru's rotation.

## Test Plan

### Manual Testing
exo.log is new. 
<img width="1992" height="706" alt="image"
src="https://github.com/user-attachments/assets/9b293993-1141-43e7-b58e-0ddd2d4eda2e"
/>
2026-02-10 18:55:06 +00:00
Jake Hillion 1f242e8eee gossipsub: stop silent message dropping and warn (#1434)
The 15-second publish_queue_duration caused messages in peer queues to
be silently dropped. When events are dropped, workers detect gaps in the
event index sequence and request missing events via the NACK path
(RequestEventLog), but this recovery is inefficient.

Removed the timeout configuration - gossipsub now uses its default
behavior without time-based eviction. If queue buildup is a concern,
queue size should be limited explicitly rather than dropping by timeout.

Split error handling to log AllQueuesFullError as a warning (indicates
peers are unresponsive) while keeping NoPeersSubscribedToTopicError
silent (expected during startup and network partitions).

Test plan:
- CI
2026-02-10 18:47:47 +00:00
rltakashige 64179c6fc1 Dont save to app directory (#1435)
## Motivation

App keeps losing Local Network permissions.

## Changes

Don't save stuff to the app directory anymore. Instead, save to .exo.

## Why It Works

<!-- Explain why your approach solves the problem -->

## Test Plan

### Manual Testing
Before:
<img width="1512" height="106" alt="image"
src="https://github.com/user-attachments/assets/544ef57e-b626-484d-941f-2472969aa208"
/>

After:
<img width="433" height="53" alt="Screenshot 2026-02-10 at 17 43 06"
src="https://github.com/user-attachments/assets/3de2856b-cdf6-4b35-aa8f-50440686344f"
/>


### Automated Testing
<!-- Describe changes to automated tests, or how existing tests cover
this change -->
<!-- - -->
2026-02-10 18:41:36 +00:00
Jake Hillion 305a3c8b70 event_log: move event log from unbounded in-memory list to disk (#1432)
The master and API event logs (list[Event]) grew unbounded in RAM for
the lifetime of the process. Events are rarely read back (only for
RequestEventLog when a new node catches up, or the dashboard /events
endpoint).

Introduced a DiskEventLog class that writes length-prefixed msgpack
records to an append-only file, using a bounded LRU cache of byte
offsets for indexed access. On close, the active file is compressed
with ZSTD and rotated into a numbered archive slot, keeping the last 5
archives (events.1.bin.zst through events.5.bin.zst). On construction,
any stale active file from a crash is rotated before opening a fresh
log. The /events API endpoint now streams the JSON array one event at a
time rather than materializing the full list in memory. Deserialization
routes msgpack through json.dumps into Pydantic's validate_json() to
get correct JSON-mode coercion (e.g. string to enum) under strict mode.

This bounds memory usage to the LRU cache (128 entries) regardless of
event volume, while still supporting efficient sequential reads from
disk when needed.

Test plan:
- CI
- New unit tests for DiskEventLog: append/read, range queries, rotation
  on close, stale file recovery, idempotent close, successive sessions,
  archive retention limit (5 max)
- Tested on a cluster with 9000 events. /events continues working.
- On disk size is 3.9MiB with ~8000 events, and the compression is very
  effective.
- Disconnected and rejoined a machine, it rejoined fine.

---------

Co-authored-by: Ryuichi Leo Takashige <leo@exolabs.net>
2026-02-10 17:27:32 +00:00
Alex Cheema ead19bea74 Always load image model cards into cache (#1421)
## Motivation

Follows up on #1408. Image models (FLUX, Qwen-Image, etc.) don't have a
`config.json` on HuggingFace. Previously, image model TOML cards were
only loaded into `_card_cache` when `EXO_ENABLE_IMAGE_MODELS=true`. When
the flag was off but an image model was requested (e.g., via
`get_placement_previews`), `ModelCard.load()` fell through to
`fetch_from_hf()` which tried to download `config.json` — causing
`FileNotFoundError` spam. #1408 added defensive error handling; this PR
fixes the root cause.

## Changes

**`model_cards.py`**: Always include `image_model_cards/` in
`CARD_SEARCH_PATH` so image model TOML cards are always loaded into
`_card_cache`. `ModelCard.load()` then finds them directly and never
falls through to `fetch_from_hf()`. The `EXO_ENABLE_IMAGE_MODELS` flag
now controls whether image models appear in `get_model_cards()` (the
listing) rather than whether they're loaded at all.

## Why It Works

`fetch_from_hf()` is designed for text models only (it hardcodes
`tasks=[ModelTask.TextGeneration]` and requires `config.json`). Image
models should never reach that path. By always having them in the cache,
the lookup succeeds immediately and `fetch_from_hf()` is never called.

## Test Plan

### Automated Testing
- `uv run basedpyright` — 0 errors
- `uv run ruff check` — passes

Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com>
Co-authored-by: rltakashige <rl.takashige@gmail.com>
2026-02-10 09:11:57 -08:00
Jake Hillion 5a83e59182 dashboard: allow typing in chat input while response is generating (#1433)
The chat textarea was fully disabled during response generation,
preventing users from drafting their next message while waiting.

Removed the `disabled={loading}` attribute from the textarea element.
Submission is still blocked during generation by the early return in
`handleSubmit()` and the submit button's own disabled state.

Test plan:
- Ran on one machine. While a model was writing a really long poem, I
typed my next response. I couldn't submit it with Enter and the button
still said "Processing" greyed out. I could send the message after
generation finished.
2026-02-10 16:12:08 +00:00
Jake Hillion 5b5577bead build-app: upload DMG to S3 for non-tagged builds (#1428)
Non-tagged builds (test-app branch, manual dispatch) only uploaded the
DMG as a GitHub artifact, which requires authentication to download.

Added an early exit path that uploads the DMG with a commit hash suffix
(EXO-<sha>.dmg) for non-tagged builds, making it publicly accessible
via S3.

Test plan:
- CI
-
https://github.com/exo-explore/exo/actions/runs/21837274032/job/63011907978
  worked as intended

Co-authored-by: rltakashige <rl.takashige@gmail.com>
2026-02-10 15:47:49 +00:00
Evan Quiney 8314a2aa78 cleaning up the todos (#1406)
kinda closes #1400 ( a bit )
2026-02-10 12:35:29 +00:00
Alex Cheema 163cf18384 Add error handling to info gatherer monitor loops (#1422)
## Motivation

If any of the `InfoGatherer` monitor loops throw an unexpected
exception, the entire monitoring task crashes and never recovers. This
can silently stop memory, network, or Thunderbolt data collection for
the lifetime of the process.

## Changes

Wrap the body of each `while True` monitor loop in a try/except that
logs the exception as a warning and continues to the next iteration. The
sleep at the end of each loop runs regardless, providing natural backoff
before retry.

Affected methods: `_monitor_misc`,
`_monitor_system_profiler_thunderbolt_data`, `_monitor_memory_usage`,
`_watch_system_info`, `_monitor_thunderbolt_bridge_status`.

`_monitor_macmon` already had its own error handling so was left as-is.

## Why It Works

A transient error (e.g., a subprocess failing, a permission issue) in
one iteration no longer kills the loop. The warning log provides
visibility while the monitor continues collecting data on subsequent
iterations.

## Test Plan

### Automated Testing
- `uv run basedpyright` — 0 errors
- `uv run ruff check` — passes

Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com>
Co-authored-by: rltakashige <rl.takashige@gmail.com>
2026-02-10 12:24:43 +00:00
rltakashige 2204f651c8 Yield from reachability checks (#1427)
## Motivation

check_reachable waits for all connection profile checks to be completed.
Since there are retries on failures, this can take around 20s to
resolve, preventing any instances from showing up. This feels very slow
for UX, and it slows down distributed testing.

## Changes

Made check_reachable an async generator.

## Test Plan

### Manual Testing
Works for me at least.
2026-02-10 12:18:45 +00:00
rltakashige 4abdaaf74b Address GPU timeouts (#1429)
## Motivation

For large prompts and/or slow machines, users are running into GPU
timeout errors very often.

## Changes

Only during prefill, we eval distributed operations. We don't do this
during decode to maintain decode performance.
Raise the prefill step size to 8192 because now we can (we see a speedup
here).
We also now see a 2x speedup in pipeline parallel prefill by disabling
an unnecessary all_gather during prefill.

## Why It Works

GPU timeout errors happen in the Metal backend when GPU operations take
too long without making progress.
By isolating distributed operations, we can allow them to run without
any timeouts.

## Test Plan

### Manual Testing
Doesn't GPU timeout on 100k tokens on Minimax anymore. Also tested on
Kimi.

### Automated Testing
Needs more exo bench, but I think this is a good step in the right
direction.
2026-02-10 11:53:23 +00:00
ciaranbor 2fbdb27bb1 Handle config.json not found (image models) (#1408)
## Motivation

When downloading image models, a missing config.json file triggers a
FileNotFoundError inside download_file_with_retry. This error was being
caught by the generic except Exception handler and retried 3 times
before failing. Then, the whole thing would be retried from the start

## Changes

- src/exo/download/download_utils.py: Added FileNotFoundError to the
list of immediately-raised exceptions in download_file_with_retry,
alongside HuggingFaceAuthenticationError. This prevents useless retries
when a file genuinely doesn't exist on the remote.
- src/exo/master/api.py: Wrapped ModelCard.load(model_id) in a
try/except that converts failures into an HTTPException(400) with a
descriptive error message, giving API consumers a clear error response.

## Why It Works


- FileNotFoundError is a deterministic error — the file won't appear on
retry, so re-raising immediately avoids 3 wasted download attempts with
exponential backoff.
- Catching ModelCard.load() failures and returning a 400 HTTP response
prevents unhandled exceptions from surfacing as opaque 500 errors in the
API.

## Test Plan

### Manual Testing

Verified an image model not in model cards does not cause an infinite
error loop
2026-02-07 03:34:58 +00:00
ciaranbor 3f57416dbf Add image lightbox (#1414)
## Motivation

No way to view generated or attached images at full resolution in the
dashboard

## Changes

- New ImageLightbox.svelte — fullscreen overlay with download, close
(click-outside/Escape), and transitions
- ChatMessages.svelte — all images (input attachments + generated) are
now clickable to open in lightbox; added expand button to generated
image hover overlay

## Why It Works

Single expandedImageSrc state variable drives the lightbox — set it to
show, null to hide.

## Test Plan

### Manual Testing

  - Click any image (attachment thumbnail or generated) → lightbox opens
  - Close via Escape, click-outside, or close button
  - Download button saves with correct extension
2026-02-07 01:30:03 +00:00
rltakashige 8f3681cf7e Synchronize before warmup (#1419)
## Motivation

Maybe addresses #1303 

## Changes

Add an mx barrier before warmup

## Why It Works

It might, it might not. Shouldn't break anything that's not already
broken though.

## Test Plan

### Manual Testing
The two machines I tested on were fine on GLM 4.7 Flash 8bit (the one in
exo.log in the issue). Obviously not definitive for anything, however.

<img width="594" height="878" alt="image"
src="https://github.com/user-attachments/assets/534d3ad6-16ef-4cb5-b823-43c8d4e1d3c6"
/>
2026-02-07 00:14:19 +00:00
ciaranbor 9dc4f786bd Ciaran/image model listing (#1417)
## Motivation

Image models (FLUX, Qwen Image) had no family grouping or quantization
metadata in the dashboard

## Changes

- Added family, quantization, base_model, and capabilities fields to all
18 image model TOML cards (FLUX.1 variants + Qwen Image variants)
  - Added FLUX and Qwen Image SVG logos to FamilyLogos.svelte
- Added "flux" and "qwen-image" families to the sidebar and family sort
order
- Added "Image Gen" and "Image Edit" capability filters in
ModelFilterPopover.svelte
  - Added image edit icon/badge to ModelPickerGroup.svelte
- Made the model category sidebar scrollable to accommodate the new
entries
  - Hidden scrollbars on model list panels

## Why It Works

Reuses the existing family/quantization grouping infrastructure that
LLMs already use, extending it to image models with appropriate metadata
and icons

## Test Plan

### Manual Testing

Verified image models behave like text models in the model list dialog

---------

Co-authored-by: Alex Cheema <41707476+AlexCheema@users.noreply.github.com>
2026-02-06 16:08:57 -08:00
rltakashige dcb4cabc15 Update the nix hash for mlx 0.30.5 (#1416)
## Motivation

<!-- Why is this change needed? What problem does it solve? -->
<!-- If it fixes an open issue, please link to the issue here -->

## Changes

<!-- Describe what you changed in detail -->

## Why It Works

<!-- Explain why your approach solves the problem -->

## Test Plan

### Manual Testing
<!-- Hardware: (e.g., MacBook Pro M1 Max 32GB, Mac Mini M2 16GB,
connected via Thunderbolt 4) -->
<!-- What you did: -->
<!-- - -->

### Automated Testing
<!-- Describe changes to automated tests, or how existing tests cover
this change -->
<!-- - -->
2026-02-06 21:27:10 +00:00
Jake Hillion d79b3a0e75 bench: make exo-bench available via nix run on all platforms (#1415)
exo-bench was gated behind isDarwin in python/parts.nix because it used
exoVenv, which pulls in MLX (Darwin-only). However, exo_bench.py is an
HTTP client that only needs loguru, transformers, huggingface-hub, and
tiktoken.

Made bench a uv workspace member with its own pyproject.toml declaring
only the minimal dependencies. Added a separate benchVenv in parts.nix
built from that workspace member, and moved exo-bench out of the
isDarwin block so it is available on all platforms.

Test plan:
- `nix run .#exo-bench -- --help` prints argparse help

---------

Co-authored-by: Ryuichi Leo Takashige <leo@exolabs.net>
Co-authored-by: rltakashige <rl.takashige@gmail.com>
2026-02-06 21:07:17 +00:00
Evan Quiney a2f1d48712 slow down catchup (#1407)
our event log request blasted the whole event log over libp2p, now it
just does the next 1000 messages - hopefully allowing nodes to catch up
a bit more consistently for long lived clusters

Co-authored-by: rltakashige <rl.takashige@gmail.com>
2026-02-06 20:45:27 +00:00
rltakashige 3b2f553a25 Fix kimi tool calling id (#1413)
## Motivation

Kimi produces its own tool id. It gets confused when we generate our own
id.

## Changes

Add id to tool call item and parse Kimi id properly.

## Test Plan

### Manual Testing
<img width="3198" height="522" alt="image"
src="https://github.com/user-attachments/assets/d71ec2be-7f57-49dc-a569-d304cc430f4d"
/>

Long running Kimi K2.5 cluster querying itself through OpenCode running
on the same Kimi K2.5 instance.
2026-02-06 11:33:51 -08:00
rltakashige 5455a97a8c Fix GLM4Moe Tensor Sharding (#1411)
## Motivation

Recent commit broke glm (non lite) sharding

## Why It Works

Assert is no longer hit, as isinstance check includes
GLM4MoeDecoderLayer.
Added type stubs to keep the type checker happy.

## Test Plan

### Manual Testing
Runs as expected without gibberish.
2026-02-06 16:53:15 +00:00
ciaranbor 6f0cb99919 Ciaran/flux1 kontext (#1394)
## Motivation

Add support for FLUX.1-Kontext-dev, an image editing variant of
FLUX.1-dev

## Changes

- New FluxKontextModelAdapter: Handles Kontext's image-to-image workflow
- encodes input image as conditioning latents with special position IDs,
generates from pure noise
- Model config: 57 transformer blocks (19 joint + 38 single), guidance
scale 4.0, ImageToImage task
- Pipeline updates: Added kontext_image_ids property to PromptData
interface, passed through diffusion runner
  - Model cards: Added TOML configs for base, 4-bit, and 8-bit variants
  - Dependency: mflux 0.15.4 → 0.15.5
- Utility: tmp/quantize_and_upload.py for quantizing and uploading
models to HuggingFace

## Test Plan

### Manual Testing

Works better than Qwen-Image-Edit
2026-02-06 16:20:31 +00:00
ciaranbor c8d3154f83 More image dimensions (#1395)
## Motivation

More dimensions for image generation

## Changes

- dashboard/src/lib/components/ImageParamsPanel.svelte: Added
"1024x1365" and "1365x1024" to the sizeOptions array
- dashboard/src/lib/stores/app.svelte.ts: Extended the size type in
ImageGenerationParams interface to include the two new dimension options
2026-02-06 15:59:06 +00:00
ciaranbor 63e9cc4fea Ciaran/num sync steps (#1396)
## Motivation

Allow users to directly configure num_sync_steps for distributed image
generation instead of deriving it from a factor of total steps.

## Changes

  - Added num_sync_steps field to AdvancedImageParams API (range 1-50)
- Changed model configs from num_sync_steps_factor: float to
num_sync_steps: int
  - Updated Flux/Qwen configs with direct values (1, 4, 7 respectively)
  - Added slider control in dashboard advanced params panel
  - Falls back to model default when not specified

## Why It Works

Decouples sync steps from inference steps, giving users direct control
over distributed inference synchronization while preserving sensible
defaults.

## Test Plan

### Manual Testing

  - Generate images with various sync step values via dashboard slider
  - Verify default behavior when parameter is unset
2026-02-06 15:51:46 +00:00
Evan Quiney 9b5cae3db6 auto bench (#1405)
runs exo_bench remotely with some nice git QoL

## usage
run tests/auto_bench.sh host1 [host2]

exo bench will be run on those hosts and its output saved to
bench/commit_hash/*.json on all models currently downloaded
2026-02-06 15:35:46 +00:00
Jake Hillion cf7201f91e pyproject: set minimum uv version
The uv.lock is churning constantly as different UV versions bounce it
between revisions. This is made worse by GitHub automatically hiding the
uv.lock changes, meaning it's hard to notice when this went wrong.

Set a minimum version for `uv` in pyproject.toml to fix this. I tried
quite a few versions (not all) and found 0.8.6 sets the revision to 3,
which I believe is the latest. This is from August 2025 so has been
around for a while.

Test plan:

```
jake@maverick:/data/users/jake/repos/exo/ > git checkout main uv.lock
jake@maverick:/data/users/jake/repos/exo/ > nix shell github:nixos/nixpkgs/3dce7f4a77812afd69efcbfe15e5223f98c5c69e#uv --command sh -c 'uv add pip --frozen && uv lock && uv remove pip --frozen && uv lock && uv --version'

Resolved 140 packages in 147ms
Added pip v26.0.1
Resolved 139 packages in 48ms
Removed pip v26.0.1
uv 0.8.6
```
2026-02-06 15:28:10 +00:00
rltakashige b315035ae0 Add minimax and fix qwen sharding strategies (#1318)
## Motivation

MiniMax tensor sharding does not provide equivalent outputs to running
it as a single node because RMSNorm weights cannot be split without
affecting the output.

Qwen3Next sharding was broken, and something with Qwen3MoE was likely
changed upstream, as several variables no longer exist.

This also ballooned into fixing prefix caching for non-standard models
as Qwen3Next was behaving weirdly.

## Changes

<!-- Describe what you changed in detail -->

## Why It Works

<!-- Explain why your approach solves the problem -->

## Test Plan

### Manual Testing
Worked for a 8 hour long eval at the same performance and a more similar
completion/reasoning token distribution.

---------

Co-authored-by: Alex Cheema <41707476+AlexCheema@users.noreply.github.com>
Co-authored-by: Alex Cheema <alexcheema123@gmail.com>
Co-authored-by: Claude Opus 4.5 <noreply@anthropic.com>
Co-authored-by: Evan <evanev7@gmail.com>
2026-02-06 13:26:59 +00:00
rltakashige c8dbbee27b skip tensor ring on bench (#1403)
## Motivation

<!-- Why is this change needed? What problem does it solve? -->
<!-- If it fixes an open issue, please link to the issue here -->

## Changes

<!-- Describe what you changed in detail -->

## Why It Works

<!-- Explain why your approach solves the problem -->

## Test Plan

### Manual Testing
<!-- Hardware: (e.g., MacBook Pro M1 Max 32GB, Mac Mini M2 16GB,
connected via Thunderbolt 4) -->
<!-- What you did: -->
<!-- - -->

### Automated Testing
<!-- Describe changes to automated tests, or how existing tests cover
this change -->
<!-- - -->
2026-02-06 13:06:59 +00:00
rltakashige f0107e9670 Fix offline no cache (#1402)
## Motivation

In offline mode, exo complains if there is no caches directory, even if
the files are there.

## Changes

Check safetensors index and the directory structure to build caches
directory.

## Test Plan

### Manual Testing
<img width="2338" height="1102" alt="image"
src="https://github.com/user-attachments/assets/ad769911-399b-4fca-ac80-aeaa046af06b"
/>
<img width="656" height="1668" alt="image"
src="https://github.com/user-attachments/assets/6080986c-3904-4600-a340-8c70f1b33266"
/>
2026-02-06 12:57:01 +00:00
Hunter Bown 9f502793c1 fix: retry downloads on transient errors instead of breaking (#1398)
## Motivation

`download_file_with_retry()` has a `break` in the generic exception
handler that exits the retry loop after the first transient failure.
This means network timeouts, connection resets, and server errors all
cause an immediate download failure — the two remaining retry attempts
never run.

## Changes

**download_utils.py**: Replaced `break` with logging and exponential
backoff in the generic exception handler, matching the existing
rate-limit handler behavior.

Before:
```python
except Exception as e:
    on_connection_lost()
    if attempt == n_attempts - 1:
        raise e
    break  # exits loop immediately
```

After:
```python
except Exception as e:
    on_connection_lost()
    if attempt == n_attempts - 1:
        raise e
    logger.error(f"Download error on attempt {attempt + 1}/{n_attempts} ...")
    logger.error(traceback.format_exc())
    await asyncio.sleep(2.0**attempt)
```

## Why It Works

The `break` statement was bypassing the retry mechanism entirely.
Replacing it with the same log-and-backoff pattern used by the
`HuggingFaceRateLimitError` handler means all 3 attempts are actually
used before giving up. The exponential backoff (1s, 2s) gives transient
issues time to resolve between attempts.

## Test Plan

### Manual Testing
- Downloads that hit transient network errors now retry instead of
failing immediately

### Automated Testing
- `uv run basedpyright` — 0 errors
- `uv run ruff check` — passes
- `uv run pytest src/exo/download/tests/ -v` — 11 tests pass

---------

Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com>
Co-authored-by: rltakashige <rl.takashige@gmail.com>
2026-02-06 11:51:54 +00:00
Evan Quiney c8371349d5 add scripts (#1401)
allow running exo-bench and the headless runner from nix
2026-02-06 11:06:40 +00:00
Evan Quiney 6b907398a4 cancel downloads for deleted instances (#1393)
after deleting an instance, if a given (node_id, model_id) pair doesn't exist in the left over instances, cancel the download of model_id on node_id.
2026-02-05 18:16:43 +00:00
Evan Quiney 572e647908 better cancellation (#1388)
a lot of our cleanup logic wasn't running leading to bad shutdown states

## changes
- added `try: except` blocks around most task groups
- made the runner shutdown code synchronous
- abandon the MpReceiver's recv_async thread on cancellation
- this only occurs during runner shutdown, the queue closing from the
other end should terminate the mp.Queue, cleaning up the thread in its
own time. i could try other methods if this is not sufficient.

## outcome
ctrl-c just works now! minus the tokio panic of course :) no more
hypercorn lifespan errors though!
2026-02-05 15:22:33 +00:00
Evan Quiney e59ebd986d set exo as the nix default package (#1391)
!!!
2026-02-05 15:15:52 +00:00
Alex Cheema 5c2f29f3f2 feat: show download availability in model picker (#1377)
## Motivation

Users browsing models in the picker need to know which models are
already downloaded and ready to run on their cluster, without having to
check the downloads page separately.

## Changes

- **ModelPickerModal.svelte**: Computes per-model download availability
by checking which nodes have `DownloadCompleted` entries and summing
their total RAM against the model's storage size. Passes availability
data to `ModelPickerGroup`. Enhances the info modal with a "Downloaded
on:" section showing node friendly names with green badges.
- **ModelPickerGroup.svelte**: Accepts new `downloadStatus` prop. Shows
a green checkmark-in-circle icon next to models that are downloaded on
sufficient nodes. Tooltip shows which nodes have the model.
- **+page.svelte**: Passes `downloadsData` and `topologyNodes` to
`ModelPickerModal`.

## Why It Works

The download state from `/state` already tracks per-node completed
downloads. The shared `getNodesWithModelDownloaded()` utility (from PR
#1375) finds nodes with `DownloadCompleted` entries for each model.
Total RAM is summed from the topology node data (using `ram_total`, not
`ram_available`) and compared to the model's `storage_size_megabytes` to
determine if there's enough aggregate memory. This is intentionally a
simple heuristic — not a full placement preview.

## Test Plan

### Manual Testing
<!-- Hardware: (e.g., MacBook Pro M1 Max 32GB, Mac Mini M2 16GB,
connected via Thunderbolt 4) -->
<!-- What you did: -->
- Open the model picker modal
- Verify downloaded models show a green checkmark icon
- Verify the checkmark appears dimmer for models downloaded on nodes
with insufficient total RAM
- Click the (i) info button on a downloaded model
- Verify "Downloaded on:" section appears with correct node names
- Verify models with no downloads show no indicator

### Automated Testing
- Dashboard builds successfully (`npm run build`)
- No new Python changes requiring type checking

> **Note:** This is a chained PR. Base branch is
`alexcheema/topology-download-indicators` (#1375).

🤖 Generated with [Claude Code](https://claude.com/claude-code)

---------

Co-authored-by: Claude Opus 4.5 <noreply@anthropic.com>
2026-02-05 14:32:53 +00:00
Alex Cheema ffe6396c91 Add Qwen3-Coder-Next model cards (#1367)
## Motivation

Qwen3-Coder-Next just dropped on mlx-community in several quantizations.
It's an 80B MoE model (Qwen3NextForCausalLM) which we already have
tensor parallelism support for via QwenShardingStrategy — just needs
model cards.

## Changes

Added model cards for all 5 available quantizations:
- `mlx-community/Qwen3-Coder-Next-4bit` (~46GB)
- `mlx-community/Qwen3-Coder-Next-5bit` (~58GB)
- `mlx-community/Qwen3-Coder-Next-6bit` (~69GB)
- `mlx-community/Qwen3-Coder-Next-8bit` (~89GB)
- `mlx-community/Qwen3-Coder-Next-bf16` (~158GB)

All with `supports_tensor = true` since the architecture is already
supported.

## Why It Works

`Qwen3NextForCausalLM` is already handled by QwenShardingStrategy in
auto_parallel.py and is in the supports_tensor allowlist in
model_cards.py. No code changes needed — just the TOML card files.

## Test Plan

### Manual Testing
<!-- n/a - model card addition only -->

### Automated Testing
- `basedpyright` — 0 errors
- `ruff check` — passes
- `nix fmt` — no changes
- `pytest` — 173 passed, 1 skipped


🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-authored-by: Claude Opus 4.5 <noreply@anthropic.com>
2026-02-05 13:37:18 +00:00
Jake Hillion 3a9baeb9db EXO: add CLI flags for root install/uninstall
The macOS app required user interaction via AppleScript prompts to
install or uninstall network configuration components, making automated
deployments difficult.

Added --install and --uninstall command line flags that execute the
network setup scripts directly when running as root, bypassing GUI
prompts. Created a new main.swift entry point that parses CLI arguments
and delegates to NetworkSetupHelper's new direct execution methods.

This enables headless installation via `sudo EXO --install` for
automated deployment scenarios while preserving the existing GUI
behavior when launched normally.

Test plan:

- Deployed to a machine that didn't have the content installed. Got
  blocked on the popup and EXO never launched.
- Relaunched EXO, confirmed it still never starts because of the popup.
- Ran `sudo /Applications/EXO.app/Contents/MacOS/EXO --install`
- Launched EXO - the API started as expected.
- Ran `sudo /Applications/EXO.app/Contents/MacOS/EXO --uninstall`
- Launched EXO - got the popup.
2026-02-05 13:27:46 +00:00
Alex Cheema 01b86a9e81 feat: add uncertainty visualization with token-level logprobs (#1180)
## Motivation

Adds uncertainty visualization to the chat interface, allowing users to
see token-level confidence scores and regenerate responses from any
point in the generation. This enables users to:
- Understand model confidence at each token
- Explore alternative completions by regenerating from uncertain tokens
- Debug and analyze model behavior

## Changes

### Uncertainty Visualization
- Add `TokenHeatmap` component showing token-level probability coloring
- Toggle uncertainty view per message with bar chart icon
- Display tooltip with probability, logprob, and top alternative tokens
on hover

### Regenerate from Token
- Add "Regenerate from here" button in token tooltip
- Use `continue_final_message` in chat template to continue within same
turn (no EOS tokens)
- Add `continue_from_prefix` flag to `ChatCompletionTaskParams`

### Request Cancellation
- Add `AbortController` to cancel in-flight requests when regenerating
mid-generation
- Handle `BrokenResourceError` server-side when client disconnects
gracefully

### Additional APIs
- Add Claude Messages API support (`/v1/messages`)
- Add OpenAI Responses API support (`/v1/responses`)

## Why It Works

- **Proper continuation**: Using `continue_final_message=True` instead
of `add_generation_prompt=True` keeps the assistant turn open, allowing
the model to continue naturally from the prefix without end-of-turn
markers
- **Clean cancellation**: AbortController aborts the HTTP request, and
server catches `BrokenResourceError` to avoid crashes
- **Stable hover during generation**: TokenHeatmap tracks hover by index
(stable across re-renders) with longer hide delay during generation

## Test Plan

### Manual Testing
<!-- Hardware: MacBook Pro M1 -->
- Send a message and verify logprobs are collected
- Enable uncertainty view and verify token coloring based on probability
- Hover over tokens to see tooltip with alternatives
- Click "Regenerate from here" on a token mid-response
- Verify the response continues naturally from that point
- Verify aborting mid-generation and regenerating works without server
crash

### Automated Testing
- Added tests for Claude Messages API adapter
- Added tests for OpenAI Responses API adapter

🤖 Generated with [Claude Code](https://claude.com/claude-code)

---------

Co-authored-by: Claude Opus 4.5 <noreply@anthropic.com>
Co-authored-by: Evan <evanev7@gmail.com>
2026-02-05 05:21:26 -08:00
rltakashige 221640a65b Acknowledge task after runner status is updated (#1381)
## Motivation

Duplicate tasks are still observed.

## Changes

Moved task acknowledgement to after the runner has changed its status.

## Why It Works

Tasks now remain pending until the runner has updated its status.

## Test Plan

### Manual Testing
Seems to work fine from manual testing. Hard to test a race condition
though.

### Automated Testing
Updated the event ordering test.
2026-02-05 12:00:37 +00:00
ciaranbor 6177550c34 Ciaran/parallel cfg (#1361)
## Motivation

Enable parallel classifier-free guidance (CFG) for Qwen image models.
CFG requires two forward passes (positive/negative prompts) - this
allows them to run on separate nodes simultaneously, reducing latency.

## Changes

  - Added uses_cfg flag to ModelCard to identify CFG-based models
- Extended PipelineShardMetadata with CFG topology fields (cfg_rank,
cfg_world_size, peer device info)
- Updated placement to create two CFG groups with reversed ordering
(places CFG peers as ring neighbors)
- Refactored DiffusionRunner to process CFG branches separately with
exchange at last pipeline stage
- Added get_cfg_branch_data() to PromptData for single-branch embeddings
  - Fixed seed handling in API for distributed consistency
  - Fixed image yield to only emit from CFG rank 0 at last stage
  - Increased num_sync_steps_factor from 0.125 to 0.25 for Qwen

## Why It Works

- 2 nodes + CFG: Both run all layers, process different CFG branches in
parallel
  - 4+ even nodes + CFG: Hybrid - 2 CFG groups × N/2 pipeline stages
  - Odd nodes or non-CFG: Falls back to pure pipeline parallelism
 
 Ring topology places CFG peers as neighbors to enable direct exchange.

## Test Plan

### Manual Testing

Verified performance gain for Qwen-Image for 2 node and 4 node cluster.
Non-CFG models still work

### Automated Testing
 
Added tests in test_placement_utils.py covering 2-node CFG parallel,
4-node hybrid, odd-node fallback, and non-CFG pipeline modes.
2026-02-04 21:16:35 +00:00
Evan Quiney 7b6cad94c6 add resources dir to nix (#1376)
add resources directory to the nix exo package, and fixes the env for the dashboard dir
2026-02-04 16:38:43 +00:00
Alex Cheema 41ed7afb3b feat: add model picker modal with grouped models and HF Hub search (#1369)
## Motivation

Reimplements the model picker modal from #1191 on top of the custom
model support branch. Replaces the inline model dropdown with a
full-featured modal that groups models by base model, supports
filtering, favorites, and HuggingFace Hub search.

## Changes

**Backend:**
- Add `family`, `quantization`, `base_model`, `capabilities` metadata
fields to `ModelCard` and all 40 TOML model cards
- Pass new fields through `ModelListModel` and `get_models()` API
response
- Add `GET /models/search` endpoint using
`huggingface_hub.list_models()`

**Dashboard (7 new files):**
- `ModelPickerModal.svelte` — Main modal with search, family filtering,
HuggingFace Hub tab
- `ModelPickerGroup.svelte` — Expandable model group row with
quantization variants
- `FamilySidebar.svelte` — Vertical sidebar with family icons (All,
Favorites, Hub, model families)
- `FamilyLogos.svelte` — SVG icons for each model family
- `ModelFilterPopover.svelte` — Capability and size range filters
- `HuggingFaceResultItem.svelte` — HF search result item with
download/like counts
- `favorites.svelte.ts` — localStorage-backed favorites store

**Integration:**
- Replace inline dropdown in `+page.svelte` with button that opens
`ModelPickerModal`
- Custom models shown in Hub tab with delete support

**Polish:**
- Real brand logos (Meta, Qwen, DeepSeek, OpenAI, GLM, MiniMax, Kimi,
HuggingFace) from Simple Icons / LobeHub
- Clean SVG stroke icons for capabilities (thinking, code, vision, image
gen)
- Consistent `border-exo-yellow/10` borders, descriptive tooltips
throughout
- Cluster memory (used/total) shown in modal header
- Selected model highlight with checkmark for both single and
multi-variant groups
- Cursor pointer on all interactive elements, fix filter popover
click-outside bug
- Custom models now appear in All tab alongside built-in models

## Bug Fix: Gemma 3 EOS tokens

Also included in this branch: fix for Gemma 3 models generating infinite
`<end_of_turn>` tokens. The tokenizer's `eos_token_ids` was missing
token ID 106 (`<end_of_turn>`), so generation never stopped. The fix
appends this token to the EOS list after loading the tokenizer. Also
handles `eos_token_ids` being a `set` (not just a `list`).

## Why It Works

Model metadata (family, capabilities, etc.) is stored directly in TOML
cards rather than derived from heuristics, ensuring accuracy. The modal
groups models by `base_model` field so quantization variants appear
together. Custom models are separated into the Hub tab since they lack
grouping metadata.

## Test Plan

### Manual Testing
- Open dashboard, click model selector to open modal
- Browse models by family sidebar, search, and filters
- Expand model groups to see quantization variants
- Star favorites and verify persistence across page reloads
- Navigate to Hub tab, search and add models
- Verify error messages shown for invalid model IDs
- Run a Gemma 3 model and verify generation stops at `<end_of_turn>`

### Automated Testing
- `uv run basedpyright` — 0 errors
- `uv run ruff check` — passes
- `nix fmt` — clean
- `uv run pytest src/` — 173 passed
- `cd dashboard && npm run build` — builds successfully

---------

Co-authored-by: Claude Opus 4.5 <noreply@anthropic.com>
2026-02-04 05:56:23 -08:00
Alex Cheema 2063278906 feat: add custom HuggingFace model support (#1368)
## Motivation

Users should be able to run any HuggingFace model, not just the ones we
ship TOML cards for. Continues the aim of #1191 with a minimal
implementation on top of the current TOML model card system.

Custom cards are saved to `~/.exo/custom_model_cards/` rather than the
bundled `resources/inference_model_cards/` because `RESOURCES_DIR` is
read-only in PyInstaller bundles (`sys._MEIPASS`). This also fixes
`fetch_from_hf` which was saving cards to the wrong path (`resources/`
root instead of `resources/inference_model_cards/`).

## Changes

- Add `EXO_CUSTOM_MODEL_CARDS_DIR` constant
(`~/.exo/custom_model_cards/`)
- Update `model_cards.py`: add custom dir to search path, fix
`save_to_custom_dir`, add `delete_custom_card`/`is_custom_card`
- Add `POST /models/add` and `DELETE /models/custom/{model_id}` API
endpoints
- Add `is_custom` field to `ModelListModel` API response
- Dashboard: add custom model input form in dropdown, delete button for
custom models, show actual API errors, auto-select newly added model

## Why It Works

Two separate directories for model cards: the bundled read-only
`resources/inference_model_cards/` for built-in cards, and user-writable
`~/.exo/custom_model_cards/` for custom cards. Both are scanned when
listing models. This works in all environments including PyInstaller
bundles where `RESOURCES_DIR` points to `sys._MEIPASS`.

## Test Plan

### Manual Testing
- Add a custom model via the dropdown (e.g.
`mlx-community/Llama-3.2-1B-Instruct-4bit`)
- Verify it appears in the model list with the delete (x) button
- Delete it and verify it disappears
- Try adding an invalid model ID and verify the actual error is shown

### Automated Testing
- `uv run basedpyright` — 0 errors
- `uv run ruff check` — passes
- `uv run pytest src/` — passes
- `cd dashboard && npm run build` — builds

---------

Co-authored-by: Claude Opus 4.5 <noreply@anthropic.com>
2026-02-04 05:06:15 -08:00
rltakashige a0f4f36355 Reduce reliance on internet (#1363)
## Motivation

Offline users currently have to wait for every retry to fail before
being able to launch a model.
For users that restart clusters often or share API keys between devices,
we also spam HuggingFace with downloads every 5 minutes.
These issues are caused by _emit_existing_download_progress being
inefficient.

## Changes

- Only query HuggingFace once while EXO is running (assumption being
that a change should only be reflected on a new EXO session)
- Only query HuggingFace when there is an internet connection (polling
connectivity every 10 seconds)
- Request download progress if we switch from no connectivity ->
connected to reduce the wait.
- Reduce download progress sleep as it's no longer expensive (queries
cache most of the time).
- Reduce retries as 30 is way too many.

## Test Plan

### Manual Testing
Manually tested the behaviour.

### Automated Testing
None, should I add any? We do have some tests for this folder, but they
are probably not too helpful.
2026-02-03 20:03:29 +00:00
Alex Cheema acb97127bf Normalize TextGenerationTaskParams.input to list[InputMessage] (#1360)
## Motivation

With the addition of the Responses API, we introduced `str |
list[InputMessage]` as the type for `TextGenerationTaskParams.input`
since the Responses API supports sending input as a plain string. But
there was no reason to leak that flexibility past the API adapter
boundary — it just meant every downstream consumer had to do `if
isinstance(messages, str):` checks, adding complexity for no benefit.

## Changes

- Changed `TextGenerationTaskParams.input` from `str |
list[InputMessage]` to `list[InputMessage]`
- Each API adapter (Chat Completions, Claude Messages, Responses) now
normalizes to `list[InputMessage]` at the boundary
- Removed `isinstance(task_params.input, str)` branches in
`utils_mlx.py` and `runner.py`
- Wrapped string inputs in `[InputMessage(role="user", content=...)]` in
the warmup path and all test files

## Why It Works

The API adapters are the only place where we deal with raw user input
formats. By normalizing there, all downstream code (worker, runner, MLX
engine) can just assume `list[InputMessage]` and skip the type-checking
branches. The type system (`basedpyright`) catches any missed call sites
at compile time.

## Test Plan

### Automated Testing
- `uv run basedpyright` — 0 errors
- `uv run ruff check` — passes
- `nix fmt` — applied
- `uv run pytest` — 174 passed, 1 skipped

Co-authored-by: Claude Opus 4.5 <noreply@anthropic.com>
2026-02-03 06:01:56 -08:00
Evan Quiney d90605f198 migrate model cards to .toml files (#1354) 2026-02-03 12:32:06 +00:00
Evan Quiney f400b4d7c5 fix InstanceViewModel.swift (#1359)
wasn't caught when we merged the API changes
2026-02-02 18:43:27 +00:00
Evan Quiney d97bca88e6 improve distributed testing (#1300)
Our distributed test now does a full query cycle for every model loaded
onto the relevant machine. This will help find bugs early, as it already
has found one with Qwen3 Next! I didn't write down what the error was
though. Gooooooood luck with that!

Co-authored-by: rltakashige <rl.takashige@gmail.com>
2026-02-02 18:25:39 +00:00
Alex Cheema dfce188d99 fix: handle unclosed tool calls and GLM arg parsing edge cases (#1344)
## Motivation

Tool-call requests can hang indefinitely when `max_tokens` truncates
generation mid-tool-call.

## Reproduction

1. Send a chat completion with `tools` and a low `max_tokens` (e.g. 65)
to Qwen3-0.6B
2. Model generates `<think>...</think>` then starts `<tool_call>` but
`max_tokens` cuts it off before `</tool_call>`
3. **Before this fix:** `parse_tool_calls` buffers tokens after
`<tool_call>`, generator exhausts, buffered tokens (including
`finish_reason`) are silently dropped → stream hangs forever
4. **After this fix:** buffered tokens are flushed as regular text with
`finish_reason` propagated → response returns normally with
`finish_reason: "length"`

Confirmed with fresh local testing: 4 unclosed tool call flushes
triggered in a single session. Also confirmed via production logs from
Jan 29 (2 occurrences).

## Changes

1. **`parse_tool_calls` unclosed tool call flush** — when the generator
exhausts inside an open `<tool_call>` block, flush buffered tokens as
regular text and propagate `finish_reason`
2. **GLM regex fix** — match literal `\n` (not escaped `\\n`) between
arg tags; handle missing `</arg_value>` via lookahead
3. **7 new unit tests** for `parse_tool_calls` covering unclosed,
closed, passthrough, and failed-parse scenarios

## Why It Works

- `parse_tool_calls` now has a post-loop check: if `in_tool_call` is
still true, it yields the buffered text with the tracked `finish_reason`
instead of silently dropping it
- The GLM regex now matches real-world output where newlines appear
between tags and `</arg_value>` may be absent

## Test Plan

### Manual Testing
- Qwen3-0.6B-4bit with `tools` + various `max_tokens` values (61-75)
- Confirmed responses return with `finish_reason: "length"` instead of
hanging
- Log output shows `"generator exhausted inside unclosed tool call,
flushing buffered text"`

### Automated Testing
- 7 new tests in `test_parse_tool_calls.py`
- Full test suite passes (`uv run pytest`)

---------

Co-authored-by: Claude Opus 4.5 <noreply@anthropic.com>
Co-authored-by: Ryuichi Leo Takashige <leo@exolabs.net>
Co-authored-by: Evan <evanev7@gmail.com>
Co-authored-by: Jake Hillion <jake@hillion.co.uk>
Co-authored-by: rltakashige <rl.takashige@gmail.com>
2026-02-02 17:45:51 +00:00
Evan Quiney 54b19879a0 create config home when checking for config file (#1353)
we didn't check before, raising a critical exception.
now we create ~/.config/exo on linux systems before touching config.toml.

this wasn't caught before since everything lives in ~/.exo on macos, and we no longer write the keypair to CONFIG_HOME, so config.toml has to do init work it avoided before.
2026-02-02 17:36:51 +00:00
ciaranbor 19965c7ba5 Ciaran/profiling (#1345)
## Motivation

Know what the hell is going on

## Changes

- Tracing library (src/exo/shared/tracing.py): trace() context manager,
Chrome Trace Format export, statistics computation
- Runner instrumentation
(src/exo/worker/engines/image/pipeline/runner.py): Wrapped sync/async
steps, compute blocks, and send/recv operations
- Trace collection: Workers send traces to master after task completion;
merged into ~/.exo/traces/trace_{task_id}.json
  - API endpoints: List, fetch, stats, and raw download at /v1/traces/*
  - Dashboard: Trace list and detail pages with Perfetto integration

## Why It Works

<img width="1236" height="767" alt="Screenshot 2026-01-30 at 19 00 09"
src="https://github.com/user-attachments/assets/73e6e46d-ba10-4e83-ba99-ff1c3f62ab05"
/>
<img width="1659" height="89" alt="Screenshot 2026-01-30 at 19 00 58"
src="https://github.com/user-attachments/assets/c0fd0e65-e4fc-4fd5-920d-b43b2887d591"
/>
2026-02-02 17:19:45 +00:00
Evan Quiney 3e27ead705 remove mdns discovered peers from appearing in state (#1312)
## motivation
eagerly discovered peers through gossipsub were added to state. this left things looking broken from one-sided connections

## changes
the worker no longer writes topology edges from these gossipsub messages
we now strictly rely on http-discovered topology, which tends to be more reflective of the actual state of the systems connectivity
2026-02-02 16:58:53 +00:00
Alex Cheema d826d309b3 chore: gitignore hosts_*.json files (#1343)
## Motivation

`hosts_*.json` files are local host configuration snapshots that
shouldn't be tracked in version control.

## Changes

Added `hosts_*.json` pattern to `.gitignore`.

## Why It Works

The glob pattern `hosts_*.json` matches any file starting with `hosts_`
and ending with `.json` in the repo root.

## Test Plan

### Manual Testing
- Verified that `hosts_*.json` files are ignored by git after this
change.

### Automated Testing
- No automated tests needed for a `.gitignore` change.

Co-authored-by: Claude Opus 4.5 <noreply@anthropic.com>
2026-02-02 16:14:11 +00:00
Alex Cheema c3537980bd feat: add Claude Messages API and OpenAI Responses API support (#1167)
## Motivation

Add support for Claude Messages API and OpenAI Responses API to allow
users to interact with exo using these popular API formats. This enables
broader compatibility with existing tooling and SDKs that expect these
API formats.

## Architecture

Adapter logic lives exclusively in the API layer
(`src/exo/master/adapters/`). On the way in, each adapter converts its
API-specific request type (`ChatCompletionRequest`,
`ClaudeMessagesRequest`, `ResponsesRequest`) into
`TextGenerationTaskParams`. On the way out, each adapter converts the
`TokenChunk` stream back into its API-specific response format.
Everything inside the application — commands, worker, runner, event
sourcing — only sees `TextGenerationTaskParams` and `TokenChunk`. No
API-specific types cross the boundary.

```
                          API layer                         │  Application internals
                                                            │
Chat Completions → [adapter] → TextGenerationTaskParams ──→ │ ──→ TextGeneration command → Runner → TokenChunk ──→ │ ──→ [adapter] → ChatCompletionResponse
Claude Messages  → [adapter] → TextGenerationTaskParams ──→ │ ──→ TextGeneration command → Runner → TokenChunk ──→ │ ──→ [adapter] → ClaudeMessagesResponse
Responses API    → [adapter] → TextGenerationTaskParams ──→ │ ──→ TextGeneration command → Runner → TokenChunk ──→ │ ──→ [adapter] → ResponsesResponse
```

## Changes

### New Files
- `src/exo/shared/types/claude_api.py` - Pydantic types for Claude
Messages API
- `src/exo/shared/types/openai_responses.py` - Pydantic types for OpenAI
Responses API
- `src/exo/shared/types/text_generation.py` - Shared
`TextGenerationTaskParams` internal type
- `src/exo/master/adapters/chat_completions.py` - Chat Completions
adapter (streaming/non-streaming)
- `src/exo/master/adapters/claude.py` - Claude Messages adapter
(streaming/non-streaming)
- `src/exo/master/adapters/responses.py` - OpenAI Responses adapter
(streaming/non-streaming)

### Modified Files
- `src/exo/master/api.py` - Refactored to use adapters uniformly for all
endpoints; extracted `_resolve_and_validate_text_model` helper to
deduplicate model validation across all text endpoints; removed ad-hoc
`try/except ValueError` blocks from non-streaming paths

### New Endpoints
- `POST /v1/messages` - Claude Messages API (streaming and
non-streaming)
- `POST /v1/responses` - OpenAI Responses API (streaming and
non-streaming)

## Why It Works

All APIs are implemented as pure conversion adapters at the edge of the
application:
1. Adapter functions in `src/exo/master/adapters/` convert incoming
requests to `TextGenerationTaskParams`
2. `api.py` wraps the params in a `TextGeneration` command and sends it
through the existing command/event flow
3. The worker, runner, and event sourcing layers only handle
`TextGenerationTaskParams` and `TokenChunk` — they have no awareness of
Chat Completions, Claude, or Responses API formats
4. On response, adapter functions convert the `TokenChunk` stream back
to the caller's expected format
5. Model validation is handled by a single shared helper
(`_resolve_and_validate_text_model`), mirroring the existing
`_validate_image_model` pattern for image endpoints

No changes to core inference logic were needed.

### Streaming Formats
- **Chat Completions**: Uses `data: {...}\n\n` with `[DONE]` terminator
- **Claude**: Uses event types `message_start`, `content_block_start`,
`content_block_delta`, `content_block_stop`, `message_delta`,
`message_stop`
- **OpenAI Responses**: Uses event types `response.created`,
`response.in_progress`, `response.output_item.added`,
`response.content_part.added`, `response.output_text.delta`,
`response.output_text.done`, `response.content_part.done`,
`response.output_item.done`, `response.completed`

## Test Plan

### Manual Testing
Hardware: MacBook Pro M3 Max

**Non-streaming tests:**
```bash
# Chat Completions API
curl -X POST http://localhost:52415/v1/chat/completions \
  -H "Content-Type: application/json" \
  -d '{"model": "llama-3.2-1b", "messages": [{"role": "user", "content": "Hello"}], "max_tokens": 20}'

# Claude Messages API
curl -X POST http://localhost:52415/v1/messages \
  -H "Content-Type: application/json" \
  -d '{"model": "llama-3.2-1b", "max_tokens": 50, "messages": [{"role": "user", "content": "Hello"}]}'

# OpenAI Responses API
curl -X POST http://localhost:52415/v1/responses \
  -H "Content-Type: application/json" \
  -d '{"model": "llama-3.2-1b", "input": "Hello", "max_output_tokens": 20}'
```

**Streaming tests:**
```bash
# Chat Completions API (streaming)
curl -N -X POST http://localhost:52415/v1/chat/completions \
  -H "Content-Type: application/json" \
  -d '{"model": "llama-3.2-1b", "messages": [{"role": "user", "content": "Hello"}], "stream": true, "max_tokens": 20}'

# Claude Messages API (streaming)
curl -N -X POST http://localhost:52415/v1/messages \
  -H "Content-Type: application/json" \
  -d '{"model": "llama-3.2-1b", "max_tokens": 50, "messages": [{"role": "user", "content": "Hello"}], "stream": true}'

# OpenAI Responses API (streaming)
curl -N -X POST http://localhost:52415/v1/responses \
  -H "Content-Type: application/json" \
  -d '{"model": "llama-3.2-1b", "input": "Hello", "stream": true, "max_output_tokens": 20}'
```

All endpoints tested successfully with proper response formats and
streaming events.

### Automated Testing
- Tests in `src/exo/master/tests/` all pass (85 tests)
- Type checker (basedpyright) passes with 0 errors
- Linter (ruff) passes

---------

Co-authored-by: Claude Opus 4.5 <noreply@anthropic.com>
Co-authored-by: Evan <evanev7@gmail.com>
2026-02-02 15:58:37 +00:00
rltakashige 21d477f1cb Update exo bench (#1357)
## Motivation

Make exo bench faster for longer prompts, lengthen default timeouts and
use pairs for pp and tg.

## Changes

- Uses binary search to find correct prompt
- Flag to force all combinations if that is desired
2026-02-02 15:46:15 +00:00
Jake Hillion b2579c78fe nix: add macmon to PATH in wrapper scripts on Darwin
`nix run .#exo` couldn't find `macmon` because the Nix wrapper scripts
didn't include it in PATH, causing `shutil.which("macmon")` to fail.

Added `--prefix PATH : ${pkgs.macmon}/bin` to the `makeWrapper` call,
conditional on Darwin via `lib.optionalString`, so macmon's binary is
available at runtime without modifying the user's system PATH.

Test plan:
- Verified `nix build .#exo` succeeds
- Checked wrapper script contains macmon store path in PATH prefix
2026-02-02 13:42:36 +00:00
Evan Quiney cd946742f7 fix skipping logic in worker plan (#1342)
the worker plan function had some skipping logic missing, leading to
double-submitting tasks.
2026-01-30 14:31:40 +00:00
rltakashige a5bc38ad1f Check all nodes to evict (#1341)
## Motivation

If nodes have uneven memory, one node may evict cache that remains on
another node. This will break prefill on some setups.

## Changes

<!-- Describe what you changed in detail -->

## Why It Works

<!-- Explain why your approach solves the problem -->

## Test Plan

### Manual Testing
<!-- Hardware: (e.g., MacBook Pro M1 Max 32GB, Mac Mini M2 16GB,
connected via Thunderbolt 4) -->
<!-- What you did: -->
<!-- - -->

### Automated Testing
<!-- Describe changes to automated tests, or how existing tests cover
this change -->
<!-- - -->
2026-01-30 13:42:09 +00:00
Evan Quiney 2a4e0d4629 make node-ids unique per-session (#1338)
we currently have no strict reuqirements that node ids persist across
sessions, so we can generate fresh nodeids each time

this avoids issues like #1332, but prevents further features such as
caching downloads or node-id dialling

Co-authored-by: rltakashige <rl.takashige@gmail.com>
2026-01-30 13:33:31 +00:00
Evan Quiney 46a14153dd switch to ModelCard.load outside of download log (#1339)
some attempts to load model cards (i.e. build_base_shard) always went
through networking rather than using downloaded model cards. we should
always default to ModelCard.load in these scenarios
2026-01-30 11:20:20 +00:00
Evan 9ba61f3733 improve log message in shard downloader
closes #1336
2026-01-30 10:35:01 +00:00
rltakashige d9eca75895 Add usage stats (#1333)
## Motivation

(Probably) the final missing piece of the Chat Completions API 

## Changes

Add UsageStats 

## Why It Works

OpenCode reviewed my PR and gave me stats:

<img width="1150" height="802" alt="image"
src="https://github.com/user-attachments/assets/ebc06bae-797f-4087-87d5-2f26cf60fc48"
/>


## Test Plan

### Automated Testing
No tests were broken.
2026-01-30 10:23:08 +00:00
rltakashige 9dabde7e57 Fix bench after recent updates (#1331)
## Motivation

A lot of changes happened without much attention to the state of exo
bench.

## Changes

Use TaggedModel for BenchChatCompletion so it serialises properly.
Don't break after gpt oss tool call to preserve parity with the rest of
the codebase.

## Why It Works

<!-- Explain why your approach solves the problem -->

## Test Plan

### Manual Testing
<img width="2856" height="678" alt="image"
src="https://github.com/user-attachments/assets/2e18cf0d-c0f8-467c-9763-1a6a59c8a327"
/>

Also tested GPT OSS tool calling in OpenCode
2026-01-29 19:14:40 +00:00
ciaranbor a31942ce12 Ciaran/image non streaming (#1328)
## Motivation

The dashboard UI attempted to parse all image generation responses as
SSE streams, even when streaming was disabled. This broke non-streaming
image generation.

## Changes

- Parse JSON responses directly when not streaming, use SSE parser only
when stream=true AND partialImages > 0
- explicitly disable partial images when not streaming

## Why It Works

Both API and dashboard now use the same condition (stream &&
partialImages > 0) to determine response format, ensuring correct
parsing.

## Test Plan

### Manual Testing

Non-streamed image generation results appear in the UI. Streamed image
generation still works
2026-01-29 17:24:32 +00:00
Alex Cheema 7cc313b22a Treat Swift/Xcode build warnings as errors (#1322)
## Motivation

Warnings that go unchecked tend to accumulate and hide real issues.
Treating them as errors ensures they are addressed immediately, both
locally during development and in CI.

## Changes

Added `SWIFT_TREAT_WARNINGS_AS_ERRORS = YES` and
`GCC_TREAT_WARNINGS_AS_ERRORS = YES` to the **project-level** Debug and
Release build configurations in `project.pbxproj`. This applies to all
targets (EXO, EXOTests, EXOUITests).

## Why It Works

Xcode's `SWIFT_TREAT_WARNINGS_AS_ERRORS` and
`GCC_TREAT_WARNINGS_AS_ERRORS` build settings promote Swift and C/ObjC
warnings to errors at compile time. Setting them at the project level
means all targets inherit the policy without needing per-target or
CI-level overrides.

## Test Plan

### Manual Testing
- Built the EXO scheme in Release configuration with `xcodebuild` — no
warning-as-error failures from Swift or C/ObjC sources.

### Automated Testing
- CI already builds with `-configuration Release`, so it will
automatically enforce warnings-as-errors via the inherited project
settings — no CI changes needed.
2026-01-29 17:15:49 +00:00
rltakashige 2837225dc7 Load pipeline layers sequentially (#1329)
## Motivation

Slightly annoyed by needing this change, but same story as for tensor
loading...
2026-01-29 17:08:38 +00:00
Jake Hillion e4c6a7dbb4 nix: add Python packaging with uv2nix
Add uv2nix to build Python packages from uv.lock. This creates a fully
Nix-managed Python environment with the Rust bindings injected via overlay.

Changes:
- Add pyproject-nix, uv2nix, and pyproject-build-systems flake inputs
- Create python/parts.nix with overlays to inject Nix-built Rust wheel
- Export packages.exo on macOS (wraps exo/exo-master/exo-worker with dashboard)
- Add checks.lint (ruff, all platforms) and checks.pytest (macOS only)
- Simplify CI typecheck job using nicknovitski/nix-develop action
- Delete .github/actions/typecheck composite action (no longer needed)
- Add no-build-package for MLX packages in pyproject.toml (use wheels)

The Python build is currently macOS-only since MLX requires Metal. Linux
support will be added once the pyproject dependencies are simplified.

Test plan:
- Run `nix flake check` on macOS to verify pytest and lint pass
- Build exo package on macOS: `nix build .#exo`
- Verify CI pipeline passes with simplified typecheck job
2026-01-29 16:35:58 +00:00
Evan b1e88a3d06 shfmt
adds shfmt, a shell formatter, and formats the bash files
2026-01-29 15:24:36 +00:00
Jake Hillion ebeddfb308 mlx: build with Nix (#1285)
In order to make testing and deployment simpler and more reproducible,
we want to provide a Nix derivation for our macOS .app build. We already
build the Rust and dashboard with Nix, but so far the Python has been
blocked because we haven't had an MLX build.

This change adds a Metal compiler derivation that uses `requireFile` to
be provided a NAR of the unfree macOS Metal compiler. It is documented
how to get this file, but effectively you have to trigger the download,
mount the DMG, and NAR the result. Once this is added to the store by
hash we can build MLX using it. The MLX build itself is quite self
explanatory.

Test plan:
- CI. We follow the instructions to grab the Metal compiler. Once this
is in Cachix we should really never do this again, and I can pin the
path too to ensure it doesn't leave.
- MLX tests run as part of the MLX derivation's build. They pass.
- `NIXPKGS_ALLOW_UNFREE=1 nix build .#mlx.passthru.tests.mlxTest
--impure --option sandbox false`

---------

Co-authored-by: Ryuichi Leo Takashige <leo@exolabs.net>
2026-01-29 14:07:00 +00:00
Alex Cheema 9111575997 Add startup delay and update network setup message (#1309)
## Summary
- Add 20-second startup delay to wait for macOS to finish network setup
after boot
- Update user-facing message to clarify the service configures local
networking, disables Thunderbolt Bridge (preventing packet storms), and
installs a Network Location

## Test plan
- [ ] Manual verification of Swift syntax
- [ ] Test network setup on macOS device after reboot

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-authored-by: Claude Opus 4.5 <noreply@anthropic.com>
Co-authored-by: rltakashige <rl.takashige@gmail.com>
2026-01-29 13:05:50 +00:00
Sami Khan ffacabe7e4 Fix uninstall button error (#1306)
## Motivation

Fix "Network setup script failed" error when clicking uninstall button
and resolve Xcode compiler warnings.

## Changes

- NetworkSetupHelper.swift: Add || true guards and explicit return 0 in
find_and_enable_thunderbolt_bridge to prevent script failures with set
-euo pipefail
- ThunderboltBridgeService.swift: Use withCString and
withUnsafeMutablePointer for Authorization API calls to fix pointer
lifetime warnings
- EXOApp.swift: Mark showNotification as nonisolated to fix main actor
isolation warning

## Why It Works

- The uninstall script's Thunderbolt re-enable function could exit
non-zero in edge cases (no bridges, no matches). Since this is a cleanup
step, failures should not abort uninstall.
- Swift requires explicit pointer lifetime management when passing
strings/structs to C APIs.
- showNotification is called from a nonisolated delegate method and uses
thread-safe APIs.

## Test Plan

### Manual Testing
Hardware: MacBook Pro

- Clicked Uninstall button, verified it completes without error
- Built in Xcode, verified no warnings   

### Automated Testing
N/A
2026-01-29 12:57:48 +00:00
rltakashige 9e58a57599 Add RDMA caveats to README.md (#1316)
## Motivation

Running RDMA from source is not well documented as is. Several
surprising things that took time to debug internally too.

App should be updated to detect MacOS versions in future.
2026-01-28 18:44:00 +00:00
Evan Quiney 748a026071 fix configdata validation for kimi-k2 (#1314)
## motivation
our shard downloader could not correctly fetch data for kimi-k2, as it
deferred some values to a text_config field.
## changes
config_data now prioritizes this field if it exists in information like
layer_count
2026-01-28 14:29:36 +00:00
Alex Cheema f1a2d054ec Update tagline to "Run frontier AI locally" (#1313)
- Update README tagline from "Run your own AI cluster at home with
everyday devices" to "Run frontier AI locally"
2026-01-28 12:38:14 +00:00
Alex Cheema b3c8f85fc8 Update MLX to 0.30.4 (#1311)
## Summary
- Bump mlx from 0.30.3 to 0.30.4

## Test plan
- [x] `uv lock` succeeds
- [x] Type checking passes (`uv run basedpyright`)
- [x] Run inference tests

🤖 Generated with [Claude Code](https://claude.com/claude-code)

---------

Co-authored-by: Claude Opus 4.5 <noreply@anthropic.com>
2026-01-28 04:30:21 -08:00
rltakashige a562114ba5 Add Kimi K2.5 support (#1302)
## Motivation

<!-- Why is this change needed? What problem does it solve? -->
<!-- If it fixes an open issue, please link to the issue here -->

## Changes

<!-- Describe what you changed in detail -->

## Why It Works

<!-- Explain why your approach solves the problem -->

## Test Plan

### Manual Testing
<!-- Hardware: (e.g., MacBook Pro M1 Max 32GB, Mac Mini M2 16GB,
connected via Thunderbolt 4) -->
<!-- What you did: -->
<!-- - -->

### Automated Testing
<!-- Describe changes to automated tests, or how existing tests cover
this change -->
<!-- - -->

---------

Co-authored-by: Alex Cheema <41707476+AlexCheema@users.noreply.github.com>
2026-01-28 05:44:19 +00:00
Evan Quiney 991d278119 replace nix fmt with treefmt in just lint (#1301)
man evaluating the nix flake is so slow. treefmt speeeedy
2026-01-27 17:03:01 +00:00
rltakashige c55cbf6739 Add mlx lm style tensor sharding for Minimax (#1299)
## Motivation

Broken right now. We'll potentially add a better one later

## Changes

<!-- Describe what you changed in detail -->

## Why It Works

<!-- Explain why your approach solves the problem -->

## Test Plan

### Manual Testing
Used for evals without any issue.

### Automated Testing
<!-- Describe changes to automated tests, or how existing tests cover
this change -->
<!-- - -->
2026-01-27 15:29:06 +00:00
Alex Cheema bd4f0bf048 Fix download speed/ETA display for re-downloads (#1294)
## Motivation

After the download verification fix, when files are re-downloaded due to
upstream changes (size mismatch), the download progress displays
correctly (completion %, bytes, file counts), but speed shows 0 B/s and
ETA shows "--" for both overall and per-file progress.

## Changes

- Modified `on_progress_wrapper` in `src/exo/download/download_utils.py`
to detect re-download scenarios
- Added re-download detection: when `curr_bytes < previous_downloaded`,
the file was deleted and download restarted
- On re-download: reset `start_time` to current time and set
`downloaded_this_session = curr_bytes`
- Added two tests to `test_download_verification.py` covering
re-download and continuing download scenarios

## Why It Works

The bug occurred because:
1. `file_progress` is initialized with the OLD local file size (e.g.,
1.5GB)
2. When `_download_file` detects size mismatch, it deletes the file and
starts fresh
3. Progress callback receives small `curr_bytes` (e.g., 8KB) but
compares against old size
4. `downloaded_this_session = 0 + (8KB - 1.5GB) = -1.5GB` (negative!)
5. Negative session bytes → 0 or negative speed → ETA shows "--"

The fix detects when `curr_bytes < previous_downloaded` (indicating
re-download started) and resets tracking to treat it as a fresh
download.

## Test Plan

### Manual Testing
<!-- Hardware: (e.g., MacBook Pro M1 Max 32GB, Mac Mini M2 16GB,
connected via Thunderbolt 4) -->
<!-- What you did: -->
- Download a model, modify a file to change its size, restart exo,
verify speed/ETA display correctly during re-download

### Automated Testing
- Added `TestProgressResetOnRedownload` class with two tests:
- `test_progress_resets_correctly_on_redownload`: Verifies progress
resets correctly when re-download starts
- `test_progress_accumulates_on_continuing_download`: Verifies
continuing downloads still accumulate correctly
- All 11 download tests pass
- Type checking (basedpyright): 0 errors
- Linting (ruff): All checks passed

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-authored-by: Claude Opus 4.5 <noreply@anthropic.com>
2026-01-26 21:56:58 +00:00
rltakashige cd8c01b7c8 Fix kv prefix cache (#1262)
## Motivation

OpenCode sends very large prompts, most of which are repeated on the
next call.

## Changes

Add prefix caching, reducing average time in prefill (in testing) from
40 seconds to 4. This massively improves user experience.

Also evicts KV caches from this prefix cache in a LRU-style manner. 

## Why It Works

We no longer prefill repeatedly but rather use kv cache stored in
memory. A future update may want to use storage to make the prefix cache
larger.

