Compare commits

..

1 Commits

Author SHA1 Message Date
Alex Cheema 928c41b13c feat: add MetaInstance declarative layer with reconciliation
Adds a declarative MetaInstance system for managing model instances with
automatic placement, retry logic (max 3 attempts), and lifecycle
management via a reconciliation loop.

- Process managers for instance health, meta-instance lifecycle, and
  node timeout detection
- Reconciliation engine driving state transitions and cascading deletes
- Dashboard UI for creating/managing MetaInstances with node selection,
  sharding config, retry status, and error feedback
- JACCL SideChannel integration for distributed inference coordination
- Comprehensive test suite (25+ edge cases)

Split from original #1519. Independent bug fixes extracted to:
  #1547 (misc fixes), #1546 (JACCL sidechannel), #1582 (download
  detection), #1580 (RDMA warning)

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
2026-02-21 13:05:10 -08:00
139 changed files with 6105 additions and 10301 deletions
-16
View File
@@ -215,22 +215,6 @@ 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.
+3 -5
View File
@@ -73,11 +73,9 @@ class GenerationResponse:
finish_reason: Optional[str] = ...
def maybe_quantize_kv_cache(
prompt_cache: Any,
quantized_kv_start: int | None,
kv_group_size: int | None,
kv_bits: int | None,
) -> None: ...
prompt_cache, quantized_kv_start, kv_group_size, kv_bits
): # -> None:
...
def generate_step(
prompt: mx.array,
model: nn.Module,
+30 -19
View File
@@ -16,7 +16,7 @@ class Cache(Protocol):
self, keys: mx.array, values: mx.array
) -> tuple[mx.array, mx.array]: ...
@property
def state(self) -> tuple[mx.array | None, mx.array | None]: ...
def state(self) -> tuple[mx.array, mx.array]: ...
@state.setter
def state(self, v) -> None: ...
@@ -92,14 +92,13 @@ class _BaseCache(Cache):
values: mx.array
offset: int
@property
def state(self) -> tuple[mx.array | None, mx.array | None]: ...
def state(self) -> tuple[mx.array, mx.array]: ...
@state.setter
def state(self, v) -> None: ...
@property
def meta_state(self) -> Literal[""]: ...
@meta_state.setter
def meta_state(self, v) -> None: ...
def trim(self, n: int) -> int: ...
def is_trimmable(self) -> Literal[False]: ...
@classmethod
def from_state(cls, state, meta_state) -> Self: ...
@@ -115,13 +114,15 @@ class ConcatenateKVCache(_BaseCache):
def update_and_fetch(self, keys, values): # -> tuple[Any | array, Any | array]:
...
@property
def state(self) -> tuple[mx.array | None, mx.array | None]: ...
def state(self): # -> tuple[Any | array | None, Any | array | None]:
...
@state.setter
def state(self, v): # -> None:
...
def is_trimmable(self): # -> Literal[True]:
...
def trim(self, n: int) -> int: ...
def trim(self, n): # -> int:
...
def make_mask(self, *args, **kwargs): # -> array | Literal['causal'] | None:
...
@@ -131,7 +132,10 @@ class QuantizedKVCache(_BaseCache):
def update_and_fetch(self, keys, values): # -> Any:
...
@property
def state(self) -> tuple[mx.array | None, mx.array | None]: ...
def state(
self,
): # -> tuple[Any | tuple[array, array, array] | None, Any | tuple[array, array, array] | None] | Any:
...
@state.setter
def state(self, v): # -> None:
...
@@ -143,7 +147,8 @@ class QuantizedKVCache(_BaseCache):
...
def is_trimmable(self): # -> Literal[True]:
...
def trim(self, n: int) -> int: ...
def trim(self, n): # -> int:
...
def make_mask(self, *args, **kwargs): # -> array | Literal['causal'] | None:
...
@@ -155,18 +160,18 @@ class KVCache(_BaseCache):
@property
def state(
self,
) -> tuple[mx.array | None, mx.array | None]: ...
) -> tuple[array, array]: ...
@state.setter
def state(self, v) -> None: ...
def is_trimmable(self): # -> Literal[True]:
...
def trim(self, n: int) -> int: ...
def trim(self, n): # -> int:
...
def to_quantized(
self, group_size: int = ..., bits: int = ...