## Test Plan

### Manual Testing
Tested speedup on OpenCode

### Automated Testing
Added a lot of tests

---------

Co-authored-by: David Hind <davehind@yahoo.co.uk>
2026-01-26 20:13:58 +00:00
rltakashige 59e991ce15 Only ignore message if actually empty (#1292)
## Motivation

<!-- Why is this change needed? What problem does it solve? -->
<!-- If it fixes an open issue, please link to the issue here -->

## Changes

<!-- Describe what you changed in detail -->

## Why It Works

<!-- Explain why your approach solves the problem -->

## Test Plan

### Manual Testing
<!-- Hardware: (e.g., MacBook Pro M1 Max 32GB, Mac Mini M2 16GB,
connected via Thunderbolt 4) -->
<!-- What you did: -->
<!-- - -->

### Automated Testing
<!-- Describe changes to automated tests, or how existing tests cover
this change -->
<!-- - -->
2026-01-26 19:33:23 +00:00
ciaranbor ffba340e70 Ciaran/image quantization (#1272)
## Motivation

Enable users to select and use quantized variants (8-bit, 4-bit) of
image models

## Changes

Use exolabs HF org for image models

## Why It Works

Quantized versions have been uploaded to exolabs HF org

## Test Plan

Loaded and ran different quantized variants. Confirmed lower memory
usage and different outputs for the same seed. Verified chat completion
still works.
2026-01-26 19:25:05 +00:00
rltakashige 9968abe816 Leo/fix basic model shard (#1291)
## Motivation

Some models, on some configurations, would have several issues that
caused the model to be stuck on loading.

## Changes

Several loading issues were with upstream mlx lm shard loading for
tensor parallel.
GLM 4.7 Flash now uses GLM 4.7 Lite.
A final portion of the issues were from mlx memory not being properly
released before calling mx.eval(model), causing the system to run out of
memory.

## Test Plan

### Manual Testing
Done a bunch (thanks @AlexCheema), hopefully exhaustive. 

### Automated Testing
A bunch of automated testing is imminent but not landed yet.

---------

Co-authored-by: Alex Cheema <alexcheema123@gmail.com>
2026-01-26 17:49:09 +00:00
Alex Cheema 0e30b0830f Fix download system for upstream file changes (#1290)
## Motivation

When upstream files change on Hugging Face, exo's download system
doesn't detect the change and downloads get stuck. The only workaround
is deleting `~/.exo/models/` and the cache.

Root causes:
1. Existing files are never re-verified against remote metadata
2. File list cache is never invalidated, causing stale sizes to be used

## Changes

1. **Verify existing files against remote size** (`_download_file`):
Before returning early for existing files, verify the local file size
matches remote. If mismatched, delete and re-download. If network fails
(offline), fall back to trusting local file.

2. **Always try fresh file list first** (`fetch_file_list_with_cache`):
Always attempt to fetch fresh data from Hugging Face. On success, update
the cache. On failure, fall back to cached data if available.

3. **Clear cache on model delete** (`delete_model`): When a model is
deleted, also delete its cache entry to prevent stale metadata.

## Why It Works

- **Online**: Stale local files are detected via size mismatch and
re-downloaded. Fresh file list is always fetched and cache is updated.
- **Offline with cache**: Existing files are trusted. Cached file list
is used as fallback.
- **Offline without cache**: Fails gracefully (can't download without
knowing what files to get).

The size check is O(1) so there's no performance impact. Hash
verification still happens after download completes (existing behavior).

## Test Plan

### Manual Testing
<!-- Hardware: (e.g., MacBook Pro M1 Max 32GB, Mac Mini M2 16GB,
connected via Thunderbolt 4) -->
<!-- What you did: -->
- Download a model, manually modify a local file's content, restart exo,
verify it re-downloads

### Automated Testing
Added 9 new tests in
`src/exo/download/tests/test_download_verification.py`:
- Re-download when file size changes upstream
- Skip download when file size matches
- Offline fallback uses local file
- Fetch fresh file list and update cache
- Fall back to cache when fetch fails
- Error propagates when no cache exists
- Model delete clears cache
- Delete when only cache exists
- Delete nonexistent model

All tests pass: `uv run pytest src/exo/download/tests/ -v`

Co-authored-by: Claude Opus 4.5 <noreply@anthropic.com>
2026-01-26 09:14:58 -08:00
Alex Cheema 44453c4c8b Remove change-detection checks from info gatherer monitors (#1283)
## Summary
- When a node times out, its info gets cleared from state. The monitor
functions only sent data when something changed, leaving no mechanism to
re-populate this info after a timeout.
- Removes change-detection checks from `_monitor_misc`,
`_monitor_system_profiler_thunderbolt_data`, `_watch_system_info`, and
`_monitor_thunderbolt_bridge_status` so data is sent periodically
regardless of whether it changed.

## Test plan
- [ ] Verify type checker passes: `uv run basedpyright`
- [ ] Verify linter passes: `uv run ruff check`
- [ ] Verify tests pass: `uv run pytest`
- [ ] Manually test that node info is re-populated after a timeout by
observing cluster behavior

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-authored-by: Claude Opus 4.5 <noreply@anthropic.com>
2026-01-26 12:23:22 +00:00
Jake Hillion 1290e8ed9f dashboard: fix prettier-svelte rebuilding on every file change
The prettier-svelte package was rebuilding whenever any file in the
repository changed because dashboardStubSrc referenced inputs.self
directly. Since inputs.self's store path hash is computed from the
entire repository contents, any file modification invalidated the
derivation.

Added dashboardLockfileSrc using lib.cleanSourceWith to filter
inputs.self to only include package.json and package-lock.json from
the dashboard directory. Updated dashboardStubSrc to reference this
filtered source instead of inputs.self directly.

This ensures prettier-svelte only rebuilds when the lockfiles actually
change, significantly improving build caching for unrelated changes.

Test plan:
- Built prettier-svelte with nix build .#prettier-svelte
- Modified src/exo/main.py and rebuilt - same store path (no rebuild)
- Modified dashboard/package.json and rebuilt - different store path (rebuild triggered)
- Ran nix flake check successfully
2026-01-26 12:02:05 +00:00
Evan Quiney d93db3d6bf re enable the evil network script (#1277)
seems like we still need the interfaces to be routable for mdns. at
least we're not dependent on this behaviour anymore.
2026-01-24 13:36:06 +00:00
Alex Cheema ff4a2022f7 Revert state compaction (#1259) (#1275)
## Summary

Reverts the state compaction feature (#1259) to investigate issues with
nodes staying as "unknown" after joining a cluster.

## Test plan

- [ ] Verify nodes properly show up after joining cluster
- [ ] Verify state catchup works correctly without compaction

🤖 Generated with [Claude Code](https://claude.com/claude-code)
2026-01-23 16:29:48 -08:00
rltakashige cee48f6f34 Parse GPT OSS tool calling (#1271)
## Motivation

<img width="3162" height="858" alt="image"
src="https://github.com/user-attachments/assets/e552f373-620a-4522-894b-6f93fd7f1e50"
/>

## Changes

OpenAI Harmony StreamableParser does parsing for us.

## Why It Works

<img width="3230" height="588" alt="image"
src="https://github.com/user-attachments/assets/81f8a43e-c04b-4bd0-9fd0-65e9b5f6ea1d"
/>
2026-01-23 20:43:53 +00:00
Evan Quiney 2b67e84a03 state compaction (#1259)
## motivation

a node joining a long-running cluster would bring down networking. this
attempts to mitigate that issue by compacting the state for catching up
new devices

## changes

introduces a new topic ("state_catchup") over which a full state can be
sent. currently the master sends the worker + api this new state, and
they update only if they have no other events applied - otherwise usual
NACK systems function

## testing

manually tested on two and eight nodes - its an improvement, not a fix

Co-authored-by: rltakashige <rl.takashige@gmail.com>
2026-01-23 20:32:49 +00:00
Alex Cheema 7204fdeb4a Restore Thunderbolt Bridge LaunchDaemon (#1270)
## Motivation

The LaunchDaemon approach for disabling Thunderbolt Bridge was removed
in commit 43f12f5d and replaced with dynamic cycle detection. However,
the LaunchDaemon runs automatically on reboot, ensuring the bridge is
always disabled before it can cause packet storms.

## Changes

- Restore `NetworkSetupHelper.promptAndInstallIfNeeded()` to install a
LaunchDaemon that disables Thunderbolt Bridge on startup
- Show user prompt explaining what will be installed before requesting
admin password
- Remove old cleanup-only logic from `EXOApp.swift`
- Installer removes any existing installation before installing fresh
(handles upgrades)

## Why It Works

The LaunchDaemon runs at boot with `RunAtLoad=true` and periodically
(every ~30 min), destroying bridge0 and disabling Thunderbolt Bridge
before it can cause packet storms. The daemon is only installed
once—`daemonAlreadyInstalled()` checks script content and plist config
match before prompting.

## Test Plan

### Manual Testing
- Run app first time → should see prompt → click Install → enter admin
password → daemon installed
- Run app again → no prompt (already installed)
- Reboot → bridge0 should be destroyed/disabled automatically
- Check daemon: `launchctl list | grep io.exo.networksetup`
- Check files: `/Library/LaunchDaemons/io.exo.networksetup.plist`,
`/Library/Application Support/EXO/disable_bridge.sh`

### Automated Testing
N/A - requires admin privileges and system-level changes

Co-authored-by: Claude Opus 4.5 <noreply@anthropic.com>
2026-01-23 20:25:37 +00:00
Evan Quiney ec345a4315 fix: deprioritise uncertain ethernet devices (#1267)
we were placing coordinators on uncertain devices (enX+) that are listed
as "USB LAN" - these could be thunderbolt ports breaking RDMA instances
2026-01-23 20:13:28 +00:00
ciaranbor 9967dfa734 Prevent conversation collision (#1266)
## Motivation

When a user switched conversations while a response was still streaming,
the streaming content would be written to the currently selected
conversation instead of the original one. For streamed image generation,
each partial image would be written to the open conversation

## Changes

Added helper methods to track and update the correct conversation during
streaming:
- updateConversationMessage() - Update a message in a specific
conversation by ID
- syncActiveMessagesIfNeeded() - Sync this.messages from target
conversation only if it's active
- conversationExists() - Check if a conversation still exists (handles
mid-stream deletion)
  - persistConversation() - Persist a specific conversation to storage
- addMessageToConversation() - Add a message directly to a specific
conversation


## Why It Works

Capturing the conversation ID at the start of the request ensures we
know which conversation to update

## Test Plan

### Manual Testing

Tested switching conversation during generation across each model type
2026-01-23 19:59:08 +00:00
ciaranbor 23fd37fe4d Add FLUX.1-Krea-dev model (#1269)
## Why It Works

Same implementation as FLUX.1-dev, just different weights
2026-01-23 19:48:24 +00:00
Alex Cheema d229df38f9 Fix placement filter to use subset matching instead of exact match (#1265)
## Motivation

When using the dashboard's instance placement filter (clicking nodes in
the topology), it was filtering to placements that use exactly the
selected nodes. This isn't the expected behavior - users want to see
placements that include all selected nodes, but may also include
additional nodes.

For example, selecting nodes [A, B] should show placements using [A, B],
[A, B, C], [A, B, C, D], etc. - not just [A, B].

## Changes

- Added `required_nodes` parameter to `place_instance()` in
`placement.py`
- Filter cycles early in placement to only those containing all required
nodes (subset matching)
- Simplified `api.py` by removing the subgraph topology filtering and
passing `required_nodes` directly to placement
- Renamed internal `node_ids` variable to `placement_node_ids` to avoid
shadowing the parameter

## Why It Works

By filtering cycles at the placement level using
`required_nodes.issubset(cycle.node_ids)`, we ensure that only cycles
containing all the user-selected nodes are considered. This happens
early in the placement algorithm, so we don't waste time computing
placements that would be filtered out later.

## Test Plan

### Manual Testing
- Select nodes in the dashboard topology view
- Verify that placements shown include all selected nodes (but may
include additional nodes)
- Verify that placements not containing the selected nodes are filtered
out

### Automated Testing
- Existing placement tests pass
- `uv run pytest src/exo/master/tests/ -v` - 37 tests pass

Co-authored-by: Claude Opus 4.5 <noreply@anthropic.com>
2026-01-23 19:40:31 +00:00
Alex Cheema 8a595fee2f Fix Thunderbolt bridge cycle detection to include 2-node cycles (#1261)
## Motivation

Packet storms occur with Thunderbolt bridge enabled on 2 machines
connected by Thunderbolt, not just 3+ node cycles as previously assumed.
The cycle detection was too conservative and missed this case.

## Changes

- Changed the minimum cycle length from >2 (3+ nodes) to >=2 (2+ nodes)
- Updated the early return threshold from `< 3` to `< 2` enabled nodes
- Updated docstring to reflect the new behavior

## Why It Works

A Thunderbolt bridge loop between just 2 machines can still create
broadcast storms when both have the bridge enabled. The previous
threshold of 3+ was based on an incorrect assumption that 2-node
connections wouldn't cause this problem.

## Test Plan

### Manual Testing
<!-- Hardware: (e.g., MacBook Pro M1 Max 32GB, Mac Mini M2 16GB,
connected via Thunderbolt 4) -->
- Tested with 2 machines connected via Thunderbolt with bridge enabled
- Confirmed packet storms occur in this configuration
- Verified the fix correctly detects and handles 2-node cycles

### Automated Testing
- Existing topology tests cover cycle detection logic

Co-authored-by: Claude Opus 4.5 <noreply@anthropic.com>
2026-01-23 19:34:48 +00:00
ciaranbor c8571a17a3 Fix guidance (#1264)
## Motivation

Previously, we only handled user-provided guidance parameter for CFG
models.

## Changes

Just pass the parameter to model setup
2026-01-23 19:13:45 +00:00
Evan Quiney 771a86331b fix instance port assignment (#1268)
we were overassigning the port 52414 to instances because of an error in placement
2026-01-23 18:37:40 +00:00
Jake Hillion 6dbbe7797b downloads: add download and delete buttons to downloads UI
The downloads page showed model download progress but provided no way
for users to trigger downloads or remove completed models from disk.

Added API endpoints (POST /download/start, DELETE /download/{node_id}/{model_id})
that send StartDownload and DeleteDownload commands via the download_command_sender.
Updated the dashboard downloads page with per-model buttons: a download button
for incomplete downloads and a delete button for completed ones.

This allows users to manage downloads directly from the UI without needing
to trigger downloads through other means.

Test plan:
- Deployed on a 3 machine cluster. Did several downloads/deletions - all
  work and the dashboard updates relatively fluently. It takes roughly 5
  seconds to render a 131GB model deletion which isn't too bad.
2026-01-23 18:11:17 +00:00
Jake Hillion 9357503c6f downloads: refactor to run at node level
The Worker previously owned the ShardDownloader directly via dependency
injection, which prevented --no-worker nodes from downloading and made
it impossible for multiple Workers to share a single downloader instance.

Moved download functionality to a new DownloadCoordinator component at
the Node level that communicates via the DOWNLOAD_COMMANDS pub/sub topic.
Workers now send StartDownload commands instead of calling the downloader
directly, and receive progress updates through the event-sourced state.

This decouples downloads from the Worker lifecycle and enables future
features like UI-triggered downloads to specific nodes and multi-worker
download sharing.

Test plan:
- Mostly tested in the next PR that adds explicit downloads/deletions to
  the dashboard.
- Started a model that isn't downloaded - it works.
2026-01-23 18:04:09 +00:00
ciaranbor ba19940828 Fix regenerate for image models (#1263)
## Motivation

The 'regenerate' button was hardcoded to chat completion. Clicking
'regenerate' for image request would result in an error after the model
is loaded

## Changes

Store request type and dispatch to appropriate request upon regeneration

## Why It Works

We make sure to repeat the same request type as was performed originally

## Test Plan

### Manual Testing

Checked 'regenerate' works for chat completion, image generation, image
editing
2026-01-23 16:33:01 +00:00
Jake Hillion f255345a1a dashboard: decouple prettier-svelte from dashboard source
The prettier-svelte formatter depended on the full dashboard build
(dashboardFull), causing the devshell to rebuild whenever any dashboard
source file changed.

Created a deps-only dream2nix derivation (deps.nix) that uses a stub
source containing only package.json, package-lock.json, and minimal
files for vite to succeed. Updated prettier-svelte to use this
derivation instead of dashboardFull.

The stub source is constant unless lockfiles change, so prettier-svelte
and the devshell no longer rebuild when dashboard source files are
modified.

Test plan:
- nix flake check passed
- nix fmt successfully formatted svelte files
2026-01-23 15:16:48 +00:00
ciaranbor a1939c89f2 Enable UI settings for image editing (#1258)
## Motivation

Image editing was missing UI controls for quality, output format, and
advanced parameters that text-to-image generation already supported.

## Changes

- Added quality, output_format, and advanced_params to image edit API
endpoints
- Extended isImageModel check to include image editing models

## Why It Works

The API now accepts and forwards these settings for image edits, and the
UI displays the appropriate controls for image editing models.

## Test Plan

### Manual Testing

Verified parameters can be set in UI and that they progagate through to
model inference
2026-01-23 13:37:25 +00:00
ciaranbor cb9c9ee55c Enable generating multiple images. Optionally stream partial images (#1251)
## Motivation

Support OpenAI API `n` setting

## Changes

- Users can select `n` to generate more than one image with the same
prompt
- each image uses a different seed -> different results
- `stream` and `partial_images` settings can be overwritten in UI
2026-01-23 11:19:58 +00:00
Alex Cheema df240f834d Fix GLM and Kimi tool calling crashes (#1255)
## Motivation

Fixes tool calling crashes with GLM-4.7-Flash and Kimi-K2 models.

Related: #1254

Two distinct issues were causing crashes:
1. **Tool parser crashes** - The upstream GLM47 and Kimi tool parsers
call `.group()` on regex matches without checking for `None`, causing
`AttributeError` when the model outputs malformed tool calls
2. **Chat template crashes** - GLM's chat template expects
`tool_calls[].function.arguments` to be a dict, but OpenAI format
provides it as a JSON string, causing `'str object' has no attribute
'items'`

## Changes

**`src/exo/worker/runner/runner.py`:**
- Add `patch_glm_tokenizer()` - fixed version of mlx_lm's glm47 parser
with None checks
- Fix `patch_kimi_tokenizer()` - add None checks before calling
`.group()` on regex matches
- Add `ValueError` and `AttributeError` to exception handling in
`parse_tool_calls()`

**`src/exo/worker/engines/mlx/utils_mlx.py`:**
- Add `_normalize_tool_calls()` - parses
`tool_calls[].function.arguments` from JSON string to dict for templates
that expect dicts (like GLM-4.7-Flash)

## Why It Works

1. **Parser fixes**: By checking if regex matches are `None` before
calling `.group()`, we can raise a proper `ValueError` instead of
crashing with `AttributeError`

2. **Template fix**: The GLM-4.7-Flash chat template iterates over
arguments with `.items()`:
   ```jinja2
   {% set _args = tc.arguments %}{% for k, v in _args.items() %}
   ```
OpenAI format has `arguments` as a JSON string.
`_normalize_tool_calls()` parses this to a dict before passing to the
template.

## Test Plan

### Manual Testing
- Hardware: Mac with GLM-4.7-Flash-4bit model
- Tested tool calling with GLM model - no longer crashes

### Automated Testing
- Existing tests pass (`uv run pytest`)
- Type checking passes (`uv run basedpyright`)
- Linting passes (`uv run ruff check`)

---------

Co-authored-by: Claude <noreply@anthropic.com>
2026-01-23 01:39:59 +00:00
ciaranbor cd125b3b8c Use icon for image editing models (#1252)
## Motivation

Visual indicator for image editing models

## Changes

Add pencil icon to edit models in model list
2026-01-22 22:37:34 +00:00
Alex Cheema b783a21399 dashboard: add placement filter by clicking topology nodes (#1248)
## Motivation

When selecting a model for placement, users often want to see placements
that utilize specific nodes in their cluster. Currently there's no way
to filter the placement previews to focus on configurations that include
particular machines.

## Changes

- **Backend**: Added `node_ids` query parameter to the
`/placement-previews` API endpoint. When provided, the endpoint filters
the topology to only include the specified nodes before generating
placements using the new `Topology.filter_to_nodes()` method.

- **Topology class**: Added `filter_to_nodes(node_ids)` method that
creates a new topology containing only the specified nodes and edges
between them.

- **App store**: Added `previewNodeFilter` state to track selected
nodes, with methods to toggle/clear the filter. Automatically cleans up
filter when nodes are removed from the cluster and re-fetches previews
when topology changes.

- **TopologyGraph component**: Added click handlers to toggle node
filter selection, hover effects to indicate clickable nodes, and visual
styling (yellow highlight for selected, dimmed for filtered-out nodes).

- **Main page**: Added filter indicator in top-right corner of topology
showing active filter count with a clear button.

## Why It Works

The filtering happens at the backend/placement generation level rather
than just filtering the results. This ensures we see all valid placement
combinations for the selected nodes, not just a subset that happened to
be generated for the full topology.

The visual feedback uses the same rendering approach as the existing
highlight system - state is tracked in Svelte and applied during render,
so it persists across data updates without flickering.

## Test Plan

### Manual Testing
<!-- Hardware: (e.g., MacBook Pro M1 Max 32GB, Mac Mini M2 16GB,
connected via Thunderbolt 4) -->
<!-- What you did: -->
- Click a node in topology → should show yellow highlight and filter
indicator
- Click another node → indicator shows "2 nodes", previews update to
show only placements using both
- Hover over nodes → subtle yellow highlight indicates they're clickable
- Click X on filter indicator → clears filter, shows all placements
again
- Disconnect a node while it's in filter → filter auto-removes that node

### Automated Testing
- Existing tests cover the Topology class; the new `filter_to_nodes`
method follows the same patterns

---------

Co-authored-by: Claude Opus 4.5 <noreply@anthropic.com>
2026-01-22 22:12:57 +00:00
Alex Cheema 43f12f5d08 Replace LaunchDaemon with dynamic Thunderbolt Bridge loop detection (#1222)
## Motivation

The previous approach installed a LaunchDaemon plist that ran
periodically to disable Thunderbolt Bridge. This required full admin
privileges upfront and ran regardless of whether a problematic loop
existed.

This change replaces that with dynamic detection - only prompting the
user when an actual TB bridge loop with 3+ machines is detected, and
using fine-grained SCPreferences authorization instead of full admin.

## Changes

**Backend (Python):**
- Added `ThunderboltBridgeStatus` model to track bridge enabled/exists
state per node
- Added `node_thunderbolt_bridge` and `thunderbolt_bridge_cycles` fields
to State
- Added `get_thunderbolt_bridge_cycles()` method to Topology class
- **Robust TB bridge detection:**
- Finds bridge network services from `-listnetworkserviceorder` (not
`-listallhardwareports` which can miss bridges)
- Checks each bridge's member interfaces via `ifconfig` to verify it
contains Thunderbolt interfaces
- Handles varying service names (e.g., "TB Bridge", "Thunderbolt
Bridge", "Bridge (bridge0)")
  - Includes `service_name` in status for correct disable commands
  - Added warning logs for all error cases in detection
- Updated `apply.py` to handle the new event type and recompute cycles
on node timeout

**Swift App:**
- New `ThunderboltBridgeService` that monitors for cycles from cluster
state
- Shows NSAlert when a cycle with >2 machines is detected
- Uses `SCPreferencesCreateWithAuthorization` with
`system.services.systemconfiguration.network` right for targeted
permissions
- **Auto-cleanup of legacy LaunchDaemon:** On app startup, checks for
and removes old plist/scripts (non-fatal if user cancels)
- **Periodic local network checking:** Re-checks every 10s so the
warning disappears when user grants permission
- **Fixed ClusterState model:** Updated to decode new granular state
fields (`nodeIdentities`, `nodeMemory`, `nodeSystem`,
`nodeThunderboltBridge`) with computed `nodeProfiles` property for
backwards compatibility
- **Fixed Topology model:** Updated to match actual JSON structure where
`nodes` is an array of strings (not objects) and `connections` is a
nested map (not flat array)
- Cleaned up `NetworkSetupHelper` by removing daemon installation code
(now only handles uninstall)

**Dashboard:**
- Added yellow warning badge on topology when TB bridge cycle detected
- On hover: highlights affected nodes in yellow on the topology graph
- Shows which machines are in the cycle with friendly names
- Provides copy-paste terminal command with the correct service name:
  ```
  sudo networksetup -setnetworkserviceenabled "<service-name>" off
  ```
- Warning appears in all topology views (full, welcome, and minimized
chat sidebar)
- **Debug mode:** Shows "TB:ON" or "TB:OFF" status next to each node in
the topology

## Why It Works

- Cycle detection happens on the backend where we have full topology
information
- Only cycles with 3+ machines are flagged (2-node connections are fine)
- TB bridge detection is robust:
- Uses `-listnetworkserviceorder` to find bridges (works on all machines
tested)
- Verifies bridge membership via `ifconfig` to confirm Thunderbolt
interfaces
  - Handles different service names across machines
- The Swift app reacts to detected cycles and prompts the user once per
cycle
- The dashboard provides visual feedback and actionable instructions
- `SCPreferencesCreateWithAuthorization` provides the minimal
permissions needed to modify network service state
- Legacy LaunchDaemon is automatically cleaned up on first launch with
this version

## Test Plan

### Manual Testing
Here EXO detected a TB bridge cycle:

#### Dashboard:
<img width="1363" height="884" alt="Screenshot 2026-01-21 at 10 07
30 PM"
src="https://github.com/user-attachments/assets/7da9c621-0c91-42c4-898e-4952188a1f61"
/>

#### Hovering the warning:
<img width="359" height="279" alt="Screenshot 2026-01-21 at 16 30 57"
src="https://github.com/user-attachments/assets/05501dcf-3d4a-4704-9f38-257748c05a53"
/>

#### macOS app warning popup:
<img width="270" height="410" alt="Screenshot 2026-01-21 at 16 29 08"
src="https://github.com/user-attachments/assets/45714427-08c3-4fb4-9e61-144925c51adf"
/>

### Which then asks for the user's password:
<img width="263" height="372" alt="Screenshot 2026-01-21 at 16 29 28"
src="https://github.com/user-attachments/assets/7502e591-596d-4128-8cf5-6a12674e27bc"
/>

Which when entered, successfully disables bridge and no longer shows the
warning on dashboard.

#### When it fails it shows the error message:
<img width="263" height="234" alt="Screenshot 2026-01-21 at 14 45 38"
src="https://github.com/user-attachments/assets/2d10b3d5-69d7-46ea-b631-d52d8651ab41"
/>

### Automated Testing
- Type checker: 0 errors (`uv run basedpyright`)
- Linter: All checks passed (`uv run ruff check`)
- Tests: 118 passed (`uv run pytest`)
- Dashboard: Builds successfully (`npm run build`)

---------

Co-authored-by: Claude Opus 4.5 <noreply@anthropic.com>
2026-01-22 21:53:05 +00:00
ciaranbor 8027d7933f Ciaran/hf token (#1250)
## Motivation

black-forest-labs models require hf auth and signup to download. We
don't handle this gracefully.
https://github.com/exo-explore/exo/issues/1242

## Changes

- Handle auth errors
- Surface error to UI and suggest resolution
- Support using HF_TOKEN env variable for auto
- Hide image functionality behind `EXO_ENABLE_IMAGE_MODELS=true` for now

## Why It Works

Users are presented with actionable feedback when issue occurs

## Test Plan

### Manual Testing

Confirmed loading black-forest-labs model in UI presents the issue in
the UI.
Confirmed both `hf auto login` and setting `HF_TOKEN` resolve the issue
2026-01-22 20:39:53 +00:00
Evan ac6efa747b add kimi tool parseing
this patches the kimi tokenizer to add tool calling - it can be reverted
once upstream support is added for kimi-k2
2026-01-22 11:49:25 +00:00
Evan 2e3c33db6d implement mlx-lm tool calling
splits up the runners generation chunks into tool calls, tokens and
errors, and writes tool call chunks when the upstream parser detects
them.
2026-01-22 11:49:25 +00:00
rltakashige fc8e6ad06b Reduce download log spam (#1249)
## Motivation

<!-- Why is this change needed? What problem does it solve? -->
<!-- If it fixes an open issue, please link to the issue here -->

## Changes

<!-- Describe what you changed in detail -->

## Why It Works

<!-- Explain why your approach solves the problem -->

## Test Plan

### Manual Testing
<!-- Hardware: (e.g., MacBook Pro M1 Max 32GB, Mac Mini M2 16GB,
connected via Thunderbolt 4) -->
<!-- What you did: -->
<!-- - -->

### Automated Testing
<!-- Describe changes to automated tests, or how existing tests cover
this change -->
<!-- - -->
2026-01-22 11:28:36 +00:00
341 changed files with 36582 additions and 9807 deletions
-12
View File
@@ -1,12 +0,0 @@
name: Type Check
description: "Run type checker"
runs:
using: "composite"
steps:
- name: Run type checker
run: |
nix --extra-experimental-features nix-command --extra-experimental-features flakes develop -c just sync
nix --extra-experimental-features nix-command --extra-experimental-features flakes develop -c just check
shell: bash
+7 -1
View File
@@ -396,7 +396,7 @@ jobs:
path: output/EXO-${{ env.RELEASE_VERSION }}.dmg
- name: Upload to S3
if: env.SPARKLE_S3_BUCKET != '' && github.ref_type == 'tag'
if: env.SPARKLE_S3_BUCKET != ''
env:
AWS_ACCESS_KEY_ID: ${{ secrets.AWS_ACCESS_KEY_ID }}
AWS_SECRET_ACCESS_KEY: ${{ secrets.AWS_SECRET_ACCESS_KEY }}
@@ -412,6 +412,12 @@ jobs:
PREFIX="${PREFIX}/"
fi
DMG_NAME="EXO-${RELEASE_VERSION}.dmg"
if [[ "${{ github.ref_type }}" != "tag" ]]; then
aws s3 cp "$DMG_NAME" "s3://${SPARKLE_S3_BUCKET}/${PREFIX}EXO-${GITHUB_SHA}.dmg"
exit 0
fi
aws s3 cp "$DMG_NAME" "s3://${SPARKLE_S3_BUCKET}/${PREFIX}${DMG_NAME}"
if [[ "$IS_ALPHA" != "true" ]]; then
aws s3 cp "$DMG_NAME" "s3://${SPARKLE_S3_BUCKET}/${PREFIX}EXO-latest.dmg"
+70 -86
View File
@@ -8,92 +8,6 @@ on:
- main
jobs:
typecheck:
runs-on: ubuntu-latest
steps:
- name: Checkout repository
uses: actions/checkout@v4
with:
lfs: false
- uses: cachix/install-nix-action@v31
with:
nix_path: nixpkgs=channel:nixos-unstable
- uses: cachix/cachix-action@v14
name: Configure Cachix
with:
name: exo
authToken: "${{ secrets.CACHIX_AUTH_TOKEN }}"
- name: Configure git user
run: |
git config --local user.email "github-actions@users.noreply.github.com"
git config --local user.name "github-actions bot"
shell: bash
- name: Pull LFS files
run: |
echo "Pulling Git LFS files..."
git lfs pull
shell: bash
- name: Setup Nix Environment
run: |
echo "Checking for nix installation..."
# Check if nix binary exists directly
if [ -f /nix/var/nix/profiles/default/bin/nix ]; then
echo "Found nix binary at /nix/var/nix/profiles/default/bin/nix"
export PATH="/nix/var/nix/profiles/default/bin:$PATH"
echo "PATH=$PATH" >> $GITHUB_ENV
nix --version
elif [ -f /nix/var/nix/profiles/default/etc/profile.d/nix-daemon.sh ]; then
echo "Found nix profile script, sourcing..."
source /nix/var/nix/profiles/default/etc/profile.d/nix-daemon.sh
nix --version
elif command -v nix >/dev/null 2>&1; then
echo "Nix already in PATH"
nix --version
else
echo "Nix not found. Debugging info:"
echo "Contents of /nix/var/nix/profiles/default/:"
ls -la /nix/var/nix/profiles/default/ 2>/dev/null || echo "Directory not found"
echo "Contents of /nix/var/nix/profiles/default/bin/:"
ls -la /nix/var/nix/profiles/default/bin/ 2>/dev/null || echo "Directory not found"
exit 1
fi
shell: bash
- name: Configure basedpyright include for local MLX
run: |
RUNNER_LABELS='${{ toJSON(runner.labels) }}'
if echo "$RUNNER_LABELS" | grep -q "local_mlx"; then
if [ -d "/Users/Shared/mlx" ]; then
echo "Updating [tool.basedpyright].include to use /Users/Shared/mlx"
awk '
BEGIN { in=0 }
/^\[tool\.basedpyright\]/ { in=1; print; next }
in && /^\[/ { in=0 } # next section
in && /^[ \t]*include[ \t]*=/ {
print "include = [\"/Users/Shared/mlx\"]"
next
}
{ print }
' pyproject.toml > pyproject.toml.tmp && mv pyproject.toml.tmp pyproject.toml
echo "New [tool.basedpyright] section:"
sed -n '/^\[tool\.basedpyright\]/,/^\[/p' pyproject.toml | sed '$d' || true
else
echo "local_mlx tag present but /Users/Shared/mlx not found; leaving pyproject unchanged."
fi
else
echo "Runner does not have 'local_mlx' tag; leaving pyproject unchanged."
fi
shell: bash
- uses: ./.github/actions/typecheck
nix:
name: Build and check (${{ matrix.system }})
runs-on: ${{ matrix.runner }}
@@ -123,6 +37,63 @@ jobs:
name: exo
authToken: "${{ secrets.CACHIX_AUTH_TOKEN }}"
- name: Build Metal packages (macOS only)
if: runner.os == 'macOS'
run: |
# Try to build metal-toolchain first (may succeed via cachix cache hit)
if nix build .#metal-toolchain 2>/dev/null; then
echo "metal-toolchain built successfully (likely cache hit)"
else
echo "metal-toolchain build failed, extracting from Xcode..."
NAR_HASH="sha256-ayR5mXN4sZAddwKEG2OszGRF93k9ZFc7H0yi2xbylQw="
NAR_NAME="metal-toolchain-17C48.nar"
# Use RUNNER_TEMP to avoid /tmp symlink issues on macOS
WORK_DIR="${RUNNER_TEMP}/metal-work"
mkdir -p "$WORK_DIR"
# Download the Metal toolchain component
xcodebuild -downloadComponent MetalToolchain
# Find and mount the DMG
DMG_PATH=$(find /System/Library/AssetsV2/com_apple_MobileAsset_MetalToolchain -name '*.dmg' 2>/dev/null | head -1)
if [ -z "$DMG_PATH" ]; then
echo "Error: Could not find Metal toolchain DMG"
exit 1
fi
echo "Found DMG at: $DMG_PATH"
hdiutil attach "$DMG_PATH" -mountpoint "${WORK_DIR}/metal-dmg"
# Copy the toolchain
cp -R "${WORK_DIR}/metal-dmg/Metal.xctoolchain" "${WORK_DIR}/metal-export"
hdiutil detach "${WORK_DIR}/metal-dmg"
# Create NAR and add to store
nix nar pack "${WORK_DIR}/metal-export" > "${WORK_DIR}/${NAR_NAME}"
STORE_PATH=$(nix store add --mode flat "${WORK_DIR}/${NAR_NAME}")
echo "Added NAR to store: $STORE_PATH"
# Verify the hash matches
ACTUAL_HASH=$(nix hash file "${WORK_DIR}/${NAR_NAME}")
if [ "$ACTUAL_HASH" != "$NAR_HASH" ]; then
echo "Warning: NAR hash mismatch!"
echo "Expected: $NAR_HASH"
echo "Actual: $ACTUAL_HASH"
echo "The metal-toolchain.nix may need updating"
fi
# Clean up
rm -rf "$WORK_DIR"
# Retry the build now that NAR is in store
nix build .#metal-toolchain
fi
# Build mlx (depends on metal-toolchain)
nix build .#mlx
- name: Build all Nix outputs
run: |
nix flake show --json | jq -r '
@@ -134,3 +105,16 @@ jobs:
- name: Run nix flake check
run: nix flake check
- name: Run pytest (macOS only)
if: runner.os == 'macOS'
run: |
# Build the test environment (requires relaxed sandbox for uv2nix on macOS)
TEST_ENV=$(nix build '.#exo-test-env' --option sandbox relaxed --print-out-paths)
# Run pytest outside sandbox (needs GPU access for MLX)
export HOME="$RUNNER_TEMP"
export EXO_TESTS=1
export EXO_DASHBOARD_DIR="$PWD/dashboard/"
export EXO_RESOURCES_DIR="$PWD/resources"
$TEST_ENV/bin/python -m pytest src -m "not slow" --import-mode=importlib
+10
View File
@@ -28,3 +28,13 @@ target/
dashboard/build/
dashboard/node_modules/
dashboard/.svelte-kit/
# host config snapshots
hosts_*.json
.swp
# bench files
bench/**/*.json
# tmp
tmp/models
@@ -0,0 +1,7 @@
"""
This type stub file was generated by pyright.
"""
from mflux.models.flux.variants.kontext.flux_kontext import Flux1Kontext
__all__ = ["Flux1Kontext"]
@@ -0,0 +1,49 @@
"""
This type stub file was generated by pyright.
"""
from pathlib import Path
from typing import Any
from mlx import nn
from mflux.models.common.config.model_config import ModelConfig
from mflux.models.flux.model.flux_text_encoder.clip_encoder.clip_encoder import (
CLIPEncoder,
)
from mflux.models.flux.model.flux_text_encoder.t5_encoder.t5_encoder import T5Encoder
from mflux.models.flux.model.flux_transformer.transformer import Transformer
from mflux.models.flux.model.flux_vae.vae import VAE
from mflux.utils.generated_image import GeneratedImage
class Flux1Kontext(nn.Module):
vae: VAE
transformer: Transformer
t5_text_encoder: T5Encoder
clip_text_encoder: CLIPEncoder
bits: int | None
lora_paths: list[str] | None
lora_scales: list[float] | None
prompt_cache: dict[str, Any]
tokenizers: dict[str, Any]
def __init__(
self,
quantize: int | None = ...,
model_path: str | None = ...,
lora_paths: list[str] | None = ...,
lora_scales: list[float] | None = ...,
model_config: ModelConfig = ...,
) -> None: ...
def generate_image(
self,
seed: int,
prompt: str,
num_inference_steps: int = ...,
height: int = ...,
width: int = ...,
guidance: float = ...,
image_path: Path | str | None = ...,
image_strength: float | None = ...,
scheduler: str = ...,
) -> GeneratedImage: ...
@@ -0,0 +1,16 @@
"""
This type stub file was generated by pyright.
"""
import mlx.core as mx
from mflux.models.flux.model.flux_vae.vae import VAE
class KontextUtil:
@staticmethod
def create_image_conditioning_latents(
vae: VAE,
height: int,
width: int,
image_path: str,
) -> tuple[mx.array, mx.array]: ...
+19 -3
View File
@@ -215,6 +215,22 @@ class StreamContext:
traceback: object | None = ...,
) -> None: ...
def device_info() -> dict[str, str | int]:
"""
Get information about the GPU device and system settings.
Currently returns:
* ``architecture``
* ``max_buffer_size``
* ``max_recommended_working_set_size``
* ``memory_size``
* ``resource_limit``
Returns:
dict: A dictionary with string keys and string or integer values.
"""
def abs(a: array, /, *, stream: Stream | Device | None = ...) -> array:
"""
Element-wise absolute value.
@@ -1139,7 +1155,7 @@ class array:
) -> array:
"""See :func:`flatten`."""
def reshape(self, *shape, stream: Stream | Device | None = ...) -> array:
def reshape(self, *shape: int, stream: Stream | Device | None = ...) -> array:
"""
Equivalent to :func:`reshape` but the shape can be passed either as a
:obj:`tuple` or as separate arguments.
@@ -1222,7 +1238,7 @@ class array:
) -> array:
"""See :func:`swapaxes`."""
def transpose(self, *axes, stream: Stream | Device | None = ...) -> array:
def transpose(self, *axes: int, stream: Stream | Device | None = ...) -> array:
"""
Equivalent to :func:`transpose` but the axes can be passed either as
a tuple or as separate arguments.
@@ -2366,7 +2382,7 @@ class custom_function:
def default_device() -> Device:
"""Get the default device."""
def default_stream(device: Device) -> Stream:
def default_stream(device: Device | DeviceType) -> Stream:
"""Get the device's default stream."""
def degrees(a: array, /, *, stream: Stream | Device | None = ...) -> array:
+1 -1
View File
@@ -200,7 +200,7 @@ class Module(dict):
) -> mx.MX_ARRAY_TREE: # -> dict[Any, Any | dict[Any, Any | dict[Any, Any] | list[Any]] | dict[Any, Any] | list[Any]]:
"""Return the submodules that do not contain other modules."""
def update(self, parameters: dict, strict: bool = ...) -> Module:
def update(self, parameters: dict[str, Any], strict: bool = ...) -> Module:
"""Replace the parameters of this Module with the provided ones in the
dict of dicts and lists.
@@ -30,6 +30,9 @@ class Conv1d(Module):
bias (bool, optional): If ``True`` add a learnable bias to the output.
Default: ``True``
"""
weight: mx.array
groups: int
def __init__(
self,
in_channels: int,
+11 -8
View File
@@ -7,7 +7,10 @@ from typing import Any, Callable, Dict, List, Optional, Tuple, Union
from mlx.core import MX_ARRAY_TREE
def tree_map(
fn: Callable, tree: Any, *rest: Any, is_leaf: Optional[Callable] = ...
fn: Callable[..., Any],
tree: Any,
*rest: Any,
is_leaf: Callable[..., bool] | None = ...,
) -> Any:
"""Applies ``fn`` to the leaves of the Python tree ``tree`` and
returns a new collection with the results.
@@ -44,11 +47,11 @@ def tree_map(
"""
def tree_map_with_path(
fn: Callable,
fn: Callable[..., Any],
tree: Any,
*rest: Any,
is_leaf: Optional[Callable] = ...,
path: Optional[Any] = ...,
is_leaf: Callable[..., bool] | None = ...,
path: str | None = ...,
) -> Any:
"""Applies ``fn`` to the path and leaves of the Python tree ``tree`` and
returns a new collection with the results.
@@ -80,9 +83,9 @@ def tree_map_with_path(
def tree_flatten(
tree: Any,
prefix: str = ...,
is_leaf: Optional[Callable] = ...,
destination: Optional[Union[List[Tuple[str, Any]], Dict[str, Any]]] = ...,
) -> Union[List[Tuple[str, Any]], Dict[str, Any]]:
is_leaf: Callable[..., bool] | None = ...,
destination: list[tuple[str, Any]] | dict[str, Any] | None = ...,
) -> list[tuple[str, Any]] | dict[str, Any]:
"""Flattens a Python tree to a list of key, value tuples.
The keys are using the dot notation to define trees of arbitrary depth and
@@ -118,7 +121,7 @@ def tree_flatten(
the Python tree.
"""
def tree_unflatten(tree: Union[List[Tuple[str, Any]], Dict[str, Any]]) -> Any:
def tree_unflatten(tree: list[tuple[str, Any]] | dict[str, Any]) -> Any:
"""Recreate a Python tree from its flat representation.
.. code-block:: python
+7 -2
View File
@@ -11,7 +11,10 @@ import mlx.core as mx
class Cache(Protocol):
keys: mx.array
values: mx.array
def update_and_fetch(self, keys: mx.array, values: mx.array) -> None: ...
offset: int
def update_and_fetch(
self, keys: mx.array, values: mx.array
) -> tuple[mx.array, mx.array]: ...
@property
def state(self) -> tuple[mx.array, mx.array]: ...
@state.setter
@@ -87,6 +90,7 @@ def create_attention_mask(
class _BaseCache(Cache):
keys: mx.array
values: mx.array
offset: int
@property
def state(self) -> tuple[mx.array, mx.array]: ...
@state.setter
@@ -245,7 +249,8 @@ class ChunkedKVCache(KVCache):
...
class CacheList(_BaseCache):
def __init__(self, *caches) -> None: ...
caches: tuple[_BaseCache, ...]
def __init__(self, *caches: _BaseCache) -> None: ...
def __getitem__(self, idx): ...
def is_trimmable(self): # -> bool:
...
+5 -1
View File
@@ -5,6 +5,7 @@ from typing import Any, Dict, Optional
import mlx.core as mx
import mlx.nn as nn
from mlx_lm.models.mla import MultiLinear
from .base import BaseModelArgs
from .switch_layers import SwitchGLU
@@ -60,7 +61,10 @@ class DeepseekV3Attention(nn.Module):
q_b_proj: nn.Linear
kv_a_proj_with_mqa: nn.Linear
kv_a_layernorm: nn.RMSNorm
kv_b_proj: nn.Linear
# kv_b_proj: nn.Linear
embed_q: MultiLinear
unembed_out: MultiLinear
o_proj: nn.Linear
rope: Any
+153
View File
@@ -0,0 +1,153 @@
from dataclasses import dataclass
from typing import Any, Dict, Optional
import mlx.core as mx
import mlx.nn as nn
from .base import BaseModelArgs
from .switch_layers import SwitchGLU
@dataclass
class ModelArgs(BaseModelArgs):
model_type: str
vocab_size: int
hidden_size: int
intermediate_size: int
max_position_embeddings: int
moe_intermediate_size: int
norm_topk_prob: bool
num_attention_heads: int
n_group: int
head_dim: int
topk_group: int
n_shared_experts: int
n_routed_experts: int
routed_scaling_factor: float
num_experts_per_tok: int
first_k_dense_replace: int
num_hidden_layers: int
num_key_value_heads: int
rms_norm_eps: float
rope_theta: float
rope_scaling: Optional[Dict[str, Any]]
use_qk_norm: bool
tie_word_embeddings: bool
attention_bias: bool
partial_rotary_factor: float
scoring_func: str
topk_method: str
class Attention(nn.Module):
n_heads: int
n_kv_heads: int
scale: float
q_proj: nn.Linear
k_proj: nn.Linear
v_proj: nn.Linear
o_proj: nn.Linear
use_qk_norm: bool
q_norm: nn.RMSNorm
k_norm: nn.RMSNorm
rope: nn.RoPE
def __init__(self, args: ModelArgs) -> None: ...
def __call__(
self,
x: mx.array,
mask: Optional[mx.array] = None,
cache: Optional[Any] = None,
) -> mx.array: ...
class MLP(nn.Module):
config: ModelArgs
hidden_size: int
intermediate_size: int
gate_proj: nn.Linear
up_proj: nn.Linear
down_proj: nn.Linear
def __init__(
self,
config: ModelArgs,
hidden_size: Optional[int] = None,
intermediate_size: Optional[int] = None,
) -> None: ...
def __call__(self, x: mx.array) -> mx.array: ...
class MoEGate(nn.Module):
config: ModelArgs
top_k: int
norm_topk_prob: bool
n_routed_experts: int
routed_scaling_factor: float
n_group: int
topk_group: int
weight: mx.array
e_score_correction_bias: mx.array
def __init__(self, config: ModelArgs) -> None: ...
def __call__(self, x: mx.array) -> tuple[mx.array, mx.array]: ...
class MoE(nn.Module):
config: ModelArgs
num_experts_per_tok: int
switch_mlp: SwitchGLU
gate: MoEGate
shared_experts: MLP
sharding_group: Optional[mx.distributed.Group]
def __init__(self, config: ModelArgs) -> None: ...
def __call__(self, x: mx.array) -> mx.array: ...
class DecoderLayer(nn.Module):
self_attn: Attention
mlp: MLP | MoE
input_layernorm: nn.RMSNorm
post_attention_layernorm: nn.RMSNorm
def __init__(self, config: ModelArgs, layer_idx: int) -> None: ...
def __call__(
self,
x: mx.array,
mask: Optional[mx.array] = None,
cache: Optional[Any] = None,
) -> mx.array: ...
class LanguageModel(nn.Module):
vocab_size: int
embed_tokens: nn.Embedding
layers: list[DecoderLayer]
norm: nn.RMSNorm
pipeline_rank: int
pipeline_size: int
start_idx: int
end_idx: Optional[int]
num_layers: int
def __init__(self, config: ModelArgs) -> None: ...
def __call__(
self,
x: mx.array,
cache: Optional[Any] = None,
) -> mx.array: ...
@property
def pipeline_layers(self) -> list[DecoderLayer]: ...
class Model(nn.Module):
args: ModelArgs
model_type: str
model: LanguageModel
lm_head: nn.Linear
def __init__(self, config: ModelArgs) -> None: ...
def __call__(
self,
inputs: mx.array,
cache: Optional[Any] = None,
) -> mx.array: ...
def sanitize(self, weights: dict[str, Any]) -> dict[str, Any]: ...
def shard(self, group: Optional[mx.distributed.Group] = None) -> None: ...
@property
def layers(self) -> list[DecoderLayer]: ...
@property
def cast_predicate(self) -> Any: ...
@@ -0,0 +1,46 @@
"""Type stubs for mlx_lm.models.glm_moe_dsa"""
from dataclasses import dataclass
from typing import Any, Dict, Optional
from .base import BaseModelArgs
from .deepseek_v32 import Model as DSV32Model
@dataclass
class ModelArgs(BaseModelArgs):
model_type: str
vocab_size: int
hidden_size: int
index_head_dim: int
index_n_heads: int
index_topk: int
intermediate_size: int
moe_intermediate_size: int
num_hidden_layers: int
num_attention_heads: int
num_key_value_heads: int
n_shared_experts: Optional[int]
n_routed_experts: Optional[int]
routed_scaling_factor: float
kv_lora_rank: int
q_lora_rank: int
qk_rope_head_dim: int
v_head_dim: int
qk_nope_head_dim: int
topk_method: str
scoring_func: str
norm_topk_prob: bool
n_group: int
topk_group: int
num_experts_per_tok: int
moe_layer_freq: int
first_k_dense_replace: int
max_position_embeddings: int
rms_norm_eps: float
rope_parameters: Dict[str, Any]
attention_bias: bool
rope_scaling: Dict[str, Any] | None
rope_theta: float | None
class Model(DSV32Model):
def __init__(self, config: ModelArgs) -> None: ...
+114
View File
@@ -0,0 +1,114 @@
"""Type stubs for mlx_lm.models.qwen3_next"""
from typing import Any, Optional
import mlx.core as mx
import mlx.nn as nn
from .switch_layers import SwitchGLU
class Qwen3NextMLP(nn.Module):
gate_proj: nn.Linear
down_proj: nn.Linear
up_proj: nn.Linear
def __init__(self, dim: int, hidden_dim: int) -> None: ...
def __call__(self, x: mx.array) -> mx.array: ...
class Qwen3NextGatedDeltaNet(nn.Module):
hidden_size: int
num_v_heads: int
num_k_heads: int
head_k_dim: int
head_v_dim: int
key_dim: int
value_dim: int
conv_kernel_size: int
conv_dim: int
conv1d: nn.Conv1d
in_proj_qkvz: nn.Linear
in_proj_ba: nn.Linear
dt_bias: mx.array
A_log: mx.array
out_proj: nn.Linear
def __init__(self, config: Any) -> None: ...
def __call__(
self,
inputs: mx.array,
mask: Optional[mx.array] = None,
cache: Optional[Any] = None,
) -> mx.array: ...
class Qwen3NextAttention(nn.Module):
num_attention_heads: int
num_key_value_heads: int
head_dim: int
scale: float
q_proj: nn.Linear
k_proj: nn.Linear
v_proj: nn.Linear
o_proj: nn.Linear
def __init__(self, args: Any) -> None: ...
def __call__(
self,
x: mx.array,
mask: Optional[mx.array] = None,
cache: Optional[Any] = None,
) -> mx.array: ...
class Qwen3NextSparseMoeBlock(nn.Module):
norm_topk_prob: bool
num_experts: int
top_k: int
gate: nn.Linear
switch_mlp: SwitchGLU
shared_expert: Qwen3NextMLP
shared_expert_gate: nn.Linear
def __init__(self, args: Any) -> None: ...
def __call__(self, x: mx.array) -> mx.array: ...
class Qwen3NextDecoderLayer(nn.Module):
is_linear: bool
linear_attn: Qwen3NextGatedDeltaNet
self_attn: Qwen3NextAttention
input_layernorm: nn.RMSNorm
post_attention_layernorm: nn.RMSNorm
mlp: Qwen3NextMLP | Qwen3NextSparseMoeBlock
def __init__(self, args: Any, layer_idx: int) -> None: ...
def __call__(
self,
x: mx.array,
mask: Optional[mx.array] = None,
cache: Optional[Any] = None,
) -> mx.array: ...
class Qwen3NextModel(nn.Module):
embed_tokens: nn.Embedding
layers: list[Qwen3NextDecoderLayer]
norm: nn.RMSNorm
def __init__(self, args: Any) -> None: ...
def __call__(
self,
inputs: mx.array,
cache: Optional[Any] = None,
) -> mx.array: ...
class Model(nn.Module):
model_type: str
model: Qwen3NextModel
lm_head: nn.Linear
def __init__(self, args: Any) -> None: ...
def __call__(
self,
inputs: mx.array,
cache: Optional[Any] = None,
) -> mx.array: ...
def sanitize(self, weights: dict[str, Any]) -> dict[str, Any]: ...
@property
def layers(self) -> list[Qwen3NextDecoderLayer]: ...
+151
View File
@@ -0,0 +1,151 @@
from dataclasses import dataclass
from typing import Any, Dict, List, Optional
import mlx.core as mx
import mlx.nn as nn
from .base import BaseModelArgs
from .switch_layers import SwitchGLU
@dataclass
class ModelArgs(BaseModelArgs):
model_type: str
hidden_size: int
num_hidden_layers: int
vocab_size: int
num_attention_heads: int
num_attention_groups: int
head_dim: int
intermediate_size: int
rms_norm_eps: float
rope_theta: float
rope_scaling: Optional[Dict[str, Any]]
max_position_embeddings: int
sliding_window: int
layer_types: Optional[List[str]]
yarn_only_types: Optional[List[str]]
partial_rotary_factors: Optional[List[float]]
attention_other_setting: Optional[Dict[str, Any]]
use_head_wise_attn_gate: bool
moe_num_experts: int
moe_top_k: int
moe_intermediate_size: int
share_expert_dim: int
moe_layers_enum: Optional[str]
moe_router_scaling_factor: float
norm_expert_weight: bool
swiglu_limits: Optional[List[float]]
swiglu_limits_shared: Optional[List[float]]
tie_word_embeddings: bool
class Step3p5MLP(nn.Module):
hidden_size: int
intermediate_size: int
gate_proj: nn.Linear
up_proj: nn.Linear
down_proj: nn.Linear
limit: Optional[float]
def __init__(
self, args: ModelArgs, intermediate_size: int, swiglu_limit: float = 0
) -> None: ...
def __call__(self, x: mx.array) -> mx.array: ...
class Step3p5MoEGate(nn.Module):
top_k: int
n_routed_experts: int
routed_scaling_factor: float
norm_topk_prob: bool
gate: nn.Linear
router_bias: mx.array
def __init__(self, args: ModelArgs) -> None: ...
def __call__(self, x: mx.array) -> tuple[mx.array, mx.array]: ...
class Step3p5MoE(nn.Module):
gate: Step3p5MoEGate
switch_mlp: SwitchGLU
share_expert: Step3p5MLP
sharding_group: Optional[mx.distributed.Group]
def __init__(self, args: ModelArgs, layer_idx: int) -> None: ...
def __call__(self, x: mx.array) -> mx.array: ...
class Step3p5Attention(nn.Module):
is_sliding: bool
num_heads: int
num_kv_heads: int
head_dim: int
scale: float
q_proj: nn.Linear
k_proj: nn.Linear
v_proj: nn.Linear
o_proj: nn.Linear
q_norm: nn.Module
k_norm: nn.Module
use_head_wise_attn_gate: bool
g_proj: nn.Linear
rope: nn.Module
def __init__(self, args: ModelArgs, layer_idx: int) -> None: ...
def __call__(
self,
x: mx.array,
mask: Optional[mx.array] = None,
cache: Optional[Any] = None,
) -> mx.array: ...
class Step3p5DecoderLayer(nn.Module):
self_attn: Step3p5Attention
is_sliding: bool
is_moe_layer: bool
mlp: Step3p5MLP | Step3p5MoE
input_layernorm: nn.Module
post_attention_layernorm: nn.Module
def __init__(self, args: ModelArgs, layer_idx: int) -> None: ...
def __call__(
self,
x: mx.array,
mask: Optional[mx.array] = None,
cache: Optional[Any] = None,
) -> mx.array: ...
class Step3p5Model(nn.Module):
args: ModelArgs
vocab_size: int
num_layers: int
embed_tokens: nn.Embedding
layers: list[Step3p5DecoderLayer]
norm: nn.Module
_swa_idx: Optional[int]
_full_idx: Optional[int]
def __init__(self, args: ModelArgs) -> None: ...
def __call__(
self,
x: mx.array,
cache: Optional[List[Any]] = None,
) -> mx.array: ...
class Model(nn.Module):
args: ModelArgs
model_type: str
model: Step3p5Model
lm_head: nn.Linear
def __init__(self, args: ModelArgs) -> None: ...
def __call__(
self,
inputs: mx.array,
cache: Optional[List[Any]] = None,
) -> mx.array: ...
def sanitize(self, weights: dict[str, Any]) -> dict[str, Any]: ...
def shard(self, group: Optional[mx.distributed.Group] = None) -> None: ...
@property
def layers(self) -> list[Step3p5DecoderLayer]: ...
def make_cache(self) -> list[Any]: ...
@property
def cast_predicate(self) -> Any: ...
@property
def quant_predicate(self) -> Any: ...
+6 -1
View File
@@ -108,16 +108,21 @@ class TokenizerWrapper:
_tokenizer: PreTrainedTokenizerFast
eos_token_id: int | None
eos_token: str | None
eos_token_ids: list[int] | set[int] | None
bos_token_id: int | None
bos_token: str | None
vocab_size: int
all_special_tokens: list[str]
think_start: str | None
think_end: str | None
think_start_id: int | None
think_end_id: int | None
def __init__(
self,
tokenizer: Any,
detokenizer_class: Any = ...,
eos_token_ids: list[int] | None = ...,
eos_token_ids: list[int] | set[int] | None = ...,
chat_template: Any = ...,
tool_parser: Any = ...,
tool_call_start: str | None = ...,
+75
View File
@@ -119,3 +119,78 @@ From .cursorrules:
## Testing
Tests use pytest-asyncio with `asyncio_mode = "auto"`. Tests are in `tests/` subdirectories alongside the code they test. The `EXO_TESTS=1` env var is set during tests.
## Dashboard UI Testing & Screenshots
### Building and Running the Dashboard
```bash
# Build the dashboard (must be done before running exo)
cd dashboard && npm install && npm run build && cd ..
# Start exo (serves the dashboard at http://localhost:52415)
uv run exo &
sleep 8 # Wait for server to start
```
### Taking Headless Screenshots with Playwright
Use Playwright with headless Chromium for programmatic screenshots — no manual browser interaction needed.
**Setup (one-time):**
```bash
npx --yes playwright install chromium
cd /tmp && npm init -y && npm install playwright
```
**Taking screenshots:**
```javascript
// Run from /tmp where playwright is installed: cd /tmp && node -e "..."
const { chromium } = require('playwright');
(async () => {
const browser = await chromium.launch({ headless: true });
const page = await browser.newPage({ viewport: { width: 1280, height: 800 } });
await page.goto('http://localhost:52415', { waitUntil: 'networkidle' });
await page.waitForTimeout(2000);
// Inject test data into localStorage if needed (e.g., recent models)
await page.evaluate(() => {
localStorage.setItem('exo-recent-models', JSON.stringify([
{ modelId: 'mlx-community/Qwen3-30B-A3B-4bit', launchedAt: Date.now() },
]));
});
await page.reload({ waitUntil: 'networkidle' });
await page.waitForTimeout(2000);
// Interact with UI elements
await page.locator('text=SELECT MODEL').click();
await page.waitForTimeout(1000);
// Take screenshot
await page.screenshot({ path: '/tmp/screenshot.png', fullPage: false });
await browser.close();
})();
```
### Uploading Images to GitHub PRs
GitHub's API doesn't support direct image upload for PR comments. Workaround:
1. **Commit images to the branch** (temporarily):
```bash
cp /tmp/screenshot.png .
git add screenshot.png
git commit -m "temp: add screenshots for PR"
git push origin <branch>
COMMIT_SHA=$(git rev-parse HEAD)
```
2. **Post PR comment** referencing the raw image URL (uses permanent commit SHA so images survive deletion):
```bash
gh pr comment <PR_NUMBER> --body "![Screenshot](https://raw.githubusercontent.com/exo-explore/exo/${COMMIT_SHA}/screenshot.png)"
```
3. **Remove the images** from the branch:
```bash
git rm screenshot.png
git commit -m "chore: remove temporary screenshot files"
git push origin <branch>
```
The images still render in the PR comment because they reference the permanent commit SHA.
Generated
+11 -225
View File
@@ -141,12 +141,6 @@ version = "0.3.9"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "76a2e8124351fda1ef8aaaa3bbd7ebbcb486bbcd4225aca0aa0d84bb2db8fecb"
[[package]]
name = "arrayvec"
version = "0.7.6"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "7c02d123df017efcdfbd739ef81735b36c5ba83ec3c59c80a9d7ecc718f92e50"
[[package]]
name = "asn1-rs"
version = "0.7.1"
@@ -304,19 +298,6 @@ version = "1.8.0"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "55248b47b0caf0546f7988906588779981c43bb1bc9d0c44087278f80cdb44ba"
[[package]]
name = "bigdecimal"
version = "0.4.9"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "560f42649de9fa436b73517378a147ec21f6c997a546581df4b4b31677828934"
dependencies = [
"autocfg",
"libm",
"num-bigint",
"num-integer",
"num-traits",
]
[[package]]
name = "bimap"
version = "0.6.3"
@@ -353,31 +334,6 @@ dependencies = [
"generic-array",
]
[[package]]
name = "bon"
version = "3.8.1"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "ebeb9aaf9329dff6ceb65c689ca3db33dbf15f324909c60e4e5eef5701ce31b1"
dependencies = [
"bon-macros",
"rustversion",
]
[[package]]
name = "bon-macros"
version = "3.8.1"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "77e9d642a7e3a318e37c2c9427b5a6a48aa1ad55dcd986f3034ab2239045a645"
dependencies = [
"darling",
"ident_case",
"prettyplease",
"proc-macro2",
"quote",
"rustversion",
"syn 2.0.111",
]
[[package]]
name = "bs58"
version = "0.5.1"
@@ -541,15 +497,6 @@ version = "0.4.3"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "2f421161cb492475f1661ddc9815a745a1c894592070661180fdec3d4872e9c3"
[[package]]
name = "convert_case"
version = "0.10.0"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "633458d4ef8c78b72454de2d54fd6ab2e60f9e02be22f3c6104cdc8a4e0fceb9"
dependencies = [
"unicode-segmentation",
]
[[package]]
name = "core-foundation"
version = "0.9.4"
@@ -700,41 +647,6 @@ dependencies = [
"syn 2.0.111",
]
[[package]]
name = "darling"
version = "0.21.3"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "9cdf337090841a411e2a7f3deb9187445851f91b309c0c0a29e05f74a00a48c0"
dependencies = [
"darling_core",
"darling_macro",
]
[[package]]
name = "darling_core"
version = "0.21.3"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "1247195ecd7e3c85f83c8d2a366e4210d588e802133e1e355180a9870b517ea4"
dependencies = [
"fnv",
"ident_case",
"proc-macro2",
"quote",
"strsim",
"syn 2.0.111",
]
[[package]]
name = "darling_macro"
version = "0.21.3"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "d38308df82d1080de0afee5d069fa14b0326a88c14f15c5ccda35b4a6c414c81"
dependencies = [
"darling_core",
"quote",
"syn 2.0.111",
]
[[package]]
name = "data-encoding"
version = "2.9.0"
@@ -806,29 +718,6 @@ dependencies = [
"powerfmt",
]
[[package]]
name = "derive_more"
version = "2.1.0"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "10b768e943bed7bf2cab53df09f4bc34bfd217cdb57d971e769874c9a6710618"
dependencies = [
"derive_more-impl",
]
[[package]]
name = "derive_more-impl"
version = "2.1.0"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "6d286bfdaf75e988b4a78e013ecd79c581e06399ab53fbacd2d916c2f904f30b"
dependencies = [
"convert_case",
"proc-macro2",
"quote",
"rustc_version",
"syn 2.0.111",
"unicode-xid",
]
[[package]]
name = "digest"
version = "0.10.7"
@@ -999,22 +888,17 @@ name = "exo_pyo3_bindings"
version = "0.0.1"
dependencies = [
"delegate",
"derive_more",
"env_logger",
"extend",
"futures",
"impl-trait-for-tuples",
"futures-lite",
"libp2p",
"log",
"networking",
"once_cell",
"pin-project",
"pyo3",
"pyo3-async-runtimes",
"pyo3-log",
"pyo3-stub-gen",
"thiserror 2.0.17",
"thread_local",
"tokio",
"util",
]
@@ -1030,6 +914,12 @@ dependencies = [
"syn 2.0.111",
]
[[package]]
name = "fastrand"
version = "2.3.0"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "37909eebbb50d72f9059c3b6d82c0463f2ff062c9e95845c43a6c9c0355411be"
[[package]]
name = "ff"
version = "0.13.1"
@@ -1138,7 +1028,10 @@ version = "2.6.1"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "f78e10609fe0e0b3f4157ffab1876319b5b0db102a2c60dc4626306dc46b44ad"
dependencies = [
"fastrand",
"futures-core",
"futures-io",
"parking",
"pin-project-lite",
]
@@ -1625,12 +1518,6 @@ dependencies = [
"zerovec",
]
[[package]]
name = "ident_case"
version = "1.0.1"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "b9e0384b61958566e926dc50660321d12159025e767c18e043daf26b70104c39"
[[package]]
name = "idna"
version = "1.1.0"
@@ -1706,17 +1593,6 @@ dependencies = [
"xmltree",
]
[[package]]
name = "impl-trait-for-tuples"
version = "0.2.3"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "a0eb5a3343abf848c0984fe4604b2b105da9539376e24fc0a3b0007411ae4fd9"
dependencies = [
"proc-macro2",
"quote",
"syn 2.0.111",
]
[[package]]
name = "indexmap"
version = "2.12.1"
@@ -1745,15 +1621,6 @@ dependencies = [
"generic-array",
]
[[package]]
name = "internment"
version = "0.8.6"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "636d4b0f6a39fd684effe2a73f5310df16a3fa7954c26d36833e98f44d1977a2"
dependencies = [
"hashbrown 0.15.5",
]
[[package]]
name = "inventory"
version = "0.3.21"
@@ -1904,12 +1771,6 @@ version = "0.2.178"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "37c93d8daa9d8a012fd8ab92f088405fb202ea0b6ab73ee2482ae66af4f42091"
[[package]]
name = "libm"
version = "0.2.15"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "f9fbbcab51052fe104eb5e5d351cf728d30a5be1fe14d9be8a3b097481fb97de"
[[package]]
name = "libp2p"
version = "0.56.0"
@@ -2899,16 +2760,13 @@ name = "networking"
version = "0.0.1"
dependencies = [
"delegate",
"derive_more",
"either",
"extend",
"futures",
"futures-lite",
"futures-timer",
"impl-trait-for-tuples",
"keccak-const",
"libp2p",
"log",
"thiserror 2.0.17",
"tokio",
"tracing-subscriber",
"util",
@@ -2993,17 +2851,6 @@ dependencies = [
"num-traits",
]
[[package]]
name = "num-rational"
version = "0.4.2"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "f83d14da390562dca69fc84082e73e548e1ad308d24accdedd2720017cb37824"
dependencies = [
"num-bigint",
"num-integer",
"num-traits",
]
[[package]]
name = "num-traits"
version = "0.2.19"
@@ -3307,16 +3154,6 @@ dependencies = [
"zerocopy",
]
[[package]]
name = "prettyplease"
version = "0.2.37"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "479ca8adacdd7ce8f1fb39ce9ecccbfe93a3f1344b3d0d97f20bc0196208f62b"
dependencies = [
"proc-macro2",
"syn 2.0.111",
]
[[package]]
name = "primeorder"
version = "0.13.6"
@@ -3364,28 +3201,14 @@ version = "0.27.2"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "ab53c047fcd1a1d2a8820fe84f05d6be69e9526be40cb03b73f86b6b03e6d87d"
dependencies = [
"bigdecimal",
"either",
"hashbrown 0.16.1",
"indexmap",
"indoc",
"inventory",
"libc",
"lock_api",
"memoffset",
"num-bigint",
"num-complex",
"num-rational",
"num-traits",
"once_cell",
"ordered-float",
"parking_lot",
"portable-atomic",
"pyo3-build-config",
"pyo3-ffi",
"pyo3-macros",
"rust_decimal",
"smallvec",
"unindent",
]
@@ -3740,12 +3563,6 @@ dependencies = [
"yasna",
]
[[package]]
name = "recursion"
version = "0.5.4"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "9dba2197bf7b1d87b4dd460c195f4edeb45a94e82e8054f8d5f317c1f0e93ca1"
[[package]]
name = "redox_syscall"
version = "0.5.18"
@@ -3832,16 +3649,6 @@ dependencies = [
"tokio",
]
[[package]]
name = "rust_decimal"
version = "1.39.0"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "35affe401787a9bd846712274d97654355d21b2a2c092a3139aabe31e9022282"
dependencies = [
"arrayvec",
"num-traits",
]
[[package]]
name = "rustc-hash"
version = "1.1.0"
@@ -4706,24 +4513,12 @@ version = "1.0.22"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "9312f7c4f6ff9069b165498234ce8be658059c6728633667c526e27dc2cf1df5"
[[package]]
name = "unicode-segmentation"
version = "1.12.0"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "f6ccf251212114b54433ec949fd6a7841275f9ada20dddd2f29e9ceea4501493"
[[package]]
name = "unicode-width"
version = "0.2.2"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "b4ac048d71ede7ee76d585517add45da530660ef4390e49b098733c6e897f254"
[[package]]
name = "unicode-xid"
version = "0.2.6"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "ebc1c04c71510c7f702b52b7c350734c9ff1295c464a03335b00bb84fc54f853"
[[package]]
name = "unicode_names2"
version = "1.3.0"
@@ -4807,15 +4602,6 @@ checksum = "06abde3611657adf66d383f00b093d7faecc7fa57071cce2578660c9f1010821"
[[package]]
name = "util"
version = "0.0.1"
dependencies = [
"bon",
"derive_more",
"extend",
"internment",
"once_cell",
"recursion",
"thiserror 2.0.17",
]
[[package]]
name = "uuid"
+1 -30
View File
@@ -26,49 +26,20 @@ opt-level = 3
networking = { path = "rust/networking" }
util = { path = "rust/util" }
# Proc-macro authoring tools
syn = "2.0"
quote = "1.0"
proc-macro2 = "1.0"
darling = "0.20"
# Macro dependecies
extend = "1.2"
delegate = "0.13"
impl-trait-for-tuples = "0.2"
clap = "4.5"
derive_more = { version = "2.0.1", features = ["display"] }
pin-project = "1"
# Utility dependencies
itertools = "0.14"
thiserror = "2"
internment = "0.8"
recursion = "0.5"
regex = "1.11"
once_cell = "1.21"
thread_local = "1.1"
bon = "3.4"
generativity = "1.1"
anyhow = "1.0"
keccak-const = "0.2"
# Functional generics/lenses frameworks
frunk_core = "0.4"
frunk = "0.4"
frunk_utils = "0.2"
frunk-enum-core = "0.3"
# Async dependencies
tokio = "1.46"
futures = "0.3"
futures-util = "0.3"
futures-lite = "2.6.1"
futures-timer = "3.0"
# Data structures
either = "1.15"
ordered-float = "5.0"
ahash = "0.8"
# Tracing/logging
log = "0.4"
+11 -11
View File
@@ -5,21 +5,21 @@
[X] Fetching download status of all models on start
[X] Deduplication of tasks in plan_step.
[X] resolve_allow_patterns should just be wildcard now.
[] no mx_barrier in genreate.py mlx_generate at the end.
[X] no mx_barrier in genreate.py mlx_generate at the end.
[] cache assertion not needed in auto_parallel.py PipelineLastLayer.
[] GPTOSS support dropped in auto_parallel.py.
[] sharding changed "all-to-sharded" became _all_to_sharded in auto_parallel.py.
[] same as above with "sharded-to-all" became _sharded_to_all in auto_parallel.py.
[] Dropped support for Ministral3Model, DeepseekV32Model, Glm4MoeModel, Qwen3NextModel, GptOssMode in auto_parallel.py.
[X] GPTOSS support dropped in auto_parallel.py.
[X] sharding changed "all-to-sharded" became _all_to_sharded in auto_parallel.py.
[X] same as above with "sharded-to-all" became _sharded_to_all in auto_parallel.py.
[X] Dropped support for Ministral3Model, DeepseekV32Model, Glm4MoeModel, Qwen3NextModel, GptOssMode in auto_parallel.py.
[] Dropped prefill/decode code in auto_parallel.py and utils_mlx.py.
[X] KV_CACHE_BITS should be None to disable quantized KV cache.
[] Dropped _set_nofile_limit in utils_mlx.py.
[] We have group optional in load_mlx_items in utils_mlx.py.
[] Dropped add_missing_chat_templates for GptOss in load_mlx_items in utils_mlx.py.
[] Dropped model.make_cache in make_kv_cache in utils_mlx.py.
[X] Dropped _set_nofile_limit in utils_mlx.py.
[X] We have group optional in load_mlx_items in utils_mlx.py.
[X] Dropped add_missing_chat_templates for GptOss in load_mlx_items in utils_mlx.py.
[X] Dropped model.make_cache in make_kv_cache in utils_mlx.py.
[X] We put cache limit back in utils_mlx.py.
[] topology.py remove_node removes the connections after checking if node is is in self._node_id_to_rx_id_map. on beta_1 it checks after, so would remove stale connections I guess?
[] Missing Glm 4.7 model cards (this isn't ready yet but should be picked up, probably create an issue... the blocker is transforemrs version doesn't support the tokenizer for Glm 4.7. rc-1 does but we can't upgrade as it breaks other things.)
[X] topology.py remove_node removes the connections after checking if node is is in self._node_id_to_rx_id_map. on beta_1 it checks after, so would remove stale connections I guess?
[X] Missing Glm 4.7 model cards (this isn't ready yet but should be picked up, probably create an issue... the blocker is transforemrs version doesn't support the tokenizer for Glm 4.7. rc-1 does but we can't upgrade as it breaks other things.)
[] try-except in _command_processor only excepts ValueError. This was silently failing leading to un-debuggable errors (we had a KeyError that was happening ). Changed this to catch Exception instead of ValueError. See exo-v2 89ae38405e0052e3c22405daf094b065878aa873 and fb99fea69b5a39017efc90c5dad0072e677455f0.
[X] In placement.py, place_instance no longer looks at model_meta.supports_tensor and check if this tensor parallel number of nodes is supported by the model's tensor dimensions.
[X] In placement.py, place_instanec, we no longer have the special case to exclude DeepSeek v3.1 pipeline parallel (it doesn't work).
+30 -4
View File
@@ -5,7 +5,7 @@
<img alt="exo logo" src="/docs/imgs/exo-logo-transparent.png" width="50%" height="50%">
</picture>
exo: Run your own AI cluster at home with everyday devices. Maintained by [exo labs](https://x.com/exolabs).
exo: Run frontier AI locally. Maintained by [exo labs](https://x.com/exolabs).
<p align="center">
<a href="https://discord.gg/TJ4P57arEm" target="_blank" rel="noopener noreferrer"><img src="https://img.shields.io/badge/Discord-Join%20Server-5865F2?logo=discord&logoColor=white" alt="Discord"></a>
@@ -72,16 +72,30 @@ There are two ways to run exo:
### Run from Source (macOS)
If you have [Nix](https://nixos.org/) installed, you can skip most of the steps below and run exo directly:
```bash
nix run .#exo
```
**Note:** To accept the Cachix binary cache (and avoid the Xcode Metal ToolChain), add to `/etc/nix/nix.conf`:
```
trusted-users = root (or your username)
experimental-features = nix-command flakes
```
Then restart the Nix daemon: `sudo launchctl kickstart -k system/org.nixos.nix-daemon`
**Prerequisites:**
- [Xcode](https://developer.apple.com/xcode/) (provides the Metal ToolChain required for MLX compilation)
- [brew](https://github.com/Homebrew/brew) (for simple package management on macOS)
```bash
/bin/bash -c "$(curl -fsSL https://raw.githubusercontent.com/Homebrew/install/HEAD/install.sh)"
```
- [uv](https://github.com/astral-sh/uv) (for Python dependency management)
- [macmon](https://github.com/vladkens/macmon) (for hardware monitoring on Apple Silicon)
- [node](https://github.com/nodejs/node) (for building the dashboard)
```bash
brew install uv macmon node
```
@@ -107,6 +121,10 @@ uv run exo
This starts the exo dashboard and API at http://localhost:52415/
*Please view the section on RDMA to enable this feature on MacOS >=26.2!*
### Run from Source (Linux)
**Prerequisites:**
@@ -230,7 +248,7 @@ This removes:
RDMA is a new capability added to macOS 26.2. It works on any Mac with Thunderbolt 5 (M4 Pro Mac Mini, M4 Max Mac Studio, M4 Max MacBook Pro, M3 Ultra Mac Studio).
Note that on Mac Studio, you cannot use the Thunderbolt 5 port next to the Ethernet port.
Please refer to the caveats for immediate troubleshooting.
To enable RDMA on macOS, follow these steps:
@@ -247,6 +265,14 @@ To enable RDMA on macOS, follow these steps:
After that, RDMA will be enabled in macOS and exo will take care of the rest.
**Important Caveats**
1. Devices that wish to be part of an RDMA cluster must be connected to all other devices in the cluster.
2. The cables must support TB5.
3. On a Mac Studio, you cannot use the Thunderbolt 5 port next to the Ethernet port.
4. If running from source, please use the script found at `tmp/set_rdma_network_config.sh`, which will disable Thunderbolt Bridge and set dhcp on each RDMA port.
5. RDMA ports may be unable to discover each other on different versions of MacOS. Please ensure that OS versions match exactly (even beta version numbers) on all devices.
---
### Using the API
-14
View File
@@ -1,28 +1,14 @@
2. Currently a lot of requests from the API are timing out, but we still process those requests internally. If an API request times out, we should cancel all corresponding tasks to that API request (why process a request with nobody listening).
3. Task cancellation. When API http request gets cancelled, it should cancel corresponding task.
4. I'd like to see profiled network latency / bandwidth.
5. I'd like to see how much bandwidth each link is using.
6. We should handle the case where one machine doesn't have the model downloaded and then other machines are waiting on it. In this case we get loads of timeout errors because the others are waiting for the one that needs to download the model.
7. Solve the problem of in continuous batching when a new prompt comes in, it will block decode of the current batch until the prefill is complete.
8. We want people to be able to copy models over to a new device without ever connecting EXO to the internet. Right now EXO require internet connection once to cache some files to check if a download is complete. Instead, we should simply check if there is a non-empty model folder locally with no .partial files. This indicates it's a fully downloaded model that can be loaded.
10. More granular control over how to deploy instances.
12. Nix is great but installing it is a pain and we have ended up in a lot of cases having PATH issues or installation issues. For example, after rebooting mike it seemed to no longer have a nix installation and needed reinstalling. It has a bunch of broken symlinks left over from nix that caused ssh to fail, making it even harder to debug. We need consistent environments (perhaps MDM) so we can guarantee nix is installed properly on each machine.
13. Memory pressure instead of memory used.
14. Show the type of each connection (TB5, Ethernet, etc.) in the UI. Refer to old exo: https://github.com/exo-explore/exo/blob/56f783b38dc6b08ce606b07a5386dc40dae00330/exo/helpers.py#L251
15. Prioritise certain connection types (or by latency). TB5 > Ethernet > WiFi. Refer to old exo: https://github.com/exo-explore/exo/blob/56f783b38dc6b08ce606b07a5386dc40dae00330/exo/helpers.py#L251
16. Dynamically switch to higher priority connection when it becomes available. Probably bring back InstanceReplacedAtomically.
17. Faster model loads by streaming model from other devices in cluster.
18. Add support for specifying the type of network connection to use in a test. Depends on 15/16.
20. Add chat completion cancellations (e.g OpenWebUI has something for cancelling an ongoing request).
23. Do we need cache_limit? We went back and forth on that a lot because we thought it might be causing issues. One problem is it sets it relative to model size. So if you have multiple models loaded in it will take the most recent model size for the cache_limit. This is problematic if you launch DeepSeek -> Llama for example.
24. further openai/lmstudio api compatibility
25. Rethink retry logic
26. Task cancellation. When API http request gets cancelled, it should cancel corresponding task.
27. Log cleanup - per-module log filters and default to DEBUG log levels
28. Validate RDMA connections with ibv_devinfo in the info gatherer
Potential refactors:
2. Topology can be simplified
Random errors we've run into:
+4
View File
@@ -342,6 +342,8 @@
SDKROOT = macosx;
SWIFT_ACTIVE_COMPILATION_CONDITIONS = "DEBUG $(inherited)";
SWIFT_OPTIMIZATION_LEVEL = "-Onone";
SWIFT_TREAT_WARNINGS_AS_ERRORS = YES;
GCC_TREAT_WARNINGS_AS_ERRORS = YES;
};
name = Debug;
};
@@ -397,6 +399,8 @@
MTL_FAST_MATH = YES;
SDKROOT = macosx;
SWIFT_COMPILATION_MODE = wholemodule;
SWIFT_TREAT_WARNINGS_AS_ERRORS = YES;
GCC_TREAT_WARNINGS_AS_ERRORS = YES;
};
name = Release;
};
+290 -136
View File
@@ -14,16 +14,28 @@ struct ContentView: View {
@EnvironmentObject private var networkStatusService: NetworkStatusService
@EnvironmentObject private var localNetworkChecker: LocalNetworkChecker
@EnvironmentObject private var updater: SparkleUpdater
@EnvironmentObject private var thunderboltBridgeService: ThunderboltBridgeService
@EnvironmentObject private var settingsWindowController: SettingsWindowController
@State private var focusedNode: NodeViewModel?
@State private var deletingInstanceIDs: Set<String> = []
@State private var showAllNodes = false
@State private var showAllInstances = false
@State private var baseURLCopied = false
@State private var showAdvanced = false
@State private var showDebugInfo = false
@State private var bugReportInFlight = false
@State private var bugReportMessage: String?
private enum BugReportPhase: Equatable {
case idle
case prompting
case sending(String)
case success(String)
case failure(String)
}
@State private var bugReportPhase: BugReportPhase = .idle
@State private var bugReportUserDescription: String = ""
@State private var uninstallInProgress = false
@State private var pendingNamespace: String = ""
@State private var pendingHFToken: String = ""
@State private var pendingEnableImageModels = false
var body: some View {
VStack(alignment: .leading, spacing: 12) {
@@ -255,96 +267,79 @@ struct ContentView: View {
VStack(alignment: .leading, spacing: 0) {
if controller.status != .stopped {
dashboardButton
baseURLRow
Divider()
.padding(.vertical, 8)
} else {
Divider()
.padding(.vertical, 4)
}
advancedSection
.padding(.bottom, 8)
controlButton(title: "Quit", tint: .secondary) {
HoverButton(
title: "Settings",
tint: .primary,
trailingSystemImage: "gear"
) {
settingsWindowController.open(
controller: controller,
updater: updater,
networkStatusService: networkStatusService,
thunderboltBridgeService: thunderboltBridgeService,
stateService: stateService
)
}
HoverButton(
title: "Check for Updates",
tint: .primary,
trailingSystemImage: "arrow.triangle.2.circlepath"
) {
updater.checkForUpdates()
}
.padding(.bottom, 8)
HoverButton(title: "Quit", tint: .secondary) {
controller.stop()
NSApplication.shared.terminate(nil)
}
}
}
private var advancedSection: some View {
VStack(alignment: .leading, spacing: 6) {
HStack {
Text("Advanced")
.font(.caption)
.foregroundColor(.secondary)
Spacer()
collapseButton(isExpanded: $showAdvanced)
}
.animation(nil, value: showAdvanced)
if showAdvanced {
VStack(alignment: .leading, spacing: 8) {
VStack(alignment: .leading, spacing: 4) {
Text("Cluster Namespace")
.font(.caption2)
.foregroundColor(.secondary)
HStack {
TextField("optional", text: $pendingNamespace)
.textFieldStyle(.roundedBorder)
.font(.caption2)
.onAppear {
pendingNamespace = controller.customNamespace
}
Button("Save & Restart") {
controller.customNamespace = pendingNamespace
if controller.status == .running || controller.status == .starting {
controller.restart()
}
}
.font(.caption2)
.disabled(pendingNamespace == controller.customNamespace)
}
}
HoverButton(title: "Check for Updates", small: true) {
updater.checkForUpdates()
}
debugSection
HoverButton(title: "Uninstall", tint: .red, small: true) {
showUninstallConfirmationAlert()
}
.disabled(uninstallInProgress)
}
.transition(.opacity)
}
}
.animation(.easeInOut(duration: 0.25), value: showAdvanced)
}
private func controlButton(title: String, tint: Color = .primary, action: @escaping () -> Void)
-> some View
{
HoverButton(title: title, tint: tint, trailingSystemImage: nil, action: action)
}
private var dashboardButton: some View {
Button {
HoverButton(
title: "Web Dashboard",
tint: .primary,
trailingSystemImage: "arrow.up.right"
) {
guard let url = URL(string: "http://localhost:52415/") else { return }
NSWorkspace.shared.open(url)
} label: {
HStack {
Image(systemName: "arrow.up.right.square")
.imageScale(.small)
Text("Dashboard")
.fontWeight(.medium)
Spacer()
}
.padding(.vertical, 8)
.padding(.horizontal, 10)
.background(
RoundedRectangle(cornerRadius: 8, style: .continuous)
.fill(Color(red: 1.0, green: 0.87, blue: 0.0).opacity(0.2))
)
}
.buttonStyle(.plain)
.padding(.bottom, 4)
}
private var baseURLRow: some View {
HStack(spacing: 6) {
Image(systemName: "link")
.imageScale(.small)
.foregroundColor(.secondary)
Text("localhost:52415/v1")
.font(.system(.caption, design: .monospaced))
.foregroundColor(.primary)
Spacer()
Button {
NSPasteboard.general.clearContents()
NSPasteboard.general.setString("http://localhost:52415/v1", forType: .string)
baseURLCopied = true
DispatchQueue.main.asyncAfter(deadline: .now() + 2) {
baseURLCopied = false
}
} label: {
Image(systemName: baseURLCopied ? "checkmark" : "doc.on.doc")
.imageScale(.small)
.foregroundColor(baseURLCopied ? .green : .secondary)
.contentTransition(.symbolEffect(.replace))
}
.buttonStyle(.plain)
.help("Copy API base URL")
}
.padding(.vertical, 4)
.padding(.horizontal, 8)
}
private func collapseButton(isExpanded: Binding<Bool>) -> some View {
@@ -423,6 +418,44 @@ struct ContentView: View {
}
}
/// Shows TB bridge status for all nodes from exo cluster state
private var clusterThunderboltBridgeView: some View {
let bridgeStatuses = stateService.latestSnapshot?.nodeThunderboltBridge ?? [:]
let localNodeId = stateService.localNodeId
let nodeProfiles = stateService.latestSnapshot?.nodeProfiles ?? [:]
return VStack(alignment: .leading, spacing: 1) {
if bridgeStatuses.isEmpty {
Text("Cluster TB Bridge: No data")
.font(.caption2)
.foregroundColor(.secondary)
} else {
Text("Cluster TB Bridge Status:")
.font(.caption2)
.foregroundColor(.secondary)
ForEach(Array(bridgeStatuses.keys.sorted()), id: \.self) { nodeId in
if let status = bridgeStatuses[nodeId] {
let nodeName =
nodeProfiles[nodeId]?.friendlyName ?? String(nodeId.prefix(8))
let isLocal = nodeId == localNodeId
let prefix = isLocal ? " \(nodeName) (local):" : " \(nodeName):"
let statusText =
!status.exists
? "N/A"
: (status.enabled ? "Enabled" : "Disabled")
let color: Color =
!status.exists
? .secondary
: (status.enabled ? .red : .green)
Text("\(prefix) \(statusText)")
.font(.caption2)
.foregroundColor(color)
}
}
}
}
}
private var interfaceIpList: some View {
let statuses = networkStatusService.status.interfaceStatuses
return VStack(alignment: .leading, spacing: 1) {
@@ -465,6 +498,7 @@ struct ContentView: View {
Text(thunderboltStatusText)
.font(.caption2)
.foregroundColor(thunderboltStatusColor)
clusterThunderboltBridgeView
interfaceIpList
rdmaStatusView
sendBugReportButton
@@ -478,21 +512,45 @@ struct ContentView: View {
}
private var rdmaStatusView: some View {
let rdma = networkStatusService.status.rdmaStatus
let rdmaStatuses = stateService.latestSnapshot?.nodeRdmaCtl ?? [:]
let localNodeId = stateService.localNodeId
let nodeProfiles = stateService.latestSnapshot?.nodeProfiles ?? [:]
let localDevices = networkStatusService.status.localRdmaDevices
let localPorts = networkStatusService.status.localRdmaActivePorts
return VStack(alignment: .leading, spacing: 1) {
Text("RDMA: \(rdmaStatusText(rdma))")
.font(.caption2)
.foregroundColor(rdmaStatusColor(rdma))
if !rdma.devices.isEmpty {
Text(" Devices: \(rdma.devices.joined(separator: ", "))")
if rdmaStatuses.isEmpty {
Text("Cluster RDMA: No data")
.font(.caption2)
.foregroundColor(.secondary)
} else {
Text("Cluster RDMA Status:")
.font(.caption2)
.foregroundColor(.secondary)
ForEach(Array(rdmaStatuses.keys.sorted()), id: \.self) { nodeId in
if let status = rdmaStatuses[nodeId] {
let nodeName =
nodeProfiles[nodeId]?.friendlyName ?? String(nodeId.prefix(8))
let isLocal = nodeId == localNodeId
let prefix = isLocal ? " \(nodeName) (local):" : " \(nodeName):"
let statusText = status.enabled ? "Enabled" : "Disabled"
let color: Color = status.enabled ? .green : .orange
Text("\(prefix) \(statusText)")
.font(.caption2)
.foregroundColor(color)
}
}
}
if !localDevices.isEmpty {
Text(" Local Devices: \(localDevices.joined(separator: ", "))")
.font(.caption2)
.foregroundColor(.secondary)
}
if !rdma.activePorts.isEmpty {
Text(" Active Ports:")
if !localPorts.isEmpty {
Text(" Local Active Ports:")
.font(.caption2)
.foregroundColor(.secondary)
ForEach(rdma.activePorts, id: \.device) { port in
ForEach(localPorts, id: \.device) { port in
Text(" \(port.device) port \(port.port): \(port.state)")
.font(.caption2)
.foregroundColor(.green)
@@ -501,62 +559,116 @@ struct ContentView: View {
}
}
private func rdmaStatusText(_ rdma: RDMAStatus) -> String {
switch rdma.rdmaCtlEnabled {
case .some(true):
return "Enabled"
case .some(false):
return "Disabled"
case nil:
return rdma.devices.isEmpty ? "Not Available" : "Available"
}
}
private func rdmaStatusColor(_ rdma: RDMAStatus) -> Color {
switch rdma.rdmaCtlEnabled {
case .some(true):
return .green
case .some(false):
return .orange
case nil:
return rdma.devices.isEmpty ? .secondary : .green
}
}
private var sendBugReportButton: some View {
VStack(alignment: .leading, spacing: 4) {
Button {
Task {
await sendBugReport()
}
} label: {
HStack {
if bugReportInFlight {
ProgressView()
.scaleEffect(0.6)
VStack(alignment: .leading, spacing: 6) {
switch bugReportPhase {
case .idle:
Button {
bugReportPhase = .prompting
bugReportUserDescription = ""
} label: {
HStack {
Text("Send Bug Report")
.font(.caption)
.fontWeight(.semibold)
Spacer()
}
Text("Send Bug Report")
.font(.caption)
.fontWeight(.semibold)
Spacer()
.padding(.vertical, 6)
.padding(.horizontal, 8)
.background(
RoundedRectangle(cornerRadius: 6)
.fill(Color.accentColor.opacity(0.12))
)
}
.padding(.vertical, 6)
.padding(.horizontal, 8)
.buttonStyle(.plain)
case .prompting:
VStack(alignment: .leading, spacing: 6) {
Text("What's the issue? (optional)")
.font(.caption2)
.foregroundColor(.secondary)
TextEditor(text: $bugReportUserDescription)
.font(.caption2)
.frame(height: 60)
.overlay(
RoundedRectangle(cornerRadius: 4)
.stroke(Color.secondary.opacity(0.3), lineWidth: 1)
)
HStack(spacing: 8) {
Button("Send") {
Task {
await sendBugReport()
}
}
.font(.caption2)
.buttonStyle(.borderedProminent)
.controlSize(.small)
Button("Cancel") {
bugReportPhase = .idle
}
.font(.caption2)
.buttonStyle(.bordered)
.controlSize(.small)
}
}
.padding(8)
.background(
RoundedRectangle(cornerRadius: 6)
.fill(Color.accentColor.opacity(0.12))
.fill(Color.accentColor.opacity(0.06))
)
}
.buttonStyle(.plain)
.disabled(bugReportInFlight)
if let message = bugReportMessage {
Text(message)
case .sending(let message):
HStack(spacing: 6) {
ProgressView()
.scaleEffect(0.6)
Text(message)
.font(.caption2)
.foregroundColor(.secondary)
}
case .success(let message):
VStack(alignment: .leading, spacing: 6) {
Text(message)
.font(.caption2)
.foregroundColor(.secondary)
.fixedSize(horizontal: false, vertical: true)
Button {
openGitHubIssue()
} label: {
HStack(spacing: 4) {
Image(systemName: "arrow.up.right.square")
.imageScale(.small)
Text("Create GitHub Issue")
.font(.caption2)
}
}
.buttonStyle(.bordered)
.controlSize(.small)
Button("Done") {
bugReportPhase = .idle
bugReportUserDescription = ""
}
.font(.caption2)
.buttonStyle(.plain)
.foregroundColor(.secondary)
.fixedSize(horizontal: false, vertical: true)
}
case .failure(let message):
VStack(alignment: .leading, spacing: 4) {
Text(message)
.font(.caption2)
.foregroundColor(.red)
.fixedSize(horizontal: false, vertical: true)
Button("Dismiss") {
bugReportPhase = .idle
}
.font(.caption2)
.buttonStyle(.plain)
.foregroundColor(.secondary)
}
}
}
.animation(.easeInOut(duration: 0.2), value: bugReportPhase)
}
private var processToggleBinding: Binding<Bool> {
@@ -600,16 +712,58 @@ struct ContentView: View {
}
private func sendBugReport() async {
bugReportInFlight = true
bugReportMessage = "Collecting logs..."
bugReportPhase = .sending("Collecting logs...")
let service = BugReportService()
let description = bugReportUserDescription.trimmingCharacters(in: .whitespacesAndNewlines)
do {
let outcome = try await service.sendReport(isManual: true)
bugReportMessage = outcome.message
let outcome = try await service.sendReport(
isManual: true,
userDescription: description.isEmpty ? nil : description
)
if outcome.success {
bugReportPhase = .success(outcome.message)
} else {
bugReportPhase = .failure(outcome.message)
}
} catch {
bugReportMessage = error.localizedDescription
bugReportPhase = .failure(error.localizedDescription)
}
}
private func openGitHubIssue() {
let description = bugReportUserDescription.trimmingCharacters(in: .whitespacesAndNewlines)
var bodyParts: [String] = []
bodyParts.append("## Describe the bug")
bodyParts.append("")
if !description.isEmpty {
bodyParts.append(description)
} else {
bodyParts.append("A clear and concise description of what the bug is.")
}
bodyParts.append("")
bodyParts.append("## Environment")
bodyParts.append("")
bodyParts.append("- macOS Version: \(ProcessInfo.processInfo.operatingSystemVersionString)")
bodyParts.append("- EXO Version: \(buildTag) (\(buildCommit))")
bodyParts.append("")
bodyParts.append("## Additional context")
bodyParts.append("")
bodyParts.append("A bug report with diagnostic logs was submitted via the app.")
let body = bodyParts.joined(separator: "\n")
var components = URLComponents(string: "https://github.com/exo-explore/exo/issues/new")!
components.queryItems = [
URLQueryItem(name: "template", value: "bug_report.md"),
URLQueryItem(name: "title", value: "[BUG] "),
URLQueryItem(name: "body", value: body),
URLQueryItem(name: "labels", value: "bug"),
]
if let url = components.url {
NSWorkspace.shared.open(url)
}
bugReportInFlight = false
}
private func showUninstallConfirmationAlert() {
+25 -6
View File
@@ -14,14 +14,16 @@ import SwiftUI
import UserNotifications
import os.log
@main
struct EXOApp: App {
@StateObject private var controller: ExoProcessController
@StateObject private var stateService: ClusterStateService
@StateObject private var networkStatusService: NetworkStatusService
@StateObject private var localNetworkChecker: LocalNetworkChecker
@StateObject private var updater: SparkleUpdater
@StateObject private var thunderboltBridgeService: ThunderboltBridgeService
@StateObject private var settingsWindowController: SettingsWindowController
private let terminationObserver: TerminationObserver
private let firstLaunchPopout = FirstLaunchPopout()
private let ciContext = CIContext(options: nil)
init() {
@@ -41,11 +43,15 @@ struct EXOApp: App {
let localNetwork = LocalNetworkChecker()
_localNetworkChecker = StateObject(wrappedValue: localNetwork)
_updater = StateObject(wrappedValue: updater)
let thunderboltBridge = ThunderboltBridgeService(clusterStateService: service)
_thunderboltBridgeService = StateObject(wrappedValue: thunderboltBridge)
_settingsWindowController = StateObject(wrappedValue: SettingsWindowController())
enableLaunchAtLoginIfNeeded()
NetworkSetupHelper.ensureLaunchDaemonInstalled()
// Check local network access BEFORE launching exo
localNetwork.check()
controller.scheduleLaunch(after: 15)
// Install LaunchDaemon to disable Thunderbolt Bridge on startup (prevents network loops)
NetworkSetupHelper.promptAndInstallIfNeeded()
// Check local network access periodically (warning disappears when user grants permission)
localNetwork.startPeriodicChecking(interval: 10)
controller.scheduleLaunch(after: 5)
service.startPolling()
networkStatus.startPolling()
}
@@ -58,8 +64,20 @@ struct EXOApp: App {
.environmentObject(networkStatusService)
.environmentObject(localNetworkChecker)
.environmentObject(updater)
.environmentObject(thunderboltBridgeService)
.environmentObject(settingsWindowController)
} label: {
menuBarIcon
.onReceive(controller.$isFirstLaunchReady) { ready in
if ready {
DispatchQueue.main.asyncAfter(deadline: .now() + 3.0) {
self.firstLaunchPopout.onComplete = { [weak controller] in
controller?.markOnboardingCompleted()
}
self.firstLaunchPopout.show()
}
}
}
}
.menuBarExtraStyle(.window)
}
@@ -130,6 +148,7 @@ struct EXOApp: App {
"Failed to register EXO for launch at login: \(error.localizedDescription)")
}
}
}
/// Helper for managing EXO's launch-at-login registration
@@ -219,7 +238,7 @@ private final class ExoUpdaterDelegate: NSObject, SPUUpdaterDelegate {
}
}
private func showNotification(title: String, body: String) {
nonisolated private func showNotification(title: String, body: String) {
let center = UNUserNotificationCenter.current()
let content = UNMutableNotificationContent()
content.title = title
+93 -5
View File
@@ -3,6 +3,10 @@ import Combine
import Foundation
private let customNamespaceKey = "EXOCustomNamespace"
private let hfTokenKey = "EXOHFToken"
private let enableImageModelsKey = "EXOEnableImageModels"
private let offlineModeKey = "EXOOfflineMode"
private let onboardingCompletedKey = "EXOOnboardingCompleted"
@MainActor
final class ExoProcessController: ObservableObject {
@@ -26,6 +30,10 @@ final class ExoProcessController: ObservableObject {
}
}
static let exoDirectoryURL: URL = {
URL(fileURLWithPath: NSHomeDirectory()).appendingPathComponent(".exo")
}()
@Published private(set) var status: Status = .stopped
@Published private(set) var lastError: String?
@Published private(set) var launchCountdownSeconds: Int?
@@ -37,6 +45,33 @@ final class ExoProcessController: ObservableObject {
UserDefaults.standard.set(customNamespace, forKey: customNamespaceKey)
}
}
@Published var hfToken: String = {
return UserDefaults.standard.string(forKey: hfTokenKey) ?? ""
}()
{
didSet {
UserDefaults.standard.set(hfToken, forKey: hfTokenKey)
}
}
@Published var enableImageModels: Bool = {
return UserDefaults.standard.bool(forKey: enableImageModelsKey)
}()
{
didSet {
UserDefaults.standard.set(enableImageModels, forKey: enableImageModelsKey)
}
}
@Published var offlineMode: Bool = {
return UserDefaults.standard.bool(forKey: offlineModeKey)
}()
{
didSet {
UserDefaults.standard.set(offlineMode, forKey: offlineModeKey)
}
}
/// Fires once when EXO transitions to `.running` for the very first time (fresh install).
@Published private(set) var isFirstLaunchReady = false
private var process: Process?
private var runtimeDirectoryURL: URL?
@@ -60,7 +95,11 @@ final class ExoProcessController: ObservableObject {
let child = Process()
child.executableURL = executableURL
child.currentDirectoryURL = runtimeURL
let exoHomeURL = Self.exoDirectoryURL
try? FileManager.default.createDirectory(
at: exoHomeURL, withIntermediateDirectories: true
)
child.currentDirectoryURL = exoHomeURL
child.environment = makeEnvironment(for: runtimeURL)
child.standardOutput = FileHandle.nullDevice
@@ -87,6 +126,9 @@ final class ExoProcessController: ObservableObject {
try child.run()
process = child
status = .running
// Show welcome popout on every launch
isFirstLaunchReady = true
} catch {
process = nil
status = .failed(message: "Launch error")
@@ -100,11 +142,37 @@ final class ExoProcessController: ObservableObject {
return
}
process.terminationHandler = nil
if process.isRunning {
process.terminate()
}
self.process = nil
status = .stopped
guard process.isRunning else {
self.process = nil
return
}
let proc = process
self.process = nil
Task.detached {
proc.interrupt()
for _ in 0..<50 {
if !proc.isRunning { return }
try? await Task.sleep(nanoseconds: 100_000_000)
}
if proc.isRunning {
proc.terminate()
}
for _ in 0..<30 {
if !proc.isRunning { return }
try? await Task.sleep(nanoseconds: 100_000_000)
}
if proc.isRunning {
kill(proc.processIdentifier, SIGKILL)
}
}
}
func restart() {
@@ -112,6 +180,17 @@ final class ExoProcessController: ObservableObject {
launch()
}
/// Mark onboarding as completed (user interacted with the welcome popout).
func markOnboardingCompleted() {
UserDefaults.standard.set(true, forKey: onboardingCompletedKey)
}
/// Reset onboarding so the welcome popout appears on next launch.
func resetOnboarding() {
UserDefaults.standard.removeObject(forKey: onboardingCompletedKey)
isFirstLaunchReady = false
}
func scheduleLaunch(after seconds: TimeInterval) {
cancelPendingLaunch()
let start = max(1, Int(ceil(seconds)))
@@ -191,6 +270,15 @@ final class ExoProcessController: ObservableObject {
var environment = ProcessInfo.processInfo.environment
environment["EXO_RUNTIME_DIR"] = runtimeURL.path
environment["EXO_LIBP2P_NAMESPACE"] = computeNamespace()
if !hfToken.isEmpty {
environment["HF_TOKEN"] = hfToken
}
if enableImageModels {
environment["EXO_ENABLE_IMAGE_MODELS"] = "true"
}
if offlineMode {
environment["EXO_OFFLINE"] = "true"
}
var paths: [String] = []
if let existing = environment["PATH"], !existing.isEmpty {
+115 -13
View File
@@ -5,17 +5,44 @@ import Foundation
struct ClusterState: Decodable {
let instances: [String: ClusterInstance]
let runners: [String: RunnerStatusSummary]
let nodeProfiles: [String: NodeProfile]
let tasks: [String: ClusterTask]
let topology: Topology?
let downloads: [String: [NodeDownloadStatus]]
let thunderboltBridgeCycles: [[String]]
// Granular node state (split from the old nodeProfiles)
let nodeIdentities: [String: NodeIdentity]
let nodeMemory: [String: MemoryInfo]
let nodeSystem: [String: SystemInfo]
let nodeThunderboltBridge: [String: ThunderboltBridgeStatus]
let nodeRdmaCtl: [String: NodeRdmaCtlStatus]
/// Computed property for backwards compatibility - merges granular state into NodeProfile
var nodeProfiles: [String: NodeProfile] {
var profiles: [String: NodeProfile] = [:]
let allNodeIds = Set(nodeIdentities.keys)
.union(nodeMemory.keys)
.union(nodeSystem.keys)
for nodeId in allNodeIds {
let identity = nodeIdentities[nodeId]
let memory = nodeMemory[nodeId]
let system = nodeSystem[nodeId]
profiles[nodeId] = NodeProfile(
modelId: identity?.modelId,
chipId: identity?.chipId,
friendlyName: identity?.friendlyName,
memory: memory,
system: system
)
}
return profiles
}
init(from decoder: Decoder) throws {
let container = try decoder.container(keyedBy: CodingKeys.self)
let rawInstances = try container.decode([String: TaggedInstance].self, forKey: .instances)
self.instances = rawInstances.mapValues(\.instance)
self.runners = try container.decode([String: RunnerStatusSummary].self, forKey: .runners)
self.nodeProfiles = try container.decode([String: NodeProfile].self, forKey: .nodeProfiles)
let rawTasks =
try container.decodeIfPresent([String: TaggedTask].self, forKey: .tasks) ?? [:]
self.tasks = rawTasks.compactMapValues(\.task)
@@ -24,15 +51,39 @@ struct ClusterState: Decodable {
try container.decodeIfPresent([String: [TaggedNodeDownload]].self, forKey: .downloads)
?? [:]
self.downloads = rawDownloads.mapValues { $0.compactMap(\.status) }
self.thunderboltBridgeCycles =
try container.decodeIfPresent([[String]].self, forKey: .thunderboltBridgeCycles) ?? []
// Granular node state
self.nodeIdentities =
try container.decodeIfPresent([String: NodeIdentity].self, forKey: .nodeIdentities)
?? [:]
self.nodeMemory =
try container.decodeIfPresent([String: MemoryInfo].self, forKey: .nodeMemory) ?? [:]
self.nodeSystem =
try container.decodeIfPresent([String: SystemInfo].self, forKey: .nodeSystem) ?? [:]
self.nodeThunderboltBridge =
try container.decodeIfPresent(
[String: ThunderboltBridgeStatus].self, forKey: .nodeThunderboltBridge
) ?? [:]
self.nodeRdmaCtl =
try container.decodeIfPresent(
[String: NodeRdmaCtlStatus].self, forKey: .nodeRdmaCtl
) ?? [:]
}
private enum CodingKeys: String, CodingKey {
case instances
case runners
case nodeProfiles
case topology
case tasks
case downloads
case thunderboltBridgeCycles
case nodeIdentities
case nodeMemory
case nodeSystem
case nodeThunderboltBridge
case nodeRdmaCtl
}
}
@@ -102,6 +153,22 @@ struct NodeProfile: Decodable {
let system: SystemInfo?
}
struct NodeIdentity: Decodable {
let modelId: String?
let chipId: String?
let friendlyName: String?
}
struct ThunderboltBridgeStatus: Decodable {
let enabled: Bool
let exists: Bool
let serviceName: String?
}
struct NodeRdmaCtlStatus: Decodable {
let enabled: Bool
}
struct MemoryInfo: Decodable {
let ramTotal: MemoryValue?
let ramAvailable: MemoryValue?
@@ -120,16 +187,51 @@ struct SystemInfo: Decodable {
}
struct Topology: Decodable {
let nodes: [TopologyNode]
let connections: [TopologyConnection]?
/// Node IDs in the topology
let nodes: [String]
/// Flattened list of connections (source -> sink pairs)
let connections: [TopologyConnection]
init(from decoder: Decoder) throws {
let container = try decoder.container(keyedBy: CodingKeys.self)
self.nodes = try container.decodeIfPresent([String].self, forKey: .nodes) ?? []
// Connections come as nested map: { source: { sink: [edges] } }
// We flatten to array of (source, sink) pairs
var flatConnections: [TopologyConnection] = []
if let nested = try container.decodeIfPresent(
[String: [String: [AnyCodable]]].self, forKey: .connections
) {
for (source, sinks) in nested {
for sink in sinks.keys {
flatConnections.append(
TopologyConnection(localNodeId: source, sendBackNodeId: sink))
}
}
}
self.connections = flatConnections
}
private enum CodingKeys: String, CodingKey {
case nodes
case connections
}
}
struct TopologyNode: Decodable {
let nodeId: String
let nodeProfile: NodeProfile
/// Placeholder for decoding arbitrary JSON values we don't need to inspect
private struct AnyCodable: Decodable {
init(from decoder: Decoder) throws {
// Just consume the value without storing it