) -> QuantizedKVCache: ...
def make_mask(
self, *args: Any, **kwargs: Any
) -> mx.array | Literal["causal"] | None: ...
def make_mask(self, *args, **kwargs): # -> array | Literal['causal'] | None:
...
class RotatingKVCache(_BaseCache):
step = ...
@@ -178,7 +183,8 @@ class RotatingKVCache(_BaseCache):
@property
def state(
self,
) -> tuple[mx.array | None, mx.array | None]: ...
): # -> tuple[Any | array, Any | array] | tuple[Any | array | None, Any | array | None]:
...
@state.setter
def state(self, v): # -> None:
...
@@ -190,7 +196,8 @@ class RotatingKVCache(_BaseCache):
...
def is_trimmable(self): # -> bool:
...
def trim(self, n: int) -> int: ...
def trim(self, n): # -> int:
...
def to_quantized(
self, group_size: int = ..., bits: int = ...
) -> QuantizedKVCache: ...
@@ -205,7 +212,8 @@ class ArraysCache(_BaseCache):
...
def __getitem__(self, idx): ...
@property
def state(self) -> tuple[mx.array | None, mx.array | None]: ...
def state(self): # -> list[Any | array] | list[array]:
...
@state.setter
def state(self, v): # -> None:
...
@@ -219,7 +227,8 @@ class ArraysCache(_BaseCache):
In-place extend this cache with the other cache.
"""
def make_mask(self, N: int) -> mx.array | None: ...
def make_mask(self, N: int): # -> array | None:
...
class MambaCache(ArraysCache):
def __init__(self, left_padding: Optional[List[int]] = ...) -> None: ...
@@ -230,7 +239,8 @@ class ChunkedKVCache(KVCache):
...
def update_and_fetch(self, keys, values): # -> tuple[array, array]:
...
def trim(self, n: int) -> int: ...
def trim(self, n): # -> int:
...
@property
def meta_state(self): # -> tuple[str, ...]:
...
@@ -243,9 +253,10 @@ class CacheList(_BaseCache):
def __getitem__(self, idx): ...
def is_trimmable(self): # -> bool:
...
def trim(self, n: int) -> int: ...
def trim(self, n): ...
@property
def state(self) -> list[tuple[mx.array | None, mx.array | None]]: ...
def state(self): # -> list[Any]:
...
@state.setter
def state(self, v): # -> None:
...
-153
View File
@@ -1,153 +0,0 @@
from dataclasses import dataclass
from typing import Any, Optional
import mlx.core as mx
import mlx.nn as nn
from .cache import ArraysCache, KVCache
from .qwen3_next import (
Qwen3NextAttention as Attention,
Qwen3NextMLP as MLP,
Qwen3NextRMSNormGated as RMSNormGated,
Qwen3NextSparseMoeBlock,
)
SparseMoeBlock = Qwen3NextSparseMoeBlock
from .switch_layers import SwitchGLU
@dataclass
class TextModelArgs:
model_type: str
hidden_size: int
intermediate_size: int
num_hidden_layers: int
num_attention_heads: int
rms_norm_eps: float
vocab_size: int
num_key_value_heads: int
max_position_embeddings: int
linear_num_value_heads: int
linear_num_key_heads: int
linear_key_head_dim: int
linear_value_head_dim: int
linear_conv_kernel_dim: int
tie_word_embeddings: bool
attention_bias: bool
head_dim: Optional[int]
full_attention_interval: int
num_experts: int
num_experts_per_tok: int
decoder_sparse_step: int
shared_expert_intermediate_size: int
moe_intermediate_size: int
norm_topk_prob: bool
rope_parameters: Optional[dict[str, Any]]
partial_rotary_factor: float
rope_theta: float
rope_scaling: Optional[dict[str, Any]]
@classmethod
def from_dict(cls, params: dict[str, Any]) -> TextModelArgs: ...
def __post_init__(self) -> None: ...
class GatedDeltaNet(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_qkv: nn.Linear
in_proj_z: nn.Linear
in_proj_b: nn.Linear
in_proj_a: nn.Linear
dt_bias: mx.array
A_log: mx.array
norm: RMSNormGated
out_proj: nn.Linear
def __init__(self, config: TextModelArgs) -> None: ...
def __call__(
self,
inputs: mx.array,
mask: Optional[mx.array] = None,
cache: Optional[Any] = None,
) -> mx.array: ...
class DecoderLayer(nn.Module):
is_linear: bool
linear_attn: GatedDeltaNet
self_attn: Attention
input_layernorm: nn.RMSNorm
post_attention_layernorm: nn.RMSNorm
mlp: MLP | SparseMoeBlock
def __init__(self, args: TextModelArgs, layer_idx: int) -> None: ...
def __call__(
self,
x: mx.array,
mask: Optional[mx.array] = None,
cache: Optional[Any] = None,
) -> mx.array: ...
class Qwen3_5TextModel(nn.Module):
embed_tokens: nn.Embedding
layers: list[DecoderLayer]
norm: nn.RMSNorm
ssm_idx: int
fa_idx: int
def __init__(self, args: TextModelArgs) -> None: ...
def __call__(
self,
inputs: mx.array,
cache: Optional[Any] = None,
input_embeddings: Optional[mx.array] = None,
) -> mx.array: ...
class TextModel(nn.Module):
args: TextModelArgs
model_type: str
model: Qwen3_5TextModel
lm_head: nn.Linear
def __init__(self, args: TextModelArgs) -> None: ...
def __call__(
self,
inputs: mx.array,
cache: Optional[Any] = None,
input_embeddings: Optional[mx.array] = None,
) -> mx.array: ...
@property
def layers(self) -> list[DecoderLayer]: ...
def make_cache(self) -> list[ArraysCache | KVCache]: ...
def sanitize(self, weights: dict[str, Any]) -> dict[str, Any]: ...
@dataclass
class ModelArgs:
model_type: str
text_config: dict[str, Any]
@classmethod
def from_dict(cls, params: dict[str, Any]) -> ModelArgs: ...
class Model(nn.Module):
args: ModelArgs
model_type: str
language_model: TextModel
def __init__(self, args: ModelArgs) -> None: ...
def __call__(
self,
inputs: mx.array,
cache: Optional[Any] = None,
input_embeddings: Optional[mx.array] = None,
) -> mx.array: ...
def sanitize(self, weights: dict[str, Any]) -> dict[str, Any]: ...
@property
def layers(self) -> list[DecoderLayer]: ...
def make_cache(self) -> list[ArraysCache | KVCache]: ...
@@ -1,19 +0,0 @@
from dataclasses import dataclass
from typing import Any, Optional
import mlx.core as mx
import mlx.nn as nn
from .cache import ArraysCache, KVCache
from .qwen3_5 import DecoderLayer, Model as Qwen3_5Model, TextModel
@dataclass
class ModelArgs:
model_type: str
text_config: dict[str, Any]
@classmethod
def from_dict(cls, params: dict[str, Any]) -> ModelArgs: ...
class Model(Qwen3_5Model):
def sanitize(self, weights: dict[str, Any]) -> dict[str, Any]: ...
@@ -7,15 +7,6 @@ import mlx.nn as nn
from .switch_layers import SwitchGLU
class Qwen3NextRMSNormGated(nn.Module):
eps: float
weight: mx.array
def __init__(self, hidden_size: int, eps: float = ...) -> None: ...
def __call__(
self, hidden_states: mx.array, gate: mx.array | None = None
) -> mx.array: ...