_ = try? decoder.singleValueContainer().decode(Bool.self)
_ = try? decoder.singleValueContainer().decode(Int.self)
_ = try? decoder.singleValueContainer().decode(Double.self)
_ = try? decoder.singleValueContainer().decode(String.self)
_ = try? decoder.singleValueContainer().decode([AnyCodable].self)
_ = try? decoder.singleValueContainer().decode([String: AnyCodable].self)
}
}
struct TopologyConnection: Decodable {
struct TopologyConnection {
let localNodeId: String
let sendBackNodeId: String
}
@@ -201,7 +303,7 @@ struct ClusterTask {
let modelName: String?
let promptPreview: String?
let errorMessage: String?
let parameters: ChatCompletionTaskParameters?
let parameters: TextGenerationTaskParameters?
var sortPriority: Int {
switch status {
@@ -238,12 +340,12 @@ struct ClusterTaskPayload: Decodable {
let taskStatus: TaskStatus?
let instanceId: String?
let commandId: String?
let taskParams: ChatCompletionTaskParameters?
let taskParams: TextGenerationTaskParameters?
let errorType: String?
let errorMessage: String?
}
struct ChatCompletionTaskParameters: Decodable, Equatable {
struct TextGenerationTaskParameters: Decodable, Equatable {
let model: String?
let messages: [ChatCompletionMessage]?
let maxTokens: Int?
@@ -282,7 +384,7 @@ extension ClusterTask {
guard let id = payload.taskId else { return nil }
let status = payload.taskStatus ?? .unknown
switch kindKey {
case "ChatCompletion":
case "TextGeneration":
self.init(
id: id,
status: status,
+100 -27
View File
@@ -38,37 +38,46 @@ struct BugReportService {
func sendReport(
baseURL: URL = URL(string: "http://127.0.0.1:52415")!,
now: Date = Date(),
isManual: Bool = false
isManual: Bool = false,
userDescription: String? = nil
) async throws -> BugReportOutcome {
let timestamp = Self.runTimestampString(now)
let dayPrefix = Self.dayPrefixString(now)
let prefix = "reports/\(dayPrefix)/\(timestamp)/"
let logData = readLog()
let logFiles = readAllLogs()
let ifconfigText = try await captureIfconfig()
let hostName = Host.current().localizedName ?? "unknown"
let debugInfo = readDebugInfo()
async let stateResult = fetch(url: baseURL.appendingPathComponent("state"))
async let eventsResult = fetch(url: baseURL.appendingPathComponent("events"))
let stateData = try await fetch(url: baseURL.appendingPathComponent("state"))
let stateData = try await stateResult
let eventsData = try await eventsResult
// Extract cluster TB bridge status from exo state
let clusterTbBridgeStatus = extractClusterTbBridgeStatus(from: stateData)
let reportJSON = makeReportJson(
timestamp: timestamp,
hostName: hostName,
ifconfig: ifconfigText,
debugInfo: debugInfo,
isManual: isManual
isManual: isManual,
clusterTbBridgeStatus: clusterTbBridgeStatus,
userDescription: userDescription
)
let uploads: [(path: String, data: Data?)] = [
("\(prefix)exo.log", logData),
let eventLogFiles = readAllEventLogs()
var uploads: [(path: String, data: Data?)] = logFiles.map { (path, data) in
("\(prefix)\(path)", data)
}
uploads.append(
contentsOf: eventLogFiles.map { (path, data) in
("\(prefix)\(path)", data as Data?)
})
uploads.append(contentsOf: [
("\(prefix)state.json", stateData),
("\(prefix)events.json", eventsData),
("\(prefix)report.json", reportJSON),
]
])
let uploadItems: [(key: String, body: Data)] = uploads.compactMap { item in
guard let body = item.data else { return nil }
@@ -145,11 +154,40 @@ struct BugReportService {
return decoded.urls
}
private func readLog() -> Data? {
let logURL = URL(fileURLWithPath: NSHomeDirectory())
private func readAllLogs() -> [(path: String, data: Data)] {
let dir = URL(fileURLWithPath: NSHomeDirectory())
.appendingPathComponent(".exo")
.appendingPathComponent("exo.log")
return try? Data(contentsOf: logURL)
.appendingPathComponent("exo_log")
var results: [(path: String, data: Data)] = []
let contents = (try? FileManager.default.contentsOfDirectory(atPath: dir.path)) ?? []
for name in contents {
if let data = try? Data(contentsOf: dir.appendingPathComponent(name)) {
results.append(("exo_log/\(name)", data))
}
}
return results
}
private func readAllEventLogs() -> [(path: String, data: Data)] {
let eventLogDir = URL(fileURLWithPath: NSHomeDirectory())
.appendingPathComponent(".exo")
.appendingPathComponent("event_log")
var results: [(path: String, data: Data)] = []
for subdir in ["master", "api"] {
let dir = eventLogDir.appendingPathComponent(subdir)
let contents =
(try? FileManager.default.contentsOfDirectory(atPath: dir.path)) ?? []
for name in contents where name.hasPrefix("events.") {
if let data = try? Data(contentsOf: dir.appendingPathComponent(name)) {
results.append(("event_log/\(subdir)/\(name)", data))
}
}
}
return results
}
private func captureIfconfig() async throws -> String {
@@ -178,18 +216,19 @@ struct BugReportService {
}
private func readThunderboltBridgeDisabled() -> Bool? {
let result = runCommand([
"/usr/sbin/networksetup", "-getnetworkserviceenabled", "Thunderbolt Bridge",
])
guard result.exitCode == 0 else { return nil }
let output = result.output.lowercased()
if output.contains("enabled") {
return false
// Dynamically find the Thunderbolt Bridge service (don't assume the name)
guard let serviceName = ThunderboltBridgeDetector.findThunderboltBridgeServiceName() else {
// No bridge containing Thunderbolt interfaces exists
return nil
}
if output.contains("disabled") {
return true
guard let isEnabled = ThunderboltBridgeDetector.isServiceEnabled(serviceName: serviceName)
else {
return nil
}
return nil
// Return true if disabled, false if enabled
return !isEnabled
}
private func readInterfaces() -> [DebugInfo.InterfaceStatus] {
@@ -268,11 +307,13 @@ struct BugReportService {
hostName: String,
ifconfig: String,
debugInfo: DebugInfo,
isManual: Bool
isManual: Bool,
clusterTbBridgeStatus: [[String: Any]]?,
userDescription: String? = nil
) -> Data? {
let system = readSystemMetadata()
let exo = readExoMetadata()
let payload: [String: Any] = [
var payload: [String: Any] = [
"timestamp": timestamp,
"host": hostName,
"ifconfig": ifconfig,
@@ -282,9 +323,41 @@ struct BugReportService {
"exo_commit": exo.commit as Any,
"report_type": isManual ? "manual" : "automated",
]
if let tbStatus = clusterTbBridgeStatus {
payload["cluster_thunderbolt_bridge"] = tbStatus
}
if let desc = userDescription, !desc.isEmpty {
payload["user_description"] = desc
}
return try? JSONSerialization.data(withJSONObject: payload, options: [.prettyPrinted])
}
/// Extracts cluster-wide Thunderbolt Bridge status from exo state JSON
private func extractClusterTbBridgeStatus(from stateData: Data?) -> [[String: Any]]? {
guard let data = stateData,
let json = try? JSONSerialization.jsonObject(with: data) as? [String: Any],
let nodeThunderboltBridge = json["node_thunderbolt_bridge"] as? [String: [String: Any]]
else {
return nil
}
var result: [[String: Any]] = []
for (nodeId, status) in nodeThunderboltBridge {
var entry: [String: Any] = ["node_id": nodeId]
if let enabled = status["enabled"] as? Bool {
entry["enabled"] = enabled
}
if let exists = status["exists"] as? Bool {
entry["exists"] = exists
}
if let serviceName = status["service_name"] as? String {
entry["service_name"] = serviceName
}
result.append(entry)
}
return result.isEmpty ? nil : result
}
private func readSystemMetadata() -> [String: Any] {
let hostname = safeRunCommand(["/bin/hostname"])
let computerName = safeRunCommand(["/usr/sbin/scutil", "--get", "ComputerName"])
+38 -4
View File
@@ -41,6 +41,7 @@ final class LocalNetworkChecker: ObservableObject {
private var connection: NWConnection?
private var checkTask: Task<Void, Never>?
private var periodicTask: Task<Void, Never>?
/// Whether we've completed at least one check (stored in UserDefaults)
private var hasCompletedInitialCheck: Bool {
@@ -48,10 +49,39 @@ final class LocalNetworkChecker: ObservableObject {
set { UserDefaults.standard.set(newValue, forKey: Self.hasCompletedInitialCheckKey) }
}
/// Checks if local network access is working.
/// Checks if local network access is working (one-time check).
func check() {
performCheck()
}
/// Starts periodic checking of local network access.
/// Re-checks every `interval` seconds so the warning disappears when user grants permission.
func startPeriodicChecking(interval: TimeInterval = 10) {
stopPeriodicChecking()
// Do an immediate check first
performCheck()
// Then schedule periodic checks
periodicTask = Task { [weak self] in
while !Task.isCancelled {
try? await Task.sleep(nanoseconds: UInt64(interval * 1_000_000_000))
guard !Task.isCancelled else { break }
self?.performCheck()
}
}
}
/// Stops periodic checking.
func stopPeriodicChecking() {
periodicTask?.cancel()
periodicTask = nil
}
private func performCheck() {
checkTask?.cancel()
status = .checking
// Only show "checking" status on first check to avoid UI flicker
if status == .unknown {
status = .checking
}
// Use longer timeout on first launch to allow time for permission prompt
let isFirstCheck = !hasCompletedInitialCheck
@@ -60,12 +90,15 @@ final class LocalNetworkChecker: ObservableObject {
checkTask = Task { [weak self] in
guard let self else { return }
Self.logger.info("Checking local network connectivity (first check: \(isFirstCheck))")
Self.logger.debug("Checking local network connectivity (first check: \(isFirstCheck))")
let result = await self.checkConnectivity(timeout: timeout)
self.status = result
self.hasCompletedInitialCheck = true
Self.logger.info("Local network check complete: \(result.displayText)")
// Only log on state changes or first check to reduce noise
if isFirstCheck || result != self.status {
Self.logger.info("Local network check: \(result.displayText)")
}
}
}
@@ -141,6 +174,7 @@ final class LocalNetworkChecker: ObservableObject {
}
func stop() {
stopPeriodicChecking()
checkTask?.cancel()
checkTask = nil
connection?.cancel()
+201 -49
View File
@@ -7,14 +7,20 @@ enum NetworkSetupHelper {
private static let daemonLabel = "io.exo.networksetup"
private static let scriptDestination =
"/Library/Application Support/EXO/disable_bridge.sh"
// Legacy script path from older versions
private static let legacyScriptDestination =
"/Library/Application Support/EXO/disable_bridge_enable_dhcp.sh"
private static let plistDestination = "/Library/LaunchDaemons/io.exo.networksetup.plist"
private static let requiredStartInterval: Int = 1791
private static let requiredStartInterval: Int = 1786
private static let setupScript = """
#!/usr/bin/env bash
set -euo pipefail
# Wait for macOS to finish network setup after boot
sleep 20
PREFS="/Library/Preferences/SystemConfiguration/preferences.plist"
# Remove bridge0 interface
@@ -28,19 +34,69 @@ enum NetworkSetupHelper {
# Remove Thunderbolt Bridge from VirtualNetworkInterfaces in preferences.plist
/usr/libexec/PlistBuddy -c "Delete :VirtualNetworkInterfaces:Bridge:bridge0" "$PREFS" 2>/dev/null || true
networksetup -listlocations | grep -q exo || {
networksetup -createlocation exo
}
networksetup -switchtolocation exo
networksetup -listallhardwareports \\
| awk -F': ' '/Hardware Port: / {print $2}' \\
| while IFS=":" read -r name; do
case "$name" in
"Ethernet Adapter"*)
;;
"Thunderbolt Bridge")
;;
"Thunderbolt "*)
networksetup -listallnetworkservices \\
| grep -q "EXO $name" \\
|| networksetup -createnetworkservice "EXO $name" "$name" 2>/dev/null \\
|| continue
networksetup -setdhcp "EXO $name"
;;
*)
networksetup -listallnetworkservices \\
| grep -q "$name" \\
|| networksetup -createnetworkservice "$name" "$name" 2>/dev/null \\
|| continue
;;
esac
done
networksetup -listnetworkservices | grep -q "Thunderbolt Bridge" && {
networksetup -setnetworkserviceenabled "Thunderbolt Bridge" off
} || true
"""
static func ensureLaunchDaemonInstalled() {
/// Prompts user and installs the LaunchDaemon if not already installed.
/// Shows an alert explaining what will be installed before requesting admin privileges.
static func promptAndInstallIfNeeded() {
// Use .utility priority to match NSAppleScript's internal QoS and avoid priority inversion
Task.detached(priority: .utility) {
// If already correctly installed, skip
if daemonAlreadyInstalled() {
return
}
// Show alert on main thread
let shouldInstall = await MainActor.run {
let alert = NSAlert()
alert.messageText = "EXO Network Configuration"
alert.informativeText =
"EXO needs to install a system service to configure local networking. This will disable Thunderbolt Bridge (preventing packet storms) and install a Network Location.\n\nYou will be prompted for your password."
alert.alertStyle = .informational
alert.addButton(withTitle: "Install")
alert.addButton(withTitle: "Not Now")
return alert.runModal() == .alertFirstButtonReturn
}
guard shouldInstall else {
logger.info("User deferred network setup daemon installation")
return
}
do {
if daemonAlreadyInstalled() {
return
}
try await installLaunchDaemon()
try installLaunchDaemon()
logger.info("Network setup launch daemon installed and started")
} catch {
logger.error(
@@ -63,48 +119,9 @@ enum NetworkSetupHelper {
static func hasInstalledComponents() -> Bool {
let manager = FileManager.default
let scriptExists = manager.fileExists(atPath: scriptDestination)
let legacyScriptExists = manager.fileExists(atPath: legacyScriptDestination)
let plistExists = manager.fileExists(atPath: plistDestination)
return scriptExists || plistExists
}
private static func makeUninstallScript() -> String {
"""
set -euo pipefail
LABEL="\(daemonLabel)"
SCRIPT_DEST="\(scriptDestination)"
PLIST_DEST="\(plistDestination)"
LOG_OUT="/var/log/\(daemonLabel).log"
LOG_ERR="/var/log/\(daemonLabel).err.log"
# Unload the LaunchDaemon if running
launchctl bootout system/"$LABEL" 2>/dev/null || true
# Remove LaunchDaemon plist
rm -f "$PLIST_DEST"
# Remove the script and parent directory if empty
rm -f "$SCRIPT_DEST"
rmdir "$(dirname "$SCRIPT_DEST")" 2>/dev/null || true
# Remove log files
rm -f "$LOG_OUT" "$LOG_ERR"
# Switch back to Automatic network location
networksetup -switchtolocation Automatic 2>/dev/null || true
# Delete the exo network location if it exists
networksetup -listlocations | grep -q '^exo$' && {
networksetup -deletelocation exo 2>/dev/null || true
} || true
# Re-enable Thunderbolt Bridge if it exists
networksetup -listnetworkservices | grep -q "Thunderbolt Bridge" && {
networksetup -setnetworkserviceenabled "Thunderbolt Bridge" on 2>/dev/null || true
} || true
echo "EXO network components removed successfully"
"""
return scriptExists || legacyScriptExists || plistExists
}
private static func daemonAlreadyInstalled() -> Bool {
@@ -140,7 +157,7 @@ enum NetworkSetupHelper {
return true
}
private static func installLaunchDaemon() async throws {
private static func installLaunchDaemon() throws {
let installerScript = makeInstallerScript()
try runShellAsAdmin(installerScript)
}
@@ -151,8 +168,19 @@ enum NetworkSetupHelper {
LABEL="\(daemonLabel)"
SCRIPT_DEST="\(scriptDestination)"
LEGACY_SCRIPT_DEST="\(legacyScriptDestination)"
PLIST_DEST="\(plistDestination)"
LOG_OUT="/var/log/\(daemonLabel).log"
LOG_ERR="/var/log/\(daemonLabel).err.log"
# First, completely remove any existing installation
launchctl bootout system/"$LABEL" 2>/dev/null || true
rm -f "$PLIST_DEST"
rm -f "$SCRIPT_DEST"
rm -f "$LEGACY_SCRIPT_DEST"
rm -f "$LOG_OUT" "$LOG_ERR"
# Install fresh
mkdir -p "$(dirname "$SCRIPT_DEST")"
cat > "$SCRIPT_DEST" <<'EOF_SCRIPT'
@@ -184,13 +212,137 @@ enum NetworkSetupHelper {
</plist>
EOF_PLIST
launchctl bootout system/"$LABEL" >/dev/null 2>&1 || true
launchctl bootstrap system "$PLIST_DEST"
launchctl enable system/"$LABEL"
launchctl kickstart -k system/"$LABEL"
"""
}
private static func makeUninstallScript() -> String {
"""
set -euo pipefail
LABEL="\(daemonLabel)"
SCRIPT_DEST="\(scriptDestination)"
LEGACY_SCRIPT_DEST="\(legacyScriptDestination)"
PLIST_DEST="\(plistDestination)"
LOG_OUT="/var/log/\(daemonLabel).log"
LOG_ERR="/var/log/\(daemonLabel).err.log"
# Unload the LaunchDaemon if running
launchctl bootout system/"$LABEL" 2>/dev/null || true
# Remove LaunchDaemon plist
rm -f "$PLIST_DEST"
# Remove the script (current and legacy paths) and parent directory if empty
rm -f "$SCRIPT_DEST"
rm -f "$LEGACY_SCRIPT_DEST"
rmdir "$(dirname "$SCRIPT_DEST")" 2>/dev/null || true
# Remove log files
rm -f "$LOG_OUT" "$LOG_ERR"
# Switch back to Automatic network location
networksetup -switchtolocation Automatic >/dev/null 2>&1 || true
# Delete the exo network location if it exists
networksetup -listlocations 2>/dev/null | grep -q '^exo$' && {
networksetup -deletelocation exo >/dev/null 2>&1 || true
} || true
# Re-enable any Thunderbolt Bridge service if it exists
# We find it dynamically by looking for bridges containing Thunderbolt interfaces
find_and_enable_thunderbolt_bridge() {
# Get Thunderbolt interface devices from hardware ports
tb_devices=$(networksetup -listallhardwareports 2>/dev/null | awk '
/^Hardware Port:/ { port = tolower(substr($0, 16)) }
/^Device:/ { if (port ~ /thunderbolt/) print substr($0, 9) }
') || true
[ -z "$tb_devices" ] && return 0
# For each bridge device, check if it contains Thunderbolt interfaces
for bridge in bridge0 bridge1 bridge2; do
members=$(ifconfig "$bridge" 2>/dev/null | awk '/member:/ {print $2}') || true
[ -z "$members" ] && continue
for tb_dev in $tb_devices; do
if echo "$members" | grep -qx "$tb_dev"; then
# Find the service name for this bridge device
service_name=$(networksetup -listnetworkserviceorder 2>/dev/null | awk -v dev="$bridge" '
/^\\([0-9*]/ { gsub(/^\\([0-9*]+\\) /, ""); svc = $0 }
/Device:/ && $0 ~ dev { print svc; exit }
') || true
if [ -n "$service_name" ]; then
networksetup -setnetworkserviceenabled "$service_name" on 2>/dev/null || true
return 0
fi
fi
done
done
return 0
}
find_and_enable_thunderbolt_bridge || true
echo "EXO network components removed successfully"
"""
}
/// Direct install without GUI (requires root).
/// Returns true on success, false on failure.
static func installDirectly() -> Bool {
let script = makeInstallerScript()
return runShellDirectly(script)
}
/// Direct uninstall without GUI (requires root).
/// Returns true on success, false on failure.
static func uninstallDirectly() -> Bool {
let script = makeUninstallScript()
return runShellDirectly(script)
}
/// Run a shell script directly via Process (no AppleScript, requires root).
/// Returns true on success, false on failure.
private static func runShellDirectly(_ script: String) -> Bool {
let process = Process()
process.executableURL = URL(fileURLWithPath: "/bin/bash")
process.arguments = ["-c", script]
let outputPipe = Pipe()
let errorPipe = Pipe()
process.standardOutput = outputPipe
process.standardError = errorPipe
do {
try process.run()
process.waitUntilExit()
let outputData = outputPipe.fileHandleForReading.readDataToEndOfFile()
let errorData = errorPipe.fileHandleForReading.readDataToEndOfFile()
if let output = String(data: outputData, encoding: .utf8), !output.isEmpty {
print(output)
}
if let errorOutput = String(data: errorData, encoding: .utf8), !errorOutput.isEmpty {
fputs(errorOutput, stderr)
}
if process.terminationStatus == 0 {
logger.info("Shell script completed successfully")
return true
} else {
logger.error("Shell script failed with exit code \(process.terminationStatus)")
return false
}
} catch {
logger.error(
"Failed to run shell script: \(error.localizedDescription, privacy: .public)")
fputs("Error: \(error.localizedDescription)\n", stderr)
return false
}
}
private static func runShellAsAdmin(_ script: String) throws {
let escapedScript =
script
+16 -50
View File
@@ -35,28 +35,18 @@ struct NetworkStatus: Equatable {
let thunderboltBridgeState: ThunderboltState?
let bridgeInactive: Bool?
let interfaceStatuses: [InterfaceIpStatus]
let rdmaStatus: RDMAStatus
let localRdmaDevices: [String]
let localRdmaActivePorts: [RDMAPort]
static let empty = NetworkStatus(
thunderboltBridgeState: nil,
bridgeInactive: nil,
interfaceStatuses: [],
rdmaStatus: .empty
localRdmaDevices: [],
localRdmaActivePorts: []
)
}
struct RDMAStatus: Equatable {
let rdmaCtlEnabled: Bool?
let devices: [String]
let activePorts: [RDMAPort]
var isAvailable: Bool {
rdmaCtlEnabled == true || !devices.isEmpty
}
static let empty = RDMAStatus(rdmaCtlEnabled: nil, devices: [], activePorts: [])
}
struct RDMAPort: Equatable {
let device: String
let port: String
@@ -80,31 +70,11 @@ private struct NetworkStatusFetcher {
thunderboltBridgeState: readThunderboltBridgeState(),
bridgeInactive: readBridgeInactive(),
interfaceStatuses: readInterfaceStatuses(),
rdmaStatus: readRDMAStatus()
localRdmaDevices: readRDMADevices(),
localRdmaActivePorts: readRDMAActivePorts()
)
}
private func readRDMAStatus() -> RDMAStatus {
let rdmaCtlEnabled = readRDMACtlEnabled()
let devices = readRDMADevices()
let activePorts = readRDMAActivePorts()
return RDMAStatus(
rdmaCtlEnabled: rdmaCtlEnabled, devices: devices, activePorts: activePorts)
}
private func readRDMACtlEnabled() -> Bool? {
let result = runCommand(["rdma_ctl", "status"])
guard result.exitCode == 0 else { return nil }
let output = result.output.lowercased().trimmingCharacters(in: .whitespacesAndNewlines)
if output.contains("enabled") {
return true
}
if output.contains("disabled") {
return false
}
return nil
}
private func readRDMADevices() -> [String] {
let result = runCommand(["ibv_devices"])
guard result.exitCode == 0 else { return [] }
@@ -153,22 +123,18 @@ private struct NetworkStatusFetcher {
}
private func readThunderboltBridgeState() -> ThunderboltState? {
let result = runCommand(["networksetup", "-getnetworkserviceenabled", "Thunderbolt Bridge"])
guard result.exitCode == 0 else {
let lower = result.output.lowercased() + result.error.lowercased()
if lower.contains("not a recognized network service") {
return .deleted
}
// Dynamically find the Thunderbolt Bridge service (don't assume the name)
guard let serviceName = ThunderboltBridgeDetector.findThunderboltBridgeServiceName() else {
// No bridge containing Thunderbolt interfaces exists
return .deleted
}
guard let isEnabled = ThunderboltBridgeDetector.isServiceEnabled(serviceName: serviceName)
else {
return nil
}
let output = result.output.lowercased()
if output.contains("enabled") {
return .enabled
}
if output.contains("disabled") {
return .disabled
}
return nil
return isEnabled ? .enabled : .disabled
}
private func readBridgeInactive() -> Bool? {
@@ -0,0 +1,194 @@
import Foundation
import os.log
/// Utility for dynamically detecting Thunderbolt Bridge network services.
/// This mirrors the Python logic in info_gatherer.py - we never assume the service
/// is named "Thunderbolt Bridge", instead we find bridges containing Thunderbolt interfaces.
enum ThunderboltBridgeDetector {
private static let logger = Logger(
subsystem: "io.exo.EXO", category: "ThunderboltBridgeDetector")
struct CommandResult {
let exitCode: Int32
let output: String
let error: String
}
/// Find the network service name of a bridge containing Thunderbolt interfaces.
/// Returns nil if no such bridge exists.
static func findThunderboltBridgeServiceName() -> String? {
// 1. Get all Thunderbolt interface devices (e.g., en2, en3)
guard let thunderboltDevices = getThunderboltDevices(), !thunderboltDevices.isEmpty else {
logger.debug("No Thunderbolt devices found")
return nil
}
logger.debug("Found Thunderbolt devices: \(thunderboltDevices.joined(separator: ", "))")
// 2. Get bridge services from network service order
guard let bridgeServices = getBridgeServices(), !bridgeServices.isEmpty else {
logger.debug("No bridge services found")
return nil
}
logger.debug("Found bridge services: \(bridgeServices.keys.joined(separator: ", "))")
// 3. Find a bridge that contains Thunderbolt interfaces
for (bridgeDevice, serviceName) in bridgeServices {
let members = getBridgeMembers(bridgeDevice: bridgeDevice)
logger.debug(
"Bridge \(bridgeDevice) (\(serviceName)) has members: \(members.joined(separator: ", "))"
)
// Check if any Thunderbolt device is a member of this bridge
if !members.isDisjoint(with: thunderboltDevices) {
logger.info(
"Found Thunderbolt Bridge service: '\(serviceName)' (device: \(bridgeDevice))")
return serviceName
}
}
logger.debug("No bridge found containing Thunderbolt interfaces")
return nil
}
/// Get Thunderbolt interface device names (e.g., en2, en3) from hardware ports.
private static func getThunderboltDevices() -> Set<String>? {
let result = runCommand(["networksetup", "-listallhardwareports"])
guard result.exitCode == 0 else {
logger.warning("networksetup -listallhardwareports failed: \(result.error)")
return nil
}
var thunderboltDevices: Set<String> = []
var currentPort: String?
for line in result.output.components(separatedBy: .newlines) {
let trimmed = line.trimmingCharacters(in: .whitespaces)
if trimmed.hasPrefix("Hardware Port:") {
currentPort = String(trimmed.dropFirst("Hardware Port:".count)).trimmingCharacters(
in: .whitespaces)
} else if trimmed.hasPrefix("Device:"), let port = currentPort {
let device = String(trimmed.dropFirst("Device:".count)).trimmingCharacters(
in: .whitespaces)
if port.lowercased().contains("thunderbolt") {
thunderboltDevices.insert(device)
}
currentPort = nil
}
}
return thunderboltDevices
}
/// Get mapping of bridge device -> service name from network service order.
private static func getBridgeServices() -> [String: String]? {
let result = runCommand(["networksetup", "-listnetworkserviceorder"])
guard result.exitCode == 0 else {
logger.warning("networksetup -listnetworkserviceorder failed: \(result.error)")
return nil
}
// Parse service order to find bridge devices and their service names
// Format: "(1) Service Name\n(Hardware Port: ..., Device: bridge0)\n"
var bridgeServices: [String: String] = [:]
var currentService: String?
for line in result.output.components(separatedBy: .newlines) {
let trimmed = line.trimmingCharacters(in: .whitespaces)
// Match "(N) Service Name" or "(*) Service Name" (disabled)
// but NOT "(Hardware Port: ...)" lines
if trimmed.hasPrefix("("), trimmed.contains(")"),
!trimmed.hasPrefix("(Hardware Port:")
{
if let parenEnd = trimmed.firstIndex(of: ")") {
let afterParen = trimmed.index(after: parenEnd)
if afterParen < trimmed.endIndex {
currentService =
String(trimmed[afterParen...])
.trimmingCharacters(in: .whitespaces)
}
}
}
// Match "(Hardware Port: ..., Device: bridgeX)"
else if let service = currentService, trimmed.contains("Device: bridge") {
// Extract device name from "..., Device: bridge0)"
if let deviceRange = trimmed.range(of: "Device: ") {
let afterDevice = trimmed[deviceRange.upperBound...]
if let parenIndex = afterDevice.firstIndex(of: ")") {
let device = String(afterDevice[..<parenIndex])
bridgeServices[device] = service
}
}
}
}
return bridgeServices
}
/// Get member interfaces of a bridge device via ifconfig.
private static func getBridgeMembers(bridgeDevice: String) -> Set<String> {
let result = runCommand(["ifconfig", bridgeDevice])
guard result.exitCode == 0 else {
logger.debug("ifconfig \(bridgeDevice) failed")
return []
}
var members: Set<String> = []
for line in result.output.components(separatedBy: .newlines) {
let trimmed = line.trimmingCharacters(in: .whitespaces)
if trimmed.hasPrefix("member:") {
let parts = trimmed.split(separator: " ")
if parts.count > 1 {
members.insert(String(parts[1]))
}
}
}
return members
}
/// Check if a network service is enabled.
static func isServiceEnabled(serviceName: String) -> Bool? {
let result = runCommand(["networksetup", "-getnetworkserviceenabled", serviceName])
guard result.exitCode == 0 else {
logger.warning("Failed to check if '\(serviceName)' is enabled: \(result.error)")
return nil
}
let output = result.output.lowercased().trimmingCharacters(in: .whitespacesAndNewlines)
if output.contains("enabled") {
return true
}
if output.contains("disabled") {
return false
}
return nil
}
private static func runCommand(_ arguments: [String]) -> CommandResult {
let process = Process()
process.launchPath = "/usr/bin/env"
process.arguments = arguments
let stdout = Pipe()
let stderr = Pipe()
process.standardOutput = stdout
process.standardError = stderr
do {
try process.run()
} catch {
return CommandResult(exitCode: -1, output: "", error: error.localizedDescription)
}
process.waitUntilExit()
let outputData = stdout.fileHandleForReading.readDataToEndOfFile()
let errorData = stderr.fileHandleForReading.readDataToEndOfFile()
return CommandResult(
exitCode: process.terminationStatus,
output: String(decoding: outputData, as: UTF8.self),
error: String(decoding: errorData, as: UTF8.self)
)
}
}
@@ -0,0 +1,261 @@
import AppKit
import Combine
import Foundation
import Security
import SystemConfiguration
import os.log
@MainActor
final class ThunderboltBridgeService: ObservableObject {
private static let logger = Logger(subsystem: "io.exo.EXO", category: "ThunderboltBridge")
@Published private(set) var detectedCycle: [String]?
@Published private(set) var hasPromptedForCurrentCycle = false
@Published private(set) var lastError: String?
private weak var clusterStateService: ClusterStateService?
private var cancellables = Set<AnyCancellable>()
private var previousCycleSignature: String?
init(clusterStateService: ClusterStateService) {
self.clusterStateService = clusterStateService
setupObserver()
}
private func setupObserver() {
guard let service = clusterStateService else { return }
service.$latestSnapshot
.compactMap { $0 }
.sink { [weak self] snapshot in
self?.checkForCycles(snapshot: snapshot)
}
.store(in: &cancellables)
}
private func checkForCycles(snapshot: ClusterState) {
let cycles = snapshot.thunderboltBridgeCycles
// Only consider cycles with more than 2 nodes
guard let firstCycle = cycles.first, firstCycle.count > 2 else {
// No problematic cycles detected, reset state
if detectedCycle != nil {
detectedCycle = nil
previousCycleSignature = nil
hasPromptedForCurrentCycle = false
}
return
}
// Create a signature for this cycle to detect if it changed
let cycleSignature = firstCycle.sorted().joined(separator: ",")
// If this is a new/different cycle, reset the prompt state
if cycleSignature != previousCycleSignature {
previousCycleSignature = cycleSignature
hasPromptedForCurrentCycle = false
}
detectedCycle = firstCycle
// Only prompt once per cycle
if !hasPromptedForCurrentCycle {
showDisableBridgePrompt(nodeIds: firstCycle)
}
}
private func showDisableBridgePrompt(nodeIds: [String]) {
hasPromptedForCurrentCycle = true
// Get friendly names for the nodes if available
let nodeNames = nodeIds.map { nodeId -> String in
if let snapshot = clusterStateService?.latestSnapshot,
let profile = snapshot.nodeProfiles[nodeId],
let friendlyName = profile.friendlyName, !friendlyName.isEmpty
{
return friendlyName
}
return String(nodeId.prefix(8)) // Use first 8 chars of node ID as fallback
}
let machineNames = nodeNames.joined(separator: ", ")
let alert = NSAlert()
alert.messageText = "Thunderbolt Bridge Loop Detected"
alert.informativeText = """
A Thunderbolt Bridge loop has been detected between \(nodeNames.count) machines: \(machineNames).
This can cause network packet storms and connectivity issues. Would you like to disable Thunderbolt Bridge on this machine to break the loop?
"""
alert.alertStyle = .warning
alert.addButton(withTitle: "Disable Bridge")
alert.addButton(withTitle: "Not Now")
let response = alert.runModal()
if response == .alertFirstButtonReturn {
Task {
await disableThunderboltBridge()
}
}
}
func disableThunderboltBridge() async {
Self.logger.info("Attempting to disable Thunderbolt Bridge via SCPreferences")
lastError = nil
do {
try await disableThunderboltBridgeWithSCPreferences()
Self.logger.info("Successfully disabled Thunderbolt Bridge")
} catch {
Self.logger.error(
"Failed to disable Thunderbolt Bridge: \(error.localizedDescription, privacy: .public)"
)
lastError = error.localizedDescription
showErrorAlert(message: error.localizedDescription)
}
}
private func disableThunderboltBridgeWithSCPreferences() async throws {
// 1. Create authorization reference
var authRef: AuthorizationRef?
var status = AuthorizationCreate(nil, nil, [], &authRef)
guard status == errAuthorizationSuccess, let authRef = authRef else {
throw ThunderboltBridgeError.authorizationFailed
}
defer { AuthorizationFree(authRef, [.destroyRights]) }
// 2. Request specific network configuration rights
let rightName = "system.services.systemconfiguration.network"
status = rightName.withCString { nameCString in
var item = AuthorizationItem(
name: nameCString,
valueLength: 0,
value: nil,
flags: 0
)
return withUnsafeMutablePointer(to: &item) { itemPointer in
var rights = AuthorizationRights(count: 1, items: itemPointer)
return AuthorizationCopyRights(
authRef,
&rights,
nil,
[.extendRights, .interactionAllowed],
nil
)
}
}
guard status == errAuthorizationSuccess else {
if status == errAuthorizationCanceled {
throw ThunderboltBridgeError.authorizationCanceled
}
throw ThunderboltBridgeError.authorizationDenied
}
// 3. Create SCPreferences with authorization
guard
let prefs = SCPreferencesCreateWithAuthorization(
kCFAllocatorDefault,
"EXO" as CFString,
nil,
authRef
)
else {
throw ThunderboltBridgeError.preferencesCreationFailed
}
// 4. Lock, modify, commit
guard SCPreferencesLock(prefs, true) else {
throw ThunderboltBridgeError.lockFailed
}
defer {
SCPreferencesUnlock(prefs)
}
// 5. Find the Thunderbolt Bridge service dynamically (don't assume the name)
guard let targetServiceName = ThunderboltBridgeDetector.findThunderboltBridgeServiceName()
else {
throw ThunderboltBridgeError.serviceNotFound
}
guard let allServices = SCNetworkServiceCopyAll(prefs) as? [SCNetworkService] else {
throw ThunderboltBridgeError.servicesNotFound
}
var found = false
for service in allServices {
if let name = SCNetworkServiceGetName(service) as String?,
name == targetServiceName
{
guard SCNetworkServiceSetEnabled(service, false) else {
throw ThunderboltBridgeError.disableFailed
}
found = true
Self.logger.info(
"Found and disabled Thunderbolt Bridge service: '\(targetServiceName)'")
break
}
}
if !found {
throw ThunderboltBridgeError.serviceNotFound
}
// 6. Commit and apply
guard SCPreferencesCommitChanges(prefs) else {
throw ThunderboltBridgeError.commitFailed
}
guard SCPreferencesApplyChanges(prefs) else {
throw ThunderboltBridgeError.applyFailed
}
}
private func showErrorAlert(message: String) {
let alert = NSAlert()
alert.messageText = "Failed to Disable Thunderbolt Bridge"
alert.informativeText = message
alert.alertStyle = .critical
alert.addButton(withTitle: "OK")
alert.runModal()
}
}
enum ThunderboltBridgeError: LocalizedError {
case authorizationFailed
case authorizationCanceled
case authorizationDenied
case preferencesCreationFailed
case lockFailed
case servicesNotFound
case serviceNotFound
case disableFailed
case commitFailed
case applyFailed
var errorDescription: String? {
switch self {
case .authorizationFailed:
return "Failed to create authorization"
case .authorizationCanceled:
return "Authorization was canceled by user"
case .authorizationDenied:
return "Authorization was denied"
case .preferencesCreationFailed:
return "Failed to access network preferences"
case .lockFailed:
return "Failed to lock network preferences for modification"
case .servicesNotFound:
return "Could not retrieve network services"
case .serviceNotFound:
return "Thunderbolt Bridge service not found"
case .disableFailed:
return "Failed to disable Thunderbolt Bridge service"
case .commitFailed:
return "Failed to save network configuration changes"
case .applyFailed:
return "Failed to apply network configuration changes"
}
}
}
@@ -216,7 +216,7 @@ struct InstanceTaskViewModel: Identifiable, Equatable {
let promptPreview: String?
let errorMessage: String?
let subtitle: String?
let parameters: ChatCompletionTaskParameters?
let parameters: TextGenerationTaskParameters?
var title: String {
switch kind {
+4 -4
View File
@@ -86,7 +86,7 @@ struct TopologyViewModel {
extension ClusterState {
func topologyViewModel(localNodeId: String?) -> TopologyViewModel? {
let topologyNodeIds = Set(topology?.nodes.map(\.nodeId) ?? [])
let topologyNodeIds = Set(topology?.nodes ?? [])
let allNodes = nodeViewModels().filter {
topologyNodeIds.isEmpty || topologyNodeIds.contains($0.id)
}
@@ -95,8 +95,8 @@ extension ClusterState {
let nodesById = Dictionary(uniqueKeysWithValues: allNodes.map { ($0.id, $0) })
var orderedNodes: [NodeViewModel] = []
if let topologyNodes = topology?.nodes {
for topoNode in topologyNodes {
if let viewModel = nodesById[topoNode.nodeId] {
for nodeId in topologyNodes {
if let viewModel = nodesById[nodeId] {
orderedNodes.append(viewModel)
}
}
@@ -116,7 +116,7 @@ extension ClusterState {
let nodeIds = Set(orderedNodes.map(\.id))
let edgesArray: [TopologyEdgeViewModel] =
topology?.connections?.compactMap { connection in
topology?.connections.compactMap { connection in
guard nodeIds.contains(connection.localNodeId),
nodeIds.contains(connection.sendBackNodeId)
else { return nil }
+200
View File
@@ -0,0 +1,200 @@
import AppKit
import SwiftUI
/// A popover callout anchored to the menu bar icon on every launch,
/// pointing the user to the web dashboard with an arrow connecting to the icon.
@MainActor
final class FirstLaunchPopout {
private var popover: NSPopover?
private var countdownTask: Task<Void, Never>?
private static let dashboardURL = "http://localhost:52415/"
/// Called when the user completes onboarding (clicks Open Dashboard or dismisses).
var onComplete: (() -> Void)?
func show() {
guard popover == nil else { return }
// The status bar button may not exist yet on first launch; retry generously.
showWithRetry(attemptsRemaining: 15)
}
private func showWithRetry(attemptsRemaining: Int) {
guard attemptsRemaining > 0 else {
// Exhausted retries fall back to just opening the dashboard directly.
openDashboard()
onComplete?()
return
}
guard let button = Self.findStatusItemButton() else {
DispatchQueue.main.asyncAfter(deadline: .now() + 1.0) { [weak self] in
self?.showWithRetry(attemptsRemaining: attemptsRemaining - 1)
}
return
}
let pop = NSPopover()
pop.behavior = .applicationDefined
pop.animates = true
pop.contentSize = NSSize(width: 280, height: 120)
pop.contentViewController = NSHostingController(
rootView: WelcomeCalloutView(
countdownDuration: 10,
onDismiss: { [weak self] in
self?.onComplete?()
self?.dismiss()
},
onOpen: { [weak self] in
self?.openDashboard()
self?.onComplete?()
self?.dismiss()
}
)
)
self.popover = pop
pop.show(relativeTo: button.bounds, of: button, preferredEdge: .minY)
// Auto-open dashboard after 10s then dismiss
countdownTask = Task {
try? await Task.sleep(nanoseconds: 10_000_000_000)
if !Task.isCancelled {
openDashboard()
onComplete?()
dismiss()
}
}
}
func dismiss() {
countdownTask?.cancel()
countdownTask = nil
guard let pop = popover else { return }
popover = nil
pop.performClose(nil)
}
private func openDashboard() {
guard let url = URL(string: Self.dashboardURL) else { return }
NSWorkspace.shared.open(url)
}
/// Finds the NSStatusBarButton created by SwiftUI's MenuBarExtra.
/// Walks the view hierarchy to find the actual button rather than the content view.
private static func findStatusItemButton() -> NSView? {
for window in NSApp.windows {
let className = NSStringFromClass(type(of: window))
// Match NSStatusBarWindow or any internal SwiftUI status bar window
guard className.contains("StatusBar") || className.contains("MenuBarExtra") else {
continue
}
if let content = window.contentView {
if let button = findButton(in: content) {
return button
}
// Fall back to the content view itself if it has a non-zero frame
if content.frame.width > 0 {
return content
}
}
}
return nil
}
/// Recursively searches the view hierarchy for an NSStatusBarButton.
private static func findButton(in view: NSView) -> NSView? {
let className = NSStringFromClass(type(of: view))
if className.contains("StatusBarButton") || className.contains("StatusItem") {
return view
}
for subview in view.subviews {
if let found = findButton(in: subview) {
return found
}
}
return nil
}
}
/// Minimal welcome callout friendly pointer, not a wall of text.
/// Rendered inside the NSPopover which provides its own chrome and arrow.
private struct WelcomeCalloutView: View {
let countdownDuration: Int
let onDismiss: () -> Void
let onOpen: () -> Void
@State private var countdown: Int
@State private var timerTask: Task<Void, Never>?
init(countdownDuration: Int, onDismiss: @escaping () -> Void, onOpen: @escaping () -> Void) {
self.countdownDuration = countdownDuration
self.onDismiss = onDismiss
self.onOpen = onOpen
self._countdown = State(initialValue: countdownDuration)
}
var body: some View {
VStack(alignment: .leading, spacing: 10) {
HStack(alignment: .top) {
Text("EXO is running")
.font(.system(.headline, design: .rounded))
.fontWeight(.semibold)
.foregroundColor(.primary)
Spacer()
Button {
onDismiss()
} label: {
Image(systemName: "xmark.circle.fill")
.font(.system(size: 14))
.foregroundStyle(.tertiary)
}
.buttonStyle(.plain)
}
Text("Run your first model here:")
.font(.system(.subheadline, design: .default))
.foregroundColor(.secondary)
HStack {
Button {
onOpen()
} label: {
Label("Open Dashboard", systemImage: "arrow.up.right.square")
.font(.system(.caption, design: .default))
.fontWeight(.medium)
}
.buttonStyle(.borderedProminent)
.tint(.accentColor)
.controlSize(.small)
Spacer()
if countdown > 0 {
Text("Launching in \(countdown) secs...")
.font(.system(.caption2, design: .default))
.foregroundColor(.secondary.opacity(0.6))
.monospacedDigit()
}
}
}
.padding(14)
.onAppear {
startCountdown()
}
.onDisappear {
timerTask?.cancel()
timerTask = nil
}
}
private func startCountdown() {
timerTask = Task {
while countdown > 0 {
try? await Task.sleep(nanoseconds: 1_000_000_000)
if !Task.isCancelled {
countdown -= 1
}
}
}
}
}
+489
View File
@@ -0,0 +1,489 @@
import AppKit
import SwiftUI
/// Native macOS Settings window following Apple HIG.
/// Organized into General, Model, Advanced, and About sections.
struct SettingsView: View {
@EnvironmentObject private var controller: ExoProcessController
@EnvironmentObject private var updater: SparkleUpdater
@EnvironmentObject private var networkStatusService: NetworkStatusService
@EnvironmentObject private var thunderboltBridgeService: ThunderboltBridgeService
@EnvironmentObject private var stateService: ClusterStateService
@State private var pendingNamespace: String = ""
@State private var pendingHFToken: String = ""
@State private var pendingEnableImageModels = false
@State private var pendingOfflineMode = false
@State private var needsRestart = false
@State private var bugReportInFlight = false
@State private var bugReportMessage: String?
@State private var uninstallInProgress = false
var body: some View {
TabView {
generalTab
.tabItem {
Label("General", systemImage: "gear")
}
modelTab
.tabItem {
Label("Model", systemImage: "cube")
}
advancedTab
.tabItem {
Label("Advanced", systemImage: "wrench.and.screwdriver")
}
aboutTab
.tabItem {
Label("About", systemImage: "info.circle")
}
}
.frame(width: 450, height: 400)
.onAppear {
pendingNamespace = controller.customNamespace
pendingHFToken = controller.hfToken
pendingEnableImageModels = controller.enableImageModels
pendingOfflineMode = controller.offlineMode
needsRestart = false
}
}
// MARK: - General Tab
private var generalTab: some View {
Form {
Section {
LabeledContent("Cluster Namespace") {
TextField("default", text: $pendingNamespace)
.textFieldStyle(.roundedBorder)
.frame(width: 200)
}
Text("Nodes with the same namespace form a cluster. Leave empty for default.")
.font(.caption)
.foregroundColor(.secondary)
}
Section {
LabeledContent("HuggingFace Token") {
SecureField("optional", text: $pendingHFToken)
.textFieldStyle(.roundedBorder)
.frame(width: 200)
}
Text("Required for gated models. Get yours at huggingface.co/settings/tokens")
.font(.caption)
.foregroundColor(.secondary)
}
Section {
Toggle("Offline Mode", isOn: $pendingOfflineMode)
Text("Skip internet checks and use only locally available models.")
.font(.caption)
.foregroundColor(.secondary)
}
Section {
HStack {
Spacer()
Button("Save & Restart") {
applyGeneralSettings()
}
.disabled(!hasGeneralChanges)
}
}
}
.formStyle(.grouped)
.padding()
}
// MARK: - Model Tab
private var modelTab: some View {
Form {
Section {
Toggle("Enable Image Models (experimental)", isOn: $pendingEnableImageModels)
Text("Allow text-to-image and image-to-image models in the model picker.")
.font(.caption)
.foregroundColor(.secondary)
}
Section {
HStack {
Spacer()
Button("Save & Restart") {
applyModelSettings()
}
.disabled(!hasModelChanges)
}
}
}
.formStyle(.grouped)
.padding()
}
// MARK: - Advanced Tab
private var advancedTab: some View {
Form {
Section("Onboarding") {
HStack {
VStack(alignment: .leading) {
Text("Reset Onboarding")
Text("Opens the dashboard and resets the onboarding wizard.")
.font(.caption)
.foregroundColor(.secondary)
}
Spacer()
Button("Reset") {
guard let url = URL(string: "http://localhost:52415/?reset-onboarding")
else { return }
NSWorkspace.shared.open(url)
}
}
}
Section("Debug Info") {
LabeledContent("Thunderbolt Bridge") {
Text(thunderboltStatusText)
.foregroundColor(thunderboltStatusColor)
}
VStack(alignment: .leading, spacing: 2) {
clusterThunderboltBridgeView
}
VStack(alignment: .leading, spacing: 2) {
interfaceIpList
}
VStack(alignment: .leading, spacing: 2) {
rdmaStatusView
}
sendBugReportButton
}
Section("Danger Zone") {
Button(role: .destructive) {
showUninstallConfirmationAlert()
} label: {
HStack {
Text("Uninstall EXO")
Spacer()
Image(systemName: "trash")
.imageScale(.small)
}
}
.disabled(uninstallInProgress)
}
}
.formStyle(.grouped)
.padding()
}
// MARK: - About Tab
private var aboutTab: some View {
Form {
Section {
LabeledContent("Version") {
Text(buildTag)
.textSelection(.enabled)
}
LabeledContent("Commit") {
Text(buildCommit)
.font(.system(.body, design: .monospaced))
.textSelection(.enabled)
}
}
Section {
Button("Check for Updates") {
updater.checkForUpdates()
}
}
}
.formStyle(.grouped)
.padding()
}
// MARK: - Debug Info Views (moved from ContentView)
private var thunderboltStatusText: String {
switch networkStatusService.status.thunderboltBridgeState {
case .some(.disabled):
return "Disabled"
case .some(.deleted):
return "Deleted"
case .some(.enabled):
return "Enabled"
case nil:
return "Unknown"
}
}
private var thunderboltStatusColor: Color {
switch networkStatusService.status.thunderboltBridgeState {
case .some(.disabled), .some(.deleted):
return .green
case .some(.enabled):
return .red
case nil:
return .secondary
}
}
private var clusterThunderboltBridgeView: some View {
let bridgeStatuses = stateService.latestSnapshot?.nodeThunderboltBridge ?? [:]
let localNodeId = stateService.localNodeId
let nodeProfiles = stateService.latestSnapshot?.nodeProfiles ?? [:]
return VStack(alignment: .leading, spacing: 1) {
if bridgeStatuses.isEmpty {
Text("Cluster TB Bridge: No data")
.font(.caption2)
.foregroundColor(.secondary)
} else {
Text("Cluster TB Bridge Status:")
.font(.caption2)
.foregroundColor(.secondary)
ForEach(Array(bridgeStatuses.keys.sorted()), id: \.self) { nodeId in
if let status = bridgeStatuses[nodeId] {
let nodeName =
nodeProfiles[nodeId]?.friendlyName ?? String(nodeId.prefix(8))
let isLocal = nodeId == localNodeId
let prefix = isLocal ? " \(nodeName) (local):" : " \(nodeName):"
let statusText =
!status.exists
? "N/A"
: (status.enabled ? "Enabled" : "Disabled")
let color: Color =
!status.exists
? .secondary
: (status.enabled ? .red : .green)
Text("\(prefix) \(statusText)")
.font(.caption2)
.foregroundColor(color)
}
}
}
}
}
private var interfaceIpList: some View {
let statuses = networkStatusService.status.interfaceStatuses
return VStack(alignment: .leading, spacing: 1) {
Text("Interfaces (en0en7):")
.font(.caption2)
.foregroundColor(.secondary)
if statuses.isEmpty {
Text(" Unknown")
.font(.caption2)
.foregroundColor(.secondary)
} else {
ForEach(statuses, id: \.interfaceName) { status in
let ipText = status.ipAddress ?? "No IP"
Text(" \(status.interfaceName): \(ipText)")
.font(.caption2)
.foregroundColor(status.ipAddress == nil ? .red : .green)
}
}
}
}
private var rdmaStatusView: some View {
let rdmaStatuses = stateService.latestSnapshot?.nodeRdmaCtl ?? [:]
let localNodeId = stateService.localNodeId
let nodeProfiles = stateService.latestSnapshot?.nodeProfiles ?? [:]
let localDevices = networkStatusService.status.localRdmaDevices
let localPorts = networkStatusService.status.localRdmaActivePorts
return VStack(alignment: .leading, spacing: 1) {
if rdmaStatuses.isEmpty {
Text("Cluster RDMA: No data")
.font(.caption2)
.foregroundColor(.secondary)
} else {
Text("Cluster RDMA Status:")
.font(.caption2)
.foregroundColor(.secondary)
ForEach(Array(rdmaStatuses.keys.sorted()), id: \.self) { nodeId in
if let status = rdmaStatuses[nodeId] {
let nodeName =
nodeProfiles[nodeId]?.friendlyName ?? String(nodeId.prefix(8))
let isLocal = nodeId == localNodeId
let prefix = isLocal ? " \(nodeName) (local):" : " \(nodeName):"
let statusText = status.enabled ? "Enabled" : "Disabled"
let color: Color = status.enabled ? .green : .orange
Text("\(prefix) \(statusText)")
.font(.caption2)
.foregroundColor(color)
}
}
}
if !localDevices.isEmpty {
Text(" Local Devices: \(localDevices.joined(separator: ", "))")
.font(.caption2)
.foregroundColor(.secondary)
}
if !localPorts.isEmpty {
Text(" Local Active Ports:")
.font(.caption2)
.foregroundColor(.secondary)
ForEach(localPorts, id: \.device) { port in
Text(" \(port.device) port \(port.port): \(port.state)")
.font(.caption2)
.foregroundColor(.green)
}
}
}
}
private var sendBugReportButton: some View {
VStack(alignment: .leading, spacing: 4) {
Button {
Task {
await sendBugReport()
}
} label: {
HStack {
if bugReportInFlight {
ProgressView()
.scaleEffect(0.6)
}
Text("Send Bug Report")
.font(.caption)
.fontWeight(.semibold)
Spacer()
}
}
.disabled(bugReportInFlight)
if let message = bugReportMessage {
Text(message)
.font(.caption2)
.foregroundColor(.secondary)
.fixedSize(horizontal: false, vertical: true)
}
}
}
// MARK: - Actions
private func sendBugReport() async {
bugReportInFlight = true
bugReportMessage = "Collecting logs..."
let service = BugReportService()
do {
let outcome = try await service.sendReport(isManual: true)
bugReportMessage = outcome.message
} catch {
bugReportMessage = error.localizedDescription
}
bugReportInFlight = false
}
private func showUninstallConfirmationAlert() {
let alert = NSAlert()
alert.messageText = "Uninstall EXO"
alert.informativeText = """
This will remove EXO and all its system components:
Network configuration daemon
Launch at login registration
EXO network location
The app will be moved to Trash.
"""
alert.alertStyle = .warning
alert.addButton(withTitle: "Uninstall")
alert.addButton(withTitle: "Cancel")
if let uninstallButton = alert.buttons.first {
uninstallButton.hasDestructiveAction = true
}
let response = alert.runModal()
if response == .alertFirstButtonReturn {
performUninstall()
}
}
private func performUninstall() {
uninstallInProgress = true
controller.cancelPendingLaunch()
controller.stop()
stateService.stopPolling()
DispatchQueue.global(qos: .utility).async {
do {
try NetworkSetupHelper.uninstall()
DispatchQueue.main.async {
LaunchAtLoginHelper.disable()
self.moveAppToTrash()
DispatchQueue.main.asyncAfter(deadline: .now() + 0.5) {
NSApplication.shared.terminate(nil)
}
}
} catch {
DispatchQueue.main.async {
let errorAlert = NSAlert()
errorAlert.messageText = "Uninstall Failed"
errorAlert.informativeText = error.localizedDescription
errorAlert.alertStyle = .critical
errorAlert.addButton(withTitle: "OK")
errorAlert.runModal()
self.uninstallInProgress = false
}
}
}
}
private func moveAppToTrash() {
guard let appURL = Bundle.main.bundleURL as URL? else { return }
do {
try FileManager.default.trashItem(at: appURL, resultingItemURL: nil)
} catch {
// If we can't trash the app, that's OK - user can do it manually
}
}
// MARK: - Helpers
private var hasGeneralChanges: Bool {
pendingNamespace != controller.customNamespace || pendingHFToken != controller.hfToken
|| pendingOfflineMode != controller.offlineMode
}
private var hasModelChanges: Bool {
pendingEnableImageModels != controller.enableImageModels
}
private func applyGeneralSettings() {
controller.customNamespace = pendingNamespace
controller.hfToken = pendingHFToken
controller.offlineMode = pendingOfflineMode
restartIfRunning()
}
private func applyModelSettings() {
controller.enableImageModels = pendingEnableImageModels
restartIfRunning()
}
private func restartIfRunning() {
if controller.status == .running || controller.status == .starting {
controller.restart()
}
}
private var buildTag: String {
Bundle.main.infoDictionary?["EXOBuildTag"] as? String ?? "unknown"
}
private var buildCommit: String {
Bundle.main.infoDictionary?["EXOBuildCommit"] as? String ?? "unknown"
}
}
@@ -0,0 +1,47 @@
import AppKit
import SwiftUI
/// Manages a standalone native macOS Settings window.
/// Ensures only one instance exists and brings it to front on repeated opens.
@MainActor
final class SettingsWindowController: ObservableObject {
private var window: NSWindow?
func open(
controller: ExoProcessController,
updater: SparkleUpdater,
networkStatusService: NetworkStatusService,
thunderboltBridgeService: ThunderboltBridgeService,
stateService: ClusterStateService
) {
if let existing = window, existing.isVisible {
existing.makeKeyAndOrderFront(nil)
NSApp.activate(ignoringOtherApps: true)
return
}
let settingsView = SettingsView()
.environmentObject(controller)
.environmentObject(updater)
.environmentObject(networkStatusService)
.environmentObject(thunderboltBridgeService)
.environmentObject(stateService)
let hostingView = NSHostingView(rootView: settingsView)
let newWindow = NSWindow(
contentRect: NSRect(x: 0, y: 0, width: 450, height: 400),
styleMask: [.titled, .closable],
backing: .buffered,
defer: false
)
newWindow.title = "EXO Settings"
newWindow.contentView = hostingView
newWindow.center()
newWindow.isReleasedWhenClosed = false
newWindow.makeKeyAndOrderFront(nil)
NSApp.activate(ignoringOtherApps: true)
window = newWindow
}
}
+85
View File
@@ -0,0 +1,85 @@
//
// main.swift
// EXO
//
// Created by Jake Hillion on 2026-02-03.
//
import Foundation
/// Command line options for the EXO app
enum CLICommand {
case install
case uninstall
case help
case none
}
/// Parse command line arguments to determine the CLI command
func parseArguments() -> CLICommand {
let args = CommandLine.arguments
if args.contains("--help") || args.contains("-h") {
return .help
}
if args.contains("--install") {
return .install
}
if args.contains("--uninstall") {
return .uninstall
}
return .none
}
/// Print usage information
func printUsage() {
let programName = (CommandLine.arguments.first as NSString?)?.lastPathComponent ?? "EXO"
print(
"""
Usage: \(programName) [OPTIONS]
Options:
--install Install EXO network configuration (requires root)
--uninstall Uninstall EXO network configuration (requires root)
--help, -h Show this help message
When run without options, starts the normal GUI application.
Examples:
sudo \(programName) --install Install network components as root
sudo \(programName) --uninstall Remove network components as root
""")
}
/// Check if running as root
func isRunningAsRoot() -> Bool {
return getuid() == 0
}
// Main entry point
let command = parseArguments()
switch command {
case .help:
printUsage()
exit(0)
case .install:
if !isRunningAsRoot() {
fputs("Error: --install requires root privileges. Run with sudo.\n", stderr)
exit(1)
}
let success = NetworkSetupHelper.installDirectly()
exit(success ? 0 : 1)
case .uninstall:
if !isRunningAsRoot() {
fputs("Error: --uninstall requires root privileges. Run with sudo.\n", stderr)
exit(1)
}
let success = NetworkSetupHelper.uninstallDirectly()
exit(success ? 0 : 1)
case .none:
// Start normal GUI application
EXOApp.main()
}
+38 -39
View File
@@ -29,21 +29,21 @@ YELLOW='\033[1;33m'
NC='\033[0m' # No Color
echo_info() {
echo -e "${GREEN}[INFO]${NC} $1"
echo -e "${GREEN}[INFO]${NC} $1"
}
echo_warn() {
echo -e "${YELLOW}[WARN]${NC} $1"
echo -e "${YELLOW}[WARN]${NC} $1"
}
echo_error() {
echo -e "${RED}[ERROR]${NC} $1"
echo -e "${RED}[ERROR]${NC} $1"
}
# Check if running as root
if [[ $EUID -ne 0 ]]; then
echo_error "This script must be run as root (use sudo)"
exit 1
echo_error "This script must be run as root (use sudo)"
exit 1
fi
echo ""
@@ -55,64 +55,64 @@ echo ""
# Unload the LaunchDaemon if running
echo_info "Stopping network setup daemon..."
if launchctl list | grep -q "$LABEL"; then
launchctl bootout system/"$LABEL" 2>/dev/null || true
echo_info "Daemon stopped"
launchctl bootout system/"$LABEL" 2>/dev/null || true
echo_info "Daemon stopped"
else
echo_warn "Daemon was not running"
echo_warn "Daemon was not running"
fi
# Remove LaunchDaemon plist
if [[ -f "$PLIST_DEST" ]]; then
rm -f "$PLIST_DEST"
echo_info "Removed LaunchDaemon plist"
if [[ -f $PLIST_DEST ]]; then
rm -f "$PLIST_DEST"
echo_info "Removed LaunchDaemon plist"
else
echo_warn "LaunchDaemon plist not found (already removed?)"
echo_warn "LaunchDaemon plist not found (already removed?)"
fi
# Remove the script and parent directory
if [[ -f "$SCRIPT_DEST" ]]; then
rm -f "$SCRIPT_DEST"
echo_info "Removed network setup script"
if [[ -f $SCRIPT_DEST ]]; then
rm -f "$SCRIPT_DEST"
echo_info "Removed network setup script"
else
echo_warn "Network setup script not found (already removed?)"
echo_warn "Network setup script not found (already removed?)"
fi
# Remove EXO directory if empty
if [[ -d "/Library/Application Support/EXO" ]]; then
rmdir "/Library/Application Support/EXO" 2>/dev/null && \
echo_info "Removed EXO support directory" || \
echo_warn "EXO support directory not empty, leaving in place"
rmdir "/Library/Application Support/EXO" 2>/dev/null &&
echo_info "Removed EXO support directory" ||
echo_warn "EXO support directory not empty, leaving in place"
fi
# Remove log files
if [[ -f "$LOG_OUT" ]] || [[ -f "$LOG_ERR" ]]; then
rm -f "$LOG_OUT" "$LOG_ERR"
echo_info "Removed log files"
if [[ -f $LOG_OUT ]] || [[ -f $LOG_ERR ]]; then
rm -f "$LOG_OUT" "$LOG_ERR"
echo_info "Removed log files"
else
echo_warn "Log files not found (already removed?)"
echo_warn "Log files not found (already removed?)"
fi
# Switch back to Automatic network location
echo_info "Restoring network configuration..."
if networksetup -listlocations | grep -q "^Automatic$"; then
networksetup -switchtolocation Automatic 2>/dev/null || true
echo_info "Switched to Automatic network location"
networksetup -switchtolocation Automatic 2>/dev/null || true
echo_info "Switched to Automatic network location"
else
echo_warn "Automatic network location not found"
echo_warn "Automatic network location not found"
fi
# Delete the exo network location if it exists
if networksetup -listlocations | grep -q "^exo$"; then
networksetup -deletelocation exo 2>/dev/null || true
echo_info "Deleted 'exo' network location"
networksetup -deletelocation exo 2>/dev/null || true
echo_info "Deleted 'exo' network location"
else
echo_warn "'exo' network location not found (already removed?)"
echo_warn "'exo' network location not found (already removed?)"
fi
# Re-enable Thunderbolt Bridge if it exists
if networksetup -listnetworkservices 2>/dev/null | grep -q "Thunderbolt Bridge"; then
networksetup -setnetworkserviceenabled "Thunderbolt Bridge" on 2>/dev/null || true
echo_info "Re-enabled Thunderbolt Bridge"
networksetup -setnetworkserviceenabled "Thunderbolt Bridge" on 2>/dev/null || true
echo_info "Re-enabled Thunderbolt Bridge"
fi
# Note about launch at login registration
@@ -124,14 +124,14 @@ echo_warn " System Settings → General → Login Items → Remove EXO"
# Check if EXO.app exists in common locations
APP_FOUND=false
for app_path in "/Applications/EXO.app" "$HOME/Applications/EXO.app"; do
if [[ -d "$app_path" ]]; then
if [[ "$APP_FOUND" == false ]]; then
echo ""
APP_FOUND=true
fi
echo_warn "EXO.app found at: $app_path"
echo_warn "You may want to move it to Trash manually."
if [[ -d $app_path ]]; then
if [[ $APP_FOUND == false ]]; then
echo ""
APP_FOUND=true
fi
echo_warn "EXO.app found at: $app_path"
echo_warn "You may want to move it to Trash manually."
fi
done
echo ""
@@ -151,4 +151,3 @@ echo ""
echo "Manual step required:"
echo " Remove EXO from Login Items in System Settings → General → Login Items"
echo ""
+7
View File
@@ -0,0 +1,7 @@
# Canary benchmark manifest
#
# Lists the suite files to include. Each file defines benchmarks
# with shared constraints, topology, and default args.
include = [
"single-m3-ultra.toml",
]
File diff suppressed because it is too large Load Diff
+213 -313
View File
@@ -1,168 +1,125 @@
# type: ignore
#!/usr/bin/env python3
# pyright: reportAny=false, reportUnknownMemberType=false, reportUnknownVariableType=false, reportUnknownArgumentType=false
"""Tool-calling eval for exo's OpenAI-compatible API.
Tests whether models correctly:
- Trigger tool calls when appropriate
- Return valid JSON arguments matching function schemas
- Handle multi-turn tool use (call -> result -> final answer)
- Avoid calling tools when unnecessary
Start exo with a model first, then run:
uv run python tool_call_eval.py --model <model-id>
uv run python tool_call_eval.py --model <model-id> --host 10.0.0.5 --port 52415
uv run python tool_call_eval.py --model <model-id> --repeat 3
uv run python tool_call_eval.py --model <model-id> --scenarios weather_simple calculator_multi_turn
"""
from __future__ import annotations
import argparse
import contextlib
import http.client
import itertools
import json
import os
import sys
import time
from collections.abc import Callable
from pathlib import Path
from statistics import mean
from typing import Any
from urllib.parse import urlencode
from harness import (
ExoClient,
ExoHttpError,
add_common_instance_args,
instance_id_from_instance,
nodes_used_in_instance,
resolve_model_short_id,
run_planning_phase,
settle_and_fetch_placements,
wait_for_instance_gone,
wait_for_instance_ready,
)
from loguru import logger
from transformers import AutoTokenizer
# Monkey-patch for transformers 5.x compatibility
# Kimi's tokenization_kimi.py imports bytes_to_unicode from the old location
# which was moved in transformers 5.0.0rc2
try:
import transformers.models.gpt2.tokenization_gpt2 as gpt2_tokenization
from transformers.convert_slow_tokenizer import bytes_to_unicode
class ExoHttpError(RuntimeError):
def __init__(self, status: int, reason: str, body_preview: str):
super().__init__(f"HTTP {status} {reason}: {body_preview}")
self.status = status
if not hasattr(gpt2_tokenization, "bytes_to_unicode"):
gpt2_tokenization.bytes_to_unicode = bytes_to_unicode # type: ignore[attr-defined]
except ImportError:
pass # transformers < 5.0 or bytes_to_unicode not available
class ExoClient:
def __init__(self, host: str, port: int, timeout_s: float = 600.0):
self.host = host
self.port = port
self.timeout_s = timeout_s
def load_tokenizer_for_bench(model_id: str) -> Any:
"""
Load tokenizer for benchmarking, with special handling for Kimi models.
def request_json(
self,
method: str,
path: str,
params: dict[str, Any] | None = None,
body: dict[str, Any] | None = None,
headers: dict[str, str] | None = None,
) -> Any:
if not path.startswith("/"):
path = "/" + path
if params:
path = path + "?" + urlencode(params)
Kimi uses a custom TikTokenTokenizer that transformers 5.x can't load via AutoTokenizer.
This function replicates the logic from utils_mlx.py for bench compatibility.
"""
model_id_lower = model_id.lower()
conn = http.client.HTTPConnection(self.host, self.port, timeout=self.timeout_s)
try:
payload: bytes | None = None
hdrs: dict[str, str] = {"Accept": "application/json"}
if "kimi-k2" in model_id_lower:
import importlib.util
import types
if body is not None:
payload = json.dumps(body).encode("utf-8")
hdrs["Content-Type"] = "application/json"
if headers:
hdrs.update(headers)
from huggingface_hub import snapshot_download
conn.request(method.upper(), path, body=payload, headers=hdrs)
resp = conn.getresponse()
raw = resp.read()
text = raw.decode("utf-8", errors="replace") if raw else ""
# Download/get the model path
model_path = Path(
snapshot_download(
model_id,
allow_patterns=["*.json", "*.py", "*.tiktoken", "*.model"],
)
)
if resp.status >= 400:
raise ExoHttpError(resp.status, resp.reason, text[:300])
sys.path.insert(0, str(model_path))
if not text:
return None
return json.loads(text)
finally:
conn.close()
# Load tool_declaration_ts first (tokenization_kimi imports it with relative import)
tool_decl_path = model_path / "tool_declaration_ts.py"
if tool_decl_path.exists():
spec = importlib.util.spec_from_file_location(
"tool_declaration_ts", tool_decl_path
)
if spec and spec.loader:
tool_decl_module = importlib.util.module_from_spec(spec)
sys.modules["tool_declaration_ts"] = tool_decl_module
spec.loader.exec_module(tool_decl_module)
def post_bench_chat_completions(self, payload: dict[str, Any]) -> dict[str, Any]:
return self.request_json("POST", "/bench/chat/completions", body=payload)
# Load tokenization_kimi with patched source (convert relative to absolute import)
tok_path = model_path / "tokenization_kimi.py"
source = tok_path.read_text()
source = source.replace("from .tool_declaration_ts", "from tool_declaration_ts")
spec = importlib.util.spec_from_file_location("tokenization_kimi", tok_path)
if spec:
tok_module = types.ModuleType("tokenization_kimi")
tok_module.__file__ = str(tok_path)
sys.modules["tokenization_kimi"] = tok_module
exec(compile(source, tok_path, "exec"), tok_module.__dict__) # noqa: S102
TikTokenTokenizer = tok_module.TikTokenTokenizer # noqa: N806
else:
from tokenization_kimi import TikTokenTokenizer # type: ignore[import-not-found] # noqa: I001
hf_tokenizer: Any = TikTokenTokenizer.from_pretrained(model_path)
def unwrap_instance(instance: dict[str, Any]) -> dict[str, Any]:
if len(instance) != 1:
raise KeyError(f"Expected 1 key, got keys={list(instance.keys())}")
# Patch encode to use internal tiktoken model directly
# transformers 5.x has a bug in the encode->pad path for slow tokenizers
def _patched_encode(text: str, **kwargs: object) -> list[int]:
# Pass allowed_special="all" to handle special tokens like <|im_user|>
return list(hf_tokenizer.model.encode(text, allowed_special="all"))
tag = next(iter(instance))
inner = instance[tag]
if not isinstance(inner, dict):
raise TypeError(f"payload for {tag} must be dict, got {type(inner)}")
return inner
hf_tokenizer.encode = _patched_encode
return hf_tokenizer
def instance_id_from_instance(instance: dict[str, Any]) -> str:
inner = unwrap_instance(instance)
return str(inner["instanceId"])
def nodes_used_in_instance(instance: dict[str, Any]) -> int:
inner = unwrap_instance(instance)
return len(inner["shardAssignments"]["nodeToRunner"])
def runner_ids_from_instance(instance: dict[str, Any]) -> list[str]:
inner = unwrap_instance(instance)
runner_to_shard = inner["shardAssignments"]["runnerToShard"]
return list(runner_to_shard.keys())
def runner_ready(runner: dict[str, Any]) -> bool:
return "RunnerReady" in runner
def runner_failed(runner: dict[str, Any]) -> bool:
return "RunnerFailed" in runner
def get_runner_failed_message(runner: dict[str, Any]) -> str | None:
if "RunnerFailed" in runner:
return runner["RunnerFailed"].get("errorMessage")
return None
def wait_for_instance_ready(
client: ExoClient, instance_id: str, timeout: float = 24000.0
) -> None:
start_time = time.time()
instance_existed = False
while time.time() - start_time < timeout:
state = client.request_json("GET", "/state")
instances = state.get("instances", {})
if instance_id not in instances:
if instance_existed:
# Instance was deleted after being created - likely due to runner failure
raise RuntimeError(
f"Instance {instance_id} was deleted (runner may have failed)"
)
time.sleep(0.1)
continue
instance_existed = True
instance = instances[instance_id]
runner_ids = runner_ids_from_instance(instance)
runners = state.get("runners", {})
# Check for failed runners first
for rid in runner_ids:
runner = runners.get(rid, {})
if runner_failed(runner):
error_msg = get_runner_failed_message(runner) or "Unknown error"
raise RuntimeError(f"Runner {rid} failed: {error_msg}")
if all(runner_ready(runners.get(rid, {})) for rid in runner_ids):
return
time.sleep(0.1)
raise TimeoutError(f"Instance {instance_id} did not become ready within {timeout=}")
def wait_for_instance_gone(
client: ExoClient, instance_id: str, timeout: float = 3.0
) -> None:
start_time = time.time()
while time.time() - start_time < timeout:
try:
client.request_json("GET", f"/instance/{instance_id}")
time.sleep(0.4)
except ExoHttpError as e:
if e.status == 404:
return
raise TimeoutError(f"Instance {instance_id} did not get deleted within {timeout=}")
# Default: use AutoTokenizer
return AutoTokenizer.from_pretrained(model_id, trust_remote_code=True)
def format_peak_memory(b: float) -> str:
@@ -180,47 +137,7 @@ def parse_int_list(values: list[str]) -> list[int]:
part = part.strip()
if part:
items.append(int(part))
seen: set[int] = set()
out: list[int] = []
for x in items:
if x not in seen:
out.append(x)
seen.add(x)
return out
def resolve_model_short_id(client: ExoClient, model_arg: str) -> tuple[str, str]:
models = client.request_json("GET", "/models") or {}
data = models.get("data") or []
for m in data:
if m.get("name").lower() == model_arg.lower():
short_id = str(m["name"])
full_id = str(m.get("hugging_face_id") or m["name"])
return short_id, full_id
for m in data:
if m.get("hugging_face_id") == model_arg:
short_id = str(m["name"])
full_id = str(m["hugging_face_id"])
return short_id, full_id
raise ValueError(f"Model not found in /models: {model_arg}")
def placement_filter(instance_meta: str, wanted: str) -> bool:
s = (instance_meta or "").lower()
if wanted == "both":
return ("ring" in s) or ("jaccl" in s)
return wanted in s
def sharding_filter(sharding: str, wanted: str) -> bool:
s = (sharding or "").lower()
if wanted == "both":
return ("pipeline" in s) or ("tensor" in s)
return wanted in s
return items
def run_one_completion(
@@ -240,7 +157,11 @@ def run_one_completion(
stats = out.get("generation_stats")
preview = (out.get("choices") or [{}])[0]["message"]["content"][:200]
# Extract preview, handling None content (common for thinking models)
choices = out.get("choices") or [{}]
message = choices[0].get("message", {}) if choices else {}
content = message.get("content") or ""
preview = content[:200] if content else ""
return {
"elapsed_s": elapsed,
@@ -277,12 +198,29 @@ class PromptSizer:
f"Target ({target}) is smaller than template overhead ({self.base_tokens})."
)
content = ""
tok = self.count_fn(content)
# Estimate tokens per atom using a sample
sample_count = 100
sample_content = self.atom * sample_count
sample_tokens = self.count_fn(sample_content) - self.base_tokens
tokens_per_atom = sample_tokens / sample_count
while tok < target:
content += self.atom
tok = self.count_fn(content)
# Estimate starting point
needed_tokens = target - self.base_tokens
estimated_atoms = int(needed_tokens / tokens_per_atom)
# Binary search to find exact atom count
low, high = 0, estimated_atoms * 2 + 100
while low < high:
mid = (low + high) // 2
tok = self.count_fn(self.atom * mid)
if tok < target:
low = mid + 1
else:
high = mid
content = self.atom * low
tok = self.count_fn(content)
logger.info(f"{tok=}")
if tok != target:
raise RuntimeError(
@@ -298,11 +236,7 @@ def main() -> int:
prog="exo-bench",
description="Benchmark exo model throughput across placement previews.",
)
ap.add_argument("--host", default=os.environ.get("EXO_HOST", "localhost"))
ap.add_argument(
"--port", type=int, default=int(os.environ.get("EXO_PORT", "52415"))
)
ap.add_argument("--model", required=True, help="Model short id or huggingface id")
add_common_instance_args(ap)
ap.add_argument(
"--pp",
nargs="+",
@@ -315,29 +249,6 @@ def main() -> int:
required=True,
help="Generation lengths (ints). Accepts commas.",
)