class Qwen3NextMLP(nn.Module):
gate_proj: nn.Linear
down_proj: nn.Linear
Generated
-24
View File
@@ -216,28 +216,6 @@ dependencies = [
"windows-sys 0.61.2",
]
[[package]]
name = "async-stream"
version = "0.3.6"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "0b5a71a6f37880a80d1d7f19efd781e4b5de42c88f0722cc13bcb6cc2cfe8476"
dependencies = [
"async-stream-impl",
"futures-core",
"pin-project-lite",
]
[[package]]
name = "async-stream-impl"
version = "0.3.6"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "c7c24de15d275a1ecfd47a380fb4d5ec9bfe0933f309ed5e705b775596a3574d"
dependencies = [
"proc-macro2",
"quote",
"syn 2.0.111",
]
[[package]]
name = "async-trait"
version = "0.1.89"
@@ -2781,7 +2759,6 @@ dependencies = [
name = "networking"
version = "0.0.1"
dependencies = [
"async-stream",
"delegate",
"either",
"extend",
@@ -2790,7 +2767,6 @@ dependencies = [
"keccak-const",
"libp2p",
"log",
"pin-project",
"tokio",
"tracing-subscriber",
"util",
+5 -2
View File
@@ -1,6 +1,10 @@
[workspace]
resolver = "3"
members = ["rust/networking", "rust/exo_pyo3_bindings", "rust/util"]
members = [
"rust/networking",
"rust/exo_pyo3_bindings",
"rust/util",
]
[workspace.package]
version = "0.0.1"
@@ -30,7 +34,6 @@ delegate = "0.13"
keccak-const = "0.2"
# Async dependencies
async-stream = "0.3"
tokio = "1.46"
futures-lite = "2.6.1"
futures-timer = "3.0"
+145 -212
View File
@@ -15,23 +15,14 @@ struct ContentView: View {
@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
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 bugReportInFlight = false
@State private var bugReportMessage: String?
@State private var uninstallInProgress = false
@State private var pendingNamespace: String = ""
@State private var pendingHFToken: String = ""
@@ -267,79 +258,139 @@ struct ContentView: View {
VStack(alignment: .leading, spacing: 0) {
if controller.status != .stopped {
dashboardButton
baseURLRow
Divider()
.padding(.vertical, 8)
} else {
Divider()
.padding(.vertical, 4)
}
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) {
advancedSection
.padding(.bottom, 8)
controlButton(title: "Quit", tint: .secondary) {
controller.stop()
NSApplication.shared.terminate(nil)
}
}
}
private var dashboardButton: some View {
HoverButton(
title: "Web Dashboard",
tint: .primary,
trailingSystemImage: "arrow.up.right"
) {
guard let url = URL(string: "http://localhost:52415/") else { return }
NSWorkspace.shared.open(url)
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)
}
}
VStack(alignment: .leading, spacing: 4) {
Text("HuggingFace Token")
.font(.caption2)
.foregroundColor(.secondary)
HStack {
SecureField("optional", text: $pendingHFToken)
.textFieldStyle(.roundedBorder)
.font(.caption2)
.onAppear {
pendingHFToken = controller.hfToken
}
Button("Save & Restart") {
controller.hfToken = pendingHFToken
if controller.status == .running || controller.status == .starting {
controller.restart()
}
}
.font(.caption2)
.disabled(pendingHFToken == controller.hfToken)
}
}
Divider()
HStack {
Toggle(
"Enable Image Models (experimental)", isOn: $pendingEnableImageModels
)
.toggleStyle(.switch)
.font(.caption2)
.onAppear {
pendingEnableImageModels = controller.enableImageModels
}
Spacer()
Button("Save & Restart") {
controller.enableImageModels = pendingEnableImageModels
if controller.status == .running || controller.status == .starting {
controller.restart()
}
}
.font(.caption2)
.disabled(pendingEnableImageModels == controller.enableImageModels)
}
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 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")
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 {
guard let url = URL(string: "http://localhost:52415/") else { return }
NSWorkspace.shared.open(url)
} label: {
HStack {
Image(systemName: "arrow.up.right.square")
.imageScale(.small)
.foregroundColor(baseURLCopied ? .green : .secondary)
.contentTransition(.symbolEffect(.replace))
Text("Dashboard")
.fontWeight(.medium)
Spacer()
}
.buttonStyle(.plain)
.help("Copy API base URL")
.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))
)
}
.padding(.vertical, 4)
.padding(.horizontal, 8)
.buttonStyle(.plain)
.padding(.bottom, 4)
}
private func collapseButton(isExpanded: Binding<Bool>) -> some View {
@@ -560,115 +611,39 @@ struct ContentView: View {
}
private var sendBugReportButton: some View {
VStack(alignment: .leading, spacing: 6) {
switch bugReportPhase {
case .idle:
Button {
bugReportPhase = .prompting
bugReportUserDescription = ""
} label: {
HStack {
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))
)
VStack(alignment: .leading, spacing: 4) {
Button {
Task {
await sendBugReport()
}
.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)
} label: {
HStack {
if bugReportInFlight {
ProgressView()
.scaleEffect(0.6)
}
Text("Send Bug Report")
.font(.caption)
.fontWeight(.semibold)
Spacer()
}
.padding(8)
.padding(.vertical, 6)
.padding(.horizontal, 8)
.background(
RoundedRectangle(cornerRadius: 6)
.fill(Color.accentColor.opacity(0.06))
.fill(Color.accentColor.opacity(0.12))
)
}
.buttonStyle(.plain)
.disabled(bugReportInFlight)
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 = ""
}
if let message = bugReportMessage {
Text(message)
.font(.caption2)
.buttonStyle(.plain)
.foregroundColor(.secondary)
}
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)
}
.fixedSize(horizontal: false, vertical: true)
}
}
.animation(.easeInOut(duration: 0.2), value: bugReportPhase)
}
private var processToggleBinding: Binding<Bool> {
@@ -712,58 +687,16 @@ struct ContentView: View {
}
private func sendBugReport() async {
bugReportPhase = .sending("Collecting logs...")