ap.add_argument(
"--max-nodes",
type=int,
default=4,
help="Only consider placements using <= this many nodes.",
)
ap.add_argument(
"--min-nodes",
type=int,
default=1,
help="Only consider placements using >= this many nodes.",
)
ap.add_argument(
"--instance-meta", choices=["ring", "jaccl", "both"], default="both"
)
ap.add_argument(
"--sharding", choices=["pipeline", "tensor", "both"], default="both"
)
ap.add_argument(
"--skip-pipeline-jaccl",
action="store_true",
help="Pipeline jaccl is often pointless, skip by default",
)
ap.add_argument(
"--repeat", type=int, default=1, help="Repetitions per (pp,tg) pair."
)
@@ -347,17 +258,20 @@ def main() -> int:
default=0,
help="Warmup runs per placement (uses first pp/tg).",
)
ap.add_argument(
"--timeout", type=float, default=600.0, help="HTTP timeout (seconds)."
)
ap.add_argument(
"--json-out",
default="bench/results.json",
help="Write raw per-run results JSON to this path.",
)
ap.add_argument("--stdout", action="store_true", help="Write results to stdout")
ap.add_argument(
"--dry-run", action="store_true", help="List selected placements and exit."
)
ap.add_argument(
"--all-combinations",
action="store_true",
help="Force all pp×tg combinations (cartesian product) even when lists have equal length.",
)
args = ap.parse_args()
pp_list = parse_int_list(args.pp)
@@ -369,18 +283,19 @@ def main() -> int:
logger.error("--repeat must be >= 1")
return 2
# Log pairing mode
use_combinations = args.all_combinations or len(pp_list) != len(tg_list)
if use_combinations:
logger.info(
f"pp/tg mode: combinations (product) - {len(pp_list) * len(tg_list)} pairs"
)
else:
logger.info(f"pp/tg mode: tandem (zip) - {len(pp_list)} pairs")
client = ExoClient(args.host, args.port, timeout_s=args.timeout)
short_id, full_model_id = resolve_model_short_id(client, args.model)
previews_resp = client.request_json(
"GET", "/instance/previews", params={"model_id": full_model_id}
)
previews = previews_resp.get("previews") or []
tokenizer = AutoTokenizer.from_pretrained(
full_model_id,
trust_remote_code=True,
)
tokenizer = load_tokenizer_for_bench(full_model_id)
if tokenizer is None:
raise RuntimeError("[exo-bench] tokenizer load failed")
@@ -391,44 +306,9 @@ def main() -> int:
logger.error("[exo-bench] tokenizer usable but prompt sizing failed")
raise
selected: list[dict[str, Any]] = []
for p in previews:
if p.get("error") is not None:
continue
if not placement_filter(str(p.get("instance_meta", "")), args.instance_meta):
continue
if not sharding_filter(str(p.get("sharding", "")), args.sharding):
continue
instance = p.get("instance")
if not isinstance(instance, dict):
continue
n = nodes_used_in_instance(instance)
# Skip tensor ring single node as it is pointless when pipeline ring
if n == 1 and (
(args.sharding == "both" and "tensor" in p.get("sharding", "").lower())
or (
args.instance_meta == "both"
and "jaccl" in p.get("instance_meta", "").lower()
)
):
continue
if (
args.skip_pipeline_jaccl
and (
args.instance_meta == "both"
and "jaccl" in p.get("instance_meta", "").lower()
)
and (
args.sharding == "both" and "pipeline" in p.get("sharding", "").lower()
)
):
continue
if args.min_nodes <= n <= args.max_nodes:
selected.append(p)
selected = settle_and_fetch_placements(
client, full_model_id, args, settle_timeout=args.settle_timeout
)
if not selected:
logger.error("No valid placements matched your filters.")
@@ -453,6 +333,24 @@ def main() -> int:
if args.dry_run:
return 0
settle_deadline = (
time.monotonic() + args.settle_timeout if args.settle_timeout > 0 else None
)
logger.info("Planning phase: checking downloads...")
download_duration_s = run_planning_phase(
client,
full_model_id,
selected[0],
args.danger_delete_downloads,
args.timeout,
settle_deadline,
)
if download_duration_s is not None:
logger.info(f"Download: {download_duration_s:.1f}s (freshly downloaded)")
else:
logger.info("Download: model already cached")
all_rows: list[dict[str, Any]] = []
for preview in selected:
@@ -486,60 +384,60 @@ def main() -> int:
)
logger.debug(f" warmup {i + 1}/{args.warmup} done")
for pp in pp_list:
# if (
# pp * n_nodes > 2048
# and "ring" in instance_meta.lower()
# and "tensor" in sharding.lower()
# ):
# model_card = MODEL_CARDS[short_id]
# if model_card.metadata.storage_size > Memory.from_gb(10):
# logger.info(
# f"Skipping tensor ring as this is too slow for model of size {model_card.metadata.storage_size} on {n_nodes=}"
# )
# continue
for tg in tg_list:
runs: list[dict[str, Any]] = []
for r in range(args.repeat):
time.sleep(3)
try:
row, actual_pp_tokens = run_one_completion(
client, full_model_id, pp, tg, prompt_sizer
)
except Exception as e:
logger.error(e)
continue
row.update(
{
"model_short_id": short_id,
"model_id": full_model_id,
"placement_sharding": sharding,
"placement_instance_meta": instance_meta,
"placement_nodes": n_nodes,
"instance_id": instance_id,
"pp_tokens": actual_pp_tokens,
"tg": tg,
"repeat_index": r,
}
)
runs.append(row)
all_rows.append(row)
# If pp and tg lists have same length, run in tandem (zip)
# Otherwise (or if --all-combinations), run all combinations (cartesian product)
if use_combinations:
pp_tg_pairs = list(itertools.product(pp_list, tg_list))
else:
pp_tg_pairs = list(zip(pp_list, tg_list, strict=True))
if runs:
prompt_tps = mean(x["stats"]["prompt_tps"] for x in runs)
gen_tps = mean(x["stats"]["generation_tps"] for x in runs)
ptok = mean(x["stats"]["prompt_tokens"] for x in runs)
gtok = mean(x["stats"]["generation_tokens"] for x in runs)
peak = mean(
x["stats"]["peak_memory_usage"]["inBytes"] for x in runs
for pp, tg in pp_tg_pairs:
runs: list[dict[str, Any]] = []
for r in range(args.repeat):
time.sleep(3)
try:
row, actual_pp_tokens = run_one_completion(
client, full_model_id, pp, tg, prompt_sizer
)
except Exception as e:
logger.error(e)
continue
row.update(
{
"model_short_id": short_id,
"model_id": full_model_id,
"placement_sharding": sharding,
"placement_instance_meta": instance_meta,
"placement_nodes": n_nodes,
"instance_id": instance_id,
"pp_tokens": actual_pp_tokens,
"tg": tg,
"repeat_index": r,
**(
{"download_duration_s": download_duration_s}
if download_duration_s is not None
else {}
),
}
)
runs.append(row)
all_rows.append(row)
logger.info(
f"prompt_tps={prompt_tps:.2f} gen_tps={gen_tps:.2f} "
f"prompt_tokens={ptok} gen_tokens={gtok} "
f"peak_memory={format_peak_memory(peak)}\n"
)
time.sleep(2)
if runs:
prompt_tps = mean(x["stats"]["prompt_tps"] for x in runs)
gen_tps = mean(x["stats"]["generation_tps"] for x in runs)
ptok = mean(x["stats"]["prompt_tokens"] for x in runs)
gtok = mean(x["stats"]["generation_tokens"] for x in runs)
peak = mean(
x["stats"]["peak_memory_usage"]["inBytes"] for x in runs
)
logger.info(
f"prompt_tps={prompt_tps:.2f} gen_tps={gen_tps:.2f} "
f"prompt_tokens={ptok} gen_tokens={gtok} "
f"peak_memory={format_peak_memory(peak)}\n"
)
time.sleep(2)
finally:
try:
client.request_json("DELETE", f"/instance/{instance_id}")
@@ -551,7 +449,9 @@ def main() -> int:
time.sleep(5)
if args.json_out:
if args.stdout:
json.dump(all_rows, sys.stdout, indent=2, ensure_ascii=False)
elif args.json_out:
with open(args.json_out, "w", encoding="utf-8") as f:
json.dump(all_rows, f, indent=2, ensure_ascii=False)
logger.debug(f"\nWrote results JSON: {args.json_out}")
+489
View File
@@ -0,0 +1,489 @@
# type: ignore
from __future__ import annotations
import argparse
import http.client
import json
import os
import time
from typing import Any
from urllib.parse import urlencode
from loguru import logger
_SETTLE_INITIAL_BACKOFF_S = 1.0
_SETTLE_MAX_BACKOFF_S = 60.0
_SETTLE_BACKOFF_MULTIPLIER = 2.0
class ExoHttpError(RuntimeError):
def __init__(self, status: int, reason: str, body_preview: str):
super().__init__(f"HTTP {status} {reason}: {body_preview}")
self.status = status
class ExoClient:
def __init__(self, host: str, port: int, timeout_s: float = 7200.0):
self.host = host
self.port = port
self.timeout_s = timeout_s
def request_json(
self,
method: str,
path: str,
params: dict[str, Any] | None = None,
body: dict[str, Any] | None = None,
headers: dict[str, str] | None = None,
) -> Any:
if not path.startswith("/"):
path = "/" + path
if params:
path = path + "?" + urlencode(params)
conn = http.client.HTTPConnection(self.host, self.port, timeout=self.timeout_s)
try:
payload: bytes | None = None
hdrs: dict[str, str] = {"Accept": "application/json"}
if body is not None:
payload = json.dumps(body).encode("utf-8")
hdrs["Content-Type"] = "application/json"
if headers:
hdrs.update(headers)
conn.request(method.upper(), path, body=payload, headers=hdrs)
resp = conn.getresponse()
raw = resp.read()
text = raw.decode("utf-8", errors="replace") if raw else ""
if resp.status >= 400:
raise ExoHttpError(resp.status, resp.reason, text[:300])
if not text:
return None
return json.loads(text)
finally:
conn.close()
def post_bench_chat_completions(self, payload: dict[str, Any]) -> dict[str, Any]:
return self.request_json("POST", "/bench/chat/completions", body=payload)
def unwrap_instance(instance: dict[str, Any]) -> dict[str, Any]:
if len(instance) != 1:
raise KeyError(f"Expected 1 key, got keys={list(instance.keys())}")
tag = next(iter(instance))
inner = instance[tag]
if not isinstance(inner, dict):
raise TypeError(f"payload for {tag} must be dict, got {type(inner)}")
return inner
def instance_id_from_instance(instance: dict[str, Any]) -> str:
inner = unwrap_instance(instance)
return str(inner["instanceId"])
def nodes_used_in_instance(instance: dict[str, Any]) -> int:
inner = unwrap_instance(instance)
return len(inner["shardAssignments"]["nodeToRunner"])
def runner_ids_from_instance(instance: dict[str, Any]) -> list[str]:
inner = unwrap_instance(instance)
runner_to_shard = inner["shardAssignments"]["runnerToShard"]
return list(runner_to_shard.keys())
def runner_ready(runner: dict[str, Any]) -> bool:
return "RunnerReady" in runner
def runner_failed(runner: dict[str, Any]) -> bool:
return "RunnerFailed" in runner
def get_runner_failed_message(runner: dict[str, Any]) -> str | None:
if "RunnerFailed" in runner:
return runner["RunnerFailed"].get("errorMessage")
return None
def wait_for_instance_ready(
client: ExoClient, instance_id: str, timeout: float = 24000.0
) -> None:
start_time = time.time()
instance_existed = False
while time.time() - start_time < timeout:
state = client.request_json("GET", "/state")
instances = state.get("instances", {})
if instance_id not in instances:
if instance_existed:
# Instance was deleted after being created - likely due to runner failure
raise RuntimeError(
f"Instance {instance_id} was deleted (runner may have failed)"
)
time.sleep(0.1)
continue
instance_existed = True
instance = instances[instance_id]
runner_ids = runner_ids_from_instance(instance)
runners = state.get("runners", {})
# Check for failed runners first
for rid in runner_ids:
runner = runners.get(rid, {})
if runner_failed(runner):
error_msg = get_runner_failed_message(runner) or "Unknown error"
raise RuntimeError(f"Runner {rid} failed: {error_msg}")
if all(runner_ready(runners.get(rid, {})) for rid in runner_ids):
return
time.sleep(0.1)
raise TimeoutError(f"Instance {instance_id} did not become ready within {timeout=}")
def wait_for_instance_gone(
client: ExoClient, instance_id: str, timeout: float = 3.0
) -> None:
start_time = time.time()
while time.time() - start_time < timeout:
try:
client.request_json("GET", f"/instance/{instance_id}")
time.sleep(0.4)
except ExoHttpError as e:
if e.status == 404:
return
raise
raise TimeoutError(f"Instance {instance_id} did not get deleted within {timeout=}")
def resolve_model_short_id(client: ExoClient, model_arg: str) -> tuple[str, str]:
models = client.request_json("GET", "/models") or {}
data = models.get("data") or []
for m in data:
if (m.get("name") or "").lower() == model_arg.lower():
short_id = str(m["name"])
full_id = str(m.get("hugging_face_id") or m["name"])
return short_id, full_id
for m in data:
if m.get("hugging_face_id") == model_arg:
short_id = str(m["name"])
full_id = str(m["hugging_face_id"])
return short_id, full_id
raise ValueError(f"Model not found in /models: {model_arg}")
def placement_filter(instance_meta: str, wanted: str) -> bool:
s = (instance_meta or "").lower()
if wanted == "both":
return ("ring" in s) or ("jaccl" in s)
return wanted in s
def sharding_filter(sharding: str, wanted: str) -> bool:
s = (sharding or "").lower()
if wanted == "both":
return ("pipeline" in s) or ("tensor" in s)
return wanted in s
def fetch_and_filter_placements(
client: ExoClient, full_model_id: str, args: argparse.Namespace
) -> list[dict[str, Any]]:
previews_resp = client.request_json(
"GET", "/instance/previews", params={"model_id": full_model_id}
)
previews = previews_resp.get("previews") or []
selected: list[dict[str, Any]] = []
for p in previews:
if p.get("error") is not None:
continue
if not placement_filter(str(p.get("instance_meta", "")), args.instance_meta):
continue
if not sharding_filter(str(p.get("sharding", "")), args.sharding):
continue
instance = p.get("instance")
if not isinstance(instance, dict):
continue
n = nodes_used_in_instance(instance)
# Skip tensor ring single node as it is pointless when pipeline ring
if n == 1 and (
(args.sharding == "both" and "tensor" in p.get("sharding", "").lower())
or (
args.instance_meta == "both"
and "jaccl" in p.get("instance_meta", "").lower()
)
):
continue
if (
args.skip_pipeline_jaccl
and (
args.instance_meta == "both"
and "jaccl" in p.get("instance_meta", "").lower()
)
and (
args.sharding == "both" and "pipeline" in p.get("sharding", "").lower()
)
):
continue
if (
args.skip_tensor_ring
and (
args.instance_meta == "both"
and "ring" in p.get("instance_meta", "").lower()
)
and (args.sharding == "both" and "tensor" in p.get("sharding", "").lower())
):
continue
if args.min_nodes <= n <= args.max_nodes:
selected.append(p)
return selected
def settle_and_fetch_placements(
client: ExoClient,
full_model_id: str,
args: argparse.Namespace,
settle_timeout: float = 0,
) -> list[dict[str, Any]]:
selected = fetch_and_filter_placements(client, full_model_id, args)
if not selected and settle_timeout > 0:
backoff = _SETTLE_INITIAL_BACKOFF_S
deadline = time.monotonic() + settle_timeout
while not selected and time.monotonic() < deadline:
remaining = deadline - time.monotonic()
logger.warning(
f"No valid placements yet (cluster may still be settling). "
f"Retrying in {backoff:.1f}s ({remaining:.0f}s remaining)..."
)
time.sleep(min(backoff, remaining))
backoff = min(backoff * _SETTLE_BACKOFF_MULTIPLIER, _SETTLE_MAX_BACKOFF_S)
selected = fetch_and_filter_placements(client, full_model_id, args)
return selected
def run_planning_phase(
client: ExoClient,
full_model_id: str,
preview: dict[str, Any],
danger_delete: bool,
timeout: float,
settle_deadline: float | None,
) -> float | None:
"""Check disk space and ensure model is downloaded before benchmarking.
Returns the wall-clock download duration in seconds if a fresh download
was needed, or None if the model was already cached on all nodes.
"""
# Get model size from /models
models = client.request_json("GET", "/models") or {}
model_bytes = 0
for m in models.get("data", []):
if m.get("hugging_face_id") == full_model_id:
model_bytes = m.get("storage_size_megabytes", 0) * 1024 * 1024
break
if not model_bytes:
logger.warning(
f"Could not determine size for {full_model_id}, skipping disk check"
)
return None
# Get nodes from preview
inner = unwrap_instance(preview["instance"])
node_ids = list(inner["shardAssignments"]["nodeToRunner"].keys())
runner_to_shard = inner["shardAssignments"]["runnerToShard"]
state = client.request_json("GET", "/state")
downloads = state.get("downloads", {})
node_disk = state.get("nodeDisk", {})
needs_download = False
for node_id in node_ids:
node_downloads = downloads.get(node_id, [])
# Check if model already downloaded on this node
already_downloaded = any(
"DownloadCompleted" in p
and unwrap_instance(p["DownloadCompleted"]["shardMetadata"])["modelCard"][
"modelId"
]
== full_model_id
for p in node_downloads
)
if already_downloaded:
continue
needs_download = True
# Wait for disk info if settle_deadline is set
disk_info = node_disk.get(node_id, {})
backoff = _SETTLE_INITIAL_BACKOFF_S
while not disk_info and settle_deadline and time.monotonic() < settle_deadline:
remaining = settle_deadline - time.monotonic()
logger.info(
f"Waiting for disk info on {node_id} ({remaining:.0f}s remaining)..."
)
time.sleep(min(backoff, remaining))
backoff = min(backoff * _SETTLE_BACKOFF_MULTIPLIER, _SETTLE_MAX_BACKOFF_S)
state = client.request_json("GET", "/state")
node_disk = state.get("nodeDisk", {})
disk_info = node_disk.get(node_id, {})
if not disk_info:
logger.warning(f"No disk info for {node_id}, skipping space check")
continue
avail = disk_info.get("available", {}).get("inBytes", 0)
if avail >= model_bytes:
continue
if not danger_delete:
raise RuntimeError(
f"Insufficient disk on {node_id}: need {model_bytes // (1024**3)}GB, "
f"have {avail // (1024**3)}GB. Use --danger-delete-downloads to free space."
)
# Delete from smallest to largest (skip read-only models from EXO_MODELS_PATH)
completed = [
(
unwrap_instance(p["DownloadCompleted"]["shardMetadata"])["modelCard"][
"modelId"
],
p["DownloadCompleted"]["total"]["inBytes"],
)
for p in node_downloads
if "DownloadCompleted" in p
and not p["DownloadCompleted"].get("readOnly", False)
]
for del_model, size in sorted(completed, key=lambda x: x[1]):
logger.info(f"Deleting {del_model} from {node_id} ({size // (1024**2)}MB)")
client.request_json("DELETE", f"/download/{node_id}/{del_model}")
avail += size
if avail >= model_bytes:
break
if avail < model_bytes:
raise RuntimeError(f"Could not free enough space on {node_id}")
# Start downloads (idempotent)
download_t0 = time.perf_counter() if needs_download else None
for node_id in node_ids:
runner_id = inner["shardAssignments"]["nodeToRunner"][node_id]
shard = runner_to_shard[runner_id]
client.request_json(
"POST",
"/download/start",
body={
"targetNodeId": node_id,
"shardMetadata": shard,
},
)
logger.info(f"Started download on {node_id}")
# Wait for downloads
start = time.time()
while time.time() - start < timeout:
state = client.request_json("GET", "/state")
downloads = state.get("downloads", {})
all_done = True
for node_id in node_ids:
done = any(
"DownloadCompleted" in p
and unwrap_instance(p["DownloadCompleted"]["shardMetadata"])[
"modelCard"
]["modelId"]
== full_model_id
for p in downloads.get(node_id, [])
)
failed = [
p["DownloadFailed"]["errorMessage"]
for p in downloads.get(node_id, [])
if "DownloadFailed" in p
and unwrap_instance(p["DownloadFailed"]["shardMetadata"])["modelCard"][
"modelId"
]
== full_model_id
]
if failed:
raise RuntimeError(f"Download failed on {node_id}: {failed[0]}")
if not done:
all_done = False
if all_done:
if download_t0 is not None:
return time.perf_counter() - download_t0
return None
time.sleep(1)
raise TimeoutError("Downloads did not complete in time")
def add_common_instance_args(ap: argparse.ArgumentParser) -> None:
ap.add_argument("--host", default=os.environ.get("EXO_HOST", "localhost"))
ap.add_argument(
"--port", type=int, default=int(os.environ.get("EXO_PORT", "52415"))
)
ap.add_argument("--model", required=True, help="Model short id or huggingface id")
ap.add_argument(
"--max-nodes",
type=int,
default=4,
help="Only consider placements using <= this many nodes.",
)
ap.add_argument(
"--min-nodes",
type=int,
default=1,
help="Only consider placements using >= this many nodes.",
)
ap.add_argument(
"--instance-meta", choices=["ring", "jaccl", "both"], default="both"
)
ap.add_argument(
"--sharding", choices=["pipeline", "tensor", "both"], default="both"
)
ap.add_argument(
"--skip-pipeline-jaccl",
action="store_true",
help="Skip pipeline+jaccl placements, as it's often pointless.",
)
ap.add_argument(
"--skip-tensor-ring",
action="store_true",
help="Skip tensor+ring placements, as it's so slow.",
)
ap.add_argument(
"--timeout", type=float, default=7200.0, help="HTTP timeout (seconds)."
)
ap.add_argument(
"--settle-timeout",
type=float,
default=0,
help="Max seconds to wait for the cluster to produce valid placements (0 = try once).",
)
ap.add_argument(
"--danger-delete-downloads",
action="store_true",
help="Delete existing models from smallest to largest to make room for benchmark model.",
)
+18
View File
@@ -0,0 +1,18 @@
[project]
name = "exo-bench"
version = "0.1.0"
description = "Benchmarking tool for exo distributed inference"
requires-python = ">=3.13"
dependencies = [
"httpx>=0.27.0",
"loguru>=0.7.3",
"transformers>=5.0.0",
"huggingface-hub>=0.33.4",
"tiktoken>=0.12.0",
"jinja2>=3.1.0",
"protobuf>=5.29.0",
]
[build-system]
requires = ["hatchling"]
build-backend = "hatchling.build"
+306
View File
@@ -0,0 +1,306 @@
# Tool definitions — each becomes an OpenAI function tool.
# All scenarios get all tools unless they specify a `tools` list.
[tools.get_current_weather]
description = "Get the current weather in a given location"
required = ["location"]
[tools.get_current_weather.properties.location]
type = "string"
description = "City and state, e.g. San Francisco, CA"
[tools.get_current_weather.properties.unit]
type = "string"
enum = ["celsius", "fahrenheit"]
description = "Temperature unit"
[tools.calculate]
description = "Evaluate a mathematical expression and return the numeric result"
required = ["expression"]
[tools.calculate.properties.expression]
type = "string"
description = "The math expression to evaluate, e.g. '2 + 3 * 4'"
[tools.search_products]
description = "Search for products in a catalog by query, category, and price"
required = ["query"]
[tools.search_products.properties.query]
type = "string"
description = "Search query string"
[tools.search_products.properties.category]
type = "string"
enum = ["electronics", "clothing", "food", "books"]
description = "Product category to filter by"
[tools.search_products.properties.max_price]
type = "number"
description = "Maximum price in USD"
[tools.create_todos]
description = "Create a structured todo list"
required = ["todos"]
[tools.create_todos.properties.todos]
type = "array"
description = "List of todo items"
[tools.create_todos.properties.todos.items]
type = "object"
required = ["content", "status", "priority"]
[tools.create_todos.properties.todos.items.properties.content]
type = "string"
description = "The todo item text"
[tools.create_todos.properties.todos.items.properties.status]
type = "string"
description = "Status: pending, in_progress, or completed"
[tools.create_todos.properties.todos.items.properties.priority]
type = "string"
description = "Priority: low, normal, or high"
# -- Should call a tool --
[[scenarios]]
name = "weather_simple"
description = "Basic weather query -> get_current_weather"
expect_tool_call = true
expected_function = "get_current_weather"
required_arg_keys = ["location"]
[[scenarios.messages]]
role = "user"
content = "What's the weather like in Tokyo right now?"
[[scenarios]]
name = "calculator_simple"
description = "Math question -> calculate"
expect_tool_call = true
expected_function = "calculate"
required_arg_keys = ["expression"]
[[scenarios.messages]]
role = "user"
content = "Use the calculator to compute 3847 * 926 + 17293"
[[scenarios]]
name = "search_with_filters"
description = "Product search with category and price filter"
expect_tool_call = true
expected_function = "search_products"
required_arg_keys = ["query"]
[[scenarios.messages]]
role = "user"
content = "Find me electronics under $50"
# -- Multi-turn: tool call then follow-up --
[[scenarios]]
name = "weather_multi_turn"
description = "Weather query -> tool result -> natural language summary"
expect_tool_call = true
expected_function = "get_current_weather"
required_arg_keys = ["location"]
[scenarios.tool_result]
temperature = "18C"
condition = "partly cloudy"
humidity = "65%"
wind = "12 km/h NW"
[[scenarios.messages]]
role = "user"
content = "What's the weather in Paris?"
[[scenarios]]
name = "calculator_multi_turn"
description = "Math query -> tool result -> model reports the answer"
expect_tool_call = true
expected_function = "calculate"
required_arg_keys = ["expression"]
[scenarios.tool_result]
result = 491682
[[scenarios.messages]]
role = "user"
content = "Use the calculator to compute 1847 * 263 + 5921"
[[scenarios]]
name = "search_multi_turn"
description = "Search query -> tool result -> model summarizes products"
expect_tool_call = true
expected_function = "search_products"
required_arg_keys = ["query"]
[[scenarios.tool_result.results]]
name = "Hands-On Machine Learning"
price = 45.99
rating = 4.8
[[scenarios.tool_result.results]]
name = "Deep Learning with Python"
price = 39.99
rating = 4.6
[[scenarios.messages]]
role = "user"
content = "Search for books about machine learning"
# -- Sequential tool calls --
[[scenarios]]
name = "chained_tool_calls_same"
description = "Thinking + weather(Tokyo) -> result -> model must call weather(London)"
expect_tool_call = true
expected_function = "get_current_weather"
required_arg_keys = ["location"]
[[scenarios.messages]]
role = "user"
content = "Compare the weather in Tokyo and London."
[[scenarios.messages]]
role = "assistant"
content = "I'll check both cities. Let me start with Tokyo."
[[scenarios.messages.tool_calls]]
id = "call_1"
name = "get_current_weather"
arguments = { location = "Tokyo" }
[[scenarios.messages]]
role = "tool"
tool_call_id = "call_1"
content = '{"temperature": "25C", "condition": "sunny"}'
[[scenarios]]
name = "chained_tool_calls_different"
description = "Thinking + weather(Berlin) -> result -> model must call calculator"
expect_tool_call = true
expected_function = "calculate"
required_arg_keys = ["expression"]
[[scenarios.messages]]
role = "user"
content = "What's the weather in Berlin, and also use the calculator to compute 4819 * 37 + 291."
[[scenarios.messages]]
role = "assistant"
content = "I'll handle both. Let me check Berlin's weather first."
[[scenarios.messages.tool_calls]]
id = "call_2"
name = "get_current_weather"
arguments = { location = "Berlin" }
[[scenarios.messages]]
role = "tool"
tool_call_id = "call_2"
content = '{"temperature": "12C", "condition": "rainy"}'
[[scenarios]]
name = "chained_tool_calls_three"
description = "Two prior thinking+tool calls -> results -> model must make a third"
expect_tool_call = true
expected_function = "get_current_weather"
required_arg_keys = ["location"]
[[scenarios.messages]]
role = "user"
content = "Compare weather in Tokyo, Paris, and London."
[[scenarios.messages]]
role = "assistant"
content = "I'll check all three cities. Starting with Tokyo."
[[scenarios.messages.tool_calls]]
id = "call_3"
name = "get_current_weather"
arguments = { location = "Tokyo" }
[[scenarios.messages]]
role = "tool"
tool_call_id = "call_3"
content = '{"temperature": "25C", "condition": "sunny"}'
[[scenarios.messages]]
role = "assistant"
content = "Got Tokyo. Now checking Paris."
[[scenarios.messages.tool_calls]]
id = "call_4"
name = "get_current_weather"
arguments = { location = "Paris" }
[[scenarios.messages]]
role = "tool"
tool_call_id = "call_4"
content = '{"temperature": "18C", "condition": "cloudy"}'
# -- Nested object schema (regression for lossy chat template rendering) --
[[scenarios]]
name = "nested_schema_tool_call"
description = "Tool call with nested object array schema -> create_todos"
expect_tool_call = true
expected_function = "create_todos"
required_arg_keys = ["todos"]
nested_array_key = "todos"
required_item_keys = ["content", "status", "priority"]
tools = ["create_todos"]
[[scenarios.messages]]
role = "user"
content = "Create a todo list with 3 items to learn Python"
# -- Tool name integrity (regression for harmony token leaking into name) --
[tools.glob]
description = "Search for files matching a glob pattern in the codebase"
required = ["pattern"]
[tools.glob.properties.pattern]
type = "string"
description = "The glob pattern to match files against, e.g. '**/*.py'"
[tools.glob.properties.path]
type = "string"
description = "The directory to search in"
[[scenarios]]
name = "tool_name_integrity"
description = "Tool name must not contain harmony tokens like <|channel|>"
expect_tool_call = true
expected_function = "glob"
required_arg_keys = ["pattern"]
tools = ["glob"]
[[scenarios.messages]]
role = "user"
content = "Find all Python files in the src directory"
# -- Should NOT call a tool --
[[scenarios]]
name = "no_tool_joke"
description = "Joke request should NOT trigger any tool"
expect_tool_call = false
[[scenarios.messages]]
role = "user"
content = "Tell me a funny joke about cats."
[[scenarios]]
name = "no_tool_factual"
description = "Factual question answerable from training data"
expect_tool_call = false
[[scenarios.messages]]
role = "user"
content = "What is the capital of Japan?"
+189
View File
@@ -0,0 +1,189 @@
# Single-node M3 Ultra benchmarks
#
# Shared constraints applied to ALL benchmarks in this file.
constraints = [
"All(MacOsBuild(=25D125))",
"Hosts(=1)",
"All(Chip(m3_ultra))",
"All(GpuCores(=80))",
]
[topology]
type = "none"
# Default args merged into each benchmark's args (benchmark-level args win).
[defaults]
pp = [512, 2048, 8192, 16384]
tg = 128
[[benchmark]]
model = "mlx-community/Meta-Llama-3.1-70B-Instruct-4bit"
extra_constraints = ["All(Memory(>=96GiB))"]
[[benchmark]]
model = "mlx-community/gpt-oss-120b-MXFP4-Q8"
extra_constraints = ["All(Memory(>=96GiB))"]
[[benchmark]]
model = "mlx-community/GLM-4.7-Flash-8bit"
extra_constraints = ["All(Memory(>=96GiB))"]
[[benchmark]]
model = "mlx-community/Qwen3-Coder-Next-6bit"
extra_constraints = ["All(Memory(>=96GiB))"]
[[benchmark]]
model = "mlx-community/Qwen3-30B-A3B-8bit"
extra_constraints = ["All(Memory(>=96GiB))"]
[[benchmark]]
model = "mlx-community/Qwen3-0.6B-4bit"
extra_constraints = ["All(Memory(>=96GiB))"]
[[benchmark]]
model = "mlx-community/Qwen3-0.6B-8bit"
extra_constraints = ["All(Memory(>=96GiB))"]
[[benchmark]]
model = "mlx-community/Llama-3.2-1B-Instruct-4bit"
extra_constraints = ["All(Memory(>=96GiB))"]
[[benchmark]]
model = "mlx-community/Llama-3.2-3B-Instruct-4bit"
extra_constraints = ["All(Memory(>=96GiB))"]
[[benchmark]]
model = "mlx-community/Llama-3.2-3B-Instruct-8bit"
extra_constraints = ["All(Memory(>=96GiB))"]
[[benchmark]]
model = "mlx-community/Meta-Llama-3.1-8B-Instruct-4bit"
extra_constraints = ["All(Memory(>=96GiB))"]
[[benchmark]]
model = "mlx-community/Meta-Llama-3.1-8B-Instruct-8bit"
extra_constraints = ["All(Memory(>=96GiB))"]
[[benchmark]]
model = "mlx-community/Meta-Llama-3.1-8B-Instruct-bf16"
extra_constraints = ["All(Memory(>=96GiB))"]
[[benchmark]]
model = "mlx-community/gpt-oss-20b-MXFP4-Q8"
extra_constraints = ["All(Memory(>=96GiB))"]
[[benchmark]]
model = "mlx-community/Qwen3-30B-A3B-4bit"
extra_constraints = ["All(Memory(>=96GiB))"]
[[benchmark]]
model = "mlx-community/GLM-4.7-Flash-4bit"
extra_constraints = ["All(Memory(>=96GiB))"]
[[benchmark]]
model = "mlx-community/GLM-4.7-Flash-5bit"
extra_constraints = ["All(Memory(>=96GiB))"]
[[benchmark]]
model = "mlx-community/GLM-4.7-Flash-6bit"
extra_constraints = ["All(Memory(>=96GiB))"]
[[benchmark]]
model = "mlx-community/Llama-3.3-70B-Instruct-4bit"
extra_constraints = ["All(Memory(>=96GiB))"]
[[benchmark]]
model = "mlx-community/Qwen3-Coder-Next-4bit"
extra_constraints = ["All(Memory(>=96GiB))"]
[[benchmark]]
model = "mlx-community/Qwen3-Coder-Next-5bit"
extra_constraints = ["All(Memory(>=96GiB))"]
[[benchmark]]
model = "mlx-community/Qwen3-Coder-Next-8bit"
extra_constraints = ["All(Memory(>=96GiB))"]
[[benchmark]]
model = "mlx-community/Qwen3-Next-80B-A3B-Instruct-4bit"
extra_constraints = ["All(Memory(>=96GiB))"]
[[benchmark]]
model = "mlx-community/Qwen3-Next-80B-A3B-Instruct-8bit"
extra_constraints = ["All(Memory(>=96GiB))"]
[[benchmark]]
model = "mlx-community/Qwen3-Next-80B-A3B-Thinking-4bit"
extra_constraints = ["All(Memory(>=96GiB))"]
[[benchmark]]
model = "mlx-community/Qwen3-Next-80B-A3B-Thinking-8bit"
extra_constraints = ["All(Memory(>=96GiB))"]
[[benchmark]]
model = "mlx-community/Llama-3.3-70B-Instruct-8bit"
extra_constraints = ["All(Memory(>=256GiB))"]
[[benchmark]]
model = "mlx-community/llama-3.3-70b-instruct-fp16"
extra_constraints = ["All(Memory(>=256GiB))"]
[[benchmark]]
model = "mlx-community/GLM-4.5-Air-8bit"
extra_constraints = ["All(Memory(>=256GiB))"]
[[benchmark]]
model = "mlx-community/GLM-4.5-Air-bf16"
extra_constraints = ["All(Memory(>=256GiB))"]
[[benchmark]]
model = "mlx-community/GLM-4.7-4bit"
extra_constraints = ["All(Memory(>=256GiB))"]
[[benchmark]]
model = "mlx-community/MiniMax-M2.1-3bit"
extra_constraints = ["All(Memory(>=256GiB))"]
[[benchmark]]
model = "mlx-community/MiniMax-M2.1-8bit"
extra_constraints = ["All(Memory(>=256GiB))"]
[[benchmark]]
model = "mlx-community/Qwen3-235B-A22B-Instruct-2507-4bit"
extra_constraints = ["All(Memory(>=256GiB))"]
[[benchmark]]
model = "mlx-community/Qwen3-Coder-Next-bf16"
extra_constraints = ["All(Memory(>=256GiB))"]
[[benchmark]]
model = "mlx-community/Step-3.5-Flash-4bit"
extra_constraints = ["All(Memory(>=256GiB))"]
[[benchmark]]
model = "mlx-community/Step-3.5-Flash-6bit"
extra_constraints = ["All(Memory(>=256GiB))"]
[[benchmark]]
model = "mlx-community/Step-3.5-Flash-8Bit"
extra_constraints = ["All(Memory(>=256GiB))"]
[[benchmark]]
model = "mlx-community/DeepSeek-V3.1-4bit"
extra_constraints = ["All(Memory(>=512GiB))"]
[[benchmark]]
model = "mlx-community/GLM-4.7-6bit"
extra_constraints = ["All(Memory(>=512GiB))"]
[[benchmark]]
model = "mlx-community/GLM-4.7-8bit-gs32"
extra_constraints = ["All(Memory(>=512GiB))"]
[[benchmark]]
model = "mlx-community/Qwen3-235B-A22B-Instruct-2507-8bit"
extra_constraints = ["All(Memory(>=512GiB))"]
[[benchmark]]
model = "mlx-community/Qwen3-Coder-480B-A35B-Instruct-4bit"
extra_constraints = ["All(Memory(>=512GiB))"]
View File
+59 -3
View File
@@ -3,6 +3,61 @@
perSystem =
{ pkgs, lib, ... }:
let
# Filter source to ONLY include package.json and package-lock.json
# This ensures prettier-svelte only rebuilds when lockfiles change
dashboardLockfileSrc = lib.cleanSourceWith {
src = inputs.self;
filter =
path: type:
let
baseName = builtins.baseNameOf path;
isDashboardDir = baseName == "dashboard" && type == "directory";
isPackageFile =
(lib.hasInfix "/dashboard/" path || lib.hasSuffix "/dashboard" (builtins.dirOf path))
&& (baseName == "package.json" || baseName == "package-lock.json");
in
isDashboardDir || isPackageFile;
};
# Stub source with lockfiles and minimal files for build to succeed
# This allows prettier-svelte to avoid rebuilding when dashboard source changes
dashboardStubSrc = pkgs.runCommand "dashboard-stub-src" { } ''
mkdir -p $out
cp ${dashboardLockfileSrc}/dashboard/package.json $out/
cp ${dashboardLockfileSrc}/dashboard/package-lock.json $out/
# Minimal files so vite build succeeds (produces empty output)
echo '<!DOCTYPE html><html><head></head><body></body></html>' > $out/index.html
mkdir -p $out/src
touch $out/src/app.html
'';
# Deps-only build using stub source (for prettier-svelte)
# Only rebuilds when package.json or package-lock.json change
dashboardDeps = inputs.dream2nix.lib.evalModules {
packageSets.nixpkgs = pkgs;
modules = [
./dashboard.nix
{
paths.projectRoot = inputs.self;
paths.projectRootFile = "flake.nix";
paths.package = inputs.self + "/dashboard";
}
{
deps.dashboardSrc = lib.mkForce dashboardStubSrc;
}
# Override build phases to skip the actual build - just need node_modules
{
mkDerivation = {
buildPhase = lib.mkForce "true";
installPhase = lib.mkForce ''
runHook preInstall
runHook postInstall
'';
};
}
];
};
# Filter source to only include dashboard directory
dashboardSrc = lib.cleanSourceWith {
src = inputs.self;
@@ -42,11 +97,12 @@
'';
# Prettier with svelte plugin for treefmt
# Uses dashboardDeps instead of dashboardFull to avoid rebuilding on source changes
packages.prettier-svelte = pkgs.writeShellScriptBin "prettier-svelte" ''
export NODE_PATH="${dashboardFull}/lib/node_modules/exo-dashboard/node_modules"
export NODE_PATH="${dashboardDeps}/lib/node_modules/exo-dashboard/node_modules"
exec ${pkgs.nodejs}/bin/node \
${dashboardFull}/lib/node_modules/exo-dashboard/node_modules/prettier/bin/prettier.cjs \
--plugin "${dashboardFull}/lib/node_modules/exo-dashboard/node_modules/prettier-plugin-svelte/plugin.js" \
${dashboardDeps}/lib/node_modules/exo-dashboard/node_modules/prettier/bin/prettier.cjs \
--plugin "${dashboardDeps}/lib/node_modules/exo-dashboard/node_modules/prettier-plugin-svelte/plugin.js" \
"$@"
'';
};
+65
View File
@@ -202,6 +202,23 @@
filter: drop-shadow(0 0 3px oklch(0.85 0.18 85 / 0.5));
}
/* Onboarding step 2: connection line between devices */
.onboarding-connection-line {
stroke: oklch(0.85 0.18 85 / 0.5);
stroke-width: 1.5px;
stroke-dasharray: 6, 6;
animation: flowAnimation 1s linear infinite;
filter: drop-shadow(0 0 4px oklch(0.85 0.18 85 / 0.4));
}
/* Onboarding step 4: red connection line for disconnect */
.onboarding-connection-line-red {
stroke: rgba(220, 38, 38, 0.7);
stroke-width: 1.5px;
stroke-dasharray: 6, 6;
filter: drop-shadow(0 0 2px rgba(220, 38, 38, 0.3));
}
.graph-link-active {
stroke: oklch(0.85 0.18 85 / 0.8);
stroke-width: 2px;
@@ -320,3 +337,51 @@ input:focus, textarea:focus {
transform: translate(400px, 400px);
}
}
/* Onboarding smooth fade-in */
@keyframes onb-fade-in {
from { opacity: 0; transform: translateY(8px); }
to { opacity: 1; transform: none; }
}
/* Opacity-only fade-in — no transform, safe for containers with fixed-position children */
@keyframes onb-fade-opacity {
from { opacity: 0; }
to { opacity: 1; }
}
/* Respect reduced motion preference */
@media (prefers-reduced-motion: reduce) {
.shooting-star,
.shooting-star::before {
animation: none !important;
opacity: 0 !important;
}
.graph-link {
animation: none;
}
.status-pulse {
animation: none;
}
.cursor-blink {
animation: none;
}
.onboarding-connection-line {
animation: none;
}
.onboarding-connection-line-red {
animation: none;
}
[style*="onb-fade-in"],
[style*="onb-fade-opacity"] {
animation: none !important;
opacity: 1 !important;
}
*,
*::before,
*::after {
transition-duration: 0.01ms !important;
animation-duration: 0.01ms !important;
animation-iteration-count: 1 !important;
}
}
+180 -236
View File
@@ -7,11 +7,12 @@
editingImage,
clearEditingImage,
selectedChatModel,
setSelectedChatModel,
instances,
ttftMs,
tps,
totalTokens,
thinkingEnabled as thinkingEnabledStore,
setConversationThinking,
stopGeneration,
} from "$lib/stores/app.svelte";
import ChatAttachments from "./ChatAttachments.svelte";
import ImageParamsPanel from "./ImageParamsPanel.svelte";
@@ -25,6 +26,20 @@
autofocus?: boolean;
showModelSelector?: boolean;
modelTasks?: Record<string, string[]>;
modelCapabilities?: Record<string, string[]>;
onSend?: () => void;
onAutoSend?: (
content: string,
files?: {
id: string;
name: string;
type: string;
textContent?: string;
preview?: string;
}[],
) => void;
onOpenModelPicker?: () => void;
modelDisplayOverride?: string;
}
let {
@@ -34,6 +49,11 @@
autofocus = true,
showModelSelector = false,
modelTasks = {},
modelCapabilities = {},
onSend,
onAutoSend,
onOpenModelPicker,
modelDisplayOverride,
}: Props = $props();
let message = $state("");
@@ -41,29 +61,15 @@
let fileInputRef: HTMLInputElement | undefined = $state();
let uploadedFiles = $state<ChatUploadedFile[]>([]);
let isDragOver = $state(false);
const thinkingEnabled = $derived(thinkingEnabledStore());
let loading = $derived(isLoading());
const currentModel = $derived(selectedChatModel());
const instanceData = $derived(instances());
const currentTtft = $derived(ttftMs());
const currentTps = $derived(tps());
const currentTokens = $derived(totalTokens());
const currentEditingImage = $derived(editingImage());
const isEditMode = $derived(currentEditingImage !== null);
// Custom dropdown state
let isModelDropdownOpen = $state(false);
let dropdownButtonRef: HTMLButtonElement | undefined = $state();
let dropdownPosition = $derived(() => {
if (!dropdownButtonRef || !isModelDropdownOpen)
return { top: 0, left: 0, width: 0 };
const rect = dropdownButtonRef.getBoundingClientRect();
return {
top: rect.top,
left: rect.left,
width: rect.width,
};
});
// Accept all supported file types
const acceptString = getAcceptString(["image", "text", "pdf"]);
@@ -89,7 +95,16 @@
const isImageModel = $derived(() => {
if (!currentModel) return false;
return modelSupportsTextToImage(currentModel);
return (
modelSupportsTextToImage(currentModel) ||
modelSupportsImageEditing(currentModel)
);
});
const modelSupportsThinking = $derived(() => {
if (!currentModel) return false;
const caps = modelCapabilities[currentModel] || [];
return caps.includes("thinking_toggle") && caps.includes("text");
});
const isEditOnlyWithoutImage = $derived(
@@ -107,73 +122,14 @@
uploadedFiles.length > 0),
);
// Extract available models from running instances
const availableModels = $derived(() => {
const models: Array<{ id: string; label: string; isImageModel: boolean }> =
[];
for (const [, instance] of Object.entries(instanceData)) {
const modelId = getInstanceModelId(instance);
if (
modelId &&
modelId !== "Unknown" &&
!models.some((m) => m.id === modelId)
) {
models.push({
id: modelId,
label: modelId.split("/").pop() || modelId,
isImageModel: modelSupportsImageGeneration(modelId),
});
}
}
return models;
});
// Track previous model IDs to detect newly added models (plain variable to avoid reactive loop)
let previousModelIds: Set<string> = new Set();
// Auto-select the first available model if none is selected, if current selection is stale, or if a new model is added
$effect(() => {
const models = availableModels();
const currentModelIds = new Set(models.map((m) => m.id));
if (models.length > 0) {
// Find newly added models (in current but not in previous)
const newModels = models.filter((m) => !previousModelIds.has(m.id));
// If no model selected, select the first available
if (!currentModel) {
setSelectedChatModel(models[0].id);
}
// If current model is stale (no longer has a running instance), reset to first available
else if (!models.some((m) => m.id === currentModel)) {
setSelectedChatModel(models[0].id);
}
// If a new model was just added, select it
else if (newModels.length > 0 && previousModelIds.size > 0) {
setSelectedChatModel(newModels[0].id);
}
} else {
// No instances running - clear the selected model
if (currentModel) {
setSelectedChatModel("");
}
}
// Update previous model IDs for next comparison
previousModelIds = currentModelIds;
});
function getInstanceModelId(instanceWrapped: unknown): string {
if (!instanceWrapped || typeof instanceWrapped !== "object") return "";
const keys = Object.keys(instanceWrapped as Record<string, unknown>);
if (keys.length === 1) {
const instance = (instanceWrapped as Record<string, unknown>)[
keys[0]
] as { shardAssignments?: { modelId?: string } };
return instance?.shardAssignments?.modelId || "";
}
return "";
}
// Short label for the currently selected model
const currentModelLabel = $derived(
currentModel
? currentModel.split("/").pop() || currentModel
: modelDisplayOverride
? modelDisplayOverride.split("/").pop() || modelDisplayOverride
: "",
);
async function handleFiles(files: File[]) {
if (files.length === 0) return;
@@ -251,6 +207,7 @@
function handleSubmit() {
if ((!message.trim() && uploadedFiles.length === 0) || loading) return;
if (isEditOnlyWithoutImage) return;
const content = message.trim();
const files = [...uploadedFiles];
@@ -259,6 +216,15 @@
uploadedFiles = [];
resetTextareaHeight();
// When onAutoSend is provided, the parent controls all send logic
// (including launching non-running models before sending)
if (onAutoSend) {
onAutoSend(content, files);
onSend?.();
setTimeout(() => textareaRef?.focus(), 10);
return;
}
// Use image editing if in edit mode
if (isEditMode && currentEditingImage && content) {
editImage(content, currentEditingImage.imageDataUrl);
@@ -275,13 +241,23 @@
if (imageFile.preview) {
editImage(content, imageFile.preview);
}
} else if (isImageModel() && content) {
} else if (
currentModel &&
modelSupportsTextToImage(currentModel) &&
content
) {
// Use image generation for text-to-image models
generateImage(content);
} else {
sendMessage(content, files);
sendMessage(
content,
files,
modelSupportsThinking() ? thinkingEnabled : null,
);
}
onSend?.();
// Refocus the textarea after sending
setTimeout(() => textareaRef?.focus(), 10);
}
@@ -392,7 +368,7 @@
{/if}
<!-- Model selector (when enabled) -->
{#if showModelSelector && availableModels().length > 0}
{#if showModelSelector}
<div
class="flex items-center justify-between gap-2 px-3 py-2 border-b border-exo-medium-gray/30"
>
@@ -401,33 +377,22 @@
class="text-xs text-exo-light-gray uppercase tracking-wider flex-shrink-0"
>MODEL:</span
>
<!-- Custom dropdown -->
<!-- Model button — opens the full model picker -->
<div class="relative flex-1 max-w-xs">
<button
bind:this={dropdownButtonRef}
type="button"
onclick={() => (isModelDropdownOpen = !isModelDropdownOpen)}
class="w-full bg-exo-medium-gray/50 border border-exo-yellow/30 rounded pl-3 pr-8 py-1.5 text-xs font-mono text-left tracking-wide cursor-pointer transition-all duration-200 hover:border-exo-yellow/50 focus:outline-none focus:border-exo-yellow/70 {isModelDropdownOpen
? 'border-exo-yellow/70'
: ''}"
onclick={() => onOpenModelPicker?.()}
class="w-full bg-exo-medium-gray/50 border border-exo-yellow/30 rounded pl-3 pr-8 py-1.5 text-xs font-mono text-left tracking-wide cursor-pointer transition-all duration-200 hover:border-exo-yellow/50 focus:outline-none focus:border-exo-yellow/70"
>
{#if availableModels().find((m) => m.id === currentModel)}
<span class="text-exo-yellow truncate"
>{availableModels().find((m) => m.id === currentModel)
?.label}</span
>
{:else if availableModels().length > 0}
<span class="text-exo-yellow truncate"
>{availableModels()[0].label}</span
{#if currentModelLabel}
<span class="text-exo-yellow truncate">{currentModelLabel}</span
>
{:else}
<span class="text-exo-light-gray/50">— SELECT MODEL —</span>
{/if}
</button>
<div
class="absolute right-2 top-1/2 -translate-y-1/2 pointer-events-none transition-transform duration-200 {isModelDropdownOpen
? 'rotate-180'
: ''}"
class="absolute right-2 top-1/2 -translate-y-1/2 pointer-events-none"
>
<svg
class="w-3 h-3 text-exo-yellow/60"
@@ -444,79 +409,36 @@
</svg>
</div>
</div>
{#if isModelDropdownOpen}
<!-- Backdrop to close dropdown -->
<button
type="button"
class="fixed inset-0 z-[9998] cursor-default"
onclick={() => (isModelDropdownOpen = false)}
aria-label="Close dropdown"
></button>
<!-- Dropdown Panel - fixed positioning to escape overflow:hidden -->
<div
class="fixed bg-exo-dark-gray border border-exo-yellow/30 rounded shadow-lg shadow-black/50 z-[9999] max-h-48 overflow-y-auto"
style="bottom: calc(100vh - {dropdownPosition()
.top}px + 4px); left: {dropdownPosition()
.left}px; width: {dropdownPosition().width}px;"
>
<div class="py-1">
{#each availableModels() as model}
<button
type="button"
onclick={() => {
setSelectedChatModel(model.id);
isModelDropdownOpen = false;
}}
class="w-full px-3 py-2 text-left text-xs font-mono tracking-wide transition-colors duration-100 flex items-center gap-2 {currentModel ===
model.id
? 'bg-transparent text-exo-yellow'
: 'text-exo-light-gray hover:text-exo-yellow'}"
>
{#if currentModel === model.id}
<svg
class="w-3 h-3 flex-shrink-0"
fill="currentColor"
viewBox="0 0 20 20"
>
<path
fill-rule="evenodd"
d="M16.707 5.293a1 1 0 010 1.414l-8 8a1 1 0 01-1.414 0l-4-4a1 1 0 011.414-1.414L8 12.586l7.293-7.293a1 1 0 011.414 0z"
clip-rule="evenodd"
/>
</svg>
{:else}
<span class="w-3"></span>
{/if}
{#if model.isImageModel}
<svg
class="w-3.5 h-3.5 flex-shrink-0 text-exo-yellow"
fill="none"
viewBox="0 0 24 24"
stroke="currentColor"
stroke-width="2"
aria-label="Image generation model"
>
<rect
x="3"
y="3"
width="18"
height="18"
rx="2"
ry="2"
/>
<circle cx="8.5" cy="8.5" r="1.5" />
<polyline points="21 15 16 10 5 21" />
</svg>
{/if}
<span class="truncate flex-1">{model.label}</span>
</button>
{/each}
</div>
</div>
{/if}
</div>
<!-- Thinking toggle -->
{#if modelSupportsThinking()}
<button
type="button"
onclick={() => setConversationThinking(!thinkingEnabled)}
class="flex items-center gap-1.5 px-2 py-1 rounded text-xs font-mono tracking-wide transition-all duration-200 flex-shrink-0 cursor-pointer border {thinkingEnabled
? 'bg-exo-yellow/15 border-exo-yellow/40 text-exo-yellow'
: 'bg-exo-medium-gray/30 border-exo-medium-gray/50 text-exo-light-gray/60 hover:text-exo-light-gray'}"
title={thinkingEnabled
? "Thinking enabled — click to disable"
: "Thinking disabled — click to enable"}
>
<svg
class="w-3.5 h-3.5"
viewBox="0 0 24 24"
fill="none"
stroke="currentColor"
stroke-width="1.5"
>
<path
d="M12 2a7 7 0 0 0-7 7c0 2.38 1.19 4.47 3 5.74V17a1 1 0 0 0 1 1h6a1 1 0 0 0 1-1v-2.26c1.81-1.27 3-3.36 3-5.74a7 7 0 0 0-7-7zM9 20h6M10 22h4"
stroke-linecap="round"
stroke-linejoin="round"
/>
</svg>
<span>{thinkingEnabled ? "THINK" : "NO THINK"}</span>
</button>
{/if}
<!-- Performance stats -->
{#if currentTtft !== null || currentTps !== null}
<div class="flex items-center gap-4 text-xs font-mono flex-shrink-0">
@@ -596,75 +518,97 @@
: isImageModel()
? "Describe the image you want to generate..."
: placeholder}
disabled={loading}
rows={1}
class="flex-1 resize-none bg-transparent text-foreground placeholder:text-exo-light-gray/60 placeholder:text-sm placeholder:tracking-[0.15em] placeholder:leading-7 focus:outline-none focus:ring-0 focus:border-none disabled:opacity-50 text-sm leading-7 font-mono"
class="flex-1 resize-none bg-transparent text-foreground placeholder:text-exo-light-gray/60 placeholder:text-sm placeholder:tracking-[0.15em] placeholder:leading-7 focus:outline-none focus:ring-0 focus:border-none text-sm leading-7 font-mono"
style="min-height: 28px; max-height: 150px;"
></textarea>
<button
type="submit"
disabled={!canSend || loading || isEditOnlyWithoutImage}
class="px-2.5 sm:px-4 py-1.5 sm:py-2 rounded text-xs sm:text-xs tracking-[0.1em] sm:tracking-[0.15em] uppercase font-medium transition-all duration-200 whitespace-nowrap
{!canSend || loading || isEditOnlyWithoutImage
? 'bg-exo-medium-gray/50 text-exo-light-gray cursor-not-allowed'
: 'bg-exo-yellow text-exo-black hover:bg-exo-yellow-darker hover:shadow-[0_0_20px_rgba(255,215,0,0.3)]'}"
aria-label={shouldShowEditMode
? "Edit image"
: isImageModel()
? "Generate image"
: "Send message"}
>
{#if loading}
{#if loading}
<button
type="button"
onclick={() => stopGeneration()}
class="px-2.5 sm:px-4 py-1.5 sm:py-2 rounded text-xs sm:text-xs tracking-[0.1em] sm:tracking-[0.15em] font-medium transition-all duration-200 whitespace-nowrap bg-exo-medium-gray/70 text-exo-light-gray hover:bg-exo-medium-gray hover:text-white"
aria-label="Stop generation"
>
<span class="inline-flex items-center gap-1 sm:gap-2">
<span
class="w-2.5 h-2.5 sm:w-3 sm:h-3 border-2 border-current border-t-transparent rounded-full animate-spin"
></span>
<span class="hidden sm:inline"
>{shouldShowEditMode
? "EDITING"
: isImageModel()
? "GENERATING"
: "PROCESSING"}</span
>
<span class="sm:hidden">...</span>
</span>
{:else if shouldShowEditMode}
<span class="inline-flex items-center gap-1.5">
<svg
class="w-3.5 h-3.5"
fill="none"
class="w-3 h-3 sm:w-3.5 sm:h-3.5"
fill="currentColor"
viewBox="0 0 24 24"
stroke="currentColor"
stroke-width="2"
>
<path
stroke-linecap="round"
stroke-linejoin="round"
d="M11 5H6a2 2 0 00-2 2v11a2 2 0 002 2h11a2 2 0 002-2v-5m-1.414-9.414a2 2 0 112.828 2.828L11.828 15H9v-2.828l8.586-8.586z"
/>
<rect x="6" y="6" width="12" height="12" rx="1" />
</svg>
<span>EDIT</span>
<span class="hidden sm:inline">Cancel</span>
</span>
{:else if isImageModel()}
<span class="inline-flex items-center gap-1.5">
<svg
class="w-3.5 h-3.5"
fill="none"
viewBox="0 0 24 24"
stroke="currentColor"
stroke-width="2"
>
<rect x="3" y="3" width="18" height="18" rx="2" ry="2" />
<circle cx="8.5" cy="8.5" r="1.5" />
<polyline points="21 15 16 10 5 21" />
</svg>
<span>GENERATE</span>
</span>
{:else}
SEND
{/if}
</button>
</button>
{:else}
<button
type="submit"
disabled={!canSend || isEditOnlyWithoutImage}
class="px-2.5 sm:px-4 py-1.5 sm:py-2 rounded text-xs sm:text-xs tracking-[0.1em] sm:tracking-[0.15em] uppercase font-medium transition-all duration-200 whitespace-nowrap
{!canSend || isEditOnlyWithoutImage
? 'bg-exo-medium-gray/50 text-exo-light-gray cursor-not-allowed'
: 'bg-exo-yellow text-exo-black hover:bg-exo-yellow-darker hover:shadow-[0_0_20px_rgba(255,215,0,0.3)]'}"
aria-label={shouldShowEditMode
? "Edit image"
: isImageModel()
? "Generate image"
: "Send message"}
>
{#if shouldShowEditMode}
<span class="inline-flex items-center gap-1.5">
<svg
class="w-3.5 h-3.5"
fill="none"
viewBox="0 0 24 24"
stroke="currentColor"
stroke-width="2"
>
<path
stroke-linecap="round"
stroke-linejoin="round"
d="M11 5H6a2 2 0 00-2 2v11a2 2 0 002 2h11a2 2 0 002-2v-5m-1.414-9.414a2 2 0 112.828 2.828L11.828 15H9v-2.828l8.586-8.586z"
/>
</svg>
<span>EDIT</span>
</span>
{:else if isEditOnlyWithoutImage}
<span class="inline-flex items-center gap-1.5">
<svg
class="w-3.5 h-3.5"
fill="none"
viewBox="0 0 24 24"
stroke="currentColor"
stroke-width="2"
>
<path
stroke-linecap="round"
stroke-linejoin="round"
d="M11 5H6a2 2 0 00-2 2v11a2 2 0 002 2h11a2 2 0 002-2v-5m-1.414-9.414a2 2 0 112.828 2.828L11.828 15H9v-2.828l8.586-8.586z"
/>
</svg>
<span>EDIT</span>
</span>
{:else if isImageModel()}
<span class="inline-flex items-center gap-1.5">
<svg
class="w-3.5 h-3.5"
fill="none"
viewBox="0 0 24 24"
stroke="currentColor"
stroke-width="2"
>
<rect x="3" y="3" width="18" height="18" rx="2" ry="2" />
<circle cx="8.5" cy="8.5" r="1.5" />
<polyline points="21 15 16 10 5 21" />
</svg>
<span>GENERATE</span>
</span>
{:else}
SEND
{/if}
</button>
{/if}
</div>
<!-- Bottom accent line -->
+133 -18
View File
@@ -3,14 +3,18 @@
messages,
currentResponse,
isLoading,
prefillProgress,
deleteMessage,
editAndRegenerate,
regenerateLastResponse,
regenerateFromToken,
setEditingImage,
} from "$lib/stores/app.svelte";
import type { Message } from "$lib/stores/app.svelte";
import type { MessageAttachment } from "$lib/stores/app.svelte";
import MarkdownContent from "./MarkdownContent.svelte";
import TokenHeatmap from "./TokenHeatmap.svelte";
import PrefillProgressBar from "./PrefillProgressBar.svelte";
import ImageLightbox from "./ImageLightbox.svelte";
interface Props {
class?: string;
@@ -22,6 +26,7 @@
const messageList = $derived(messages());
const response = $derived(currentResponse());
const loading = $derived(isLoading());
const prefill = $derived(prefillProgress());
// Scroll management - user controls scroll, show button when not at bottom
const SCROLL_THRESHOLD = 100;
@@ -99,6 +104,26 @@
let copiedMessageId = $state<string | null>(null);
let expandedThinkingMessageIds = $state<Set<string>>(new Set());
// Lightbox state
let expandedImageSrc = $state<string | null>(null);
// Uncertainty heatmap toggle
let heatmapMessageIds = $state<Set<string>>(new Set());
function toggleHeatmap(messageId: string) {
const next = new Set(heatmapMessageIds);
if (next.has(messageId)) {
next.delete(messageId);
} else {
next.add(messageId);
}
heatmapMessageIds = next;
}
function isHeatmapVisible(messageId: string): boolean {
return heatmapMessageIds.has(messageId);
}
function formatTimestamp(timestamp: number): string {
return new Date(timestamp).toLocaleTimeString("en-US", {
hour12: false,
@@ -202,6 +227,7 @@
}
function handleDeleteClick(messageId: string) {
if (loading) return;
deleteConfirmId = messageId;
}
@@ -232,7 +258,7 @@
</script>
<div class="flex flex-col gap-4 sm:gap-6 {className}">
{#each messageList as message (message.id)}
{#each messageList as message, i (message.id)}
<div
class="group flex {message.role === 'user'
? 'justify-end'
@@ -294,9 +320,11 @@
<!-- Delete confirmation -->
<div class="bg-red-500/10 border border-red-500/30 rounded-lg p-3">
<p class="text-xs text-red-400 mb-3">
Delete this message{message.role === "user"
? " and all responses after it"
: ""}?
{#if i === messageList.length - 1}
Delete this message?
{:else}
Delete this message and all messages after it?
{/if}
</p>
<div class="flex gap-2 justify-end">
<button
@@ -370,10 +398,15 @@
class="flex items-center gap-2 bg-exo-dark-gray/60 border border-exo-yellow/20 rounded px-2 py-1 text-xs font-mono"
>
{#if attachment.type === "image" && attachment.preview}
<!-- svelte-ignore a11y_no_noninteractive_element_interactions, a11y_click_events_have_key_events -->
<img
src={attachment.preview}
alt={attachment.name}
class="w-12 h-12 object-cover rounded border border-exo-yellow/20"
class="w-12 h-12 object-cover rounded border border-exo-yellow/20 cursor-pointer hover:border-exo-yellow/50 transition-colors"
onclick={() => {
if (attachment.preview)
expandedImageSrc = attachment.preview;
}}
/>
{:else}
<span>{getAttachmentIcon(attachment)}</span>
@@ -397,6 +430,9 @@
{:else}
<!-- Assistant message styling -->
<div class="p-3 sm:p-4">
{#if loading && isLastAssistantMessage(message.id) && prefill && !message.content}
<PrefillProgressBar progress={prefill} class="mb-3" />
{/if}
{#if message.thinking && message.thinking.trim().length > 0}
<div
class="mb-3 rounded border border-exo-yellow/20 bg-exo-black/40"
@@ -447,15 +483,44 @@
<div class="mb-3">
{#each message.attachments.filter((a) => a.type === "generated-image") as attachment}
<div class="relative group/img inline-block">
<!-- svelte-ignore a11y_no_noninteractive_element_interactions, a11y_click_events_have_key_events -->
<img
src={attachment.preview}
alt=""
class="max-w-full max-h-[512px] rounded-lg border border-exo-yellow/20 shadow-lg shadow-black/20"
class="max-w-full max-h-[512px] rounded-lg border border-exo-yellow/20 shadow-lg shadow-black/20 cursor-pointer"
onclick={() => {
if (attachment.preview)
expandedImageSrc = attachment.preview;
}}
/>
<!-- Button overlay -->
<div
class="absolute top-2 right-2 flex gap-1 opacity-0 group-hover/img:opacity-100 transition-opacity"
>
<!-- Expand button -->
<button
type="button"
class="p-2 rounded-lg bg-exo-dark-gray/80 border border-exo-yellow/30 text-exo-yellow hover:bg-exo-dark-gray hover:border-exo-yellow/50 cursor-pointer"
onclick={() => {
if (attachment.preview)
expandedImageSrc = attachment.preview;
}}
title="Expand image"
>
<svg
class="w-4 h-4"
fill="none"
viewBox="0 0 24 24"
stroke="currentColor"
stroke-width="2"
>
<path
stroke-linecap="round"
stroke-linejoin="round"
d="M4 8V4m0 0h4M4 4l5 5m11-1V4m0 0h-4m4 0l-5 5M4 16v4m0 0h4m-4 0l5-5m11 5l-5-5m5 5v-4m0 4h-4"
/>
</svg>
</button>
<!-- Edit button -->
<button
type="button"
@@ -548,13 +613,23 @@
>
</div>
{:else if message.content || (loading && !message.attachments?.some((a) => a.type === "generated-image"))}
<MarkdownContent
content={message.content || (loading ? response : "")}
/>
{#if loading && !message.content}
<span
class="inline-block w-2 h-4 bg-exo-yellow/70 ml-1 cursor-blink"
></span>
{#if isHeatmapVisible(message.id) && message.tokens && message.tokens.length > 0}
<TokenHeatmap
tokens={message.tokens}
isGenerating={loading &&
isLastAssistantMessage(message.id)}
onRegenerateFrom={(tokenIndex) =>
regenerateFromToken(message.id, tokenIndex)}
/>
{:else}
<MarkdownContent
content={message.content || (loading ? response : "")}
/>
{#if loading && !message.content}
<span
class="inline-block w-2 h-4 bg-exo-yellow/70 ml-1 cursor-blink"
></span>
{/if}
{/if}
{/if}
</div>
@@ -629,6 +704,35 @@
</button>
{/if}
<!-- Uncertainty heatmap toggle (assistant messages with tokens) -->
{#if message.role === "assistant" && message.tokens && message.tokens.length > 0}
<button
onclick={() => toggleHeatmap(message.id)}
class="p-1.5 transition-colors rounded cursor-pointer {isHeatmapVisible(
message.id,
)
? 'text-exo-yellow'
: 'text-exo-light-gray hover:text-exo-yellow'}"
title={isHeatmapVisible(message.id)
? "Hide uncertainty heatmap"
: "Show uncertainty heatmap"}
>
<svg
class="w-3.5 h-3.5"
fill="none"
viewBox="0 0 24 24"
stroke="currentColor"
>
<path
stroke-linecap="round"
stroke-linejoin="round"
stroke-width="2"
d="M9 19v-6a2 2 0 00-2-2H5a2 2 0 00-2 2v6a2 2 0 002 2h2a2 2 0 002-2zm0 0V9a2 2 0 012-2h2a2 2 0 012 2v10m-6 0a2 2 0 002 2h2a2 2 0 002-2m0 0V5a2 2 0 012-2h2a2 2 0 012 2v14a2 2 0 01-2 2h-2a2 2 0 01-2-2z"
/>
</svg>
</button>
{/if}
<!-- Regenerate button (last assistant message only) -->
{#if message.role === "assistant" && isLastAssistantMessage(message.id) && !loading}
<button
@@ -655,8 +759,13 @@
<!-- Delete button -->
<button
onclick={() => handleDeleteClick(message.id)}
class="p-1.5 text-exo-light-gray hover:text-red-400 transition-colors rounded hover:bg-red-500/10 cursor-pointer"
title="Delete message"
disabled={loading}
class="p-1.5 transition-colors rounded {loading
? 'text-exo-light-gray/30 cursor-not-allowed'
: 'text-exo-light-gray hover:text-red-400 hover:bg-red-500/10 cursor-pointer'}"
title={loading
? "Cannot delete while generating"
: "Delete message"}
>
<svg
class="w-3.5 h-3.5"
@@ -698,8 +807,8 @@
>
AWAITING INPUT
</p>
<p class="text-sm sm:text-xs text-exo-light-gray tracking-wider mt-1">
ENTER A QUERY TO BEGIN
<p class="text-xs text-white/30 tracking-wider mt-1.5 font-mono">
Type a message below &middot; Shift+Enter for newline
</p>
</div>
{/if}
@@ -714,6 +823,7 @@
onclick={scrollToBottom}
class="sticky bottom-4 left-1/2 -translate-x-1/2 w-10 h-10 rounded-full bg-exo-dark-gray/90 border border-exo-medium-gray/50 flex items-center justify-center text-exo-light-gray hover:text-exo-yellow hover:border-exo-yellow/50 transition-all shadow-lg cursor-pointer z-10"
title="Scroll to bottom"
aria-label="Scroll to bottom of messages"
>
<svg
class="w-5 h-5"
@@ -731,3 +841,8 @@
</button>
{/if}
</div>
<ImageLightbox
src={expandedImageSrc}
onclose={() => (expandedImageSrc = null)}
/>
@@ -0,0 +1,401 @@
<script lang="ts" module>
export interface ChatModelInfo {
id: string;
name: string;
base_model: string;
storage_size_megabytes: number;
capabilities: string[];
family: string;
quantization: string;
}
// Auto mode tier list (for when user just starts typing)
export const AUTO_TIERS: string[][] = [
// Tier 1 (frontier)
["DeepSeek V3.1", "GLM-5", "Kimi K2.5", "Qwen3 Coder Next"],
// Tier 2 (excellent)
[
"Kimi K2",
"Qwen3 235B",
"MiniMax M2.5",
"Step 3.5 Flash",
"Qwen3 Next 80B",
],
// Tier 3 (great)
[
"GLM 4.7",
"MiniMax M2.1",
"Qwen3 Coder 480B",
"GLM 4.5 Air",
"Llama 3.3 70B",
],
// Tier 4 (good)
["GPT-OSS 120B", "Qwen3 30B", "Llama 3.1 70B", "GLM 4.7 Flash"],
// Tier 5 (small/fast)
[
"Llama 3.1 8B",
"GPT-OSS 20B",
"Llama 3.2 3B",
"Qwen3 0.6B",
"Llama 3.2 1B",
],
];
/** Return the tier index (0 = best) for a base_model name. */
export function getAutoTierIndex(baseModel: string): number {
for (let i = 0; i < AUTO_TIERS.length; i++) {
if (AUTO_TIERS[i].includes(baseModel)) return i;
}
return AUTO_TIERS.length; // not in any tier → lowest priority
}
/** Auto mode: walk tiers top-down, pick biggest fitting variant from highest tier. */
export function pickAutoModel(
modelList: ChatModelInfo[],
memoryGB: number,
): ChatModelInfo | null {
for (const tier of AUTO_TIERS) {
const candidates: ChatModelInfo[] = [];
for (const baseModel of tier) {
const variants = modelList
.filter(
(m) =>
m.base_model === baseModel &&
(m.storage_size_megabytes || 0) / 1024 <= memoryGB &&
(m.storage_size_megabytes || 0) > 0,
)
.sort(
(a, b) =>
(b.storage_size_megabytes || 0) - (a.storage_size_megabytes || 0),
);
if (variants[0]) candidates.push(variants[0]);
}
if (candidates.length > 0) {
candidates.sort(
(a, b) =>
(b.storage_size_megabytes || 0) - (a.storage_size_megabytes || 0),
);
return candidates[0];
}
}
return null;
}
</script>
<script lang="ts">
interface CategoryRecommendation {
category: "coding" | "writing" | "agentic" | "biggest";
label: string;
model: ChatModelInfo | null;
tooltip: string;
}
interface Props {
models: ChatModelInfo[];
clusterLabel: string;
totalMemoryGB: number;
onSelect: (modelId: string, category: string) => void;
onAddModel: () => void;
class?: string;
}
let {
models,
clusterLabel,
totalMemoryGB,
onSelect,
onAddModel,
class: className = "",
}: Props = $props();
// --- Hardcoded Rankings ---
const CODING_RANKING = [
"Qwen3 Coder Next",
"Qwen3 Coder 480B",
"Qwen3 30B",
"GPT-OSS 20B",
"Llama 3.1 8B",
"Llama 3.2 3B",
"Qwen3 0.6B",
];
const WRITING_RANKING = [
"Kimi K2.5",
"Kimi K2",
"Qwen3 Next 80B",
"Llama 3.3 70B",
"MiniMax M2.5",
"GLM 4.5 Air",
"GLM 4.7 Flash",
"GPT-OSS 20B",
"Llama 3.1 8B",
"Llama 3.2 3B",
"Qwen3 0.6B",
];
const AGENTIC_RANKING = [
"DeepSeek V3.1",
"GLM-5",
"Qwen3 235B",
"Step 3.5 Flash",
"GLM 4.7",
"MiniMax M2.1",
"GPT-OSS 120B",
"Llama 3.3 70B",
"Llama 3.1 70B",
"GLM 4.7 Flash",
"GPT-OSS 20B",
"Qwen3 30B",
"Llama 3.1 8B",
"Llama 3.2 3B",
"Qwen3 0.6B",
];
function getModelSizeGB(m: ChatModelInfo): number {
return (m.storage_size_megabytes || 0) / 1024;
}
function fitsInMemory(m: ChatModelInfo): boolean {
return getModelSizeGB(m) <= totalMemoryGB && getModelSizeGB(m) > 0;
}
/** For a given base_model name, find the biggest quant variant that fits in memory. */
function pickBestVariant(baseModel: string): ChatModelInfo | null {
const variants = models
.filter((m) => m.base_model === baseModel && fitsInMemory(m))
.sort((a, b) => getModelSizeGB(b) - getModelSizeGB(a));
return variants[0] ?? null;
}
/** Walk a ranked list of base_model names, return the first that has a fitting variant. */
function pickFromRanking(ranking: string[]): ChatModelInfo | null {
for (const baseModel of ranking) {
const pick = pickBestVariant(baseModel);
if (pick) return pick;
}
return null;
}
/** Pick the single biggest model that fits. */
function pickBiggest(): ChatModelInfo | null {
const fitting = models
.filter((m) => fitsInMemory(m))
.sort((a, b) => getModelSizeGB(b) - getModelSizeGB(a));
return fitting[0] ?? null;
}
const recommendations = $derived.by((): CategoryRecommendation[] => {
return [
{
category: "coding",
label: "Best for Coding",
model: pickFromRanking(CODING_RANKING),
tooltip:
"Ranked by coding benchmark performance (LiveCodeBench, SWE-bench)",
},
{
category: "writing",
label: "Best for Writing",
model: pickFromRanking(WRITING_RANKING),
tooltip: "Ranked by creative writing quality and instruction following",
},
{
category: "agentic",
label: "Best Agentic",
model: pickFromRanking(AGENTIC_RANKING),
tooltip: "Ranked by reasoning, planning, and tool-use capability",
},
{
category: "biggest",
label: "Biggest",
model: pickBiggest(),
tooltip: "Largest model that fits in your available memory",
},
];
});
function formatSize(mb: number): string {
const gb = mb / 1024;
if (gb >= 100) return `${Math.round(gb)} GB`;
return `${gb.toFixed(1)} GB`;
}
// Category icons (SVG paths)
const categoryIcons: Record<string, string> = {
coding:
"M8 9l3 3-3 3m5 0h3M5 20h14a2 2 0 002-2V6a2 2 0 00-2-2H5a2 2 0 00-2 2v12a2 2 0 002 2z",
writing:
"M11 5H6a2 2 0 00-2 2v11a2 2 0 002 2h11a2 2 0 002-2v-5m-1.414-9.414a2 2 0 112.828 2.828L11.828 15H9v-2.828l8.586-8.586z",
agentic:
"M9.663 17h4.673M12 3v1m6.364 1.636l-.707.707M21 12h-1M4 12H3m3.343-5.657l-.707-.707m2.828 9.9a5 5 0 117.072 0l-.548.547A3.374 3.374 0 0014 18.469V19a2 2 0 11-4 0v-.531c0-.895-.356-1.754-.988-2.386l-.548-.547z",
biggest:
"M19 11H5m14 0a2 2 0 012 2v6a2 2 0 01-2 2H5a2 2 0 01-2-2v-6a2 2 0 012-2m14 0V9a2 2 0 00-2-2M5 11V9a2 2 0 012-2m0 0V5a2 2 0 012-2h6a2 2 0 012 2v2M7 7h10",
};
let hoveredTooltip = $state<string | null>(null);
let tooltipAnchor = $state<{ x: number; y: number } | null>(null);
function showTooltip(category: string, e: MouseEvent | FocusEvent) {
hoveredTooltip = category;
const target = e.currentTarget as HTMLElement;
const rect = target.getBoundingClientRect();
tooltipAnchor = { x: rect.left + rect.width / 2, y: rect.top };
}
function hideTooltip() {
hoveredTooltip = null;
tooltipAnchor = null;
}
</script>
<div class="flex flex-col items-center justify-center gap-6 {className}">
<!-- Header -->
<div class="text-center">
<p class="text-xs text-exo-light-gray uppercase tracking-[0.2em] mb-1">
Recommended for your
</p>
<p class="text-sm text-white font-mono tracking-wide">{clusterLabel}</p>
</div>
<!-- Category Cards Grid -->
<div class="grid grid-cols-2 gap-3 w-full max-w-md">
{#each recommendations as rec}
{#if rec.model}
<button
type="button"
onclick={() => rec.model && onSelect(rec.model.id, rec.category)}
class="group relative flex flex-col items-start gap-2 p-4 rounded-lg border border-exo-medium-gray/50 bg-exo-dark-gray/50 hover:border-exo-yellow/40 hover:bg-exo-dark-gray transition-all duration-200 cursor-pointer text-left"
>
<!-- Category icon + label -->
<div class="flex items-center gap-2 w-full">
<svg
class="w-4 h-4 text-exo-yellow/70 group-hover:text-exo-yellow transition-colors flex-shrink-0"
fill="none"
viewBox="0 0 24 24"
stroke="currentColor"
stroke-width="1.5"
>
<path
stroke-linecap="round"
stroke-linejoin="round"
d={categoryIcons[rec.category]}
/>
</svg>
<span
class="text-xs font-mono uppercase tracking-wider text-exo-light-gray group-hover:text-white transition-colors"
>
{rec.label}
</span>
<!-- Info tooltip -->
<div class="ml-auto flex-shrink-0">
<span
role="button"
tabindex="-1"
class="text-exo-light-gray/40 hover:text-exo-light-gray transition-colors cursor-help inline-flex"
onmouseenter={(e: MouseEvent) => showTooltip(rec.category, e)}
onmouseleave={() => hideTooltip()}
onclick={(e: MouseEvent) => {
e.stopPropagation();
if (hoveredTooltip === rec.category) {
hideTooltip();
} else {
showTooltip(rec.category, e);
}
}}
>
<svg
class="w-3.5 h-3.5"
fill="none"
viewBox="0 0 24 24"
stroke="currentColor"
stroke-width="2"
>
<circle cx="12" cy="12" r="10" />
<path d="M12 16v-4m0-4h.01" />
</svg>
</span>
</div>
</div>
<!-- Model name + size -->
<div class="w-full">
<p class="text-sm text-white font-mono truncate">
{rec.model.base_model}
</p>
<p class="text-xs text-exo-light-gray/60 font-mono mt-0.5">
{formatSize(rec.model.storage_size_megabytes)}
{#if rec.model.quantization}
<span class="text-exo-light-gray/40"
>&middot; {rec.model.quantization}</span
>
{/if}
</p>
</div>
</button>
{:else}
<!-- No model fits for this category -->
<div
class="flex flex-col items-start gap-2 p-4 rounded-lg border border-exo-medium-gray/30 bg-exo-dark-gray/30 opacity-50"
>
<div class="flex items-center gap-2">
<svg
class="w-4 h-4 text-exo-light-gray/40 flex-shrink-0"
fill="none"
viewBox="0 0 24 24"
stroke="currentColor"
stroke-width="1.5"
>
<path
stroke-linecap="round"
stroke-linejoin="round"
d={categoryIcons[rec.category]}
/>
</svg>
<span
class="text-xs font-mono uppercase tracking-wider text-exo-light-gray/50"
>{rec.label}</span
>
</div>
<p class="text-xs text-exo-light-gray/40 font-mono">No model fits</p>
</div>
{/if}
{/each}
</div>
<!-- Add Model Button -->
<button
type="button"
onclick={onAddModel}
class="flex items-center gap-2 px-4 py-2 text-xs font-mono uppercase tracking-wider text-exo-light-gray hover:text-exo-yellow border border-exo-medium-gray/30 hover:border-exo-yellow/30 rounded-lg transition-all duration-200 cursor-pointer"
>
<svg
class="w-3.5 h-3.5"
fill="none"
viewBox="0 0 24 24"
stroke="currentColor"
stroke-width="2"
>
<path stroke-linecap="round" stroke-linejoin="round" d="M12 4v16m8-8H4" />
</svg>
Add Model
</button>
<!-- Auto hint -->
<p class="text-xs text-exo-light-gray/40 font-mono tracking-wide text-center">
Or just start typing &mdash; we'll pick the best model automatically
</p>
</div>
<!-- Fixed-position tooltip (escapes overflow-hidden ancestors) -->
{#if hoveredTooltip && tooltipAnchor}
{@const rec = recommendations.find((r) => r.category === hoveredTooltip)}
{#if rec}
<div
class="fixed z-[9999] px-3 py-2 bg-exo-black border border-exo-medium-gray/50 rounded text-xs text-exo-light-gray whitespace-nowrap shadow-lg pointer-events-none"
style="left: {tooltipAnchor.x}px; top: {tooltipAnchor.y -
8}px; transform: translate(-50%, -100%);"
>
{rec.tooltip}
</div>
{/if}
{/if}
+29 -29
View File