bugReportInFlight = true
bugReportMessage = "Collecting logs..."
let service = BugReportService()
let description = bugReportUserDescription.trimmingCharacters(in: .whitespacesAndNewlines)
do {
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)
}
let outcome = try await service.sendReport(isManual: true)
bugReportMessage = outcome.message
} catch {
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)
bugReportMessage = error.localizedDescription
}
bugReportInFlight = false
}
private func showUninstallConfirmationAlert() {
+1 -15
View File
@@ -21,9 +21,7 @@ struct EXOApp: App {
@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() {
@@ -45,13 +43,12 @@ struct EXOApp: App {
_updater = StateObject(wrappedValue: updater)
let thunderboltBridge = ThunderboltBridgeService(clusterStateService: service)
_thunderboltBridgeService = StateObject(wrappedValue: thunderboltBridge)
_settingsWindowController = StateObject(wrappedValue: SettingsWindowController())
enableLaunchAtLoginIfNeeded()
// 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)
controller.scheduleLaunch(after: 15)
service.startPolling()
networkStatus.startPolling()
}
@@ -65,19 +62,8 @@ struct EXOApp: App {
.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)
}
-30
View File
@@ -5,8 +5,6 @@ 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 {
@@ -61,17 +59,6 @@ final class ExoProcessController: ObservableObject {
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?
@@ -126,9 +113,6 @@ 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")
@@ -180,17 +164,6 @@ 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)))
@@ -276,9 +249,6 @@ final class ExoProcessController: ObservableObject {
if enableImageModels {
environment["EXO_ENABLE_IMAGE_MODELS"] = "true"
}
if offlineMode {
environment["EXO_OFFLINE"] = "true"
}
var paths: [String] = []
if let existing = environment["PATH"], !existing.isEmpty {
+3 -9
View File
@@ -38,8 +38,7 @@ struct BugReportService {
func sendReport(
baseURL: URL = URL(string: "http://127.0.0.1:52415")!,
now: Date = Date(),
isManual: Bool = false,
userDescription: String? = nil
isManual: Bool = false
) async throws -> BugReportOutcome {
let timestamp = Self.runTimestampString(now)
let dayPrefix = Self.dayPrefixString(now)
@@ -61,8 +60,7 @@ struct BugReportService {
ifconfig: ifconfigText,
debugInfo: debugInfo,
isManual: isManual,
clusterTbBridgeStatus: clusterTbBridgeStatus,
userDescription: userDescription
clusterTbBridgeStatus: clusterTbBridgeStatus
)
let eventLogFiles = readAllEventLogs()
@@ -308,8 +306,7 @@ struct BugReportService {
ifconfig: String,
debugInfo: DebugInfo,
isManual: Bool,
clusterTbBridgeStatus: [[String: Any]]?,
userDescription: String? = nil
clusterTbBridgeStatus: [[String: Any]]?
) -> Data? {
let system = readSystemMetadata()
let exo = readExoMetadata()
@@ -326,9 +323,6 @@ struct BugReportService {
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])
}
-200
View File
@@ -1,200 +0,0 @@
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
@@ -1,489 +0,0 @@
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"
}
}
@@ -1,47 +0,0 @@
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
}
}
+3 -1
View File
@@ -2,4 +2,6 @@
#
# Lists the suite files to include. Each file defines benchmarks
# with shared constraints, topology, and default args.
include = ["single-m3-ultra.toml"]
include = [
"single-m3-ultra.toml",
]
+1 -1
View File
@@ -75,7 +75,7 @@ def load_tokenizer_for_bench(model_id: str) -> Any:
model_path = Path(
snapshot_download(
model_id,
allow_patterns=["*.json", "*.py", "*.tiktoken", "*.model"],
allow_patterns=["*.json", "*.py", "*.tiktoken"],
)
)
+1 -1
View File
@@ -371,7 +371,7 @@ def run_planning_phase(
unwrap_instance(p["DownloadCompleted"]["shardMetadata"])["modelCard"][
"modelId"
],
p["DownloadCompleted"]["total"]["inBytes"],
p["DownloadCompleted"]["totalBytes"]["inBytes"],
)
for p in node_downloads
if "DownloadCompleted" in p
+6 -7
View File
@@ -4,13 +4,12 @@ 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",
"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",
]
[build-system]
+4 -4
View File
@@ -2,10 +2,10 @@
#
# Shared constraints applied to ALL benchmarks in this file.
constraints = [
"All(MacOsBuild(=25D125))",
"Hosts(=1)",
"All(Chip(m3_ultra))",
"All(GpuCores(=80))",
"All(MacOsBuild(=25D125))",
"Hosts(=1)",
"All(Chip(m3_ultra))",
"All(GpuCores(=80))",
]
[topology]
-65
View File
@@ -202,23 +202,6 @@
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;
@@ -337,51 +320,3 @@ 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;
}
}
+174 -43
View File
@@ -7,6 +7,8 @@
editingImage,
clearEditingImage,
selectedChatModel,
setSelectedChatModel,
instances,
ttftMs,
tps,
totalTokens,
@@ -27,19 +29,6 @@
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 {
@@ -50,10 +39,6 @@
showModelSelector = false,
modelTasks = {},
modelCapabilities = {},
onSend,
onAutoSend,
onOpenModelPicker,
modelDisplayOverride,
}: Props = $props();
let message = $state("");
@@ -64,12 +49,27 @@
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"]);
@@ -122,14 +122,73 @@
uploadedFiles.length > 0),
);
// Short label for the currently selected model
const currentModelLabel = $derived(
currentModel
? currentModel.split("/").pop() || currentModel
: modelDisplayOverride
? modelDisplayOverride.split("/").pop() || modelDisplayOverride
: "",
);
// 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 "";
}
async function handleFiles(files: File[]) {
if (files.length === 0) return;
@@ -216,15 +275,6 @@
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);
@@ -256,8 +306,6 @@
);
}
onSend?.();
// Refocus the textarea after sending
setTimeout(() => textareaRef?.focus(), 10);
}
@@ -368,7 +416,7 @@
{/if}
<!-- Model selector (when enabled) -->
{#if showModelSelector}
{#if showModelSelector && availableModels().length > 0}
<div
class="flex items-center justify-between gap-2 px-3 py-2 border-b border-exo-medium-gray/30"
>
@@ -377,22 +425,33 @@
class="text-xs text-exo-light-gray uppercase tracking-wider flex-shrink-0"
>MODEL:</span
>
<!-- Model button — opens the full model picker -->
<!-- Custom dropdown -->
<div class="relative flex-1 max-w-xs">
<button
bind:this={dropdownButtonRef}
type="button"
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"
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'
: ''}"
>
{#if currentModelLabel}
<span class="text-exo-yellow truncate">{currentModelLabel}</span
{#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
>
{: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"
class="absolute right-2 top-1/2 -translate-y-1/2 pointer-events-none transition-transform duration-200 {isModelDropdownOpen
? 'rotate-180'
: ''}"
>
<svg
class="w-3 h-3 text-exo-yellow/60"
@@ -409,6 +468,78 @@
</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()}
@@ -807,8 +807,8 @@
>
AWAITING INPUT
</p>
<p class="text-xs text-white/30 tracking-wider mt-1.5 font-mono">