@@ -2,7 +2,6 @@
import {
conversations,
activeConversationId,
createConversation,
loadConversation,
deleteConversation,
deleteAllConversations,
@@ -17,9 +16,15 @@
interface Props {
class?: string;
onNewChat?: () => void;
onSelectConversation?: () => void;
}
let { class: className = "" }: Props = $props();
let {
class: className = "",
onNewChat,
onSelectConversation,
}: Props = $props();
const conversationList = $derived(conversations());
const activeId = $derived(activeConversationId());
@@ -42,10 +47,11 @@
);
function handleNewChat() {
createConversation();
onNewChat?.();
}
function handleSelectConversation(id: string) {
onSelectConversation?.();
loadConversation(id);
}
@@ -185,11 +191,7 @@
let instanceType: string | null = null;
if (instanceTag === "MlxRingInstance") instanceType = "MLX Ring";
else if (
instanceTag === "MlxIbvInstance" ||
instanceTag === "MlxJacclInstance"
)
instanceType = "MLX RDMA";
else if (instanceTag === "MlxJacclInstance") instanceType = "MLX RDMA";
let sharding: string | null = null;
const inst = instance as {
@@ -307,7 +309,7 @@
<div class="py-2">
<div class="px-4 py-2">
<span
class="text-sm text-white/70 font-mono tracking-wider uppercase"
class="text-xs text-exo-light-gray font-mono tracking-wider uppercase"
>
{searchQuery ? "SEARCH RESULTS" : "CONVERSATIONS"}
</span>
@@ -376,39 +378,37 @@
onkeydown={(e) =>
e.key === "Enter" &&
handleSelectConversation(conversation.id)}
class="group w-full flex items-center justify-between p-2 rounded mb-1 transition-all text-left cursor-pointer
class="group w-full flex items-center justify-between p-2.5 rounded-lg mb-1 transition-all text-left cursor-pointer
{activeId === conversation.id
? 'bg-transparent border border-exo-yellow/30'
: 'hover:border-exo-yellow/20 border border-transparent'}"
? 'bg-exo-yellow/5 border border-exo-yellow/30'
: 'hover:bg-white/[0.03] hover:border-white/10 border border-transparent'}"
>
<div class="flex-1 min-w-0 pr-2">
<div
class="text-sm truncate {activeId === conversation.id
class="text-sm font-medium truncate {activeId ===
conversation.id
? 'text-exo-yellow'
: 'text-white/90'}"
: 'text-white'}"
>
{conversation.name}
</div>
<div class="text-sm text-white/50 mt-0.5">
<div class="text-xs text-white/60 mt-0.5">
{formatDate(conversation.updatedAt)}
</div>
<div class="text-sm text-white/70 truncate">
<div class="text-xs text-exo-light-gray truncate">
{info.modelLabel}
</div>
<div class="text-xs text-white/60 font-mono">
Strategy: <span class="text-white/80"
>{info.strategyLabel}</span
>
</div>
{#if stats}
<div class="text-xs text-white/60 font-mono mt-1">
{#if stats.ttftMs}<span class="text-white/40">TTFT</span>
{stats.ttftMs.toFixed(
0,
)}ms{/if}{#if stats.ttftMs && stats.tps}<span
class="text-white/30 mx-1.5"></span
>{/if}{#if stats.tps}{stats.tps.toFixed(1)}
<span class="text-white/40">tok/s</span>{/if}
<div class="text-xs text-white/70 font-mono mt-1">
{#if stats.ttftMs}<span class="text-white/50">TTFT</span>
<span class="text-exo-yellow/80"
>{stats.ttftMs.toFixed(0)}ms</span
>{/if}{#if stats.ttftMs && stats.tps}<span
class="text-white/30 mx-1.5">·</span
>{/if}{#if stats.tps}<span class="text-exo-yellow/80"
>{stats.tps.toFixed(1)}</span
>
<span class="text-white/50">tok/s</span>{/if}
</div>
{/if}
</div>
@@ -0,0 +1,20 @@
<script lang="ts">
import { isConnected } from "$lib/stores/app.svelte";
import { slide } from "svelte/transition";
const connected = $derived(isConnected());
</script>
{#if !connected}
<div
transition:slide={{ duration: 200 }}
class="relative z-50 flex items-center justify-center gap-2 px-4 py-2 bg-red-950/80 border-b border-red-500/30"
role="alert"
aria-live="assertive"
>
<div class="w-2 h-2 bg-red-500 rounded-full animate-pulse"></div>
<span class="text-xs font-mono text-red-300 tracking-wider uppercase">
Connection lost &mdash; Reconnecting to backend&hellip;
</span>
</div>
{/if}
@@ -0,0 +1,277 @@
<script lang="ts">
/**
* DeviceIcon — renders a device icon as an SVG <g> element.
* Uses the exact same proportional math as TopologyGraph.svelte
* so that devices look identical in both the topology view and
* the onboarding animation.
*
* Must be placed inside an <svg> element.
*/
interface Props {
/** "macbook pro" | "mac studio" | "mac mini" etc. */
deviceType: string;
/** Center X coordinate in SVG space */
cx: number;
/** Center Y coordinate in SVG space */
cy: number;
/** Base sizing factor (equivalent to TopologyGraph's nodeRadius) */
size?: number;
/** RAM usage 0100 */
ramPercent?: number;
/** Unique id suffix for clip-path ids */
uid?: string;
}
let {
deviceType,
cx,
cy,
size = 60,
ramPercent = 60,
uid = "dev",
}: Props = $props();
// Apple logo path — same constant used by TopologyGraph
const APPLE_LOGO_PATH =
"M788.1 340.9c-5.8 4.5-108.2 62.2-108.2 190.5 0 148.4 130.3 200.9 134.2 202.2-.6 3.2-20.7 71.9-68.7 141.9-42.8 61.6-87.5 123.1-155.5 123.1s-85.5-39.5-164-39.5c-76.5 0-103.7 40.8-165.9 40.8s-105.6-57-155.5-127C46.7 790.7 0 663 0 541.8c0-194.4 126.4-297.5 250.8-297.5 66.1 0 121.2 43.4 162.7 43.4 39.5 0 101.1-46 176.3-46 28.5 0 130.9 2.6 198.3 99.2zm-234-181.5c31.1-36.9 53.1-88.1 53.1-139.3 0-7.1-.6-14.3-1.9-20.1-50.6 1.9-110.8 33.7-147.1 75.8-28.5 32.4-55.1 83.6-55.1 135.5 0 7.8 1.3 15.6 1.9 18.1 3.2.6 8.4 1.3 13.6 1.3 45.4 0 102.5-30.4 135.5-71.3z";
const LOGO_NATIVE_WIDTH = 814;
const LOGO_NATIVE_HEIGHT = 1000;
const wireColor = "rgba(179,179,179,0.8)";
const strokeWidth = 1.5;
const modelLower = $derived(deviceType.toLowerCase());
// ── Mac Studio dimensions (same ratios as TopologyGraph) ──
const studioW = $derived(size * 1.25);
const studioH = $derived(size * 0.85);
const studioX = $derived(cx - studioW / 2);
const studioY = $derived(cy - studioH / 2);
const studioCorner = 4;
const studioTopH = $derived(studioH * 0.15);
// Studio front panel details
const studioSlotH = $derived(studioH * 0.14);
const studioVSlotW = $derived(studioW * 0.05);
const studioVSlotY = $derived(
studioY + studioTopH + (studioH - studioTopH) * 0.6,
);
const studioVSlot1X = $derived(studioX + studioW * 0.18);
const studioVSlot2X = $derived(studioX + studioW * 0.28);
const studioHSlotW = $derived(studioW * 0.2);
const studioHSlotX = $derived(studioX + studioW * 0.5 - studioHSlotW / 2);
// Studio memory fill
const studioMemTotalH = $derived(studioH - studioTopH);
const studioMemH = $derived((ramPercent / 100) * studioMemTotalH);
// ── MacBook dimensions (same ratios as TopologyGraph) ──
const mbW = $derived((size * 1.6 * 0.85) / 1.15);
const mbH = $derived(size * 0.85);
const mbX = $derived(cx - mbW / 2);
const mbY = $derived(cy - mbH / 2);
const mbScreenH = $derived(mbH * 0.7);
const mbBaseH = $derived(mbH * 0.3);
const mbScreenW = $derived(mbW * 0.85);
const mbScreenX = $derived(cx - mbScreenW / 2);
const mbBezel = 3;
// MacBook memory fill
const mbMemTotalH = $derived(mbScreenH - mbBezel * 2);
const mbMemH = $derived((ramPercent / 100) * mbMemTotalH);
// Apple logo sizing
const mbLogoTargetH = $derived(mbScreenH * 0.22);
const mbLogoScale = $derived(mbLogoTargetH / LOGO_NATIVE_HEIGHT);
const mbLogoX = $derived(cx - (LOGO_NATIVE_WIDTH * mbLogoScale) / 2);
const mbLogoY = $derived(
mbY + mbScreenH / 2 - (LOGO_NATIVE_HEIGHT * mbLogoScale) / 2,
);
// MacBook base (trapezoidal)
const mbBaseY = $derived(mbY + mbScreenH);
const mbBaseTopW = $derived(mbScreenW);
const mbBaseBottomW = $derived(mbW);
const mbBaseTopX = $derived(cx - mbBaseTopW / 2);
const mbBaseBottomX = $derived(cx - mbBaseBottomW / 2);
// Keyboard
const mbKbX = $derived(mbBaseTopX + 6);
const mbKbY = $derived(mbBaseY + 3);
const mbKbW = $derived(mbBaseTopW - 12);
const mbKbH = $derived(mbBaseH * 0.55);
// Trackpad
const mbTpW = $derived(mbBaseTopW * 0.4);
const mbTpX = $derived(cx - mbTpW / 2);
const mbTpY = $derived(mbBaseY + mbKbH + 5);
const mbTpH = $derived(mbBaseH * 0.3);
// Clip IDs
const screenClipId = $derived(`di-screen-${uid}`);
const studioClipId = $derived(`di-studio-${uid}`);
</script>
{#if modelLower === "mac studio" || modelLower === "mac mini"}
<!-- Mac Studio / Mac Mini -->
<defs>
<clipPath id={studioClipId}>
<rect
x={studioX}
y={studioY + studioTopH}
width={studioW}
height={studioH - studioTopH}
rx={studioCorner - 1}
/>
</clipPath>
</defs>
<!-- Main body -->
<rect
x={studioX}
y={studioY}
width={studioW}
height={studioH}
rx={studioCorner}
fill="#1a1a1a"
stroke={wireColor}
stroke-width={strokeWidth}
/>
<!-- Memory fill -->
{#if ramPercent > 0}
<rect
x={studioX}
y={studioY + studioTopH + (studioMemTotalH - studioMemH)}
width={studioW}
height={studioMemH}
fill="rgba(255,215,0,0.75)"
clip-path="url(#{studioClipId})"
/>
{/if}
<!-- Top surface divider -->
<line
x1={studioX}
y1={studioY + studioTopH}
x2={studioX + studioW}
y2={studioY + studioTopH}
stroke="rgba(179,179,179,0.3)"
stroke-width="0.5"
/>
<!-- Front panel: vertical slots -->
<rect
x={studioVSlot1X - studioVSlotW / 2}
y={studioVSlotY}
width={studioVSlotW}
height={studioSlotH}
fill="rgba(0,0,0,0.35)"
rx="1.5"
/>
<rect
x={studioVSlot2X - studioVSlotW / 2}
y={studioVSlotY}
width={studioVSlotW}
height={studioSlotH}
fill="rgba(0,0,0,0.35)"
rx="1.5"
/>
<!-- Horizontal slot (SD card) -->
<rect
x={studioHSlotX}
y={studioVSlotY}
width={studioHSlotW}
height={studioSlotH * 0.6}
fill="rgba(0,0,0,0.35)"
rx="1"
/>
{:else}
<!-- MacBook Pro -->
<defs>
<clipPath id={screenClipId}>
<rect
x={mbScreenX + mbBezel}
y={mbY + mbBezel}
width={mbScreenW - mbBezel * 2}
height={mbScreenH - mbBezel * 2}
rx="2"
/>
</clipPath>
</defs>
<!-- Screen outer frame -->
<rect
x={mbScreenX}
y={mbY}
width={mbScreenW}
height={mbScreenH}
rx="3"
fill="#1a1a1a"
stroke={wireColor}
stroke-width={strokeWidth}
/>
<!-- Screen inner (dark) -->
<rect
x={mbScreenX + mbBezel}
y={mbY + mbBezel}
width={mbScreenW - mbBezel * 2}
height={mbScreenH - mbBezel * 2}
rx="2"
fill="#0a0a12"
/>
<!-- Memory fill on screen -->
{#if ramPercent > 0}
<rect
x={mbScreenX + mbBezel}
y={mbY + mbBezel + (mbMemTotalH - mbMemH)}
width={mbScreenW - mbBezel * 2}
height={mbMemH}
fill="rgba(255,215,0,0.85)"
clip-path="url(#{screenClipId})"
/>
{/if}
<!-- Apple logo -->
<path
d={APPLE_LOGO_PATH}
transform="translate({mbLogoX}, {mbLogoY}) scale({mbLogoScale})"
fill="#FFFFFF"
opacity="0.9"
/>
<!-- Keyboard base (trapezoidal) -->
<path
d="M {mbBaseTopX} {mbBaseY} L {mbBaseTopX +
mbBaseTopW} {mbBaseY} L {mbBaseBottomX + mbBaseBottomW} {mbBaseY +
mbBaseH} L {mbBaseBottomX} {mbBaseY + mbBaseH} Z"
fill="#2c2c2c"
stroke={wireColor}
stroke-width="1"
/>
<!-- Keyboard area -->
<rect
x={mbKbX}
y={mbKbY}
width={mbKbW}
height={mbKbH}
fill="rgba(0,0,0,0.2)"
rx="2"
/>
<!-- Trackpad -->
<rect
x={mbTpX}
y={mbTpY}
width={mbTpW}
height={mbTpH}
fill="rgba(255,255,255,0.08)"
rx="2"
/>
{/if}
@@ -0,0 +1,91 @@
<script lang="ts">
type FamilyLogoProps = {
family: string;
class?: string;
};
let { family, class: className = "" }: FamilyLogoProps = $props();
</script>
{#if family === "favorites"}
<svg class="w-6 h-6 {className}" viewBox="0 0 24 24" fill="currentColor">
<path
d="M12 2l3.09 6.26L22 9.27l-5 4.87 1.18 6.88L12 17.77l-6.18 3.25L7 14.14 2 9.27l6.91-1.01L12 2z"
/>
</svg>
{:else if family === "recents"}
<svg class="w-6 h-6 {className}" viewBox="0 0 24 24" fill="currentColor">
<path
d="M13 3a9 9 0 0 0-9 9H1l3.89 3.89.07.14L9 12H6c0-3.87 3.13-7 7-7s7 3.13 7 7-3.13 7-7 7c-1.93 0-3.68-.79-4.94-2.06l-1.42 1.42A8.954 8.954 0 0 0 13 21a9 9 0 0 0 0-18zm-1 5v5l4.28 2.54.72-1.21-3.5-2.08V8H12z"
/>
</svg>
{:else if family === "llama" || family === "meta"}
<svg class="w-6 h-6 {className}" viewBox="0 0 24 24" fill="currentColor">
<path
d="M6.915 4.03c-1.968 0-3.683 1.28-4.871 3.113C.704 9.208 0 11.883 0 14.449c0 .706.07 1.369.21 1.973a6.624 6.624 0 0 0 .265.86 5.297 5.297 0 0 0 .371.761c.696 1.159 1.818 1.927 3.593 1.927 1.497 0 2.633-.671 3.965-2.444.76-1.012 1.144-1.626 2.663-4.32l.756-1.339.186-.325c.061.1.121.196.183.3l2.152 3.595c.724 1.21 1.665 2.556 2.47 3.314 1.046.987 1.992 1.22 3.06 1.22 1.075 0 1.876-.355 2.455-.843a3.743 3.743 0 0 0 .81-.973c.542-.939.861-2.127.861-3.745 0-2.72-.681-5.357-2.084-7.45-1.282-1.912-2.957-2.93-4.716-2.93-1.047 0-2.088.467-3.053 1.308-.652.57-1.257 1.29-1.82 2.05-.69-.875-1.335-1.547-1.958-2.056-1.182-.966-2.315-1.303-3.454-1.303zm10.16 2.053c1.147 0 2.188.758 2.992 1.999 1.132 1.748 1.647 4.195 1.647 6.4 0 1.548-.368 2.9-1.839 2.9-.58 0-1.027-.23-1.664-1.004-.496-.601-1.343-1.878-2.832-4.358l-.617-1.028a44.908 44.908 0 0 0-1.255-1.98c.07-.109.141-.224.211-.327 1.12-1.667 2.118-2.602 3.358-2.602zm-10.201.553c1.265 0 2.058.791 2.675 1.446.307.327.737.871 1.234 1.579l-1.02 1.566c-.757 1.163-1.882 3.017-2.837 4.338-1.191 1.649-1.81 1.817-2.486 1.817-.524 0-1.038-.237-1.383-.794-.263-.426-.464-1.13-.464-2.046 0-2.221.63-4.535 1.66-6.088.454-.687.964-1.226 1.533-1.533a2.264 2.264 0 0 1 1.088-.285z"
/>
</svg>
{:else if family === "qwen"}
<svg class="w-6 h-6 {className}" viewBox="0 0 24 24" fill="currentColor">
<path
d="M12.604 1.34c.393.69.784 1.382 1.174 2.075a.18.18 0 00.157.091h5.552c.174 0 .322.11.446.327l1.454 2.57c.19.337.24.478.024.837-.26.43-.513.864-.76 1.3l-.367.658c-.106.196-.223.28-.04.512l2.652 4.637c.172.301.111.494-.043.77-.437.785-.882 1.564-1.335 2.34-.159.272-.352.375-.68.37-.777-.016-1.552-.01-2.327.016a.099.099 0 00-.081.05 575.097 575.097 0 01-2.705 4.74c-.169.293-.38.363-.725.364-.997.003-2.002.004-3.017.002a.537.537 0 01-.465-.271l-1.335-2.323a.09.09 0 00-.083-.049H4.982c-.285.03-.553-.001-.805-.092l-1.603-2.77a.543.543 0 01-.002-.54l1.207-2.12a.198.198 0 000-.197 550.951 550.951 0 01-1.875-3.272l-.79-1.395c-.16-.31-.173-.496.095-.965.465-.813.927-1.625 1.387-2.436.132-.234.304-.334.584-.335a338.3 338.3 0 012.589-.001.124.124 0 00.107-.063l2.806-4.895a.488.488 0 01.422-.246c.524-.001 1.053 0 1.583-.006L11.704 1c.341-.003.724.032.9.34zm-3.432.403a.06.06 0 00-.052.03L6.254 6.788a.157.157 0 01-.135.078H3.253c-.056 0-.07.025-.041.074l5.81 10.156c.025.042.013.062-.034.063l-2.795.015a.218.218 0 00-.2.116l-1.32 2.31c-.044.078-.021.118.068.118l5.716.008c.046 0 .08.02.104.061l1.403 2.454c.046.081.092.082.139 0l5.006-8.76.783-1.382a.055.055 0 01.096 0l1.424 2.53a.122.122 0 00.107.062l2.763-.02a.04.04 0 00.035-.02.041.041 0 000-.04l-2.9-5.086a.108.108 0 010-.113l.293-.507 1.12-1.977c.024-.041.012-.062-.035-.062H9.2c-.059 0-.073-.026-.043-.077l1.434-2.505a.107.107 0 000-.114L9.225 1.774a.06.06 0 00-.053-.031zm6.29 8.02c.046 0 .058.02.034.06l-.832 1.465-2.613 4.585a.056.056 0 01-.05.029.058.058 0 01-.05-.029L8.498 9.841c-.02-.034-.01-.052.028-.054l.216-.012 6.722-.012z"
/>
</svg>
{:else if family === "deepseek"}
<svg class="w-6 h-6 {className}" viewBox="0 0 24 24" fill="currentColor">
<path
d="M23.748 4.482c-.254-.124-.364.113-.512.234-.051.039-.094.09-.137.136-.372.397-.806.657-1.373.626-.829-.046-1.537.214-2.163.848-.133-.782-.575-1.248-1.247-1.548-.352-.156-.708-.311-.955-.65-.172-.241-.219-.51-.305-.774-.055-.16-.11-.323-.293-.35-.2-.031-.278.136-.356.276-.313.572-.434 1.202-.422 1.84.027 1.436.633 2.58 1.838 3.393.137.093.172.187.129.323-.082.28-.18.552-.266.833-.055.179-.137.217-.329.14a5.526 5.526 0 01-1.736-1.18c-.857-.828-1.631-1.742-2.597-2.458a11.365 11.365 0 00-.689-.471c-.985-.957.13-1.743.388-1.836.27-.098.093-.432-.779-.428-.872.004-1.67.295-2.687.684a3.055 3.055 0 01-.465.137 9.597 9.597 0 00-2.883-.102c-1.885.21-3.39 1.102-4.497 2.623C.082 8.606-.231 10.684.152 12.85c.403 2.284 1.569 4.175 3.36 5.653 1.858 1.533 3.997 2.284 6.438 2.14 1.482-.085 3.133-.284 4.994-1.86.47.234.962.327 1.78.397.63.059 1.236-.03 1.705-.128.735-.156.684-.837.419-.961-2.155-1.004-1.682-.595-2.113-.926 1.096-1.296 2.746-2.642 3.392-7.003.05-.347.007-.565 0-.845-.004-.17.035-.237.23-.256a4.173 4.173 0 001.545-.475c1.396-.763 1.96-2.015 2.093-3.517.02-.23-.004-.467-.247-.588zM11.581 18c-2.089-1.642-3.102-2.183-3.52-2.16-.392.024-.321.471-.235.763.09.288.207.486.371.739.114.167.192.416-.113.603-.673.416-1.842-.14-1.897-.167-1.361-.802-2.5-1.86-3.301-3.307-.774-1.393-1.224-2.887-1.298-4.482-.02-.386.093-.522.477-.592a4.696 4.696 0 011.529-.039c2.132.312 3.946 1.265 5.468 2.774.868.86 1.525 1.887 2.202 2.891.72 1.066 1.494 2.082 2.48 2.914.348.292.625.514.891.677-.802.09-2.14.11-3.054-.614zm1-6.44a.306.306 0 01.415-.287.302.302 0 01.2.288.306.306 0 01-.31.307.303.303 0 01-.304-.308zm3.11 1.596c-.2.081-.399.151-.59.16a1.245 1.245 0 01-.798-.254c-.274-.23-.47-.358-.552-.758a1.73 1.73 0 01.016-.588c.07-.327-.008-.537-.239-.727-.187-.156-.426-.199-.688-.199a.559.559 0 01-.254-.078c-.11-.054-.2-.19-.114-.358.028-.054.16-.186.192-.21.356-.202.767-.136 1.146.016.352.144.618.408 1.001.782.391.451.462.576.685.914.176.265.336.537.445.848.067.195-.019.354-.25.452z"
/>
</svg>
{:else if family === "openai" || family === "gpt-oss"}
<svg class="w-6 h-6 {className}" viewBox="0 0 24 24" fill="currentColor">
<path
d="M22.2819 9.8211a5.9847 5.9847 0 0 0-.5157-4.9108 6.0462 6.0462 0 0 0-6.5098-2.9A6.0651 6.0651 0 0 0 4.9807 4.1818a5.9847 5.9847 0 0 0-3.9977 2.9 6.0462 6.0462 0 0 0 .7427 7.0966 5.98 5.98 0 0 0 .511 4.9107 6.051 6.051 0 0 0 6.5146 2.9001A5.9847 5.9847 0 0 0 13.2599 24a6.0557 6.0557 0 0 0 5.7718-4.2058 5.9894 5.9894 0 0 0 3.9977-2.9001 6.0557 6.0557 0 0 0-.7475-7.0729zm-9.022 12.6081a4.4755 4.4755 0 0 1-2.8764-1.0408l.1419-.0804 4.7783-2.7582a.7948.7948 0 0 0 .3927-.6813v-6.7369l2.02 1.1686a.071.071 0 0 1 .038.052v5.5826a4.504 4.504 0 0 1-4.4945 4.4944zm-9.6607-4.1254a4.4708 4.4708 0 0 1-.5346-3.0137l.142.0852 4.783 2.7582a.7712.7712 0 0 0 .7806 0l5.8428-3.3685v2.3324a.0804.0804 0 0 1-.0332.0615L9.74 19.9502a4.4992 4.4992 0 0 1-6.1408-1.6464zM2.3408 7.8956a4.485 4.485 0 0 1 2.3655-1.9728V11.6a.7664.7664 0 0 0 .3879.6765l5.8144 3.3543-2.0201 1.1685a.0757.0757 0 0 1-.071 0l-4.8303-2.7865A4.504 4.504 0 0 1 2.3408 7.872zm16.5963 3.8558L13.1038 8.364 15.1192 7.2a.0757.0757 0 0 1 .071 0l4.8303 2.7913a4.4944 4.4944 0 0 1-.6765 8.1042v-5.6772a.79.79 0 0 0-.407-.667zm2.0107-3.0231l-.142-.0852-4.7735-2.7818a.7759.7759 0 0 0-.7854 0L9.409 9.2297V6.8974a.0662.0662 0 0 1 .0284-.0615l4.8303-2.7866a4.4992 4.4992 0 0 1 6.6802 4.66zM8.3065 12.863l-2.02-1.1638a.0804.0804 0 0 1-.038-.0567V6.0742a4.4992 4.4992 0 0 1 7.3757-3.4537l-.142.0805L8.704 5.459a.7948.7948 0 0 0-.3927.6813zm1.0976-2.3654l2.602-1.4998 2.6069 1.4998v2.9994l-2.5974 1.4997-2.6067-1.4997Z"
/>
</svg>
{:else if family === "glm"}
<svg class="w-6 h-6 {className}" viewBox="0 0 24 24" fill="currentColor">
<path
d="M11.991 23.503a.24.24 0 00-.244.248.24.24 0 00.244.249.24.24 0 00.245-.249.24.24 0 00-.22-.247l-.025-.001zM9.671 5.365a1.697 1.697 0 011.099 2.132l-.071.172-.016.04-.018.054c-.07.16-.104.32-.104.498-.035.71.47 1.279 1.186 1.314h.366c1.309.053 2.338 1.173 2.286 2.523-.052 1.332-1.152 2.38-2.478 2.327h-.174c-.715.018-1.274.64-1.239 1.368 0 .124.018.23.053.337.209.373.54.658.96.8.75.23 1.517-.125 1.9-.782l.018-.035c.402-.64 1.17-.96 1.92-.711.854.284 1.378 1.226 1.099 2.167a1.661 1.661 0 01-2.077 1.102 1.711 1.711 0 01-.907-.711l-.017-.035c-.2-.323-.463-.58-.851-.711l-.056-.018a1.646 1.646 0 00-1.954.746 1.66 1.66 0 01-1.065.764 1.677 1.677 0 01-1.989-1.279c-.209-.906.332-1.83 1.257-2.043a1.51 1.51 0 01.296-.035h.018c.68-.071 1.151-.622 1.116-1.333a1.307 1.307 0 00-.227-.693 2.515 2.515 0 01-.366-1.403 2.39 2.39 0 01.366-1.208c.14-.195.21-.444.227-.693.018-.71-.506-1.261-1.186-1.332l-.07-.018a1.43 1.43 0 01-.299-.07l-.05-.019a1.7 1.7 0 01-1.047-2.114 1.68 1.68 0 012.094-1.101zm-5.575 10.11c.26-.264.639-.367.994-.27.355.096.633.379.728.74.095.362-.007.748-.267 1.013-.402.41-1.053.41-1.455 0a1.062 1.062 0 010-1.482zm14.845-.294c.359-.09.738.024.992.297.254.274.344.665.237 1.025-.107.36-.396.634-.756.718-.551.128-1.1-.22-1.23-.781a1.05 1.05 0 01.757-1.26zm-.064-4.39c.314.32.49.753.49 1.206 0 .452-.176.886-.49 1.206-.315.32-.74.5-1.185.5-.444 0-.87-.18-1.184-.5a1.727 1.727 0 010-2.412 1.654 1.654 0 012.369 0zm-11.243.163c.364.484.447 1.128.218 1.691a1.665 1.665 0 01-2.188.923c-.855-.36-1.26-1.358-.907-2.228a1.68 1.68 0 011.33-1.038c.593-.08 1.183.169 1.547.652zm11.545-4.221c.368 0 .708.2.892.524.184.324.184.724 0 1.048a1.026 1.026 0 01-.892.524c-.568 0-1.03-.47-1.03-1.048 0-.579.462-1.048 1.03-1.048zm-14.358 0c.368 0 .707.2.891.524.184.324.184.724 0 1.048a1.026 1.026 0 01-.891.524c-.569 0-1.03-.47-1.03-1.048 0-.579.461-1.048 1.03-1.048zm10.031-1.475c.925 0 1.675.764 1.675 1.706s-.75 1.705-1.675 1.705-1.674-.763-1.674-1.705c0-.942.75-1.706 1.674-1.706zm-2.626-.684c.362-.082.653-.356.761-.718a1.062 1.062 0 00-.238-1.028 1.017 1.017 0 00-.996-.294c-.547.14-.881.7-.752 1.257.13.558.675.907 1.225.783zm0 16.876c.359-.087.644-.36.75-.72a1.062 1.062 0 00-.237-1.019 1.018 1.018 0 00-.985-.301 1.037 1.037 0 00-.762.717c-.108.361-.017.754.239 1.028.245.263.606.377.953.305l.043-.01zM17.19 3.5a.631.631 0 00.628-.64c0-.355-.279-.64-.628-.64a.631.631 0 00-.628.64c0 .355.28.64.628.64zm-10.38 0a.631.631 0 00.628-.64c0-.355-.28-.64-.628-.64a.631.631 0 00-.628.64c0 .355.279.64.628.64zm-5.182 7.852a.631.631 0 00-.628.64c0 .354.28.639.628.639a.63.63 0 00.627-.606l.001-.034a.62.62 0 00-.628-.64zm5.182 9.13a.631.631 0 00-.628.64c0 .355.279.64.628.64a.631.631 0 00.628-.64c0-.355-.28-.64-.628-.64zm10.38.018a.631.631 0 00-.628.64c0 .355.28.64.628.64a.631.631 0 00.628-.64c0-.355-.279-.64-.628-.64zm5.182-9.148a.631.631 0 00-.628.64c0 .354.279.639.628.639a.631.631 0 00.628-.64c0-.355-.28-.64-.628-.64zm-.384-4.992a.24.24 0 00.244-.249.24.24 0 00-.244-.249.24.24 0 00-.244.249c0 .142.122.249.244.249zM11.991.497a.24.24 0 00.245-.248A.24.24 0 0011.99 0a.24.24 0 00-.244.249c0 .133.108.236.223.247l.021.001zM2.011 6.36a.24.24 0 00.245-.249.24.24 0 00-.244-.249.24.24 0 00-.244.249.24.24 0 00.244.249zm0 11.263a.24.24 0 00-.243.248.24.24 0 00.244.249.24.24 0 00.244-.249.252.252 0 00-.244-.248zm19.995-.018a.24.24 0 00-.245.248.24.24 0 00.245.25.24.24 0 00.244-.25.252.252 0 00-.244-.248z"
/>
</svg>
{:else if family === "minimax"}
<svg class="w-6 h-6 {className}" viewBox="0 0 24 24" fill="currentColor">
<path
d="M16.278 2c1.156 0 2.093.927 2.093 2.07v12.501a.74.74 0 00.744.709.74.74 0 00.743-.709V9.099a2.06 2.06 0 012.071-2.049A2.06 2.06 0 0124 9.1v6.561a.649.649 0 01-.652.645.649.649 0 01-.653-.645V9.1a.762.762 0 00-.766-.758.762.762 0 00-.766.758v7.472a2.037 2.037 0 01-2.048 2.026 2.037 2.037 0 01-2.048-2.026v-12.5a.785.785 0 00-.788-.753.785.785 0 00-.789.752l-.001 15.904A2.037 2.037 0 0113.441 22a2.037 2.037 0 01-2.048-2.026V18.04c0-.356.292-.645.652-.645.36 0 .652.289.652.645v1.934c0 .263.142.506.372.638.23.131.514.131.744 0a.734.734 0 00.372-.638V4.07c0-1.143.937-2.07 2.093-2.07zm-5.674 0c1.156 0 2.093.927 2.093 2.07v11.523a.648.648 0 01-.652.645.648.648 0 01-.652-.645V4.07a.785.785 0 00-.789-.78.785.785 0 00-.789.78v14.013a2.06 2.06 0 01-2.07 2.048 2.06 2.06 0 01-2.071-2.048V9.1a.762.762 0 00-.766-.758.762.762 0 00-.766.758v3.8a2.06 2.06 0 01-2.071 2.049A2.06 2.06 0 010 12.9v-1.378c0-.357.292-.646.652-.646.36 0 .653.29.653.646V12.9c0 .418.343.757.766.757s.766-.339.766-.757V9.099a2.06 2.06 0 012.07-2.048 2.06 2.06 0 012.071 2.048v8.984c0 .419.343.758.767.758.423 0 .766-.339.766-.758V4.07c0-1.143.937-2.07 2.093-2.07z"
/>
</svg>
{:else if family === "kimi"}
<svg class="w-6 h-6 {className}" viewBox="0 0 24 24" fill="currentColor">
<path
d="M19.738 5.776c.163-.209.306-.4.457-.585.07-.087.064-.153-.004-.244-.655-.861-.717-1.817-.34-2.787.283-.73.909-1.072 1.674-1.145.477-.045.945.004 1.379.236.57.305.902.77 1.01 1.412.086.512.07 1.012-.075 1.508-.257.878-.888 1.333-1.753 1.448-.718.096-1.446.108-2.17.157-.056.004-.113 0-.178 0z"
/>
<path
d="M17.962 1.844h-4.326l-3.425 7.81H5.369V1.878H1.5V22h3.87v-8.477h6.824a3.025 3.025 0 002.743-1.75V22h3.87v-8.477a3.87 3.87 0 00-3.588-3.86v-.01h-2.125a3.94 3.94 0 002.323-2.12l2.545-5.689z"
/>
</svg>
{:else if family === "flux"}
<svg class="w-6 h-6 {className}" viewBox="0 0 24 24" fill="currentColor">
<path
d="M12 2L2 19h7.5l2.5-4.5L14.5 19H22L12 2zm0 4.5L16.5 17h-3l-1.5-2.7L10.5 17h-3L12 6.5z"
/>
</svg>
{:else if family === "qwen-image"}
<svg class="w-6 h-6 {className}" viewBox="0 0 24 24" fill="currentColor">
<path
d="M12.604 1.34c.393.69.784 1.382 1.174 2.075a.18.18 0 00.157.091h5.552c.174 0 .322.11.446.327l1.454 2.57c.19.337.24.478.024.837-.26.43-.513.864-.76 1.3l-.367.658c-.106.196-.223.28-.04.512l2.652 4.637c.172.301.111.494-.043.77-.437.785-.882 1.564-1.335 2.34-.159.272-.352.375-.68.37-.777-.016-1.552-.01-2.327.016a.099.099 0 00-.081.05 575.097 575.097 0 01-2.705 4.74c-.169.293-.38.363-.725.364-.997.003-2.002.004-3.017.002a.537.537 0 01-.465-.271l-1.335-2.323a.09.09 0 00-.083-.049H4.982c-.285.03-.553-.001-.805-.092l-1.603-2.77a.543.543 0 01-.002-.54l1.207-2.12a.198.198 0 000-.197 550.951 550.951 0 01-1.875-3.272l-.79-1.395c-.16-.31-.173-.496.095-.965.465-.813.927-1.625 1.387-2.436.132-.234.304-.334.584-.335a338.3 338.3 0 012.589-.001.124.124 0 00.107-.063l2.806-4.895a.488.488 0 01.422-.246c.524-.001 1.053 0 1.583-.006L11.704 1c.341-.003.724.032.9.34zm-3.432.403a.06.06 0 00-.052.03L6.254 6.788a.157.157 0 01-.135.078H3.253c-.056 0-.07.025-.041.074l5.81 10.156c.025.042.013.062-.034.063l-2.795.015a.218.218 0 00-.2.116l-1.32 2.31c-.044.078-.021.118.068.118l5.716.008c.046 0 .08.02.104.061l1.403 2.454c.046.081.092.082.139 0l5.006-8.76.783-1.382a.055.055 0 01.096 0l1.424 2.53a.122.122 0 00.107.062l2.763-.02a.04.04 0 00.035-.02.041.041 0 000-.04l-2.9-5.086a.108.108 0 010-.113l.293-.507 1.12-1.977c.024-.041.012-.062-.035-.062H9.2c-.059 0-.073-.026-.043-.077l1.434-2.505a.107.107 0 000-.114L9.225 1.774a.06.06 0 00-.053-.031zm6.29 8.02c.046 0 .058.02.034.06l-.832 1.465-2.613 4.585a.056.056 0 01-.05.029.058.058 0 01-.05-.029L8.498 9.841c-.02-.034-.01-.052.028-.054l.216-.012 6.722-.012z"
/>
</svg>
{:else if family === "huggingface"}
<svg class="w-6 h-6 {className}" viewBox="0 0 24 24" fill="currentColor">
<path
d="M12.025 1.13c-5.77 0-10.449 4.647-10.449 10.378 0 1.112.178 2.181.503 3.185.064-.222.203-.444.416-.577a.96.96 0 0 1 .524-.15c.293 0 .584.124.84.284.278.173.48.408.71.694.226.282.458.611.684.951v-.014c.017-.324.106-.622.264-.874s.403-.487.762-.543c.3-.047.596.06.787.203s.31.313.4.467c.15.257.212.468.233.542.01.026.653 1.552 1.657 2.54.616.605 1.01 1.223 1.082 1.912.055.537-.096 1.059-.38 1.572.637.121 1.294.187 1.967.187.657 0 1.298-.063 1.921-.178-.287-.517-.44-1.041-.384-1.581.07-.69.465-1.307 1.081-1.913 1.004-.987 1.647-2.513 1.657-2.539.021-.074.083-.285.233-.542.09-.154.208-.323.4-.467a1.08 1.08 0 0 1 .787-.203c.359.056.604.29.762.543s.247.55.265.874v.015c.225-.34.457-.67.683-.952.23-.286.432-.52.71-.694.257-.16.547-.284.84-.285a.97.97 0 0 1 .524.151c.228.143.373.388.43.625l.006.04a10.3 10.3 0 0 0 .534-3.273c0-5.731-4.678-10.378-10.449-10.378M8.327 6.583a1.5 1.5 0 0 1 .713.174 1.487 1.487 0 0 1 .617 2.013c-.183.343-.762-.214-1.102-.094-.38.134-.532.914-.917.71a1.487 1.487 0 0 1 .69-2.803m7.486 0a1.487 1.487 0 0 1 .689 2.803c-.385.204-.536-.576-.916-.71-.34-.12-.92.437-1.103.094a1.487 1.487 0 0 1 .617-2.013 1.5 1.5 0 0 1 .713-.174m-10.68 1.55a.96.96 0 1 1 0 1.921.96.96 0 0 1 0-1.92m13.838 0a.96.96 0 1 1 0 1.92.96.96 0 0 1 0-1.92M8.489 11.458c.588.01 1.965 1.157 3.572 1.164 1.607-.007 2.984-1.155 3.572-1.164.196-.003.305.12.305.454 0 .886-.424 2.328-1.563 3.202-.22-.756-1.396-1.366-1.63-1.32q-.011.001-.02.006l-.044.026-.01.008-.03.024q-.018.017-.035.036l-.032.04a1 1 0 0 0-.058.09l-.014.025q-.049.088-.11.19a1 1 0 0 1-.083.116 1.2 1.2 0 0 1-.173.18q-.035.029-.075.058a1.3 1.3 0 0 1-.251-.243 1 1 0 0 1-.076-.107c-.124-.193-.177-.363-.337-.444-.034-.016-.104-.008-.2.022q-.094.03-.216.087-.06.028-.125.063l-.13.074q-.067.04-.136.086a3 3 0 0 0-.135.096 3 3 0 0 0-.26.219 2 2 0 0 0-.12.121 2 2 0 0 0-.106.128l-.002.002a2 2 0 0 0-.09.132l-.001.001a1.2 1.2 0 0 0-.105.212q-.013.036-.024.073c-1.139-.875-1.563-2.317-1.563-3.203 0-.334.109-.457.305-.454m.836 10.354c.824-1.19.766-2.082-.365-3.194-1.13-1.112-1.789-2.738-1.789-2.738s-.246-.945-.806-.858-.97 1.499.202 2.362c1.173.864-.233 1.45-.685.64-.45-.812-1.683-2.896-2.322-3.295s-1.089-.175-.938.647 2.822 2.813 2.562 3.244-1.176-.506-1.176-.506-2.866-2.567-3.49-1.898.473 1.23 2.037 2.16c1.564.932 1.686 1.178 1.464 1.53s-3.675-2.511-4-1.297c-.323 1.214 3.524 1.567 3.287 2.405-.238.839-2.71-1.587-3.216-.642-.506.946 3.49 2.056 3.522 2.064 1.29.33 4.568 1.028 5.713-.624m5.349 0c-.824-1.19-.766-2.082.365-3.194 1.13-1.112 1.789-2.738 1.789-2.738s.246-.945.806-.858.97 1.499-.202 2.362c-1.173.864.233 1.45.685.64.451-.812 1.683-2.896 2.322-3.295s1.089-.175.938.647-2.822 2.813-2.562 3.244 1.176-.506 1.176-.506 2.866-2.567 3.49-1.898-.473 1.23-2.037 2.16c-1.564.932-1.686 1.178-1.464 1.53s3.675-2.511 4-1.297c.323 1.214-3.524 1.567-3.287 2.405.238.839 2.71-1.587 3.216-.642.506.946-3.49 2.056-3.522 2.064-1.29.33-4.568 1.028-5.713-.624"
/>
</svg>
{:else}
<svg class="w-6 h-6 {className}" viewBox="0 0 24 24" fill="currentColor">
<path
d="M12 2C6.48 2 2 6.48 2 12s4.48 10 10 10 10-4.48 10-10S17.52 2 12 2zm-2 15l-5-5 1.41-1.41L10 14.17l7.59-7.59L19 8l-9 9z"
/>
</svg>
{/if}
@@ -0,0 +1,176 @@
<script lang="ts">
import FamilyLogos from "./FamilyLogos.svelte";
type FamilySidebarProps = {
families: string[];
selectedFamily: string | null;
hasFavorites: boolean;
hasRecents: boolean;
onSelect: (family: string | null) => void;
};
let {
families,
selectedFamily,
hasFavorites,
hasRecents,
onSelect,
}: FamilySidebarProps = $props();
// Family display names
const familyNames: Record<string, string> = {
favorites: "Favorites",
recents: "Recent",
huggingface: "Hub",
llama: "Meta",
qwen: "Qwen",
deepseek: "DeepSeek",
"gpt-oss": "OpenAI",
glm: "GLM",
minimax: "MiniMax",
kimi: "Kimi",
flux: "FLUX",
"qwen-image": "Qwen Img",
};
function getFamilyName(family: string): string {
return (
familyNames[family] || family.charAt(0).toUpperCase() + family.slice(1)
);
}
</script>
<div
class="flex flex-col gap-1 py-2 px-1 border-r border-exo-yellow/10 bg-exo-medium-gray/30 min-w-[64px] overflow-y-auto scrollbar-hide"
>
<!-- All models (no filter) -->
<button
type="button"
onclick={() => onSelect(null)}
class="group flex flex-col items-center justify-center p-2 rounded transition-all duration-200 cursor-pointer {selectedFamily ===
null
? 'bg-exo-yellow/20 border-l-2 border-exo-yellow'
: 'hover:bg-white/5 border-l-2 border-transparent'}"
title="All models"
>
<svg
class="w-5 h-5 {selectedFamily === null
? 'text-exo-yellow'
: 'text-white/50 group-hover:text-white/70'}"
viewBox="0 0 24 24"
fill="currentColor"
>
<path
d="M4 8h4V4H4v4zm6 12h4v-4h-4v4zm-6 0h4v-4H4v4zm0-6h4v-4H4v4zm6 0h4v-4h-4v4zm6-10v4h4V4h-4zm-6 4h4V4h-4v4zm6 6h4v-4h-4v4zm0 6h4v-4h-4v4z"
/>
</svg>
<span
class="text-[9px] font-mono mt-0.5 {selectedFamily === null
? 'text-exo-yellow'
: 'text-white/40 group-hover:text-white/60'}">All</span
>
</button>
<!-- Favorites (only show if has favorites) -->
{#if hasFavorites}
<button
type="button"
onclick={() => onSelect("favorites")}
class="group flex flex-col items-center justify-center p-2 rounded transition-all duration-200 cursor-pointer {selectedFamily ===
'favorites'
? 'bg-exo-yellow/20 border-l-2 border-exo-yellow'
: 'hover:bg-white/5 border-l-2 border-transparent'}"
title="Show favorited models"
>
<FamilyLogos
family="favorites"
class={selectedFamily === "favorites"
? "text-amber-400"
: "text-white/50 group-hover:text-amber-400/70"}
/>
<span
class="text-[9px] font-mono mt-0.5 {selectedFamily === 'favorites'
? 'text-amber-400'
: 'text-white/40 group-hover:text-white/60'}">Faves</span
>
</button>
{/if}
<!-- Recent (only show if has recent models) -->
{#if hasRecents}
<button
type="button"
onclick={() => onSelect("recents")}
class="group flex flex-col items-center justify-center p-2 rounded transition-all duration-200 cursor-pointer {selectedFamily ===
'recents'
? 'bg-exo-yellow/20 border-l-2 border-exo-yellow'
: 'hover:bg-white/5 border-l-2 border-transparent'}"
title="Recently launched models"
>
<FamilyLogos
family="recents"
class={selectedFamily === "recents"
? "text-exo-yellow"
: "text-white/50 group-hover:text-white/70"}
/>
<span
class="text-[9px] font-mono mt-0.5 {selectedFamily === 'recents'
? 'text-exo-yellow'
: 'text-white/40 group-hover:text-white/60'}">Recent</span
>
</button>
{/if}
<!-- HuggingFace Hub -->
<button
type="button"
onclick={() => onSelect("huggingface")}
class="group flex flex-col items-center justify-center p-2 rounded transition-all duration-200 cursor-pointer {selectedFamily ===
'huggingface'
? 'bg-orange-500/20 border-l-2 border-orange-400'
: 'hover:bg-white/5 border-l-2 border-transparent'}"
title="Browse and add models from Hugging Face"
>
<FamilyLogos
family="huggingface"
class={selectedFamily === "huggingface"
? "text-orange-400"
: "text-white/50 group-hover:text-orange-400/70"}
/>
<span
class="text-[9px] font-mono mt-0.5 {selectedFamily === 'huggingface'
? 'text-orange-400'
: 'text-white/40 group-hover:text-white/60'}">Hub</span
>
</button>
<div class="h-px bg-exo-yellow/10 my-1"></div>
<!-- Model families -->
{#each families as family}
<button
type="button"
onclick={() => onSelect(family)}
class="group flex flex-col items-center justify-center p-2 rounded transition-all duration-200 cursor-pointer {selectedFamily ===
family
? 'bg-exo-yellow/20 border-l-2 border-exo-yellow'
: 'hover:bg-white/5 border-l-2 border-transparent'}"
title={getFamilyName(family)}
>
<FamilyLogos
{family}
class={selectedFamily === family
? "text-exo-yellow"
: "text-white/50 group-hover:text-white/70"}
/>
<span
class="text-[9px] font-mono mt-0.5 truncate max-w-full {selectedFamily ===
family
? 'text-exo-yellow'
: 'text-white/40 group-hover:text-white/60'}"
>
{getFamilyName(family)}
</span>
</button>
{/each}
</div>
+84 -14
View File
@@ -6,6 +6,10 @@
export let showSidebarToggle = false;
export let sidebarVisible = true;
export let onToggleSidebar: (() => void) | null = null;
export let downloadProgress: {
count: number;
percentage: number;
} | null = null;
function handleHome(): void {
if (onHome) {
@@ -33,13 +37,17 @@
<div class="absolute left-6 top-1/2 -translate-y-1/2">
<button
onclick={handleToggleSidebar}
class="p-2 rounded border border-exo-medium-gray/40 hover:border-exo-yellow/50 transition-colors cursor-pointer"
class="p-2 rounded border border-exo-light-gray/30 hover:border-exo-yellow/50 hover:bg-exo-medium-gray/30 transition-colors cursor-pointer"
title={sidebarVisible ? "Hide sidebar" : "Show sidebar"}
aria-label={sidebarVisible
? "Hide conversation sidebar"
: "Show conversation sidebar"}
aria-pressed={sidebarVisible}
>
<svg
class="w-5 h-5 {sidebarVisible
? 'text-exo-yellow'
: 'text-exo-medium-gray'}"
: 'text-exo-light-gray'}"
fill="none"
viewBox="0 0 24 24"
stroke="currentColor"
@@ -75,18 +83,19 @@
<img
src="/exo-logo.png"
alt="EXO"
class="h-18 drop-shadow-[0_0_20px_rgba(255,215,0,0.5)]"
class="h-18 drop-shadow-[0_0_4px_rgba(255,215,0,0.3)]"
/>
</button>
<!-- Right: Home + Downloads -->
<div
<nav
class="absolute right-6 top-1/2 -translate-y-1/2 flex items-center gap-4"
aria-label="Main navigation"
>
{#if showHome}
<button
onclick={handleHome}
class="text-sm text-exo-light-gray hover:text-exo-yellow transition-colors tracking-wider uppercase flex items-center gap-2 cursor-pointer"
class="text-sm text-white/70 hover:text-exo-yellow transition-colors tracking-wider uppercase flex items-center gap-2 cursor-pointer"
title="Back to topology view"
>
<svg
@@ -107,23 +116,84 @@
{/if}
<a
href="/#/downloads"
class="text-sm text-exo-light-gray hover:text-exo-yellow transition-colors tracking-wider uppercase flex items-center gap-2 cursor-pointer"
class="text-sm text-white/70 hover:text-exo-yellow transition-colors tracking-wider uppercase flex items-center gap-2 cursor-pointer"
title="View downloads overview"
>
{#if downloadProgress}
<!-- Compact download progress indicator -->
<div class="relative w-4 h-4 flex-shrink-0">
<svg class="w-4 h-4 -rotate-90" viewBox="0 0 20 20">
<circle
cx="10"
cy="10"
r="8"
fill="none"
stroke="currentColor"
stroke-width="2"
opacity="0.2"
/>
<circle
cx="10"
cy="10"
r="8"
fill="none"
stroke="currentColor"
stroke-width="2"
stroke-dasharray={2 * Math.PI * 8}
stroke-dashoffset={2 *
Math.PI *
8 *
(1 - downloadProgress.percentage / 100)}
class="text-blue-400 transition-all duration-300"
/>
</svg>
<div
class="absolute inset-0 flex items-center justify-center text-[6px] font-mono text-blue-400"
>
{downloadProgress.count}
</div>
</div>
{:else}
<svg
class="w-4 h-4"
viewBox="0 0 24 24"
fill="none"
stroke="currentColor"
stroke-width="2"
stroke-linecap="round"
stroke-linejoin="round"
>
<path d="M12 3v12" />
<path d="M7 12l5 5 5-5" />
<path d="M5 21h14" />
</svg>
{/if}
Downloads
</a>
<a
href="/#/settings"
class="text-sm text-white/70 hover:text-exo-yellow transition-colors tracking-wider uppercase flex items-center gap-2 cursor-pointer"
title="Settings"
>
<svg
class="w-4 h-4"
viewBox="0 0 24 24"
fill="none"
viewBox="0 0 24 24"
stroke="currentColor"
stroke-width="2"
stroke-linecap="round"
stroke-linejoin="round"
>
<path d="M12 3v12" />
<path d="M7 12l5 5 5-5" />
<path d="M5 21h14" />
<path
stroke-linecap="round"
stroke-linejoin="round"
d="M10.325 4.317c.426-1.756 2.924-1.756 3.35 0a1.724 1.724 0 002.573 1.066c1.543-.94 3.31.826 2.37 2.37a1.724 1.724 0 001.065 2.572c1.756.426 1.756 2.924 0 3.35a1.724 1.724 0 00-1.066 2.573c.94 1.543-.826 3.31-2.37 2.37a1.724 1.724 0 00-2.572 1.065c-.426 1.756-2.924 1.756-3.35 0a1.724 1.724 0 00-2.573-1.066c-1.543.94-3.31-.826-2.37-2.37a1.724 1.724 0 00-1.065-2.572c-1.756-.426-1.756-2.924 0-3.35a1.724 1.724 0 001.066-2.573c-.94-1.543.826-3.31 2.37-2.37.996.608 2.296.07 2.572-1.065z"
/>
<path
stroke-linecap="round"
stroke-linejoin="round"
d="M15 12a3 3 0 11-6 0 3 3 0 016 0z"
/>
</svg>
Downloads
Settings
</a>
</div>
</nav>
</header>
@@ -0,0 +1,152 @@
<script lang="ts">
interface HuggingFaceModel {
id: string;
author: string;
downloads: number;
likes: number;
last_modified: string;
tags: string[];
}
type HuggingFaceResultItemProps = {
model: HuggingFaceModel;
isAdded: boolean;
isAdding: boolean;
onAdd: () => void;
onSelect: () => void;
downloadedOnNodes?: string[];
};
let {
model,
isAdded,
isAdding,
onAdd,
onSelect,
downloadedOnNodes = [],
}: HuggingFaceResultItemProps = $props();
function formatNumber(num: number | undefined): string {
if (num == null) return "0";
if (num >= 1000000) {
return `${(num / 1000000).toFixed(1)}M`;
} else if (num >= 1000) {
return `${(num / 1000).toFixed(1)}k`;
}
return num.toString();
}
// Extract model name from full ID (e.g., "mlx-community/Llama-3.2-1B" -> "Llama-3.2-1B")
const modelName = $derived(model.id.split("/").pop() || model.id);
</script>
<div
class="flex items-center justify-between gap-3 px-3 py-2.5 hover:bg-white/5 transition-colors border-b border-white/5 last:border-b-0"
>
<div class="flex-1 min-w-0">
<div class="flex items-center gap-2">
<span class="text-sm font-mono text-white truncate" title={model.id}
>{modelName}</span
>
{#if downloadedOnNodes.length > 0}
<span
class="flex-shrink-0"
title={`Downloaded on ${downloadedOnNodes.join(", ")}`}
>
<svg
class="w-4 h-4"
viewBox="0 0 24 24"
fill="none"
stroke="currentColor"
stroke-width="2"
stroke-linecap="round"
stroke-linejoin="round"
>
<path
class="text-white/40"
d="M20 20a2 2 0 0 0 2-2V8a2 2 0 0 0-2-2h-7.9a2 2 0 0 1-1.69-.9L9.6 3.9A2 2 0 0 0 7.93 3H4a2 2 0 0 0-2 2v13a2 2 0 0 0 2 2Z"
/>
<path class="text-green-400" d="m9 13 2 2 4-4" />
</svg>
</span>
{/if}
{#if isAdded}
<span
class="px-1.5 py-0.5 text-[10px] font-mono bg-green-500/20 text-green-400 rounded"
>Added</span
>
{/if}
</div>
<div class="flex items-center gap-3 mt-0.5 text-xs text-white/40">
<span class="truncate">{model.author}</span>
<span
class="flex items-center gap-1 shrink-0"
title="Downloads in the last 30 days"
>
<svg
class="w-3 h-3"
fill="none"
stroke="currentColor"
viewBox="0 0 24 24"
>
<path
stroke-linecap="round"
stroke-linejoin="round"
stroke-width="2"
d="M4 16v1a3 3 0 003 3h10a3 3 0 003-3v-1m-4-4l-4 4m0 0l-4-4m4 4V4"
/>
</svg>
{formatNumber(model.downloads)}
</span>
<span
class="flex items-center gap-1 shrink-0"
title="Community likes on Hugging Face"
>
<svg
class="w-3 h-3"
fill="none"
stroke="currentColor"
viewBox="0 0 24 24"
>
<path
stroke-linecap="round"
stroke-linejoin="round"
stroke-width="2"
d="M4.318 6.318a4.5 4.5 0 000 6.364L12 20.364l7.682-7.682a4.5 4.5 0 00-6.364-6.364L12 7.636l-1.318-1.318a4.5 4.5 0 00-6.364 0z"
/>
</svg>
{formatNumber(model.likes)}
</span>
</div>
</div>
<div class="flex items-center gap-2 shrink-0">
{#if isAdded}
<button
type="button"
onclick={onSelect}
class="px-3 py-1.5 text-xs font-mono tracking-wider uppercase bg-exo-yellow/10 text-exo-yellow border border-exo-yellow/30 hover:bg-exo-yellow/20 transition-colors rounded cursor-pointer"
>
Select
</button>
{:else}
<button
type="button"
onclick={onAdd}
disabled={isAdding}
class="px-3 py-1.5 text-xs font-mono tracking-wider uppercase bg-orange-500/10 text-orange-400 border border-orange-400/30 hover:bg-orange-500/20 transition-colors rounded cursor-pointer disabled:opacity-50 disabled:cursor-not-allowed"
>
{#if isAdding}
<span class="flex items-center gap-1.5">
<span
class="w-3 h-3 border-2 border-orange-400 border-t-transparent rounded-full animate-spin"
></span>
Adding...
</span>
{:else}
+ Add
{/if}
</button>
{/if}
</div>
</div>
@@ -0,0 +1,96 @@
<script lang="ts">
import { fade, fly } from "svelte/transition";
import { cubicOut } from "svelte/easing";
interface Props {
src: string | null;
onclose: () => void;
}
let { src, onclose }: Props = $props();
function handleKeydown(e: KeyboardEvent) {
if (e.key === "Escape") {
onclose();
}
}
function extensionFromSrc(dataSrc: string): string {
const match = dataSrc.match(/^data:image\/(\w+)/);
if (match) return match[1] === "jpeg" ? "jpg" : match[1];
const urlMatch = dataSrc.match(/\.(\w+)(?:\?|$)/);
if (urlMatch) return urlMatch[1];
return "png";
}
function handleDownload(e: MouseEvent) {
e.stopPropagation();
if (!src) return;
const link = document.createElement("a");
link.href = src;
link.download = `image-${Date.now()}.${extensionFromSrc(src)}`;
link.click();
}
function handleClose(e: MouseEvent) {
e.stopPropagation();
onclose();
}
</script>
<svelte:window onkeydown={src ? handleKeydown : undefined} />
{#if src}
<div
class="fixed inset-0 z-50 bg-black/90 backdrop-blur-sm flex items-center justify-center"
transition:fade={{ duration: 200 }}
onclick={onclose}
role="presentation"
onintrostart={() => (document.body.style.overflow = "hidden")}
onoutroend={() => (document.body.style.overflow = "")}
>
<div class="absolute top-4 right-4 flex gap-2 z-10">
<button
type="button"
class="p-2 rounded-lg bg-exo-dark-gray/80 border border-exo-yellow/30 text-exo-yellow hover:bg-exo-dark-gray hover:border-exo-yellow/50 cursor-pointer transition-colors"
onclick={handleDownload}
title="Download image"
>
<svg
class="w-5 h-5"
fill="none"
viewBox="0 0 24 24"
stroke="currentColor"
stroke-width="2"
>
<path
stroke-linecap="round"
stroke-linejoin="round"
d="M4 16v1a3 3 0 003 3h10a3 3 0 003-3v-1m-4-4l-4 4m0 0l-4-4m4 4V4"
/>
</svg>
</button>
<button
type="button"
class="p-2 rounded-lg bg-exo-dark-gray/80 border border-exo-yellow/30 text-exo-yellow hover:bg-exo-dark-gray hover:border-exo-yellow/50 cursor-pointer transition-colors"
onclick={handleClose}
title="Close"
>
<svg class="w-5 h-5" viewBox="0 0 24 24" fill="currentColor">
<path
d="M19 6.41L17.59 5 12 10.59 6.41 5 5 6.41 10.59 12 5 17.59 6.41 19 12 13.41 17.59 19 19 17.59 13.41 12 19 6.41z"
/>
</svg>
</button>
</div>
<!-- svelte-ignore a11y_no_noninteractive_element_interactions, a11y_click_events_have_key_events -->
<img
{src}
alt=""
class="max-w-[90vw] max-h-[90vh] object-contain rounded-lg shadow-2xl"
transition:fly={{ y: 20, duration: 300, easing: cubicOut }}
onclick={(e) => e.stopPropagation()}
/>
</div>
{/if}
@@ -59,11 +59,14 @@
}
const sizeOptions: ImageGenerationParams["size"][] = [
"auto",
"512x512",
"768x768",
"1024x1024",
"1024x768",
"768x1024",
"1024x1536",
"1536x1024",
];
const qualityOptions: ImageGenerationParams["quality"][] = [
@@ -110,6 +113,36 @@
setImageGenerationParams({ negativePrompt: value || null });
}
function handleNumImagesChange(event: Event) {
const input = event.target as HTMLInputElement;
const value = input.value.trim();
if (value === "") {
setImageGenerationParams({ numImages: 1 });
} else {
const num = parseInt(value, 10);
if (!isNaN(num) && num >= 1) {
setImageGenerationParams({ numImages: num });
}
}
}
function handleStreamChange(enabled: boolean) {
setImageGenerationParams({ stream: enabled });
}
function handlePartialImagesChange(event: Event) {
const input = event.target as HTMLInputElement;
const value = input.value.trim();
if (value === "") {
setImageGenerationParams({ partialImages: 0 });
} else {
const num = parseInt(value, 10);
if (!isNaN(num) && num >= 0) {
setImageGenerationParams({ partialImages: num });
}
}
}
function clearSteps() {
setImageGenerationParams({ numInferenceSteps: null });
}
@@ -118,6 +151,15 @@
setImageGenerationParams({ guidance: null });
}
function handleNumSyncStepsChange(event: Event) {
const value = parseInt((event.target as HTMLInputElement).value, 10);
setImageGenerationParams({ numSyncSteps: value });
}
function clearNumSyncSteps() {
setImageGenerationParams({ numSyncSteps: null });
}
function handleReset() {
resetImageGenerationParams();
showAdvanced = false;
@@ -127,7 +169,8 @@
params.seed !== null ||
params.numInferenceSteps !== null ||
params.guidance !== null ||
(params.negativePrompt !== null && params.negativePrompt.trim() !== ""),
(params.negativePrompt !== null && params.negativePrompt.trim() !== "") ||
params.numSyncSteps !== null,
);
</script>
@@ -148,7 +191,7 @@
? 'border-exo-yellow/70'
: ''}"
>
{params.size}
{params.size.toUpperCase()}
</button>
<div
class="absolute right-1.5 top-1/2 -translate-y-1/2 pointer-events-none transition-transform duration-200 {isSizeDropdownOpen
@@ -182,7 +225,7 @@
<!-- Dropdown Panel - fixed positioning to escape overflow:hidden -->
<div
class="fixed bg-exo-dark-gray border border-exo-yellow/30 rounded shadow-lg shadow-black/50 z-[9999] max-h-48 overflow-y-auto min-w-max"
class="fixed bg-exo-dark-gray border border-exo-yellow/30 rounded shadow-lg shadow-black/50 z-[9999] max-h-48 overflow-y-auto overflow-x-hidden min-w-max"
style="bottom: calc(100vh - {sizeDropdownPosition()
.top}px + 4px); left: {sizeDropdownPosition().left}px;"
>
@@ -211,7 +254,7 @@
{:else}
<span class="w-3"></span>
{/if}
<span>{size}</span>
<span>{size.toUpperCase()}</span>
</button>
{/each}
</div>
@@ -267,7 +310,7 @@
<!-- Dropdown Panel - fixed positioning to escape overflow:hidden -->
<div
class="fixed bg-exo-dark-gray border border-exo-yellow/30 rounded shadow-lg shadow-black/50 z-[9999] max-h-48 overflow-y-auto min-w-max"
class="fixed bg-exo-dark-gray border border-exo-yellow/30 rounded shadow-lg shadow-black/50 z-[9999] max-h-48 overflow-y-auto overflow-x-hidden min-w-max"
style="bottom: calc(100vh - {qualityDropdownPosition()
.top}px + 4px); left: {qualityDropdownPosition().left}px;"
>
@@ -325,6 +368,59 @@
</div>
</div>
<!-- Number of Images (not in edit mode) -->
{#if !isEditMode}
<div class="flex items-center gap-1.5">
<span class="text-xs text-exo-light-gray uppercase tracking-wider"
>IMAGES:</span
>
<input
type="number"
min="1"
value={params.numImages}
oninput={handleNumImagesChange}
class="w-12 bg-exo-medium-gray/50 border border-exo-yellow/30 rounded px-2 py-1 text-xs font-mono text-exo-yellow text-center transition-all duration-200 hover:border-exo-yellow/50 focus:outline-none focus:border-exo-yellow/70"
/>
</div>
{/if}
<!-- Stream toggle -->
<div class="flex items-center gap-1.5">
<span class="text-xs text-exo-light-gray uppercase tracking-wider"
>STREAM:</span
>
<button
type="button"
onclick={() => handleStreamChange(!params.stream)}
class="w-8 h-4 rounded-full transition-all duration-200 cursor-pointer relative {params.stream
? 'bg-exo-yellow'
: 'bg-exo-medium-gray/50 border border-exo-yellow/30'}"
title={params.stream ? "Streaming enabled" : "Streaming disabled"}
>
<div
class="absolute top-0.5 w-3 h-3 rounded-full transition-all duration-200 {params.stream
? 'right-0.5 bg-exo-black'
: 'left-0.5 bg-exo-light-gray'}"
></div>
</button>
</div>
<!-- Partial Images (only when streaming) -->
{#if params.stream}
<div class="flex items-center gap-1.5">
<span class="text-xs text-exo-light-gray uppercase tracking-wider"
>PARTIALS:</span
>
<input
type="number"
min="0"
value={params.partialImages}
oninput={handlePartialImagesChange}
class="w-12 bg-exo-medium-gray/50 border border-exo-yellow/30 rounded px-2 py-1 text-xs font-mono text-exo-yellow text-center transition-all duration-200 hover:border-exo-yellow/50 focus:outline-none focus:border-exo-yellow/70"
/>
</div>
{/if}
<!-- Input Fidelity (edit mode only) -->
{#if isEditMode}
<div class="flex items-center gap-1.5">
@@ -493,7 +589,50 @@
</div>
</div>
<!-- Row 3: Negative Prompt -->
<!-- Row 3: Sync Steps -->
<div class="flex items-center gap-1.5">
<span
class="text-xs text-exo-light-gray uppercase tracking-wider whitespace-nowrap"
>SYNC STEPS:</span
>
<div class="flex items-center gap-2 flex-1 max-w-xs">
<input
type="range"
min="1"
max="100"
value={params.numSyncSteps ?? 1}
oninput={handleNumSyncStepsChange}
class="flex-1 h-1 bg-exo-medium-gray/50 rounded appearance-none cursor-pointer accent-exo-yellow"
/>
<span class="text-xs font-mono text-exo-yellow w-8 text-right">
{params.numSyncSteps ?? "--"}
</span>
{#if params.numSyncSteps !== null}
<button
type="button"
onclick={clearNumSyncSteps}
class="text-exo-light-gray hover:text-exo-yellow transition-colors"
title="Clear"
>
<svg
class="w-3 h-3"
fill="none"
viewBox="0 0 24 24"
stroke="currentColor"
>
<path
stroke-linecap="round"
stroke-linejoin="round"
stroke-width="2"
d="M6 18L18 6M6 6l12 12"
/>
</svg>
</button>
{/if}
</div>
</div>
<!-- Row 4: Negative Prompt -->
<div class="flex flex-col gap-1.5">
<span class="text-xs text-exo-light-gray uppercase tracking-wider"
>NEGATIVE PROMPT:</span
+39 -6
View File
@@ -21,7 +21,7 @@
} | null;
nodes?: Record<string, NodeInfo>;
sharding?: "Pipeline" | "Tensor";
runtime?: "MlxRing" | "MlxIbv" | "MlxJaccl";
runtime?: "MlxRing" | "MlxJaccl";
onLaunch?: () => void;
tags?: string[];
apiPreview?: PlacementPreview | null;
@@ -348,7 +348,7 @@
// Debug mode state
const isDebugMode = $derived(debugMode());
const topology = $derived(topologyData());
const isRdma = $derived(runtime === "MlxIbv" || runtime === "MlxJaccl");
const isRdma = $derived(runtime === "MlxJaccl");
// Get interface name for an IP from node data
function getInterfaceForIp(nodeId: string, ip?: string): string | null {
@@ -422,9 +422,16 @@
const bToACandidates: Array<{ ip: string; iface: string | null }> = [];
for (const edge of topology.edges) {
const ip = edge.sendBackIp || "?";
const iface =
edge.sendBackInterface || getInterfaceForIp(edge.source, ip);
let ip: string;
let iface: string | null;
if (edge.sourceRdmaIface || edge.sinkRdmaIface) {
ip = "RDMA";
iface = `${edge.sourceRdmaIface || "?"} \u2192 ${edge.sinkRdmaIface || "?"}`;
} else {
ip = edge.sendBackIp || "?";
iface = edge.sendBackInterface || getInterfaceForIp(edge.source, ip);
}
if (edge.source === nodeId1 && edge.target === nodeId2) {
aToBCandidates.push({ ip, iface });
@@ -560,20 +567,46 @@
<div class="flex items-center gap-1.5 mb-2">
<span
class="px-1.5 py-0.5 text-xs font-mono tracking-wider uppercase bg-exo-medium-gray/30 text-exo-light-gray border border-exo-medium-gray/40"
title={sharding === "Pipeline"
? "Pipeline: splits model into sequential stages across devices. Lower network overhead."
: "Tensor: splits each layer across devices. Best with high-bandwidth connections (Thunderbolt)."}
>
{sharding}
</span>
<span
class="px-1.5 py-0.5 text-xs font-mono tracking-wider uppercase bg-exo-medium-gray/30 text-exo-light-gray border border-exo-medium-gray/40"
title={runtime === "MlxRing"
? "Ring: standard networking. Works over any connection (Wi-Fi, Ethernet, Thunderbolt)."
: "RDMA: direct memory access over Thunderbolt. Significantly faster for multi-device inference."}
>
{runtime === "MlxRing"
? "MLX Ring"
: runtime === "MlxIbv" || runtime === "MlxJaccl"
: runtime === "MlxJaccl"
? "MLX RDMA"
: runtime}
</span>
</div>
<!-- Download Status -->
{#if isDownloading && progress}
<div class="mb-2 space-y-1">
<div class="flex items-center justify-between text-xs font-mono">
<span class="text-blue-400 tracking-wider uppercase">Downloading</span
>
<span class="text-white/60"
>{percentage.toFixed(1)}% &middot; {formatSpeed(progress.speed)}
&middot; {formatEta(progress.etaMs)}</span
>
</div>
<div class="h-1 bg-exo-medium-gray/30 rounded overflow-hidden">
<div
class="h-full bg-blue-500/70 transition-all duration-300"
style="width: {percentage}%"
></div>
</div>
</div>
{/if}
<!-- Mini Topology Preview -->
{#if placementPreview().nodes.length > 0}
{@const preview = placementPreview()}
@@ -0,0 +1,279 @@
<script lang="ts">
import { fly } from "svelte/transition";
import { cubicOut } from "svelte/easing";
interface FilterState {
capabilities: string[];
sizeRange: { min: number; max: number } | null;
downloadedOnly: boolean;
readyOnly: boolean;
}
type ModelFilterPopoverProps = {
filters: FilterState;
onChange: (filters: FilterState) => void;
onClear: () => void;
onClose: () => void;
};
let { filters, onChange, onClear, onClose }: ModelFilterPopoverProps =
$props();
// Available capabilities
const availableCapabilities = [
{ id: "text", label: "Text" },
{ id: "thinking", label: "Thinking" },
{ id: "code", label: "Code" },
{ id: "vision", label: "Vision" },
{ id: "image_gen", label: "Image Gen" },
{ id: "image_edit", label: "Image Edit" },
];
// Size ranges
const sizeRanges = [
{ label: "< 10GB", min: 0, max: 10 },
{ label: "10-50GB", min: 10, max: 50 },
{ label: "50-200GB", min: 50, max: 200 },
{ label: "> 200GB", min: 200, max: 10000 },
];
function toggleCapability(cap: string) {
const next = filters.capabilities.includes(cap)
? filters.capabilities.filter((c) => c !== cap)
: [...filters.capabilities, cap];
onChange({ ...filters, capabilities: next });
}
function selectSizeRange(range: { min: number; max: number } | null) {
// Toggle off if same range is clicked
if (
filters.sizeRange &&
range &&
filters.sizeRange.min === range.min &&
filters.sizeRange.max === range.max
) {
onChange({ ...filters, sizeRange: null });
} else {
onChange({ ...filters, sizeRange: range });
}
}
function handleClickOutside(e: MouseEvent) {
const target = e.target as HTMLElement;
if (
!target.closest(".filter-popover") &&
!target.closest(".filter-toggle")
) {
onClose();
}
}
</script>
<svelte:window onclick={handleClickOutside} />
<!-- svelte-ignore a11y_no_static_element_interactions -->
<div
class="filter-popover absolute right-0 top-full mt-2 w-64 bg-exo-dark-gray border border-exo-yellow/10 rounded-lg shadow-xl z-10"
transition:fly={{ y: -10, duration: 200, easing: cubicOut }}
onclick={(e) => e.stopPropagation()}
role="dialog"
aria-label="Filter options"
>
<div class="p-3 space-y-4">
<!-- Capabilities -->
<div>
<h4 class="text-xs font-mono text-white/50 mb-2">Capabilities</h4>
<div class="flex flex-wrap gap-1.5">
{#each availableCapabilities as cap}
{@const isSelected = filters.capabilities.includes(cap.id)}
<button
type="button"
class="px-2 py-1 text-xs font-mono rounded transition-colors {isSelected
? 'bg-exo-yellow/20 text-exo-yellow border border-exo-yellow/30'
: 'bg-white/5 text-white/60 hover:bg-white/10 border border-transparent'}"
onclick={() => toggleCapability(cap.id)}
>
{#if cap.id === "text"}
<svg
class="w-3.5 h-3.5 inline-block"
viewBox="0 0 24 24"
fill="none"
stroke="currentColor"
stroke-width="1.5"
><path
d="M21 15a2 2 0 0 1-2 2H7l-4 4V5a2 2 0 0 1 2-2h14a2 2 0 0 1 2 2z"
stroke-linecap="round"
stroke-linejoin="round"
/></svg
>
{:else if cap.id === "thinking"}
<svg
class="w-3.5 h-3.5 inline-block"
viewBox="0 0 24 24"
fill="none"
stroke="currentColor"
stroke-width="1.5"
><path
d="M12 2a7 7 0 0 0-7 7c0 2.38 1.19 4.47 3 5.74V17a1 1 0 0 0 1 1h6a1 1 0 0 0 1-1v-2.26c1.81-1.27 3-3.36 3-5.74a7 7 0 0 0-7-7zM9 20h6M10 22h4"
stroke-linecap="round"
stroke-linejoin="round"
/></svg
>
{:else if cap.id === "code"}
<svg
class="w-3.5 h-3.5 inline-block"
viewBox="0 0 24 24"
fill="none"
stroke="currentColor"
stroke-width="1.5"
><path
d="M16 18l6-6-6-6M8 6l-6 6 6 6"
stroke-linecap="round"
stroke-linejoin="round"
/></svg
>
{:else if cap.id === "vision"}
<svg
class="w-3.5 h-3.5 inline-block"
viewBox="0 0 24 24"
fill="none"
stroke="currentColor"
stroke-width="1.5"
><path
d="M1 12s4-8 11-8 11 8 11 8-4 8-11 8-11-8-11-8z"
stroke-linecap="round"
stroke-linejoin="round"
/><circle cx="12" cy="12" r="3" /></svg
>
{:else if cap.id === "image_gen"}
<svg
class="w-3.5 h-3.5 inline-block"
viewBox="0 0 24 24"
fill="none"
stroke="currentColor"
stroke-width="1.5"
><rect
x="3"
y="3"
width="18"
height="18"
rx="2"
ry="2"
stroke-linecap="round"
stroke-linejoin="round"
/><circle cx="8.5" cy="8.5" r="1.5" /><path
d="M21 15l-5-5L5 21"
stroke-linecap="round"
stroke-linejoin="round"
/></svg
>
{:else if cap.id === "image_edit"}
<svg
class="w-3.5 h-3.5 inline-block"
viewBox="0 0 24 24"
fill="none"
stroke="currentColor"
stroke-width="1.5"
><path
d="M11 4H4a2 2 0 0 0-2 2v14a2 2 0 0 0 2 2h14a2 2 0 0 0 2-2v-7"
stroke-linecap="round"
stroke-linejoin="round"
/><path
d="M18.5 2.5a2.121 2.121 0 0 1 3 3L12 15l-4 1 1-4 9.5-9.5z"
stroke-linecap="round"
stroke-linejoin="round"
/></svg
>
{/if}
<span class="ml-1">{cap.label}</span>
</button>
{/each}
</div>
</div>
<!-- Availability filters -->
<div>
<h4 class="text-xs font-mono text-white/50 mb-2">Availability</h4>
<div class="flex flex-wrap gap-1.5">
<button
type="button"
class="px-2 py-1 text-xs font-mono rounded transition-colors {filters.downloadedOnly
? 'bg-green-500/20 text-green-400 border border-green-500/30'
: 'bg-white/5 text-white/60 hover:bg-white/10 border border-transparent'}"
onclick={() =>
onChange({ ...filters, downloadedOnly: !filters.downloadedOnly })}
>
<svg
class="w-3.5 h-3.5 inline-block"
viewBox="0 0 24 24"
fill="none"
stroke="currentColor"
stroke-width="2"
stroke-linecap="round"
stroke-linejoin="round"
>
<path
class="text-white/40"
d="M20 20a2 2 0 0 0 2-2V8a2 2 0 0 0-2-2h-7.9a2 2 0 0 1-1.69-.9L9.6 3.9A2 2 0 0 0 7.93 3H4a2 2 0 0 0-2 2v13a2 2 0 0 0 2 2Z"
/>
<path class="text-green-400" d="m9 13 2 2 4-4" />
</svg>
<span class="ml-1">Downloaded</span>
</button>
<button
type="button"
class="px-2 py-1 text-xs font-mono rounded transition-colors {filters.readyOnly
? 'bg-green-500/20 text-green-400 border border-green-500/30'
: 'bg-white/5 text-white/60 hover:bg-white/10 border border-transparent'}"
onclick={() =>
onChange({ ...filters, readyOnly: !filters.readyOnly })}
>
<svg
class="w-3.5 h-3.5 inline-block"
viewBox="0 0 24 24"
fill="none"
stroke="currentColor"
stroke-width="2"
stroke-linecap="round"
stroke-linejoin="round"
>
<circle cx="12" cy="12" r="10" />
<path d="m9 12 2 2 4-4" />
</svg>
<span class="ml-1">Ready</span>
</button>
</div>
</div>
<!-- Size range -->
<div>
<h4 class="text-xs font-mono text-white/50 mb-2">Model Size</h4>
<div class="flex flex-wrap gap-1.5">
{#each sizeRanges as range}
{@const isSelected =
filters.sizeRange &&
filters.sizeRange.min === range.min &&
filters.sizeRange.max === range.max}
<button
type="button"
class="px-2 py-1 text-xs font-mono rounded transition-colors {isSelected
? 'bg-exo-yellow/20 text-exo-yellow border border-exo-yellow/30'
: 'bg-white/5 text-white/60 hover:bg-white/10 border border-transparent'}"
onclick={() => selectSizeRange(range)}
>
{range.label}
</button>
{/each}
</div>
</div>
<!-- Clear button -->
<button
type="button"
class="w-full py-1.5 text-xs font-mono text-white/50 hover:text-white/70 hover:bg-white/5 rounded transition-colors"
onclick={onClear}
>
Clear all filters
</button>
</div>
</div>
@@ -0,0 +1,646 @@
<script lang="ts">
interface ModelInfo {
id: string;
name?: string;
storage_size_megabytes?: number;
base_model?: string;
quantization?: string;
supports_tensor?: boolean;
capabilities?: string[];
family?: string;
is_custom?: boolean;
}
interface ModelGroup {
id: string;
name: string;
capabilities: string[];
family: string;
variants: ModelInfo[];
smallestVariant: ModelInfo;
hasMultipleVariants: boolean;
}
type DownloadAvailability = {
available: boolean;
nodeNames: string[];
nodeIds: string[];
};
type ModelFitStatus = "fits_now" | "fits_cluster_capacity" | "too_large";
type ModelPickerGroupProps = {
group: ModelGroup;
isExpanded: boolean;
isFavorite: boolean;
selectedModelId: string | null;
canModelFit: (id: string) => boolean;
getModelFitStatus: (id: string) => ModelFitStatus;
onToggleExpand: () => void;
onSelectModel: (modelId: string) => void;
onToggleFavorite: (baseModelId: string) => void;
onShowInfo: (group: ModelGroup) => void;
downloadStatusMap?: Map<string, DownloadAvailability>;
launchedAt?: number;
instanceStatuses?: Record<string, { status: string; statusClass: string }>;
};
let {
group,
isExpanded,
isFavorite,
selectedModelId,
canModelFit,
getModelFitStatus,
onToggleExpand,
onSelectModel,
onToggleFavorite,
onShowInfo,
downloadStatusMap,
launchedAt,
instanceStatuses = {},
}: ModelPickerGroupProps = $props();
// Group-level download status: show if any variant is downloaded
const groupDownloadStatus = $derived.by(() => {
if (!downloadStatusMap || downloadStatusMap.size === 0) return undefined;
// Return the first available entry (prefer "available" ones)
for (const avail of downloadStatusMap.values()) {
if (avail.available) return avail;
}
return downloadStatusMap.values().next().value;
});
// Format storage size
function formatSize(mb: number | undefined): string {
if (!mb) return "";
if (mb >= 1024) {
return `${(mb / 1024).toFixed(0)}GB`;
}
return `${mb}MB`;
}
function timeAgo(ts: number): string {
const seconds = Math.floor((Date.now() - ts) / 1000);
if (seconds < 60) return "just now";
const minutes = Math.floor(seconds / 60);
if (minutes < 60) return `${minutes}m ago`;
const hours = Math.floor(minutes / 60);
if (hours < 24) return `${hours}h ago`;
const days = Math.floor(hours / 24);
return `${days}d ago`;
}
// Check if any variant can fit
const anyVariantFits = $derived(
group.variants.some((v) => canModelFit(v.id)),
);
// Check if any variant has an active instance (ready, loading, downloading)
const anyVariantHasInstance = $derived(
instanceStatuses
? group.variants.some((v) => instanceStatuses[v.id] != null)
: false,
);
const groupFitStatus = $derived.by((): ModelFitStatus => {
let hasClusterCapacityOnly = false;
for (const variant of group.variants) {
const fitStatus = getModelFitStatus(variant.id);
if (fitStatus === "fits_now") {
return "fits_now";
}
if (fitStatus === "fits_cluster_capacity") {
hasClusterCapacityOnly = true;
}
}
return hasClusterCapacityOnly ? "fits_cluster_capacity" : "too_large";
});
function getSizeClassForFitStatus(fitStatus: ModelFitStatus): string {
switch (fitStatus) {
case "fits_now":
return "text-white/40";
case "fits_cluster_capacity":
return "text-orange-400/80";
case "too_large":
return "text-red-400/70";
}
}
// Check if this group's model is currently selected (for single-variant groups)
const isMainSelected = $derived(
!group.hasMultipleVariants &&
group.variants.some((v) => v.id === selectedModelId),
);
// Group-level instance status: show the "best" status across all variants
const groupInstanceStatus = $derived.by(() => {
if (!instanceStatuses) return null;
const readyStatuses = ["READY", "LOADED", "RUNNING"];
const loadingStatuses = ["LOADING", "WARMING UP"];
let bestStatus: { status: string; statusClass: string } | null = null;
for (const variant of group.variants) {
const s = instanceStatuses[variant.id];
if (!s) continue;
if (readyStatuses.includes(s.status)) return s; // Ready is best
if (loadingStatuses.includes(s.status) || s.status === "DOWNLOADING") {
bestStatus = s;
}
}
return bestStatus;
});
</script>
<div
class="border-b border-white/5 last:border-b-0 {!anyVariantFits &&
!anyVariantHasInstance
? 'opacity-50'
: ''}"
>
<!-- Main row -->
<div
class="flex items-center gap-2 px-3 py-2.5 transition-colors {anyVariantFits ||
anyVariantHasInstance
? 'hover:bg-white/5 cursor-pointer'
: 'cursor-not-allowed'} {isMainSelected
? 'bg-exo-yellow/10 border-l-2 border-exo-yellow'
: 'border-l-2 border-transparent'}"
onclick={() => {
if (group.hasMultipleVariants) {
onToggleExpand();
} else {
const modelId = group.variants[0]?.id;
if (modelId && (canModelFit(modelId) || instanceStatuses[modelId])) {
onSelectModel(modelId);
}
}
}}
role="button"
tabindex="0"
onkeydown={(e) => {
if (e.key === "Enter" || e.key === " ") {
e.preventDefault();
if (group.hasMultipleVariants) {
onToggleExpand();
} else {
const modelId = group.variants[0]?.id;
if (modelId && (canModelFit(modelId) || instanceStatuses[modelId])) {
onSelectModel(modelId);
}
}
}
}}
>
<!-- Expand/collapse chevron (for groups with variants) -->
{#if group.hasMultipleVariants}
<svg
class="w-4 h-4 text-white/40 transition-transform duration-200 flex-shrink-0 {isExpanded
? 'rotate-90'
: ''}"
viewBox="0 0 24 24"
fill="currentColor"
>
<path d="M8.59 16.59L13.17 12 8.59 7.41 10 6l6 6-6 6-1.41-1.41z" />
</svg>
{:else}
<div class="w-4 flex-shrink-0"></div>
{/if}
<!-- Model name -->
<div class="flex-1 min-w-0">
<div class="flex items-center gap-2">
<span class="font-mono text-sm text-white truncate">
{group.name}
</span>
<!-- Capability icons -->
{#each group.capabilities.filter((c) => c !== "text") as cap}
{#if cap === "thinking"}
<svg
class="w-3.5 h-3.5 text-white/40 flex-shrink-0"
viewBox="0 0 24 24"
fill="none"
stroke="currentColor"
stroke-width="1.5"
title="Supports Thinking"
>
<path
d="M12 2a7 7 0 0 0-7 7c0 2.38 1.19 4.47 3 5.74V17a1 1 0 0 0 1 1h6a1 1 0 0 0 1-1v-2.26c1.81-1.27 3-3.36 3-5.74a7 7 0 0 0-7-7zM9 20h6M10 22h4"
stroke-linecap="round"
stroke-linejoin="round"
/>
</svg>
{:else if cap === "code"}
<svg
class="w-3.5 h-3.5 text-white/40 flex-shrink-0"
viewBox="0 0 24 24"
fill="none"
stroke="currentColor"
stroke-width="1.5"
title="Supports code generation"
>
<path
d="M16 18l6-6-6-6M8 6l-6 6 6 6"
stroke-linecap="round"
stroke-linejoin="round"
/>
</svg>
{:else if cap === "vision"}
<svg
class="w-3.5 h-3.5 text-white/40 flex-shrink-0"
viewBox="0 0 24 24"
fill="none"
stroke="currentColor"
stroke-width="1.5"
title="Supports image input"
>
<path
d="M1 12s4-8 11-8 11 8 11 8-4 8-11 8-11-8-11-8z"
stroke-linecap="round"
stroke-linejoin="round"
/>
<circle cx="12" cy="12" r="3" />
</svg>
{:else if cap === "image_gen"}
<svg
class="w-3.5 h-3.5 text-white/40 flex-shrink-0"
viewBox="0 0 24 24"
fill="none"
stroke="currentColor"
stroke-width="1.5"
title="Supports image generation"
>
<rect
x="3"
y="3"
width="18"
height="18"
rx="2"
ry="2"
stroke-linecap="round"
stroke-linejoin="round"
/>
<circle cx="8.5" cy="8.5" r="1.5" />
<path
d="M21 15l-5-5L5 21"
stroke-linecap="round"
stroke-linejoin="round"
/>
</svg>
{:else if cap === "image_edit"}
<svg
class="w-3.5 h-3.5 text-white/40 flex-shrink-0"
viewBox="0 0 24 24"
fill="none"
stroke="currentColor"
stroke-width="1.5"
title="Supports image editing"
>
<path
d="M11 4H4a2 2 0 0 0-2 2v14a2 2 0 0 0 2 2h14a2 2 0 0 0 2-2v-7"
stroke-linecap="round"
stroke-linejoin="round"
/>
<path
d="M18.5 2.5a2.121 2.121 0 0 1 3 3L12 15l-4 1 1-4 9.5-9.5z"
stroke-linecap="round"
stroke-linejoin="round"
/>
</svg>
{/if}
{/each}
</div>
</div>
<!-- Size indicator (smallest variant) -->
{#if !group.hasMultipleVariants && group.smallestVariant?.storage_size_megabytes}
{@const singleVariantFitStatus = getModelFitStatus(
group.smallestVariant.id,
)}
<span
class="text-xs font-mono flex-shrink-0 {getSizeClassForFitStatus(
singleVariantFitStatus,
)}"
>
{formatSize(group.smallestVariant.storage_size_megabytes)}
</span>
{/if}
<!-- Variant count with size range -->
{#if group.hasMultipleVariants}
{@const sizes = group.variants
.map((v) => v.storage_size_megabytes || 0)
.filter((s) => s > 0)
.sort((a, b) => a - b)}
<span
class="text-xs font-mono flex-shrink-0 {getSizeClassForFitStatus(
groupFitStatus,
)}"
>
{group.variants.length} variants{#if sizes.length >= 2}{" "}({formatSize(
sizes[0],
)}-{formatSize(sizes[sizes.length - 1])}){/if}
</span>
{/if}
<!-- Time ago (for recent models) -->
{#if launchedAt}
<span class="text-xs font-mono text-white/20 flex-shrink-0">
{timeAgo(launchedAt)}
</span>
{/if}
<!-- Download availability indicator -->
{#if groupDownloadStatus && groupDownloadStatus.nodeIds.length > 0}
<span
class="flex-shrink-0"
title={groupDownloadStatus.available
? `Ready — downloaded on ${groupDownloadStatus.nodeNames.join(", ")}`
: `Downloaded on ${groupDownloadStatus.nodeNames.join(", ")} (may need more nodes)`}
>
<svg
class="w-4 h-4"
viewBox="0 0 24 24"
fill="none"
stroke="currentColor"
stroke-width="2"
stroke-linecap="round"
stroke-linejoin="round"
>
<path
class="text-white/40"
d="M20 20a2 2 0 0 0 2-2V8a2 2 0 0 0-2-2h-7.9a2 2 0 0 1-1.69-.9L9.6 3.9A2 2 0 0 0 7.93 3H4a2 2 0 0 0-2 2v13a2 2 0 0 0 2 2Z"
/>
<path class="text-green-400" d="m9 13 2 2 4-4" />
</svg>
</span>
{/if}
<!-- Instance status badge -->
{#if groupInstanceStatus}
{#if groupInstanceStatus.status === "READY" || groupInstanceStatus.status === "LOADED" || groupInstanceStatus.status === "RUNNING"}
<span class="flex-shrink-0" title="Running">
<svg
class="w-3 h-3 text-green-400"
viewBox="0 0 12 12"
fill="currentColor"
>
<circle cx="6" cy="6" r="5" />
</svg>
</span>
{:else if groupInstanceStatus.status === "DOWNLOADING"}
<span class="flex-shrink-0 animate-pulse" title="Downloading">
<svg
class="w-3.5 h-3.5 text-blue-400"
viewBox="0 0 24 24"
fill="none"
stroke="currentColor"
stroke-width="2"
stroke-linecap="round"
stroke-linejoin="round"
>
<path d="M21 15v4a2 2 0 01-2 2H5a2 2 0 01-2-2v-4" />
<polyline points="7 10 12 15 17 10" />
<line x1="12" y1="15" x2="12" y2="3" />
</svg>
</span>
{:else if groupInstanceStatus.status === "LOADING" || groupInstanceStatus.status === "WARMING UP"}
<span class="flex-shrink-0 animate-pulse" title="Loading">
<svg
class="w-3 h-3 text-yellow-400"
viewBox="0 0 12 12"
fill="currentColor"
>
<circle cx="6" cy="6" r="5" />
</svg>
</span>
{/if}
{/if}
<!-- Check mark if selected (single-variant) -->
{#if isMainSelected}
<svg
class="w-4 h-4 text-exo-yellow flex-shrink-0"
viewBox="0 0 24 24"
fill="currentColor"
>
<path d="M9 16.17L4.83 12l-1.42 1.41L9 19 21 7l-1.41-1.41L9 16.17z" />
</svg>
{/if}
<!-- Favorite star -->
<button
type="button"
class="p-1 rounded hover:bg-white/10 transition-colors flex-shrink-0"
onclick={(e) => {
e.stopPropagation();
onToggleFavorite(group.id);
}}
title={isFavorite ? "Remove from favorites" : "Add to favorites"}
>
{#if isFavorite}
<svg
class="w-4 h-4 text-amber-400"
viewBox="0 0 24 24"
fill="currentColor"
>
<path
d="M12 2l3.09 6.26L22 9.27l-5 4.87 1.18 6.88L12 17.77l-6.18 3.25L7 14.14 2 9.27l6.91-1.01L12 2z"
/>
</svg>
{:else}
<svg
class="w-4 h-4 text-white/30 hover:text-white/50"
viewBox="0 0 24 24"
fill="none"
stroke="currentColor"
stroke-width="2"
>
<path
d="M12 2l3.09 6.26L22 9.27l-5 4.87 1.18 6.88L12 17.77l-6.18 3.25L7 14.14 2 9.27l6.91-1.01L12 2z"
/>
</svg>
{/if}
</button>
<!-- Info button -->
<button
type="button"
class="p-1 rounded hover:bg-white/10 transition-colors flex-shrink-0"
onclick={(e) => {
e.stopPropagation();
onShowInfo(group);
}}
title="View model details"
>
<svg
class="w-4 h-4 text-white/30 hover:text-white/50"
viewBox="0 0 24 24"
fill="currentColor"
>
<path
d="M12 2C6.48 2 2 6.48 2 12s4.48 10 10 10 10-4.48 10-10S17.52 2 12 2zm1 15h-2v-6h2v6zm0-8h-2V7h2v2z"
/>
</svg>
</button>
</div>
<!-- Expanded variants -->
{#if isExpanded && group.hasMultipleVariants}
<div class="bg-black/20 border-t border-white/5">
{#each group.variants as variant}
{@const fitStatus = getModelFitStatus(variant.id)}
{@const modelCanFit = canModelFit(variant.id)}
{@const variantHasInstance = instanceStatuses[variant.id] != null}
{@const isSelected = selectedModelId === variant.id}
<div
class="w-full flex items-center gap-3 px-3 py-2 pl-10 hover:bg-white/5 transition-colors text-left {!modelCanFit &&
!variantHasInstance
? 'opacity-50 cursor-not-allowed'
: 'cursor-pointer'} {isSelected
? 'bg-exo-yellow/10 border-l-2 border-exo-yellow'
: 'border-l-2 border-transparent'}"
role="button"
tabindex="0"
onclick={() => {
if (modelCanFit || variantHasInstance) {
onSelectModel(variant.id);
}
}}
onkeydown={(e) => {
if (e.key === "Enter" || e.key === " ") {
e.preventDefault();
if (modelCanFit) {
onSelectModel(variant.id);
}
}
}}
>
<!-- Quantization badge -->
<span
class="text-xs font-mono px-1.5 py-0.5 rounded bg-white/10 text-white/70 flex-shrink-0"
>
{variant.quantization || "default"}
</span>
<!-- Size -->
<span
class="text-xs font-mono flex-1 {getSizeClassForFitStatus(
fitStatus,
)}"
>
{formatSize(variant.storage_size_megabytes)}
</span>
<!-- Download indicator for this variant -->
{#if downloadStatusMap?.get(variant.id)}
{@const variantDl = downloadStatusMap.get(variant.id)}
{#if variantDl}
<span
class="flex-shrink-0"
title={`Downloaded on ${variantDl.nodeNames.join(", ")}`}
>
<svg
class="w-3.5 h-3.5"
viewBox="0 0 24 24"
fill="none"
stroke="currentColor"
stroke-width="2"
stroke-linecap="round"
stroke-linejoin="round"
>
<path
class="text-white/40"
d="M20 20a2 2 0 0 0 2-2V8a2 2 0 0 0-2-2h-7.9a2 2 0 0 1-1.69-.9L9.6 3.9A2 2 0 0 0 7.93 3H4a2 2 0 0 0-2 2v13a2 2 0 0 0 2 2Z"
/>
<path class="text-green-400" d="m9 13 2 2 4-4" />
</svg>
</span>
{/if}
{/if}
<!-- Instance status badge -->
{#if instanceStatuses[variant.id]}
{@const instStatus = instanceStatuses[variant.id]}
{#if instStatus.status === "READY" || instStatus.status === "LOADED" || instStatus.status === "RUNNING"}
<span class="flex-shrink-0" title="Running">
<svg
class="w-3 h-3 text-green-400"
viewBox="0 0 12 12"
fill="currentColor"
>
<circle cx="6" cy="6" r="5" />
</svg>
</span>
{:else if instStatus.status === "DOWNLOADING"}
<span class="flex-shrink-0 animate-pulse" title="Downloading">
<svg
class="w-3.5 h-3.5 text-blue-400"
viewBox="0 0 24 24"
fill="none"
stroke="currentColor"
stroke-width="2"
stroke-linecap="round"
stroke-linejoin="round"
>
<path d="M21 15v4a2 2 0 01-2 2H5a2 2 0 01-2-2v-4" />
<polyline points="7 10 12 15 17 10" />
<line x1="12" y1="15" x2="12" y2="3" />
</svg>
</span>
{:else if instStatus.status === "LOADING" || instStatus.status === "WARMING UP"}
<span class="flex-shrink-0 animate-pulse" title="Loading">
<svg
class="w-3 h-3 text-yellow-400"
viewBox="0 0 12 12"
fill="currentColor"
>
<circle cx="6" cy="6" r="5" />
</svg>
</span>
{/if}
{/if}
<!-- Check mark if selected -->
{#if isSelected}
<svg
class="w-4 h-4 text-exo-yellow"
viewBox="0 0 24 24"
fill="currentColor"
>
<path
d="M9 16.17L4.83 12l-1.42 1.41L9 19 21 7l-1.41-1.41L9 16.17z"
/>
</svg>
{/if}
<!-- Info button -->
<button
type="button"
class="p-1 rounded hover:bg-white/10 transition-colors flex-shrink-0"
onclick={(e) => {
e.stopPropagation();
onShowInfo({
id: variant.id,
name: variant.name || variant.id,
capabilities: group.capabilities,
family: group.family,
variants: [variant],
smallestVariant: variant,
hasMultipleVariants: false,
});
}}
title="View variant details"
>
<svg
class="w-4 h-4 text-white/30 hover:text-white/50"
viewBox="0 0 24 24"
fill="currentColor"
>
<path
d="M12 2C6.48 2 2 6.48 2 12s4.48 10 10 10 10-4.48 10-10S17.52 2 12 2zm1 15h-2v-6h2v6zm0-8h-2V7h2v2z"
/>
</svg>
</button>
</div>
{/each}
</div>
{/if}
</div>
File diff suppressed because it is too large Load Diff
@@ -0,0 +1,70 @@
<script lang="ts">
import type { PrefillProgress } from "$lib/stores/app.svelte";
interface Props {
progress: PrefillProgress;
class?: string;
}
let { progress, class: className = "" }: Props = $props();
const percentage = $derived(
progress.total > 0
? Math.round((progress.processed / progress.total) * 100)
: 0,
);
const etaText = $derived.by(() => {
if (progress.processed <= 0 || progress.total <= 0) return null;
const elapsedMs = performance.now() - progress.startedAt;
if (elapsedMs < 200) return null; // need a minimum sample window
const tokensPerMs = progress.processed / elapsedMs;
const remainingTokens = progress.total - progress.processed;
const remainingMs = remainingTokens / tokensPerMs;
const remainingSec = Math.ceil(remainingMs / 1000);
if (remainingSec <= 0) return null;
if (remainingSec < 60) return `~${remainingSec}s remaining`;
const mins = Math.floor(remainingSec / 60);
const secs = remainingSec % 60;
return `~${mins}m ${secs}s remaining`;
});
function formatTokenCount(count: number | undefined): string {
if (count == null) return "0";
if (count >= 1000) {
return `${(count / 1000).toFixed(1)}k`;
}
return count.toString();
}
</script>
<div class="prefill-progress {className}">
<div
class="flex items-center justify-between text-xs text-exo-light-gray mb-1"
>
<span>Processing prompt</span>
<span class="font-mono">
{formatTokenCount(progress.processed)} / {formatTokenCount(
progress.total,
)} tokens
</span>
</div>
<div class="h-1.5 bg-exo-black/60 rounded-full overflow-hidden">
<div
class="h-full bg-exo-yellow rounded-full transition-all duration-150 ease-out"
style="width: {percentage}%"
></div>
</div>
<div
class="flex items-center justify-between text-xs text-exo-light-gray/70 mt-0.5 font-mono"
>
<span>{etaText ?? ""}</span>
<span>{percentage}%</span>
</div>
</div>
<style>
.prefill-progress {
width: 100%;
}
</style>
@@ -0,0 +1,117 @@
<script lang="ts">
import { toasts, dismissToast, type Toast } from "$lib/stores/toast.svelte";
import { fly, fade } from "svelte/transition";
import { flip } from "svelte/animate";
const items = $derived(toasts());
const typeStyles: Record<
Toast["type"],
{ border: string; icon: string; iconColor: string }
> = {
success: {
border: "border-l-green-500",
icon: "M9 12.75L11.25 15 15 9.75M21 12a9 9 0 11-18 0 9 9 0 0118 0z",
iconColor: "text-green-400",
},
error: {
border: "border-l-red-500",
icon: "M12 9v3.75m9-.75a9 9 0 11-18 0 9 9 0 0118 0zm-9 3.75h.008v.008H12v-.008z",
iconColor: "text-red-400",
},
warning: {
border: "border-l-yellow-500",
icon: "M12 9v3.75m-9.303 3.376c-.866 1.5.217 3.374 1.948 3.374h14.71c1.73 0 2.813-1.874 1.948-3.374L13.949 3.378c-.866-1.5-3.032-1.5-3.898 0L2.697 16.126z",
iconColor: "text-yellow-400",
},
info: {
border: "border-l-blue-500",
icon: "M11.25 11.25l.041-.02a.75.75 0 011.063.852l-.708 2.836a.75.75 0 001.063.853l.041-.021M21 12a9 9 0 11-18 0 9 9 0 0118 0zm-9-3.75h.008v.008H12V8.25z",
iconColor: "text-blue-400",
},
};
</script>
{#if items.length > 0}
<div
class="fixed bottom-6 right-6 z-[9999] flex flex-col gap-2 pointer-events-none"
role="log"
aria-live="polite"
aria-label="Notifications"
>
{#each items as toast (toast.id)}
{@const style = typeStyles[toast.type]}
<div
class="pointer-events-auto max-w-sm w-80 bg-exo-dark-gray/95 backdrop-blur-sm border border-exo-medium-gray/60 border-l-[3px] {style.border} rounded shadow-lg shadow-black/40"
in:fly={{ x: 80, duration: 250 }}
out:fade={{ duration: 150 }}
animate:flip={{ duration: 200 }}
role="alert"
>
<div class="flex items-start gap-3 px-4 py-3">
<!-- Icon -->
<svg
class="w-5 h-5 flex-shrink-0 mt-0.5 {style.iconColor}"
fill="none"
viewBox="0 0 24 24"
stroke-width="1.5"
stroke="currentColor"
>
<path
stroke-linecap="round"
stroke-linejoin="round"
d={style.icon}
/>
</svg>
<!-- Message -->
<p class="flex-1 text-sm text-white/90 font-mono leading-snug">
{toast.message}
</p>
<!-- Dismiss button -->
<button
onclick={() => dismissToast(toast.id)}
class="flex-shrink-0 p-0.5 text-white/40 hover:text-white/80 transition-colors cursor-pointer"
aria-label="Dismiss notification"
>
<svg
class="w-4 h-4"
fill="none"
viewBox="0 0 24 24"
stroke-width="2"
stroke="currentColor"
>
<path
stroke-linecap="round"
stroke-linejoin="round"
d="M6 18L18 6M6 6l12 12"
/>
</svg>
</button>
</div>
<!-- Auto-dismiss progress bar -->
{#if toast.duration > 0}
<div class="h-0.5 bg-white/5 rounded-b overflow-hidden">
<div
class="h-full {style.border.replace('border-l-', 'bg-')}/60"
style="animation: shrink {toast.duration}ms linear forwards"
></div>
</div>
{/if}
</div>
{/each}
</div>
{/if}
<style>
@keyframes shrink {
from {
width: 100%;
}
to {
width: 0%;
}
}
</style>
@@ -0,0 +1,236 @@
<script lang="ts">
import type { TokenData } from "$lib/stores/app.svelte";
interface Props {
tokens: TokenData[];
class?: string;
isGenerating?: boolean;
onRegenerateFrom?: (tokenIndex: number) => void;
}
let {
tokens,
class: className = "",
isGenerating = false,
onRegenerateFrom,
}: Props = $props();
// Tooltip state - track both token data and index
let hoveredTokenIndex = $state<number | null>(null);
let hoveredPosition = $state<{ x: number; y: number } | null>(null);
let isTooltipHovered = $state(false);
let hideTimeoutId: ReturnType<typeof setTimeout> | null = null;
// Derive the hovered token from the index (stable across re-renders)
const hoveredToken = $derived(
hoveredTokenIndex !== null && hoveredPosition && tokens[hoveredTokenIndex]
? {
token: tokens[hoveredTokenIndex],
index: hoveredTokenIndex,
...hoveredPosition,
}
: null,
);
/**
* Get confidence styling based on probability.
* Following Apple design principles: high confidence tokens blend in,
* only uncertainty draws attention.
*/
function getConfidenceClass(probability: number): string {
if (probability > 0.8) return "text-inherit"; // Expected tokens - blend in
if (probability > 0.5) return "bg-gray-500/10 text-inherit"; // Slight hint
if (probability > 0.2) return "bg-amber-500/15 text-amber-200/90"; // Subtle warmth
return "bg-red-500/20 text-red-200/90"; // Draws attention
}
/**
* Get border/underline styling for uncertain tokens
*/
function getBorderClass(probability: number): string {
if (probability > 0.8) return "border-transparent"; // No border for expected
if (probability > 0.5) return "border-gray-500/20";
if (probability > 0.2) return "border-amber-500/30";
return "border-red-500/40";
}
function clearHideTimeout() {
if (hideTimeoutId) {
clearTimeout(hideTimeoutId);
hideTimeoutId = null;
}
}
function handleMouseEnter(
event: MouseEvent,
token: TokenData,
index: number,
) {
clearHideTimeout();
const rects = (event.target as HTMLElement).getClientRects();
let rect = rects[0];
for (let j = 0; j < rects.length; j++) {
if (event.clientY >= rects[j].top && event.clientY <= rects[j].bottom) {
rect = rects[j];
break;
}
}
hoveredTokenIndex = index;
hoveredPosition = {
x: rect.left + rect.width / 2,
y: rect.top - 10,
};
}
function handleMouseLeave() {
clearHideTimeout();
// Use longer delay during generation to account for re-renders
const delay = isGenerating ? 300 : 200;
hideTimeoutId = setTimeout(() => {
if (!isTooltipHovered) {
hoveredTokenIndex = null;
hoveredPosition = null;
}
}, delay);
}
function handleTooltipEnter() {
clearHideTimeout();
isTooltipHovered = true;
}
function handleTooltipLeave() {
isTooltipHovered = false;
hoveredTokenIndex = null;
hoveredPosition = null;
}
function handleRegenerate() {
if (hoveredToken && onRegenerateFrom) {
const indexToRegenerate = hoveredToken.index;
// Clear hover state immediately
hoveredTokenIndex = null;
hoveredPosition = null;
isTooltipHovered = false;
// Call regenerate
onRegenerateFrom(indexToRegenerate);
}
}
function formatProbability(prob: number): string {
return (prob * 100).toFixed(1) + "%";
}
function formatLogprob(logprob: number): string {
return logprob.toFixed(3);
}
function getProbabilityColor(probability: number): string {
if (probability > 0.8) return "text-gray-300";
if (probability > 0.5) return "text-gray-400";
if (probability > 0.2) return "text-amber-400";
return "text-red-400";
}
</script>
<div class="token-heatmap leading-relaxed {className}">
{#each tokens as tokenData, i (i)}
<span
role="button"
tabindex="0"
class="token-span inline rounded px-0.5 py-0.5 cursor-pointer transition-all duration-150 border {getConfidenceClass(
tokenData.probability,
)} {getBorderClass(tokenData.probability)} hover:opacity-80"
onmouseenter={(e) => handleMouseEnter(e, tokenData, i)}
onmouseleave={handleMouseLeave}>{tokenData.token}</span
>
{/each}
</div>
<!-- Tooltip -->
{#if hoveredToken}
<div
class="fixed z-50 pb-2"
style="left: {hoveredToken.x}px; top: {hoveredToken.y}px; transform: translate(-50%, -100%);"
onmouseenter={handleTooltipEnter}
onmouseleave={handleTooltipLeave}
>
<div
class="bg-gray-900/95 backdrop-blur-sm border border-gray-700/50 rounded-xl shadow-xl p-3 text-sm min-w-48"
>
<!-- Token info -->
<div class="mb-2">
<span class="text-gray-500 text-xs">Token:</span>
<span class="text-white font-mono ml-1"
>"{hoveredToken.token.token}"</span
>
<span class="{getProbabilityColor(hoveredToken.token.probability)} ml-2"
>{formatProbability(hoveredToken.token.probability)}</span
>
</div>
<div class="text-gray-400 text-xs mb-1">
logprob: <span class="text-gray-300 font-mono"
>{formatLogprob(hoveredToken.token.logprob)}</span
>
</div>
<!-- Top alternatives -->
{#if hoveredToken.token.topLogprobs.length > 0}
<div class="border-t border-gray-700/50 mt-2 pt-2">
<div class="text-gray-500 text-xs mb-1">Alternatives:</div>
{#each hoveredToken.token.topLogprobs.slice(0, 5) as alt, idx (idx)}
{@const altProb = Math.exp(alt.logprob)}
<div class="flex justify-between items-center text-xs py-0.5">
<span class="text-gray-300 font-mono truncate max-w-24"
>"{alt.token}"</span
>
<span class="text-gray-400 ml-2"
>{formatProbability(altProb)}</span
>
</div>
{/each}
</div>
{/if}
<!-- Regenerate button -->
{#if onRegenerateFrom}
<button
onclick={handleRegenerate}
class="w-full mt-2 pt-2 border-t border-gray-700/50 flex items-center justify-center gap-1.5 text-xs text-gray-400 hover:text-white transition-colors cursor-pointer"
>
<svg
class="w-3 h-3"
fill="none"
viewBox="0 0 24 24"
stroke="currentColor"
>
<path
stroke-linecap="round"
stroke-linejoin="round"
stroke-width="2"
d="M4 4v5h.582m15.356 2A8.001 8.001 0 004.582 9m0 0H9m11 11v-5h-.581m0 0a8.003 8.003 0 01-15.357-2m15.357 2H15"
/>
</svg>
Regenerate from here
</button>
{/if}
</div>
<!-- Arrow -->
<div class="absolute left-1/2 -translate-x-1/2 top-full">
<div class="border-8 border-transparent border-t-gray-900"></div>
</div>
</div>
{/if}
<style>
.token-heatmap {
word-wrap: break-word;
white-space: pre-wrap;
}
.token-span {
margin: 0;
border-width: 1px;
}
</style>
+168 -34
View File
@@ -5,22 +5,36 @@
topologyData,
isTopologyMinimized,
debugMode,
nodeThunderboltBridge,
nodeRdmaCtl,
nodeIdentities,
type NodeInfo,
} from "$lib/stores/app.svelte";
interface Props {
class?: string;
highlightedNodes?: Set<string>;
filteredNodes?: Set<string>;
onNodeClick?: (nodeId: string) => void;
}
let { class: className = "", highlightedNodes = new Set() }: Props = $props();
let {
class: className = "",
highlightedNodes = new Set(),
filteredNodes = new Set(),
onNodeClick,
}: Props = $props();
let svgContainer: SVGSVGElement | undefined = $state();
let resizeObserver: ResizeObserver | undefined;
let hoveredNodeId = $state<string | null>(null);
const isMinimized = $derived(isTopologyMinimized());
const data = $derived(topologyData());
const debugEnabled = $derived(debugMode());
const tbBridgeData = $derived(nodeThunderboltBridge());
const rdmaCtlData = $derived(nodeRdmaCtl());
const identitiesData = $derived(nodeIdentities());
function getNodeLabel(nodeId: string): string {
const node = data?.nodes?.[nodeId];
@@ -323,14 +337,27 @@
if (edge.source === a) entry.aToB = true;
else entry.bToA = true;
const ip = edge.sendBackIp || "?";
const ifaceInfo = getInterfaceLabel(edge.source, ip);
let ip: string;
let ifaceLabel: string;
let missingIface: boolean;
if (edge.sourceRdmaIface || edge.sinkRdmaIface) {
ip = "RDMA";
ifaceLabel = `${edge.sourceRdmaIface || "?"} \u2192 ${edge.sinkRdmaIface || "?"}`;
missingIface = false;
} else {
ip = edge.sendBackIp || "?";
const ifaceInfo = getInterfaceLabel(edge.source, ip);
ifaceLabel = ifaceInfo.label;
missingIface = ifaceInfo.missing;
}
entry.connections.push({
from: edge.source,
to: edge.target,
ip,
ifaceLabel: ifaceInfo.label,
missingIface: ifaceInfo.missing,
ifaceLabel,
missingIface,
});
pairMap.set(key, entry);
});
@@ -522,10 +549,72 @@
}
}
let iconBaseWidth = nodeRadius * 1.2;
let iconBaseHeight = nodeRadius * 1.0;
const clipPathId = `clip-${nodeInfo.id.replace(/[^a-zA-Z0-9]/g, "-")}`;
const modelLower = modelId.toLowerCase();
// Check node states for styling
const isHighlighted = highlightedNodes.has(nodeInfo.id);
const isInFilter =
filteredNodes.size > 0 && filteredNodes.has(nodeInfo.id);
const isFilteredOut =
filteredNodes.size > 0 && !filteredNodes.has(nodeInfo.id);
const isHovered = hoveredNodeId === nodeInfo.id && !isInFilter;
// Holographic wireframe colors - bright yellow for filter, subtle yellow for hover, grey for filtered out
const wireColor = isInFilter
? "rgba(255,215,0,1)" // Bright yellow for filter selection
: isHovered
? "rgba(255,215,0,0.7)" // Subtle yellow for hover
: isHighlighted
? "rgba(255,215,0,0.9)" // Yellow for instance highlight
: isFilteredOut
? "rgba(140,140,140,0.6)" // Grey for filtered out
: "rgba(179,179,179,0.8)"; // Default
const wireColorBright = "rgba(255,255,255,0.9)";
const fillColor = isInFilter
? "rgba(255,215,0,0.25)"
: isHovered
? "rgba(255,215,0,0.12)"
: isHighlighted
? "rgba(255,215,0,0.15)"
: "rgba(255,215,0,0.08)";
const strokeWidth = isInFilter
? 3
: isHovered
? 2
: isHighlighted
? 2.5
: 1.5;
const screenFill = "rgba(0,20,40,0.9)";
const glowColor = "rgba(255,215,0,0.3)";
const nodeG = nodesGroup
.append("g")
.attr("class", "graph-node")
.style("cursor", "pointer");
.style("cursor", onNodeClick ? "pointer" : "default")
.style("opacity", isFilteredOut ? 0.5 : 1);
// Add click and hover handlers - hover just updates state, styling is applied during render
nodeG
.on("click", (event: MouseEvent) => {
if (onNodeClick) {
event.stopPropagation();
onNodeClick(nodeInfo.id);
}
})
.on("mouseenter", () => {
if (onNodeClick) {
hoveredNodeId = nodeInfo.id;
}
})
.on("mouseleave", () => {
if (hoveredNodeId === nodeInfo.id) {
hoveredNodeId = null;
}
});
// Add tooltip
nodeG
@@ -534,27 +623,6 @@
`${friendlyName}\nID: ${nodeInfo.id.slice(-8)}\nMemory: ${formatBytes(ramUsed)}/${formatBytes(ramTotal)}`,
);
let iconBaseWidth = nodeRadius * 1.2;
let iconBaseHeight = nodeRadius * 1.0;
const clipPathId = `clip-${nodeInfo.id.replace(/[^a-zA-Z0-9]/g, "-")}`;
const modelLower = modelId.toLowerCase();
// Check if this node should be highlighted (from hovered instance)
const isHighlighted = highlightedNodes.has(nodeInfo.id);
// Holographic wireframe colors - yellow border when highlighted
const wireColor = isHighlighted
? "rgba(255,215,0,0.9)"
: "rgba(179,179,179,0.8)";
const wireColorBright = "rgba(255,255,255,0.9)";
const fillColor = isHighlighted
? "rgba(255,215,0,0.15)"
: "rgba(255,215,0,0.08)";
const strokeWidth = isHighlighted ? 2.5 : 1.5;
const screenFill = "rgba(0,20,40,0.9)";
const glowColor = "rgba(255,215,0,0.3)";
if (modelLower === "mac studio") {
// Mac Studio - classic cube with memory fill
iconBaseWidth = nodeRadius * 1.25;
@@ -579,6 +647,7 @@
// Main body (uniform color)
nodeG
.append("rect")
.attr("class", "node-outline")
.attr("x", x)
.attr("y", y)
.attr("width", iconBaseWidth)
@@ -661,6 +730,7 @@
// Main body (uniform color)
nodeG
.append("rect")
.attr("class", "node-outline")
.attr("x", x)
.attr("y", y)
.attr("width", iconBaseWidth)
@@ -738,6 +808,7 @@
// Screen outer frame
nodeG
.append("rect")
.attr("class", "node-outline")
.attr("x", screenX)
.attr("y", y)
.attr("width", screenWidth)
@@ -846,6 +917,7 @@
// Main shape
nodeG
.append("polygon")
.attr("class", "node-outline")
.attr("points", hexPoints)
.attr("fill", fillColor)
.attr("stroke", wireColor)
@@ -1064,11 +1136,77 @@
.attr("fill", "rgba(179,179,179,0.7)")
.text(` (${ramUsagePercent.toFixed(0)}%)`);
}
// Debug mode: Show TB bridge and RDMA status
if (debugEnabled) {
let debugLabelY =
nodeInfo.y +
iconBaseHeight / 2 +
(showFullLabels ? 32 : showCompactLabels ? 26 : 22);
const debugFontSize = showFullLabels ? 9 : 7;
const debugLineHeight = showFullLabels ? 11 : 9;
const tbStatus = tbBridgeData[nodeInfo.id];
if (tbStatus) {
const tbColor = tbStatus.enabled
? "rgba(234,179,8,0.9)"
: "rgba(100,100,100,0.7)";
const tbText = tbStatus.enabled ? "TB:ON" : "TB:OFF";
nodeG
.append("text")
.attr("x", nodeInfo.x)
.attr("y", debugLabelY)
.attr("text-anchor", "middle")
.attr("fill", tbColor)
.attr("font-size", debugFontSize)
.attr("font-family", "SF Mono, Monaco, monospace")
.text(tbText);
debugLabelY += debugLineHeight;
}
const rdmaStatus = rdmaCtlData[nodeInfo.id];
if (rdmaStatus !== undefined) {
const rdmaColor = rdmaStatus.enabled
? "rgba(74,222,128,0.9)"
: "rgba(100,100,100,0.7)";
const rdmaText = rdmaStatus.enabled ? "RDMA:ON" : "RDMA:OFF";
nodeG
.append("text")
.attr("x", nodeInfo.x)
.attr("y", debugLabelY)
.attr("text-anchor", "middle")
.attr("fill", rdmaColor)
.attr("font-size", debugFontSize)
.attr("font-family", "SF Mono, Monaco, monospace")
.text(rdmaText);
debugLabelY += debugLineHeight;
}
const identity = identitiesData[nodeInfo.id];
if (identity?.osVersion) {
nodeG
.append("text")
.attr("x", nodeInfo.x)
.attr("y", debugLabelY)
.attr("text-anchor", "middle")
.attr("fill", "rgba(179,179,179,0.7)")
.attr("font-size", debugFontSize)
.attr("font-family", "SF Mono, Monaco, monospace")
.text(
`macOS ${identity.osVersion}${identity.osBuildVersion ? ` (${identity.osBuildVersion})` : ""}`,
);
}
}
});
}
$effect(() => {
if (data) {
// Track all reactive dependencies that affect rendering
const _data = data;
const _hoveredNodeId = hoveredNodeId;
const _filteredNodes = filteredNodes;
const _highlightedNodes = highlightedNodes;
if (_data) {
renderGraph();
}
});
@@ -1091,12 +1229,8 @@
<style>
:global(.graph-node) {
transition:
transform 0.2s ease,
opacity 0.2s ease;
}
:global(.graph-node:hover) {
filter: brightness(1.1);
/* Only transition opacity for filtered-out nodes, no transition on hover stroke changes */
transition: opacity 0.2s ease;
}
:global(.graph-link) {
stroke: var(--exo-light-gray, #b3b3b3);
+7
View File
@@ -6,3 +6,10 @@ export { default as ChatSidebar } from "./ChatSidebar.svelte";
export { default as ModelCard } from "./ModelCard.svelte";
export { default as MarkdownContent } from "./MarkdownContent.svelte";
export { default as ImageParamsPanel } from "./ImageParamsPanel.svelte";
export { default as FamilyLogos } from "./FamilyLogos.svelte";
export { default as FamilySidebar } from "./FamilySidebar.svelte";
export { default as HuggingFaceResultItem } from "./HuggingFaceResultItem.svelte";
export { default as ModelFilterPopover } from "./ModelFilterPopover.svelte";
export { default as ModelPickerGroup } from "./ModelPickerGroup.svelte";
export { default as ModelPickerModal } from "./ModelPickerModal.svelte";
export { default as ChatModelSelector } from "./ChatModelSelector.svelte";
File diff suppressed because it is too large Load Diff
@@ -0,0 +1,97 @@
/**
* FavoritesStore - Manages favorite models with localStorage persistence
*/
import { browser } from "$app/environment";
const FAVORITES_KEY = "exo-favorite-models";
class FavoritesStore {
favorites = $state<Set<string>>(new Set());
constructor() {
if (browser) {
this.loadFromStorage();
}
}
private loadFromStorage() {
try {
const stored = localStorage.getItem(FAVORITES_KEY);
if (stored) {
const parsed = JSON.parse(stored) as string[];
this.favorites = new Set(parsed);
}
} catch (error) {
console.error("Failed to load favorites:", error);
}
}
private saveToStorage() {
try {
const array = Array.from(this.favorites);
localStorage.setItem(FAVORITES_KEY, JSON.stringify(array));
} catch (error) {
console.error("Failed to save favorites:", error);
}
}
add(baseModelId: string) {
const next = new Set(this.favorites);
next.add(baseModelId);
this.favorites = next;
this.saveToStorage();
}
remove(baseModelId: string) {
const next = new Set(this.favorites);
next.delete(baseModelId);
this.favorites = next;
this.saveToStorage();
}
toggle(baseModelId: string) {
if (this.favorites.has(baseModelId)) {
this.remove(baseModelId);
} else {
this.add(baseModelId);
}
}
isFavorite(baseModelId: string): boolean {
return this.favorites.has(baseModelId);
}
getAll(): string[] {
return Array.from(this.favorites);
}
getSet(): Set<string> {
return new Set(this.favorites);
}
hasAny(): boolean {
return this.favorites.size > 0;
}
clearAll() {
this.favorites = new Set();
this.saveToStorage();
}
}
export const favoritesStore = new FavoritesStore();
export const favorites = () => favoritesStore.favorites;
export const hasFavorites = () => favoritesStore.hasAny();
export const isFavorite = (baseModelId: string) =>
favoritesStore.isFavorite(baseModelId);
export const toggleFavorite = (baseModelId: string) =>
favoritesStore.toggle(baseModelId);
export const addFavorite = (baseModelId: string) =>
favoritesStore.add(baseModelId);
export const removeFavorite = (baseModelId: string) =>
favoritesStore.remove(baseModelId);
export const getFavorites = () => favoritesStore.getAll();
export const getFavoritesSet = () => favoritesStore.getSet();
export const clearFavorites = () => favoritesStore.clearAll();
@@ -0,0 +1,75 @@
/**
* RecentsStore - Manages recently launched models with localStorage persistence
*/
import { browser } from "$app/environment";
const RECENTS_KEY = "exo-recent-models";
const MAX_RECENT_MODELS = 20;
interface RecentEntry {
modelId: string;
launchedAt: number;
}
class RecentsStore {
recents = $state<RecentEntry[]>([]);
constructor() {
if (browser) {
this.loadFromStorage();
}
}
private loadFromStorage() {
try {
const stored = localStorage.getItem(RECENTS_KEY);
if (stored) {
const parsed = JSON.parse(stored) as RecentEntry[];
this.recents = parsed;
}
} catch (error) {
console.error("Failed to load recent models:", error);
}
}
private saveToStorage() {
try {
localStorage.setItem(RECENTS_KEY, JSON.stringify(this.recents));
} catch (error) {
console.error("Failed to save recent models:", error);
}
}
recordLaunch(modelId: string) {
// Remove existing entry for this model (if any) to move it to top
const filtered = this.recents.filter((r) => r.modelId !== modelId);
// Prepend new entry
const next = [{ modelId, launchedAt: Date.now() }, ...filtered];
// Cap at max
this.recents = next.slice(0, MAX_RECENT_MODELS);
this.saveToStorage();
}
getRecentModelIds(): string[] {
return this.recents.map((r) => r.modelId);
}
hasAny(): boolean {
return this.recents.length > 0;
}
clearAll() {
this.recents = [];
this.saveToStorage();
}
}
export const recentsStore = new RecentsStore();
export const hasRecents = () => recentsStore.hasAny();
export const getRecentModelIds = () => recentsStore.getRecentModelIds();
export const getRecentEntries = () => recentsStore.recents;
export const recordRecentLaunch = (modelId: string) =>
recentsStore.recordLaunch(modelId);
export const clearRecents = () => recentsStore.clearAll();
@@ -0,0 +1,87 @@
/**
* SettingsStore - Manages exo runtime settings via the /settings API.
*/
export interface MemorySettings {
oom_prevention: boolean;
memory_threshold: number;
memory_floor_gb: number;
}
export interface GenerationSettings {
prefill_step_size: number;
max_tokens: number;
kv_cache_bits: 4 | 8 | null;
}
export interface ExoSettings {
memory: MemorySettings;
generation: GenerationSettings;
}
function defaultSettings(): ExoSettings {
return {
memory: {
oom_prevention: false,
memory_threshold: 0.8,
memory_floor_gb: 5.0,
},
generation: {
prefill_step_size: 4096,
max_tokens: 32168,
kv_cache_bits: null,
},
};
}
class SettingsStore {
settings = $state<ExoSettings>(defaultSettings());
loading = $state(false);
error = $state<string | null>(null);
async load(): Promise<void> {
this.loading = true;
this.error = null;
try {
const response = await fetch("/settings");
if (!response.ok) {
throw new Error(`Failed to fetch settings: ${response.status}`);
}
this.settings = (await response.json()) as ExoSettings;
} catch (err) {
console.error("Failed to load settings:", err);
this.error = err instanceof Error ? err.message : "Unknown error";
} finally {
this.loading = false;
}
}
async save(updated: ExoSettings): Promise<boolean> {
this.loading = true;
this.error = null;
try {
const response = await fetch("/settings", {
method: "POST",
headers: { "Content-Type": "application/json" },
body: JSON.stringify(updated),
});
if (!response.ok) {
throw new Error(`Failed to save settings: ${response.status}`);
}
this.settings = (await response.json()) as ExoSettings;
return true;
} catch (err) {
console.error("Failed to save settings:", err);
this.error = err instanceof Error ? err.message : "Unknown error";
return false;
} finally {
this.loading = false;
}
}
resetToDefaults(): ExoSettings {
return defaultSettings();
}
}
export const settingsStore = new SettingsStore();
+87
View File
@@ -0,0 +1,87 @@
/**
* Toast notification store - Global notification system for the EXO dashboard.
*
* Usage:
* import { addToast, dismissToast, toasts } from "$lib/stores/toast.svelte";
* addToast({ type: "success", message: "Model launched" });
* addToast({ type: "error", message: "Connection lost", persistent: true });
*/
type ToastType = "success" | "error" | "warning" | "info";
export interface Toast {
id: string;
type: ToastType;
message: string;
/** Auto-dismiss after this many ms. 0 = persistent (must be dismissed manually). */
duration: number;
createdAt: number;
}
interface ToastInput {
type: ToastType;
message: string;
/** If true, toast stays until manually dismissed. Default: false. */
persistent?: boolean;
/** Auto-dismiss duration in ms. Default: 4000 for success/info, 6000 for error/warning. */
duration?: number;
}
const DEFAULT_DURATIONS: Record<ToastType, number> = {
success: 4000,
info: 4000,
warning: 6000,
error: 6000,
};
let toastList = $state<Toast[]>([]);
const timers = new Map<string, ReturnType<typeof setTimeout>>();
function generateId(): string {
return `toast-${Date.now()}-${Math.random().toString(36).slice(2, 8)}`;
}
export function addToast(input: ToastInput): string {
const id = generateId();
const duration = input.persistent
? 0
: (input.duration ?? DEFAULT_DURATIONS[input.type]);
const toast: Toast = {
id,
type: input.type,
message: input.message,
duration,
createdAt: Date.now(),
};
toastList = [...toastList, toast];
if (duration > 0) {
const timer = setTimeout(() => dismissToast(id), duration);
timers.set(id, timer);
}
return id;
}
export function dismissToast(id: string): void {
const timer = timers.get(id);
if (timer) {
clearTimeout(timer);
timers.delete(id);
}
toastList = toastList.filter((t) => t.id !== id);
}
/** Dismiss all toasts matching a message (useful for dedup). */
export function dismissByMessage(message: string): void {
const matching = toastList.filter((t) => t.message === message);
for (const t of matching) {
dismissToast(t.id);
}
}
export function toasts(): Toast[] {
return toastList;
}
+152
View File
@@ -0,0 +1,152 @@
/**
* Shared utilities for parsing and querying download state.
*
* The download state from `/state` is shaped as:
* Record<NodeId, Array<TaggedDownloadEntry>>
*
* Each entry is a tagged union object like:
* { "DownloadCompleted": { shard_metadata: { "PipelineShardMetadata": { model_card: { model_id: "..." }, ... } }, ... } }
*/
/** Unwrap one level of tagged-union envelope, returning [tag, payload]. */
function unwrapTagged(
obj: Record<string, unknown>,
): [string, Record<string, unknown>] | null {
const keys = Object.keys(obj);
if (keys.length !== 1) return null;
const tag = keys[0];
const payload = obj[tag];
if (!payload || typeof payload !== "object") return null;
return [tag, payload as Record<string, unknown>];
}
/** Extract the model ID string from a download entry's nested shard_metadata. */
export function extractModelIdFromDownload(
downloadPayload: Record<string, unknown>,
): string | null {
const shardMetadata =
downloadPayload.shard_metadata ?? downloadPayload.shardMetadata;
if (!shardMetadata || typeof shardMetadata !== "object") return null;
const unwrapped = unwrapTagged(shardMetadata as Record<string, unknown>);
if (!unwrapped) return null;
const [, shardData] = unwrapped;
const modelMeta = shardData.model_card ?? shardData.modelCard;
if (!modelMeta || typeof modelMeta !== "object") return null;
const meta = modelMeta as Record<string, unknown>;
return (meta.model_id as string) ?? (meta.modelId as string) ?? null;
}
/** Extract the shard_metadata object from a download entry payload. */
export function extractShardMetadata(
downloadPayload: Record<string, unknown>,
): Record<string, unknown> | null {
const shardMetadata =
downloadPayload.shard_metadata ?? downloadPayload.shardMetadata;
if (!shardMetadata || typeof shardMetadata !== "object") return null;
return shardMetadata as Record<string, unknown>;
}
/** Get the download tag (DownloadCompleted, DownloadOngoing, etc.) from a wrapped entry. */
export function getDownloadTag(
entry: unknown,
): [string, Record<string, unknown>] | null {
if (!entry || typeof entry !== "object") return null;
return unwrapTagged(entry as Record<string, unknown>);
}
/**
* Iterate over all download entries for a given node, yielding [tag, payload, modelId].
*/
function* iterNodeDownloads(
nodeDownloads: unknown[],
): Generator<[string, Record<string, unknown>, string]> {
for (const entry of nodeDownloads) {
const tagged = getDownloadTag(entry);
if (!tagged) continue;
const [tag, payload] = tagged;
const modelId = extractModelIdFromDownload(payload);
if (!modelId) continue;
yield [tag, payload, modelId];
}
}
/** Check if a specific model is fully downloaded (DownloadCompleted) on a specific node. */
export function isModelDownloadedOnNode(
downloadsData: Record<string, unknown[]>,
nodeId: string,
modelId: string,
): boolean {
const nodeDownloads = downloadsData[nodeId];
if (!Array.isArray(nodeDownloads)) return false;
for (const [tag, , entryModelId] of iterNodeDownloads(nodeDownloads)) {
if (tag === "DownloadCompleted" && entryModelId === modelId) return true;
}
return false;
}
/** Get all node IDs where a model is fully downloaded (DownloadCompleted). */
export function getNodesWithModelDownloaded(
downloadsData: Record<string, unknown[]>,
modelId: string,
): string[] {
const result: string[] = [];
for (const nodeId of Object.keys(downloadsData)) {
if (isModelDownloadedOnNode(downloadsData, nodeId, modelId)) {
result.push(nodeId);
}
}
return result;
}
/**
* Find shard metadata for a model from any download entry across all nodes.
* Returns the first match found (completed entries are preferred).
*/
export function getShardMetadataForModel(
downloadsData: Record<string, unknown[]>,
modelId: string,
): Record<string, unknown> | null {
let fallback: Record<string, unknown> | null = null;
for (const nodeDownloads of Object.values(downloadsData)) {
if (!Array.isArray(nodeDownloads)) continue;
for (const [tag, payload, entryModelId] of iterNodeDownloads(
nodeDownloads,
)) {
if (entryModelId !== modelId) continue;
const shard = extractShardMetadata(payload);
if (!shard) continue;
if (tag === "DownloadCompleted") return shard;
if (!fallback) fallback = shard;
}
}
return fallback;
}
/**
* Get the download status tag for a specific model on a specific node.
* Returns the "best" status: DownloadCompleted > DownloadOngoing > others.
*/
export function getModelDownloadStatus(
downloadsData: Record<string, unknown[]>,
nodeId: string,
modelId: string,
): string | null {
const nodeDownloads = downloadsData[nodeId];
if (!Array.isArray(nodeDownloads)) return null;
let best: string | null = null;
for (const [tag, , entryModelId] of iterNodeDownloads(nodeDownloads)) {
if (entryModelId !== modelId) continue;
if (tag === "DownloadCompleted") return tag;
if (tag === "DownloadOngoing") best = tag;
else if (!best) best = tag;
}
return best;
}
+4
View File
@@ -1,5 +1,7 @@
<script lang="ts">
import "../app.css";
import ToastContainer from "$lib/components/ToastContainer.svelte";
import ConnectionBanner from "$lib/components/ConnectionBanner.svelte";
let { children } = $props();
</script>
@@ -10,5 +12,7 @@
</svelte:head>
<div class="min-h-screen bg-background text-foreground">
<ConnectionBanner />
{@render children?.()}
<ToastContainer />
</div>
File diff suppressed because it is too large Load Diff
File diff suppressed because it is too large Load Diff
+193
View File
@@ -0,0 +1,193 @@
<script lang="ts">
import { onMount } from "svelte";
import { fade } from "svelte/transition";
import HeaderNav from "$lib/components/HeaderNav.svelte";
import { settingsStore, type ExoSettings } from "$lib/stores/settings.svelte";
import { addToast } from "$lib/stores/toast.svelte";
let draft = $state<ExoSettings | null>(null);
const loading = $derived(settingsStore.loading);
onMount(async () => {
await settingsStore.load();
draft = structuredClone(settingsStore.settings);
});
async function handleSave() {
if (!draft) return;
const ok = await settingsStore.save(draft);
if (ok) {
addToast({ type: "success", message: "Settings saved" });
} else {
addToast({ type: "error", message: settingsStore.error ?? "Failed to save settings" });
}
}
function handleReset() {
draft = settingsStore.resetToDefaults();
}
const KV_OPTIONS: { label: string; value: 4 | 8 | null }[] = [
{ label: "None (full precision)", value: null },
{ label: "4-bit", value: 4 },
{ label: "8-bit", value: 8 },
];
</script>
<HeaderNav showHome={true} />
{#if draft}
<div class="min-h-screen bg-background text-foreground" in:fade={{ duration: 200 }}>
<div class="max-w-2xl mx-auto px-6 py-8">
<h1 class="text-2xl font-bold text-exo-yellow tracking-wider uppercase mb-8">Settings</h1>
<!-- Memory / Safety -->
<section class="mb-10">
<h2 class="text-sm font-semibold text-white/50 tracking-widest uppercase mb-4">Memory / Safety</h2>
<div class="space-y-5">
<!-- OOM Prevention Toggle -->
<div class="flex items-center justify-between">
<div>
<div class="text-sm text-white/90">OOM Prevention</div>
<div class="text-xs text-white/40 mt-0.5">Stop generation when memory is low</div>
</div>
<button
onclick={() => { if (draft) draft.memory.oom_prevention = !draft.memory.oom_prevention; }}
class="relative w-11 h-6 rounded-full transition-colors duration-200 cursor-pointer {draft.memory.oom_prevention ? 'bg-exo-yellow' : 'bg-exo-medium-gray'}"
role="switch"
aria-checked={draft.memory.oom_prevention}
>
<span
class="absolute top-0.5 left-0.5 w-5 h-5 rounded-full bg-white shadow transition-transform duration-200 {draft.memory.oom_prevention ? 'translate-x-5' : 'translate-x-0'}"
></span>
</button>
</div>
<!-- Memory Threshold Slider -->
<div>
<div class="flex items-center justify-between mb-1.5">
<div>
<div class="text-sm text-white/90">Memory Threshold</div>
<div class="text-xs text-white/40 mt-0.5">KV cache eviction triggers above this level</div>
</div>
<span class="text-sm font-mono text-exo-yellow">{(draft.memory.memory_threshold * 100).toFixed(0)}%</span>
</div>
<input
type="range"
min="0.5"
max="0.99"
step="0.01"
bind:value={draft.memory.memory_threshold}
class="w-full h-1.5 rounded-full appearance-none cursor-pointer bg-exo-medium-gray accent-exo-yellow"
/>
</div>
<!-- Memory Floor -->
<div>
<div class="flex items-center justify-between mb-1.5">
<div>
<div class="text-sm text-white/90">Memory Floor</div>
<div class="text-xs text-white/40 mt-0.5">Minimum free memory to reserve (GB)</div>
</div>
<span class="text-sm font-mono text-exo-yellow">{draft.memory.memory_floor_gb.toFixed(1)} GB</span>
</div>
<input
type="number"
min="0"
max="64"
step="0.5"
bind:value={draft.memory.memory_floor_gb}
class="w-full bg-exo-medium-gray border border-exo-light-gray/20 rounded px-3 py-1.5 text-sm text-white/90 font-mono focus:outline-none focus:border-exo-yellow/50"
/>
</div>
</div>
</section>
<!-- Generation / Performance -->
<section class="mb-10">
<h2 class="text-sm font-semibold text-white/50 tracking-widest uppercase mb-4">Generation / Performance</h2>
<div class="space-y-5">
<!-- Prefill Step Size -->
<div>
<div class="flex items-center justify-between mb-1.5">
<div>
<div class="text-sm text-white/90">Prefill Step Size</div>
<div class="text-xs text-white/40 mt-0.5">Token chunk size during prompt processing</div>
</div>
<span class="text-sm font-mono text-exo-yellow">{draft.generation.prefill_step_size.toLocaleString()}</span>
</div>
<input
type="number"
min="128"
max="32768"
step="128"
bind:value={draft.generation.prefill_step_size}
class="w-full bg-exo-medium-gray border border-exo-light-gray/20 rounded px-3 py-1.5 text-sm text-white/90 font-mono focus:outline-none focus:border-exo-yellow/50"
/>
</div>
<!-- Max Tokens -->
<div>
<div class="flex items-center justify-between mb-1.5">
<div>
<div class="text-sm text-white/90">Max Tokens</div>
<div class="text-xs text-white/40 mt-0.5">Maximum generation length per response</div>
</div>
<span class="text-sm font-mono text-exo-yellow">{draft.generation.max_tokens.toLocaleString()}</span>
</div>
<input
type="number"
min="1"
max="131072"
step="1024"
bind:value={draft.generation.max_tokens}
class="w-full bg-exo-medium-gray border border-exo-light-gray/20 rounded px-3 py-1.5 text-sm text-white/90 font-mono focus:outline-none focus:border-exo-yellow/50"
/>
</div>
<!-- KV Cache Bits -->
<div>
<div class="mb-1.5">
<div class="text-sm text-white/90">KV Cache Quantization</div>
<div class="text-xs text-white/40 mt-0.5">Lower bits save memory at slight quality cost</div>
</div>
<select
bind:value={draft.generation.kv_cache_bits}
class="w-full bg-exo-medium-gray border border-exo-light-gray/20 rounded px-3 py-1.5 text-sm text-white/90 font-mono focus:outline-none focus:border-exo-yellow/50 cursor-pointer"
>
{#each KV_OPTIONS as opt}
<option value={opt.value}>{opt.label}</option>
{/each}
</select>
</div>
</div>
</section>
<!-- Action Buttons -->
<div class="flex items-center gap-3">
<button
onclick={handleSave}
disabled={loading}
class="px-5 py-2 rounded text-sm font-semibold tracking-wider uppercase transition-colors cursor-pointer
bg-exo-yellow text-exo-black hover:bg-exo-yellow-darker
disabled:opacity-50 disabled:cursor-not-allowed"
>
{loading ? "Saving..." : "Save"}
</button>
<button
onclick={handleReset}
disabled={loading}
class="px-5 py-2 rounded text-sm font-semibold tracking-wider uppercase transition-colors cursor-pointer
border border-exo-light-gray/30 text-white/70 hover:border-exo-yellow/50 hover:text-exo-yellow
disabled:opacity-50 disabled:cursor-not-allowed"
>
Reset to Defaults
</button>
</div>
</div>
</div>
{:else}
<div class="min-h-screen bg-background flex items-center justify-center">
<div class="text-white/40 text-sm">Loading settings...</div>
</div>
{/if}
+190
View File
@@ -0,0 +1,190 @@
<script lang="ts">
import { onMount } from "svelte";
import {
listTraces,
getTraceRawUrl,
type TraceListItem,
} from "$lib/stores/app.svelte";
import HeaderNav from "$lib/components/HeaderNav.svelte";
let traces = $state<TraceListItem[]>([]);
let loading = $state(true);
let error = $state<string | null>(null);
function formatBytes(bytes: number): string {
if (!bytes || bytes <= 0) return "0B";
const units = ["B", "KB", "MB", "GB"];
const i = Math.min(
Math.floor(Math.log(bytes) / Math.log(1024)),
units.length - 1,
);
const val = bytes / Math.pow(1024, i);
return `${val.toFixed(val >= 10 ? 0 : 1)}${units[i]}`;
}
function formatDate(isoString: string): string {
const date = new Date(isoString);
return date.toLocaleString();
}
async function downloadTrace(taskId: string) {
const response = await fetch(getTraceRawUrl(taskId));
const blob = await response.blob();
const url = URL.createObjectURL(blob);
const a = document.createElement("a");
a.href = url;
a.download = `trace_${taskId}.json`;
a.click();
URL.revokeObjectURL(url);
}
async function openInPerfetto(taskId: string) {
// Fetch trace data from our local API
const response = await fetch(getTraceRawUrl(taskId));
const traceData = await response.arrayBuffer();
// Open Perfetto UI
const perfettoWindow = window.open("https://ui.perfetto.dev");
if (!perfettoWindow) {
alert("Failed to open Perfetto. Please allow popups.");
return;
}
// Wait for Perfetto to be ready, then send trace via postMessage
const onMessage = (e: MessageEvent) => {
if (e.data === "PONG") {
window.removeEventListener("message", onMessage);
perfettoWindow.postMessage(
{
perfetto: {
buffer: traceData,
title: `Trace ${taskId}`,
},
},
"https://ui.perfetto.dev",
);
}
};
window.addEventListener("message", onMessage);
// Ping Perfetto until it responds
const pingInterval = setInterval(() => {
perfettoWindow.postMessage("PING", "https://ui.perfetto.dev");
}, 50);
// Clean up after 10 seconds
setTimeout(() => {
clearInterval(pingInterval);
window.removeEventListener("message", onMessage);
}, 10000);
}
async function refresh() {
loading = true;
error = null;
try {
const response = await listTraces();
traces = response.traces;
} catch (e) {
error = e instanceof Error ? e.message : "Failed to load traces";
} finally {
loading = false;
}
}
onMount(() => {
refresh();
});
</script>
<div class="min-h-screen bg-exo-dark-gray text-white">
<HeaderNav showHome={true} />
<div class="max-w-7xl mx-auto px-4 lg:px-8 py-6 space-y-6">
<div class="flex items-center justify-between gap-4 flex-wrap">
<div>
<h1
class="text-2xl font-mono tracking-[0.2em] uppercase text-exo-yellow"
>
Traces
</h1>
</div>
<div class="flex items-center gap-3">
<button
type="button"
class="text-xs font-mono text-exo-light-gray hover:text-exo-yellow transition-colors uppercase border border-exo-medium-gray/40 px-2 py-1 rounded"
onclick={refresh}
disabled={loading}
>
Refresh
</button>
</div>
</div>
{#if loading}
<div
class="rounded border border-exo-medium-gray/30 bg-exo-black/30 p-6 text-center text-exo-light-gray"
>
<div class="text-sm">Loading traces...</div>
</div>
{:else if error}
<div
class="rounded border border-red-500/30 bg-red-500/10 p-6 text-center text-red-400"
>
<div class="text-sm">{error}</div>
</div>
{:else if traces.length === 0}
<div
class="rounded border border-exo-medium-gray/30 bg-exo-black/30 p-6 text-center text-exo-light-gray space-y-2"
>
<div class="text-sm">No traces found.</div>
<div class="text-xs text-exo-light-gray/70">
Run exo with EXO_TRACING_ENABLED=1 to collect traces.
</div>
</div>
{:else}
<div class="space-y-3">
{#each traces as trace}
<div
class="rounded border border-exo-medium-gray/30 bg-exo-black/30 p-4 flex items-center justify-between gap-4"
>
<div class="min-w-0 flex-1">
<a
href="#/traces/{trace.taskId}"
class="text-sm font-mono text-white hover:text-exo-yellow transition-colors truncate block"
>
{trace.taskId}
</a>
<div class="text-xs text-exo-light-gray font-mono mt-1">
{formatDate(trace.createdAt)} &bull; {formatBytes(
trace.fileSize,
)}
</div>
</div>
<div class="flex items-center gap-2 shrink-0">
<a
href="#/traces/{trace.taskId}"
class="text-xs font-mono text-exo-light-gray hover:text-exo-yellow transition-colors uppercase border border-exo-medium-gray/40 px-2 py-1 rounded"
>
View Stats
</a>
<button
type="button"
class="text-xs font-mono text-exo-light-gray hover:text-exo-yellow transition-colors uppercase border border-exo-medium-gray/40 px-2 py-1 rounded"
onclick={() => downloadTrace(trace.taskId)}
>
Download
</button>
<button
type="button"
class="text-xs font-mono text-exo-dark-gray bg-exo-yellow hover:bg-exo-yellow/90 transition-colors uppercase px-2 py-1 rounded font-semibold"
onclick={() => openInPerfetto(trace.taskId)}
>
View Trace
</button>
</div>
</div>
{/each}
</div>
{/if}
</div>
</div>
@@ -0,0 +1,367 @@
<script lang="ts">
import { page } from "$app/stores";
import { onMount } from "svelte";
import {
fetchTraceStats,
getTraceRawUrl,
type TraceStatsResponse,
type TraceCategoryStats,
} from "$lib/stores/app.svelte";
import HeaderNav from "$lib/components/HeaderNav.svelte";
const taskId = $derived($page.params.taskId);
let stats = $state<TraceStatsResponse | null>(null);
let loading = $state(true);
let error = $state<string | null>(null);
function formatDuration(us: number): string {
if (us < 1000) return `${us.toFixed(0)}us`;
if (us < 1_000_000) return `${(us / 1000).toFixed(2)}ms`;
return `${(us / 1_000_000).toFixed(2)}s`;
}
function formatPercentage(part: number, total: number): string {
if (total === 0) return "0.0%";
return `${((part / total) * 100).toFixed(1)}%`;
}
// Parse hierarchical categories like "sync/compute" into phases
type PhaseData = {
name: string;
subcategories: { name: string; stats: TraceCategoryStats }[];
totalUs: number; // From outer span (e.g., "sync" category)
stepCount: number; // Count of outer span events
};
function parsePhases(
byCategory: Record<string, TraceCategoryStats>,
): PhaseData[] {
const phases = new Map<
string,
{
subcats: Map<string, TraceCategoryStats>;
outerStats: TraceCategoryStats | null;
}
>();
for (const [category, catStats] of Object.entries(byCategory)) {
if (category.includes("/")) {
const [phase, subcat] = category.split("/", 2);
if (!phases.has(phase)) {
phases.set(phase, { subcats: new Map(), outerStats: null });
}
phases.get(phase)!.subcats.set(subcat, catStats);
} else {
// Outer span - this IS the phase total
if (!phases.has(category)) {
phases.set(category, { subcats: new Map(), outerStats: null });
}
phases.get(category)!.outerStats = catStats;
}
}
return Array.from(phases.entries())
.filter(([_, data]) => data.outerStats !== null) // Only phases with outer spans
.map(([name, data]) => ({
name,
subcategories: Array.from(data.subcats.entries())
.map(([subName, subStats]) => ({ name: subName, stats: subStats }))
.sort((a, b) => b.stats.totalUs - a.stats.totalUs),
totalUs: data.outerStats!.totalUs, // Outer span total
stepCount: data.outerStats!.count, // Number of steps
}))
.sort((a, b) => b.totalUs - a.totalUs);
}
async function downloadTrace() {
if (!taskId) return;
const response = await fetch(getTraceRawUrl(taskId));
const blob = await response.blob();
const url = URL.createObjectURL(blob);
const a = document.createElement("a");
a.href = url;
a.download = `trace_${taskId}.json`;
a.click();
URL.revokeObjectURL(url);
}
async function openInPerfetto() {
if (!taskId) return;
// Fetch trace data from our local API
const response = await fetch(getTraceRawUrl(taskId));
const traceData = await response.arrayBuffer();
// Open Perfetto UI
const perfettoWindow = window.open("https://ui.perfetto.dev");
if (!perfettoWindow) {
alert("Failed to open Perfetto. Please allow popups.");
return;
}
// Wait for Perfetto to be ready, then send trace via postMessage
const onMessage = (e: MessageEvent) => {
if (e.data === "PONG") {
window.removeEventListener("message", onMessage);
perfettoWindow.postMessage(
{
perfetto: {
buffer: traceData,
title: `Trace ${taskId}`,
},
},
"https://ui.perfetto.dev",
);
}
};
window.addEventListener("message", onMessage);
// Ping Perfetto until it responds
const pingInterval = setInterval(() => {
perfettoWindow.postMessage("PING", "https://ui.perfetto.dev");
}, 50);
// Clean up after 10 seconds
setTimeout(() => {
clearInterval(pingInterval);
window.removeEventListener("message", onMessage);
}, 10000);
}
onMount(async () => {
if (!taskId) {
error = "No task ID provided";
loading = false;
return;
}
try {
stats = await fetchTraceStats(taskId);
} catch (e) {
error = e instanceof Error ? e.message : "Failed to load trace";
} finally {
loading = false;
}
});
const phases = $derived(stats ? parsePhases(stats.byCategory) : []);
const sortedRanks = $derived(
stats
? Object.keys(stats.byRank)
.map(Number)
.sort((a, b) => a - b)
: [],
);
const nodeCount = $derived(sortedRanks.length || 1);
</script>
<div class="min-h-screen bg-exo-dark-gray text-white">
<HeaderNav showHome={true} />
<div class="max-w-7xl mx-auto px-4 lg:px-8 py-6 space-y-6">
<div class="flex items-center justify-between gap-4 flex-wrap">
<div>
<h1
class="text-2xl font-mono tracking-[0.2em] uppercase text-exo-yellow"
>
Trace
</h1>
<p class="text-sm text-exo-light-gray font-mono truncate max-w-lg">
{taskId}
</p>
</div>
<div class="flex items-center gap-3">
<a
href="#/traces"
class="text-xs font-mono text-exo-light-gray hover:text-exo-yellow transition-colors uppercase border border-exo-medium-gray/40 px-3 py-1.5 rounded"
>
All Traces
</a>
<button
type="button"
class="text-xs font-mono text-exo-light-gray hover:text-exo-yellow transition-colors uppercase border border-exo-medium-gray/40 px-3 py-1.5 rounded"
onclick={downloadTrace}
disabled={loading || !!error}
>
Download
</button>
<button
type="button"
class="text-xs font-mono text-exo-dark-gray bg-exo-yellow hover:bg-exo-yellow/90 transition-colors uppercase px-3 py-1.5 rounded font-semibold"
onclick={openInPerfetto}
disabled={loading || !!error}
>
View Trace
</button>
</div>
</div>
{#if loading}
<div
class="rounded border border-exo-medium-gray/30 bg-exo-black/30 p-6 text-center text-exo-light-gray"
>
<div class="text-sm">Loading trace data...</div>
</div>
{:else if error}
<div
class="rounded border border-red-500/30 bg-red-500/10 p-6 text-center text-red-400"
>
<div class="text-sm">{error}</div>
</div>
{:else if stats}
<!-- Wall Time Summary -->
<div
class="rounded border border-exo-medium-gray/30 bg-exo-black/30 p-4 space-y-2"
>
<h2
class="text-sm font-mono uppercase tracking-wider text-exo-light-gray"
>
Summary
</h2>
<div class="text-3xl font-mono text-exo-yellow">
{formatDuration(stats.totalWallTimeUs)}
</div>
<div class="text-xs text-exo-light-gray">Total wall time</div>
</div>
<!-- By Phase -->
{#if phases.length > 0}
<div
class="rounded border border-exo-medium-gray/30 bg-exo-black/30 p-4 space-y-4"
>
<h2
class="text-sm font-mono uppercase tracking-wider text-exo-light-gray"
>
By Phase <span class="text-exo-light-gray/50">(avg per node)</span>
</h2>
<div class="space-y-4">
{#each phases as phase}
{@const normalizedTotal = phase.totalUs / nodeCount}
{@const normalizedStepCount = phase.stepCount / nodeCount}
<div class="space-y-2">
<div class="flex items-center justify-between">
<span class="text-sm font-mono text-white">{phase.name}</span>
<span class="text-sm font-mono">
<span class="text-exo-yellow"
>{formatDuration(normalizedTotal)}</span
>
<span class="text-exo-light-gray ml-2">
({normalizedStepCount} steps, {formatDuration(
normalizedTotal / normalizedStepCount,
)}/step)
</span>
</span>
</div>
{#if phase.subcategories.length > 0}
<div class="pl-4 space-y-1.5">
{#each phase.subcategories as subcat}
{@const normalizedSubcat =
subcat.stats.totalUs / nodeCount}
{@const pct = formatPercentage(
normalizedSubcat,
normalizedTotal,
)}
{@const perStep = normalizedSubcat / normalizedStepCount}
<div
class="flex items-center justify-between text-xs font-mono"
>
<span class="text-exo-light-gray">{subcat.name}</span>
<span class="text-white">
{formatDuration(normalizedSubcat)}
<span class="text-exo-light-gray ml-2">({pct})</span>
<span class="text-exo-light-gray/60 ml-2"
>{formatDuration(perStep)}/step</span
>
</span>
</div>
<!-- Progress bar -->
<div
class="relative h-1.5 bg-exo-black/60 rounded-sm overflow-hidden"
>
<div
class="absolute inset-y-0 left-0 bg-gradient-to-r from-exo-yellow to-exo-yellow/70 transition-all duration-300"
style="width: {pct}"
></div>
</div>
{/each}
</div>
{/if}
</div>
{/each}
</div>
</div>
{/if}
<!-- By Rank -->
{#if sortedRanks.length > 0}
<div
class="rounded border border-exo-medium-gray/30 bg-exo-black/30 p-4 space-y-4"
>
<h2
class="text-sm font-mono uppercase tracking-wider text-exo-light-gray"
>
By Rank
</h2>
<div class="grid grid-cols-1 md:grid-cols-2 lg:grid-cols-3 gap-4">
{#each sortedRanks as rank}
{@const rankStats = stats.byRank[rank]}
{@const rankPhases = parsePhases(rankStats.byCategory)}
<div
class="rounded border border-exo-medium-gray/20 bg-exo-dark-gray/60 p-3 space-y-3"
>
<div class="text-sm font-mono text-exo-yellow">
Rank {rank}
</div>
<div class="space-y-2">
{#each rankPhases as phase}
<div class="space-y-1">
<div class="flex items-center justify-between text-xs">
<span class="font-mono text-exo-light-gray"
>{phase.name}</span
>
<span class="font-mono text-white">
{formatDuration(phase.totalUs)}
<span class="text-exo-light-gray/50 ml-1">
({phase.stepCount}x)
</span>
</span>
</div>
{#if phase.subcategories.length > 0}
<div class="pl-2 space-y-0.5">
{#each phase.subcategories as subcat}
{@const pct = formatPercentage(
subcat.stats.totalUs,
phase.totalUs,
)}
{@const perStep =
subcat.stats.totalUs / phase.stepCount}
<div
class="flex items-center justify-between text-[10px] font-mono"
>
<span class="text-exo-light-gray/70"
>{subcat.name}</span
>
<span class="text-exo-light-gray">
{formatDuration(subcat.stats.totalUs)}
<span class="text-exo-light-gray/50"
>({pct})</span
>
<span class="text-exo-light-gray/30 ml-1"
>{formatDuration(perStep)}/step</span
>
</span>
</div>
{/each}
</div>
{/if}
</div>
{/each}
</div>
</div>
{/each}
</div>
</div>
{/if}
{/if}
</div>
</div>
Generated
+59 -6
View File
@@ -21,7 +21,9 @@
"nixpkgs"
],
"purescript-overlay": "purescript-overlay",
"pyproject-nix": "pyproject-nix"
"pyproject-nix": [
"pyproject-nix"
]
},
"locked": {
"lastModified": 1765953015,
@@ -149,19 +151,44 @@
"type": "github"
}
},
"pyproject-build-systems": {
"inputs": {
"nixpkgs": [
"nixpkgs"
],
"pyproject-nix": [
"pyproject-nix"
],
"uv2nix": [
"uv2nix"
]
},
"locked": {
"lastModified": 1763662255,
"narHash": "sha256-4bocaOyLa3AfiS8KrWjZQYu+IAta05u3gYZzZ6zXbT0=",
"owner": "pyproject-nix",
"repo": "build-system-pkgs",
"rev": "042904167604c681a090c07eb6967b4dd4dae88c",
"type": "github"
},
"original": {
"owner": "pyproject-nix",
"repo": "build-system-pkgs",
"type": "github"
}
},
"pyproject-nix": {
"inputs": {
"nixpkgs": [
"dream2nix",
"nixpkgs"
]
},
"locked": {
"lastModified": 1763017646,
"narHash": "sha256-Z+R2lveIp6Skn1VPH3taQIuMhABg1IizJd8oVdmdHsQ=",
"lastModified": 1764134915,
"narHash": "sha256-xaKvtPx6YAnA3HQVp5LwyYG1MaN4LLehpQI8xEdBvBY=",
"owner": "pyproject-nix",
"repo": "pyproject.nix",
"rev": "47bd6f296502842643078d66128f7b5e5370790c",
"rev": "2c8df1383b32e5443c921f61224b198a2282a657",
"type": "github"
},
"original": {
@@ -178,7 +205,10 @@
"flake-parts": "flake-parts",
"nixpkgs": "nixpkgs",
"nixpkgs-swift": "nixpkgs-swift",
"treefmt-nix": "treefmt-nix"
"pyproject-build-systems": "pyproject-build-systems",
"pyproject-nix": "pyproject-nix",
"treefmt-nix": "treefmt-nix",
"uv2nix": "uv2nix"
}
},
"rust-analyzer-src": {
@@ -239,6 +269,29 @@
"repo": "treefmt-nix",
"type": "github"
}
},
"uv2nix": {
"inputs": {
"nixpkgs": [
"nixpkgs"
],
"pyproject-nix": [
"pyproject-nix"
]
},
"locked": {
"lastModified": 1767701098,
"narHash": "sha256-CJhKZnWb3gumR9oTRjFvCg/6lYTGbZRU7xtvcyWIRwU=",
"owner": "pyproject-nix",
"repo": "uv2nix",
"rev": "9d357f0d2ce6f5f35ec7959d7e704452352eb4da",
"type": "github"
},
"original": {
"owner": "pyproject-nix",
"repo": "uv2nix",
"type": "github"
}
}
},
"root": "root",
+45 -6
View File
@@ -24,6 +24,26 @@
dream2nix = {
url = "github:nix-community/dream2nix";
inputs.nixpkgs.follows = "nixpkgs";
inputs.pyproject-nix.follows = "pyproject-nix";
};
# Python packaging with uv2nix
pyproject-nix = {
url = "github:pyproject-nix/pyproject.nix";
inputs.nixpkgs.follows = "nixpkgs";
};
uv2nix = {
url = "github:pyproject-nix/uv2nix";
inputs.pyproject-nix.follows = "pyproject-nix";
inputs.nixpkgs.follows = "nixpkgs";
};
pyproject-build-systems = {
url = "github:pyproject-nix/build-system-pkgs";
inputs.pyproject-nix.follows = "pyproject-nix";
inputs.uv2nix.follows = "uv2nix";
inputs.nixpkgs.follows = "nixpkgs";
};
# Pinned nixpkgs for swift-format (swift is broken on x86_64-linux in newer nixpkgs)
@@ -48,16 +68,24 @@
inputs.treefmt-nix.flakeModule
./dashboard/parts.nix
./rust/parts.nix
./python/parts.nix
];
perSystem =
{ config, self', inputs', pkgs, lib, system, ... }:
let
fenixToolchain = inputs'.fenix.packages.complete;
# Use pinned nixpkgs for swift-format (swift is broken on x86_64-linux in newer nixpkgs)
pkgsSwift = import inputs.nixpkgs-swift { inherit system; };
in
{
# Allow unfree for metal-toolchain (needed for Darwin Metal packages)
_module.args.pkgs = import inputs.nixpkgs {
inherit system;
config.allowUnfreePredicate = pkg: (pkg.pname or "") == "metal-toolchain";
overlays = [
(import ./nix/apple-sdk-overlay.nix)
];
};
treefmt = {
projectRootFile = "flake.nix";
programs = {
@@ -79,14 +107,25 @@
enable = true;
package = pkgsSwift.swiftPackages.swift-format;
};
shfmt.enable = true;
};
};
checks.lint = pkgs.runCommand "lint-check" { } ''
export RUFF_CACHE_DIR="$TMPDIR/ruff-cache"
${pkgs.ruff}/bin/ruff check ${inputs.self}/
touch $out
'';
packages = lib.optionalAttrs pkgs.stdenv.hostPlatform.isDarwin (
let
uvLock = builtins.fromTOML (builtins.readFile ./uv.lock);
mlxPackage = builtins.head (builtins.filter (p: p.name == "mlx" && p.source ? git) uvLock.package);
uvLockMlxVersion = mlxPackage.version;
in
{
metal-toolchain = pkgs.callPackage ./nix/metal-toolchain.nix { };
mlx = pkgs.callPackage ./nix/mlx.nix {
inherit (self'.packages) metal-toolchain;
inherit uvLockMlxVersion;
};
default = self'.packages.exo;
}
);
devShells.default = with pkgs; pkgs.mkShell {
inputsFrom = [ self'.checks.cargo-build ];
+2 -2
View File
@@ -1,7 +1,7 @@
export NIX_CONFIG := "extra-experimental-features = nix-command flakes"
fmt:
nix fmt
treefmt || nix fmt
lint:
uv run ruff check --fix
@@ -20,7 +20,7 @@ sync-clean:
rust-rebuild:
cargo run --bin stub_gen
just sync-clean
uv sync --reinstall-package exo_pyo3_bindings
build-dashboard:
#!/usr/bin/env bash
+18
View File
@@ -0,0 +1,18 @@
# Overlay that builds apple-sdk with a custom versions.json (for SDK 26.2).
# The upstream nixpkgs package reads versions.json at eval time via a relative
# path, so we can't override it through callPackage args. Instead, we copy
# the upstream source and patch the one file.
final: _prev:
let
upstreamSrc = final.path + "/pkgs/by-name/ap/apple-sdk";
patchedSrc = final.runCommandLocal "apple-sdk-src-patched" { } ''
cp -r ${upstreamSrc} $out
chmod -R u+w $out
cp ${./apple-sdk/metadata/versions.json} $out/metadata/versions.json
'';
in
{
apple-sdk_26 = final.callPackage (patchedSrc + "/package.nix") {
darwinSdkMajorVersion = "26";
};
}
+26
View File
@@ -0,0 +1,26 @@
{
"14": {
"urls": [
"https://swcdn.apple.com/content/downloads/14/48/052-59890-A_I0F5YGAY0Y/p9n40hio7892gou31o1v031ng6fnm9sb3c/CLTools_macOSNMOS_SDK.pkg",
"https://web.archive.org/web/20250211001355/https://swcdn.apple.com/content/downloads/14/48/052-59890-A_I0F5YGAY0Y/p9n40hio7892gou31o1v031ng6fnm9sb3c/CLTools_macOSNMOS_SDK.pkg"
],
"version": "14.4",
"hash": "sha256-QozDiwY0Czc0g45vPD7G4v4Ra+3DujCJbSads3fJjjM="
},
"15": {
"urls": [
"https://swcdn.apple.com/content/downloads/52/01/082-41241-A_0747ZN8FHV/dectd075r63pppkkzsb75qk61s0lfee22j/CLTools_macOSNMOS_SDK.pkg",
"https://web.archive.org/web/20250530132510/https://swcdn.apple.com/content/downloads/52/01/082-41241-A_0747ZN8FHV/dectd075r63pppkkzsb75qk61s0lfee22j/CLTools_macOSNMOS_SDK.pkg"
],
"version": "15.5",
"hash": "sha256-HBiSJuw1XBUK5R/8Sj65c3rftSEvQl/O9ZZVp/g1Amo="
},
"26": {
"urls": [
"https://swcdn.apple.com/content/downloads/60/22/089-71960-A_W8BL1RUJJ6/5zkyplomhk1cm7z6xja2ktgapnhhti6wwd/CLTools_macOSNMOS_SDK.pkg",
"https://web.archive.org/web/20250915230423/https://swcdn.apple.com/content/downloads/60/22/089-71960-A_W8BL1RUJJ6/5zkyplomhk1cm7z6xja2ktgapnhhti6wwd/CLTools_macOSNMOS_SDK.pkg"
],
"version": "26.2",
"hash": "sha256-hXRlMieVv0smna5uiWRwq87IWOaPWtAjAldbi+wQXcw="
}
}
+79
View File
@@ -0,0 +1,79 @@
diff --git a/CMakeLists.txt b/CMakeLists.txt
index 0ed30932..d8528132 100644
--- a/CMakeLists.txt
+++ b/CMakeLists.txt
@@ -177,11 +177,7 @@ if(MLX_BUILD_METAL)
add_compile_definitions(MLX_METAL_DEBUG)
endif()
- # Throw an error if xcrun not found
- execute_process(
- COMMAND zsh "-c" "/usr/bin/xcrun -sdk macosx --show-sdk-version"
- OUTPUT_VARIABLE MACOS_SDK_VERSION
- OUTPUT_STRIP_TRAILING_WHITESPACE COMMAND_ERROR_IS_FATAL ANY)
+ set(MACOS_SDK_VERSION @sdkVersion@)
if(${MACOS_SDK_VERSION} LESS 14.0)
message(
@@ -199,11 +195,8 @@ if(MLX_BUILD_METAL)
endif()
set(XCRUN_FLAGS "-mmacosx-version-min=${CMAKE_OSX_DEPLOYMENT_TARGET}")
endif()
- execute_process(
- COMMAND
- zsh "-c"
- "echo \"__METAL_VERSION__\" | xcrun -sdk macosx metal ${XCRUN_FLAGS} -E -x metal -P - | tail -1 | tr -d '\n'"
- OUTPUT_VARIABLE MLX_METAL_VERSION COMMAND_ERROR_IS_FATAL ANY)
+ set(
+ MLX_METAL_VERSION @metalVersion@)
FetchContent_Declare(metal_cpp URL ${METAL_CPP_URL})
FetchContent_MakeAvailable(metal_cpp)
target_include_directories(
diff --git a/cmake/extension.cmake b/cmake/extension.cmake
index 13db804a..5b385132 100644
--- a/cmake/extension.cmake
+++ b/cmake/extension.cmake
@@ -36,7 +36,7 @@ macro(mlx_build_metallib)
add_custom_command(
OUTPUT ${MTLLIB_BUILD_TARGET}
COMMAND
- xcrun -sdk macosx metal
+ metal -fmodules-cache-path=${CMAKE_BINARY_DIR}/metal-cache
"$<LIST:TRANSFORM,${MTLLIB_INCLUDE_DIRS},PREPEND,-I>"
${MTLLIB_COMPILE_OPTIONS} ${MTLLIB_SOURCES} -o ${MTLLIB_BUILD_TARGET}
DEPENDS ${MTLLIB_DEPS} ${MTLLIB_SOURCES}
diff --git a/mlx/backend/metal/kernels/CMakeLists.txt b/mlx/backend/metal/kernels/CMakeLists.txt
index 262b0495..5c7446ad 100644
--- a/mlx/backend/metal/kernels/CMakeLists.txt
+++ b/mlx/backend/metal/kernels/CMakeLists.txt
@@ -29,7 +29,7 @@ function(build_kernel_base TARGET SRCFILE DEPS)
"-mmacosx-version-min=${CMAKE_OSX_DEPLOYMENT_TARGET}")
endif()
add_custom_command(
- COMMAND xcrun -sdk macosx metal ${METAL_FLAGS} -c ${SRCFILE}
+ COMMAND metal -fmodules-cache-path=${CMAKE_BINARY_DIR}/metal-cache ${METAL_FLAGS} -c ${SRCFILE}
-I${PROJECT_SOURCE_DIR} -o ${TARGET}.air
DEPENDS ${SRCFILE} ${DEPS} ${BASE_HEADERS}
OUTPUT ${TARGET}.air
@@ -170,7 +170,7 @@ endif()
add_custom_command(
OUTPUT ${MLX_METAL_PATH}/mlx.metallib
- COMMAND xcrun -sdk macosx metallib ${KERNEL_AIR} -o
+ COMMAND metallib ${KERNEL_AIR} -o
${MLX_METAL_PATH}/mlx.metallib
DEPENDS ${KERNEL_AIR}
COMMENT "Building mlx.metallib"
diff --git a/mlx/backend/metal/make_compiled_preamble.sh b/mlx/backend/metal/make_compiled_preamble.sh
index bb55ed3a..94ea7dd7 100644
--- a/mlx/backend/metal/make_compiled_preamble.sh
+++ b/mlx/backend/metal/make_compiled_preamble.sh
@@ -31,7 +31,7 @@ OUTPUT_FILE=${OUTPUT_DIR}/${SRC_NAME}.cpp
mkdir -p "$OUTPUT_DIR"
# Use the metal compiler to get a list of headers (with depth)
-CCC="xcrun -sdk macosx metal -x metal"
+CCC="metal -x metal -fmodules-cache-path=${OUTPUT_DIR}/metal-cache"
HDRS=$( $CCC -I"$SRC_DIR" -I"$JIT_INCLUDES" -DMLX_METAL_JIT -E -P -CC -C -H "$INPUT_FILE" $CFLAGS -w 2>&1 1>/dev/null )
# Remove any included system frameworks (for MetalPerformancePrimitive headers)
+56
View File
@@ -0,0 +1,56 @@
{ lib, stdenvNoCC, requireFile, nix }:
let
narFile = requireFile {
name = "metal-toolchain-17C48.nar";
message = ''
The Metal Toolchain NAR must be available.
If you have cachix configured for exo.cachix.org, this should be automatic.
Otherwise:
1. Install Xcode 26+ from the App Store
2. Run: xcodebuild -downloadComponent MetalToolchain
3. Export the toolchain:
hdiutil attach "$(find /System/Library/AssetsV2/com_apple_MobileAsset_MetalToolchain -name '*.dmg' | head -1)" -mountpoint /tmp/metal-dmg
cp -R /tmp/metal-dmg/Metal.xctoolchain /tmp/metal-export
hdiutil detach /tmp/metal-dmg
4. Create NAR and add to store:
nix nar pack /tmp/metal-export > /tmp/metal-toolchain-17C48.nar
nix store add --mode flat /tmp/metal-toolchain-17C48.nar
'';
hash = "sha256-ayR5mXN4sZAddwKEG2OszGRF93k9ZFc7H0yi2xbylQw=";
};
in
stdenvNoCC.mkDerivation {
pname = "metal-toolchain";
version = "17C48";
dontUnpack = true;
dontBuild = true;
dontFixup = true;
nativeBuildInputs = [ nix ];
installPhase = ''
runHook preInstall
nix-store --restore $out < ${narFile}
# Create bin directory with symlinks for PATH
mkdir -p $out/bin
ln -s $out/usr/bin/metal $out/bin/metal
ln -s $out/usr/bin/metallib $out/bin/metallib
runHook postInstall
'';
# Metal language version for CMake (from: echo __METAL_VERSION__ | metal -E -x metal -P -)
passthru.metalVersion = "400";
meta = {
description = "Apple Metal compiler toolchain";
platforms = [ "aarch64-darwin" ];
license = lib.licenses.unfree;
};
}
+159
View File
@@ -0,0 +1,159 @@
{ stdenv
, lib
, fetchFromGitHub
, replaceVars
, fetchzip
, cmake
, nlohmann_json
, apple-sdk_26
, metal-toolchain
, runCommand
, fmt
, python313Packages
, uvLockMlxVersion
}:
assert stdenv.isDarwin;
let
python = python313Packages.python;
# Static dependencies included directly during compilation
gguf-tools = fetchFromGitHub {
owner = "antirez";
repo = "gguf-tools";
rev = "8fa6eb65236618e28fd7710a0fba565f7faa1848";
hash = "sha256-15FvyPOFqTOr5vdWQoPnZz+mYH919++EtghjozDlnSA=";
};
metal_cpp = fetchzip {
url = "https://developer.apple.com/metal/cpp/files/metal-cpp_26.zip";
hash = "sha256-7n2eI2lw/S+Us6l7YPAATKwcIbRRpaQ8VmES7S8ZjY8=";
};
nanobind = fetchFromGitHub {
owner = "wjakob";
repo = "nanobind";
rev = "v2.10.2";
hash = "sha256-io44YhN+VpfHFWyvvLWSanRgbzA0whK8WlDNRi3hahU=";
fetchSubmodules = true;
};
mlx = stdenv.mkDerivation rec {
pname = "mlx";
version = let v = "0.30.7.dev20260225+257d5692"; in
assert v == uvLockMlxVersion || throw "MLX version mismatch: nix/mlx.nix has ${v} but uv.lock has ${uvLockMlxVersion}. Update both the version and hash in nix/mlx.nix.";
v;
pyproject = true;
src = fetchFromGitHub {
owner = "rltakashige";
repo = "mlx-jaccl-fix-small-recv";
rev = "257d5692fc7af6bba3b8afaeb63c549b7d1e43d5";
hash = "sha256-GosFIWxIB48Egb1MqJrR3xhsUsQeWdRk5rV93USY6wQ=";
};
patches = [
(replaceVars ./darwin-build-fixes.patch {
sdkVersion = apple-sdk_26.version;
metalVersion = metal-toolchain.metalVersion;
})
];
postPatch = ''
substituteInPlace mlx/backend/cpu/jit_compiler.cpp \
--replace-fail "g++" "$CXX"
'';
dontUseCmakeConfigure = true;
enableParallelBuilding = true;
# Allows multiple cores to be used in Python builds.
postUnpack = ''
export MAKEFLAGS+="''${enableParallelBuilding:+-j$NIX_BUILD_CORES}"
'';
# Updates the wrong fetcher rev attribute
passthru.skipBulkUpdate = true;
env = {
DEV_RELEASE = 1;
CMAKE_ARGS = toString [
(lib.cmakeBool "USE_SYSTEM_FMT" true)
(lib.cmakeOptionType "filepath" "FETCHCONTENT_SOURCE_DIR_GGUFLIB" "${gguf-tools}")
(lib.cmakeOptionType "filepath" "FETCHCONTENT_SOURCE_DIR_JSON" "${nlohmann_json.src}")
(lib.cmakeOptionType "filepath" "FETCHCONTENT_SOURCE_DIR_NANOBIND" "${nanobind}")
(lib.cmakeBool "FETCHCONTENT_FULLY_DISCONNECTED" true)
(lib.cmakeBool "MLX_BUILD_CPU" true)
(lib.cmakeBool "MLX_BUILD_METAL" true)
(lib.cmakeOptionType "filepath" "FETCHCONTENT_SOURCE_DIR_METAL_CPP" "${metal_cpp}")
(lib.cmakeOptionType "string" "CMAKE_OSX_DEPLOYMENT_TARGET" "${apple-sdk_26.version}")
(lib.cmakeOptionType "filepath" "CMAKE_OSX_SYSROOT" "${apple-sdk_26.passthru.sdkroot}")
];
SDKROOT = apple-sdk_26.passthru.sdkroot;
MACOSX_DEPLOYMENT_TARGET = apple-sdk_26.version;
};
build-system = [
python313Packages.setuptools
];
nativeBuildInputs = [
cmake
metal-toolchain
python313Packages.pypaBuildHook
python313Packages.pypaInstallHook
python313Packages.setuptools
python313Packages.typing-extensions
python313Packages.wheel
python313Packages.cmake
python313Packages.ninja
];
buildInputs = [
fmt
gguf-tools
python313Packages.nanobind
python313Packages.pybind11
apple-sdk_26
];
# Tests require Metal GPU access which isn't available in the Nix sandbox.
# To run tests, build with: nix build --option sandbox false .#mlx.passthru.tests.mlxTest
doCheck = false;
pythonImportsCheck = [ "mlx" ];
passthru.tests = {
# Runs example scripts to verify MLX works. Requires --option sandbox false
# since Metal GPU access is needed.
mlxTest =
runCommand "run-mlx-examples"
{
buildInputs = [ mlx ];
nativeBuildInputs = [ python ];
}
''
cp ${src}/examples/python/logistic_regression.py .
${python.interpreter} logistic_regression.py
rm logistic_regression.py
cp ${src}/examples/python/linear_regression.py .
${python.interpreter} linear_regression.py
rm linear_regression.py
touch $out
'';
};
meta = {
homepage = "https://github.com/ml-explore/mlx";
description = "Array framework for Apple silicon";
changelog = "https://github.com/ml-explore/mlx/releases/tag/${src.tag}";
license = lib.licenses.mit;
platforms = [ "aarch64-darwin" ];
};
};
in
mlx
Binary file not shown.

After

Width:  |  Height:  |  Size: 6.8 KiB

+112
View File
@@ -0,0 +1,112 @@
#!/usr/bin/env bash
# create-dmg.sh — Build a polished macOS DMG installer for EXO
#
# Usage:
# ./packaging/dmg/create-dmg.sh <app-path> <output-dmg> [volume-name]
#
# Example:
# ./packaging/dmg/create-dmg.sh output/EXO.app EXO-1.0.0.dmg "EXO"
#
# Creates a DMG with:
# - Custom background image with drag-to-Applications arrow
# - App icon on left, Applications alias on right
# - Proper window size and icon positioning
set -euo pipefail
APP_PATH="${1:?Usage: create-dmg.sh <app-path> <output-dmg> [volume-name]}"
OUTPUT_DMG="${2:?Usage: create-dmg.sh <app-path> <output-dmg> [volume-name]}"
VOLUME_NAME="${3:-EXO}"
SCRIPT_DIR="$(cd "$(dirname "$0")" && pwd)"
BACKGROUND_SCRIPT="${SCRIPT_DIR}/generate-background.py"
TEMP_DIR="$(mktemp -d)"
DMG_STAGING="${TEMP_DIR}/dmg-root"
TEMP_DMG="${TEMP_DIR}/temp.dmg"
BACKGROUND_PNG="${TEMP_DIR}/background.png"
cleanup() { rm -rf "$TEMP_DIR"; }
trap cleanup EXIT
echo "==> Creating DMG installer for ${VOLUME_NAME}"
# ── Step 1: Generate background image ────────────────────────────────────────
if command -v python3 &>/dev/null; then
python3 "$BACKGROUND_SCRIPT" "$BACKGROUND_PNG"
echo " Background image generated"
else
echo " Warning: python3 not found, skipping custom background"
BACKGROUND_PNG=""
fi
# ── Step 2: Prepare staging directory ─────────────────────────────────────────
mkdir -p "$DMG_STAGING"
cp -R "$APP_PATH" "$DMG_STAGING/"
ln -s /Applications "$DMG_STAGING/Applications"
# ── Step 3: Create writable DMG ──────────────────────────────────────────────
# Calculate required size (app size + 20MB headroom)
APP_SIZE_KB=$(du -sk "$APP_PATH" | cut -f1)
DMG_SIZE_KB=$((APP_SIZE_KB + 20480))
hdiutil create \
-volname "$VOLUME_NAME" \
-size "${DMG_SIZE_KB}k" \
-fs HFS+ \
-layout SPUD \
"$TEMP_DMG"
# ── Step 4: Mount and configure ──────────────────────────────────────────────
MOUNT_DIR=$(hdiutil attach "$TEMP_DMG" -readwrite -noverify | awk -F'\t' '/Apple_HFS/ {gsub(/^[[:space:]]+|[[:space:]]+$/, "", $NF); print $NF}')
echo " Mounted at: $MOUNT_DIR"
# Copy contents
cp -R "$DMG_STAGING/"* "$MOUNT_DIR/"
# Add background image
if [[ -n $BACKGROUND_PNG && -f $BACKGROUND_PNG ]]; then
mkdir -p "$MOUNT_DIR/.background"
cp "$BACKGROUND_PNG" "$MOUNT_DIR/.background/background.png"
fi
# ── Step 5: Configure window appearance via AppleScript ──────────────────────
# Window: 800×400, app icon on left, Applications on right (matches Ollama layout)
# Background image is 1600×740 (2× retina for 800×400 logical window).
APP_NAME="$(basename "$APP_PATH")"
osascript <<APPLESCRIPT
tell application "Finder"
tell disk "$VOLUME_NAME"
open
set current view of container window to icon view
set toolbar visible of container window to false
set statusbar visible of container window to false
set bounds of container window to {200, 120, 1000, 520}
set opts to icon view options of container window
set icon size of opts to 128
set text size of opts to 12
set arrangement of opts to not arranged
if exists file ".background:background.png" then
set background picture of opts to file ".background:background.png"
end if
set position of item "$APP_NAME" of container window to {200, 190}
set position of item "Applications" of container window to {600, 190}
close
open
update without registering applications
delay 1
close
end tell
end tell
APPLESCRIPT
echo " Window layout configured"
# Ensure Finder updates are flushed
sync
# ── Step 6: Finalise ─────────────────────────────────────────────────────────
hdiutil detach "$MOUNT_DIR" -quiet
hdiutil convert "$TEMP_DMG" -format UDZO -imagekey zlib-level=9 -o "$OUTPUT_DMG"
echo "==> DMG created: $OUTPUT_DMG"
echo " Size: $(du -h "$OUTPUT_DMG" | cut -f1)"
+91
View File
@@ -0,0 +1,91 @@
#!/usr/bin/env python3
"""Generate the DMG background image with a centered drag-to-Applications arrow.
The output is a 1600×740 retina PNG (2× for 800×400 logical window).
Icons are positioned at (200, 190) and (600, 190) in logical coordinates;
the arrow is drawn centered between them.
Usage:
python3 generate-background.py [output.png]
If no output path is given, overwrites the bundled background.png in-place.
"""
from __future__ import annotations
import math
import sys
from pathlib import Path
from PIL import Image, ImageDraw
# Retina dimensions (2× logical 800×400)
WIDTH = 1600
HEIGHT = 740
# Icon positions in logical coords → retina coords
# App icon at (200, 190), Applications at (600, 190)
APP_X = 200 * 2 # 400
APPS_X = 600 * 2 # 1200
ICON_Y = 190 * 2 # 380
# Arrow drawn between icons, slightly above icon center
ARROW_START_X = APP_X + 160 # past the icon
ARROW_END_X = APPS_X - 160 # before the Applications icon
ARROW_Y = ICON_Y # same height as icons
ARROW_RISE = 120 # upward arc height
def draw_arrow(draw: ImageDraw.ImageDraw) -> None:
"""Draw a hand-drawn-style curved arrow from app icon toward Applications."""
color = (30, 30, 30)
line_width = 8
# Compute bezier curve points for a gentle upward arc
points: list[tuple[float, float]] = []
steps = 80
for i in range(steps + 1):
t = i / steps
# Quadratic bezier: start → control → end
cx = (ARROW_START_X + ARROW_END_X) / 2
cy = ARROW_Y - ARROW_RISE
x = (1 - t) ** 2 * ARROW_START_X + 2 * (1 - t) * t * cx + t**2 * ARROW_END_X
y = (1 - t) ** 2 * ARROW_Y + 2 * (1 - t) * t * cy + t**2 * ARROW_Y
points.append((x, y))
# Draw the curve as connected line segments
for i in range(len(points) - 1):
draw.line([points[i], points[i + 1]], fill=color, width=line_width)
# Arrowhead at the end
end_x, end_y = points[-1]
# Direction from second-to-last to last point
prev_x, prev_y = points[-3]
angle = math.atan2(end_y - prev_y, end_x - prev_x)
head_len = 36
head_angle = math.radians(25)
left_x = end_x - head_len * math.cos(angle - head_angle)
left_y = end_y - head_len * math.sin(angle - head_angle)
right_x = end_x - head_len * math.cos(angle + head_angle)
right_y = end_y - head_len * math.sin(angle + head_angle)
draw.polygon(
[(end_x, end_y), (left_x, left_y), (right_x, right_y)],
fill=color,
)
def generate_background(output_path: str) -> None:
"""Generate a white DMG background with a centered arrow."""
img = Image.new("RGBA", (WIDTH, HEIGHT), (255, 255, 255, 255))
draw = ImageDraw.Draw(img)
draw_arrow(draw)
img.save(output_path, "PNG")
if __name__ == "__main__":
default_output = str(Path(__file__).parent / "background.png")
out = sys.argv[1] if len(sys.argv) >= 2 else default_output
generate_background(out)
print(f"Background image written to {out}")
+5
View File
@@ -10,6 +10,7 @@ PROJECT_ROOT = Path.cwd()
SOURCE_ROOT = PROJECT_ROOT / "src"
ENTRYPOINT = SOURCE_ROOT / "exo" / "__main__.py"
DASHBOARD_DIR = PROJECT_ROOT / "dashboard" / "build"
RESOURCES_DIR = PROJECT_ROOT / "resources"
EXO_SHARED_MODELS_DIR = SOURCE_ROOT / "exo" / "shared" / "models"
if not ENTRYPOINT.is_file():
@@ -18,6 +19,9 @@ if not ENTRYPOINT.is_file():
if not DASHBOARD_DIR.is_dir():
raise SystemExit(f"Dashboard assets are missing: {DASHBOARD_DIR}")
if not RESOURCES_DIR.is_dir():
raise SystemExit(f"Resource assets are missing: {RESOURCES_DIR}")
if not EXO_SHARED_MODELS_DIR.is_dir():
raise SystemExit(f"Shared model assets are missing: {EXO_SHARED_MODELS_DIR}")
@@ -58,6 +62,7 @@ HIDDEN_IMPORTS = sorted(
DATAS: list[tuple[str, str]] = [
(str(DASHBOARD_DIR), "dashboard"),
(str(RESOURCES_DIR), "resources"),
(str(MLX_LIB_DIR), "mlx/lib"),
(str(EXO_SHARED_MODELS_DIR), "exo/shared/models"),
]
+12 -9
View File
@@ -1,6 +1,6 @@
[project]
name = "exo"
version = "0.3.0"
version = "0.3.68"
description = "Exo"
readme = "README.md"
requires-python = ">=3.13"
@@ -17,24 +17,23 @@ dependencies = [
"loguru>=0.7.3",
"exo_pyo3_bindings", # rust bindings
"anyio==4.11.0",
"mlx==0.30.3; sys_platform == 'darwin'",
"mlx[cpu]==0.30.3; sys_platform == 'linux'",
"mlx-lm @ git+https://github.com/AlexCheema/mlx-lm.git@fix-transformers-5.0.0rc2",
"mlx; sys_platform == 'darwin'",
"mlx[cpu]==0.30.6; sys_platform == 'linux'",
"mlx-lm==0.30.7",
"tiktoken>=0.12.0", # required for kimi k2 tokenizer
"hypercorn>=0.18.0",
"openai-harmony>=0.0.8",
"httpx>=0.28.1",
"tomlkit>=0.14.0",
"pillow>=11.0,<12.0", # compatibility with mflux
"mflux>=0.14.2",
"mflux==0.15.5",
"python-multipart>=0.0.21",
"msgspec>=0.19.0",
"zstandard>=0.23.0",
]
[project.scripts]
exo-master = "exo.master.main:main"
exo-worker = "exo.worker.main:main"
exo = "exo.main:main"
exo-rsh = "exo.rsh.client:main"
# dependencies only required for development
[dependency-groups]
@@ -60,10 +59,13 @@ dev = [
[tool.uv.workspace]
members = [
"rust/exo_pyo3_bindings",
"bench",
]
[tool.uv.sources]
exo_pyo3_bindings = { workspace = true }
mlx = { git = "https://github.com/rltakashige/mlx-jaccl-fix-small-recv.git", branch = "address-rdma-gpu-locks", marker = "sys_platform == 'darwin'" }
#mlx-lm = { git = "https://github.com/davidmcc73/mlx-lm", branch = "stable" }
# Uncomment to use local mlx/mlx-lm development versions:
# mlx = { path = "/Users/Shared/mlx", editable=true }
# mlx-lm = { path = "/Users/Shared/mlx-lm", editable=true }
@@ -105,6 +107,7 @@ root = "src"
# supported platforms for this project
[tool.uv]
required-version = ">=0.8.6"
prerelease = "allow"
environments = [
"sys_platform == 'darwin'",
@@ -130,7 +133,7 @@ markers = [
env = [
"EXO_TESTS=1"
]
addopts = "-m 'not slow'"
addopts = "-m 'not slow' --ignore=tests/start_distributed_test.py"
filterwarnings = [
"ignore:builtin type Swig:DeprecationWarning",
]
+189
View File
@@ -0,0 +1,189 @@
{ inputs, ... }:
{
perSystem =
{ config, self', pkgs, lib, system, ... }:
let
# Load workspace from uv.lock
workspace = inputs.uv2nix.lib.workspace.loadWorkspace {
workspaceRoot = inputs.self;
};
# Create overlay from workspace
# Use wheels from PyPI for most packages; we override mlx with our pure Nix Metal build
overlay = workspace.mkPyprojectOverlay { sourcePreference = "wheel"; };
# Override overlay to inject Nix-built components
exoOverlay = final: prev: {
# Replace workspace exo_pyo3_bindings with Nix-built wheel.
# Preserve passthru so mkVirtualEnv can resolve dependency groups.
# Copy .pyi stub + py.typed marker so basedpyright can find the types.
exo-pyo3-bindings = pkgs.stdenv.mkDerivation {
pname = "exo-pyo3-bindings";
version = "0.1.0";
src = self'.packages.exo_pyo3_bindings;
# Install from pre-built wheel
nativeBuildInputs = [ final.pyprojectWheelHook ];
dontStrip = true;
passthru = prev.exo-pyo3-bindings.passthru or { };
postInstall = ''
local siteDir=$out/${final.python.sitePackages}/exo_pyo3_bindings
cp ${inputs.self}/rust/exo_pyo3_bindings/exo_pyo3_bindings.pyi $siteDir/
touch $siteDir/py.typed
'';
};
};
python = pkgs.python313;
# Overlay to provide build systems and custom packages
buildSystemsOverlay = final: prev: {
# mlx-lm is a git dependency that needs setuptools
mlx-lm = prev.mlx-lm.overrideAttrs (old: {
nativeBuildInputs = (old.nativeBuildInputs or [ ]) ++ [
final.setuptools
];
});
} // lib.optionalAttrs pkgs.stdenv.hostPlatform.isDarwin {
# Use our pure Nix-built MLX with Metal support (macOS only)
mlx = self'.packages.mlx;
};
# Additional overlay for Linux-specific fixes (type checking env).
# Native wheels have shared lib dependencies we don't need at type-check time.
linuxOverlay = final: prev:
let
ignoreMissing = drv: drv.overrideAttrs { autoPatchelfIgnoreMissingDeps = [ "*" ]; };
nvidiaPackages = lib.filterAttrs (name: _: lib.hasPrefix "nvidia-" name) prev;
in
lib.optionalAttrs pkgs.stdenv.hostPlatform.isLinux (
(lib.mapAttrs (_: ignoreMissing) nvidiaPackages) // {
mlx = ignoreMissing prev.mlx;
mlx-cuda-13 = prev.mlx-cuda-13.overrideAttrs (old: {
buildInputs = (old.buildInputs or [ ]) ++ [
final.nvidia-cublas
final.nvidia-cuda-nvrtc
final.nvidia-cudnn-cu13
final.nvidia-nccl-cu13
];
preFixup = ''
addAutoPatchelfSearchPath ${final.nvidia-cublas}
addAutoPatchelfSearchPath ${final.nvidia-cuda-nvrtc}
addAutoPatchelfSearchPath ${final.nvidia-cudnn-cu13}
addAutoPatchelfSearchPath ${final.nvidia-nccl-cu13}
'';
autoPatchelfIgnoreMissingDeps = [ "libcuda.so.1" ];
});
torch = ignoreMissing prev.torch;
triton = ignoreMissing prev.triton;
}
);
pythonSet = (pkgs.callPackage inputs.pyproject-nix.build.packages {
inherit python;
}).overrideScope (
lib.composeManyExtensions [
inputs.pyproject-build-systems.overlays.default
overlay
exoOverlay
buildSystemsOverlay
linuxOverlay
]
);
# mlx-cpu and mlx-cuda-13 both ship mlx/ site-packages files; keep first.
# mlx-cpu/mlx-cuda-13 and nvidia-cudnn-cu12/cu13 ship overlapping files.
venvCollisionPaths = lib.optionals pkgs.stdenv.hostPlatform.isLinux [
"lib/python3.13/site-packages/mlx*"
"lib/python3.13/site-packages/nvidia*"
];
exoVenv = (pythonSet.mkVirtualEnv "exo-env" workspace.deps.default).overrideAttrs {
venvIgnoreCollisions = venvCollisionPaths;
};
# Virtual environment with dev dependencies for testing
testVenv = (pythonSet.mkVirtualEnv "exo-test-env" (
workspace.deps.default // {
exo = [ "dev" ]; # Include pytest, pytest-asyncio, pytest-env
}
)).overrideAttrs {
venvIgnoreCollisions = venvCollisionPaths;
};
mkPythonScript = name: path: pkgs.writeShellApplication {
inherit name;
runtimeInputs = [ exoVenv ];
runtimeEnv = {
EXO_DASHBOARD_DIR = self'.packages.dashboard;
EXO_RESOURCES_DIR = inputs.self + /resources;
};
text = ''exec python ${path} "$@"'';
};
benchVenv = pythonSet.mkVirtualEnv "exo-bench-env" {
exo-bench = [ ];
};
mkBenchScript = name: path: pkgs.writeShellApplication {
inherit name;
runtimeInputs = [ benchVenv ];
text = ''exec python ${path} "$@"'';
};
mkSimplePythonScript = name: path: pkgs.writeShellApplication {
inherit name;
runtimeInputs = [ pkgs.python313 ];
text = ''exec python ${path} "$@"'';
};
exoPackage = pkgs.runCommand "exo"
{
nativeBuildInputs = [ pkgs.makeWrapper ];
}
''
mkdir -p $out/bin
# Create wrapper script
makeWrapper ${exoVenv}/bin/exo $out/bin/exo \
--set EXO_DASHBOARD_DIR ${self'.packages.dashboard} \
--set EXO_RESOURCES_DIR ${inputs.self + /resources} \
${lib.optionalString pkgs.stdenv.hostPlatform.isDarwin "--prefix PATH : ${pkgs.macmon}/bin"}
'';
in
{
# Python package only available on macOS (requires MLX/Metal)
packages = lib.optionalAttrs pkgs.stdenv.hostPlatform.isDarwin
{
exo = exoPackage;
# Test environment for running pytest outside of Nix sandbox (needs GPU access)
exo-test-env = testVenv;
} // {
exo-bench = mkBenchScript "exo-bench" (inputs.self + /bench/exo_bench.py);
exo-eval-tool-calls = mkBenchScript "exo-eval-tool-calls" (inputs.self + /bench/eval_tool_calls.py);
exo-get-all-models-on-cluster = mkSimplePythonScript "exo-get-all-models-on-cluster" (inputs.self + /tests/get_all_models_on_cluster.py);
};
checks = {
# Ruff linting (works on all platforms)
lint = pkgs.runCommand "ruff-lint" { } ''
export RUFF_CACHE_DIR="$TMPDIR/ruff-cache"
${pkgs.ruff}/bin/ruff check ${inputs.self}
touch $out
'';
# Hermetic basedpyright type checking
typecheck = pkgs.runCommand "typecheck"
{
nativeBuildInputs = [
testVenv
pkgs.basedpyright
];
}
''
cd ${inputs.self}
export HOME=$TMPDIR
basedpyright --pythonpath ${testVenv}/bin/python
touch $out
'';
};
};
}
@@ -0,0 +1,49 @@
model_id = "exolabs/FLUX.1-Kontext-dev-4bit"
n_layers = 57
hidden_size = 1
supports_tensor = false
tasks = ["ImageToImage"]
family = "flux"
quantization = "4bit"
base_model = "FLUX.1 Kontext"
capabilities = ["image_edit"]
[storage_size]
in_bytes = 15475325472
[[components]]
component_name = "text_encoder"
component_path = "text_encoder/"
n_layers = 12
can_shard = false
[components.storage_size]
in_bytes = 0
[[components]]
component_name = "text_encoder_2"
component_path = "text_encoder_2/"
n_layers = 24
can_shard = false
safetensors_index_filename = "model.safetensors.index.json"
[components.storage_size]
in_bytes = 9524621312
[[components]]
component_name = "transformer"
component_path = "transformer/"
n_layers = 57
can_shard = true
safetensors_index_filename = "diffusion_pytorch_model.safetensors.index.json"
[components.storage_size]
in_bytes = 5950704160
[[components]]
component_name = "vae"
component_path = "vae/"
can_shard = false
[components.storage_size]
in_bytes = 0

Some files were not shown because too many files have changed in this diff Show More