Type a message below &middot; Shift+Enter for newline
<p class="text-sm sm:text-xs text-exo-light-gray tracking-wider mt-1">
ENTER A QUERY TO BEGIN
</p>
</div>
{/if}
@@ -823,7 +823,6 @@
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"
@@ -1,401 +0,0 @@
<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
@@ -2,6 +2,7 @@
import {
conversations,
activeConversationId,
createConversation,
loadConversation,
deleteConversation,
deleteAllConversations,
@@ -16,15 +17,9 @@
interface Props {
class?: string;
onNewChat?: () => void;
onSelectConversation?: () => void;
}
let {
class: className = "",
onNewChat,
onSelectConversation,
}: Props = $props();
let { class: className = "" }: Props = $props();
const conversationList = $derived(conversations());
const activeId = $derived(activeConversationId());
@@ -47,11 +42,10 @@
);
function handleNewChat() {
onNewChat?.();
createConversation();
}
function handleSelectConversation(id: string) {
onSelectConversation?.();
loadConversation(id);
}
@@ -191,7 +185,11 @@
let instanceType: string | null = null;
if (instanceTag === "MlxRingInstance") instanceType = "MLX Ring";
else if (instanceTag === "MlxJacclInstance") instanceType = "MLX RDMA";
else if (
instanceTag === "MlxIbvInstance" ||
instanceTag === "MlxJacclInstance"
)
instanceType = "MLX RDMA";
let sharding: string | null = null;
const inst = instance as {
@@ -309,7 +307,7 @@
<div class="py-2">
<div class="px-4 py-2">
<span
class="text-xs text-exo-light-gray font-mono tracking-wider uppercase"
class="text-sm text-white/70 font-mono tracking-wider uppercase"
>
{searchQuery ? "SEARCH RESULTS" : "CONVERSATIONS"}
</span>
@@ -378,37 +376,39 @@
onkeydown={(e) =>
e.key === "Enter" &&
handleSelectConversation(conversation.id)}
class="group w-full flex items-center justify-between p-2.5 rounded-lg mb-1 transition-all text-left cursor-pointer
class="group w-full flex items-center justify-between p-2 rounded mb-1 transition-all text-left cursor-pointer
{activeId === conversation.id
? 'bg-exo-yellow/5 border border-exo-yellow/30'
: 'hover:bg-white/[0.03] hover:border-white/10 border border-transparent'}"
? 'bg-transparent border border-exo-yellow/30'
: 'hover:border-exo-yellow/20 border border-transparent'}"
>
<div class="flex-1 min-w-0 pr-2">
<div
class="text-sm font-medium truncate {activeId ===
conversation.id
class="text-sm truncate {activeId === conversation.id
? 'text-exo-yellow'
: 'text-white'}"
: 'text-white/90'}"
>
{conversation.name}
</div>
<div class="text-xs text-white/60 mt-0.5">
<div class="text-sm text-white/50 mt-0.5">
{formatDate(conversation.updatedAt)}
</div>
<div class="text-xs text-exo-light-gray truncate">
<div class="text-sm text-white/70 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/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 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>
{/if}
</div>
@@ -1,20 +0,0 @@
<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}
@@ -1,277 +0,0 @@
<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}
+20 -65
View File
@@ -6,10 +6,6 @@
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) {
@@ -37,17 +33,13 @@
<div class="absolute left-6 top-1/2 -translate-y-1/2">
<button
onclick={handleToggleSidebar}
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"
class="p-2 rounded border border-exo-medium-gray/40 hover:border-exo-yellow/50 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-light-gray'}"
: 'text-exo-medium-gray'}"
fill="none"
viewBox="0 0 24 24"
stroke="currentColor"
@@ -83,19 +75,18 @@
<img
src="/exo-logo.png"
alt="EXO"
class="h-18 drop-shadow-[0_0_4px_rgba(255,215,0,0.3)]"
class="h-18 drop-shadow-[0_0_20px_rgba(255,215,0,0.5)]"
/>
</button>
<!-- Right: Home + Downloads -->
<nav
<div
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-white/70 hover:text-exo-yellow transition-colors tracking-wider uppercase flex items-center gap-2 cursor-pointer"
class="text-sm text-exo-light-gray hover:text-exo-yellow transition-colors tracking-wider uppercase flex items-center gap-2 cursor-pointer"
title="Back to topology view"
>
<svg
@@ -116,59 +107,23 @@
{/if}
<a
href="/#/downloads"
class="text-sm text-white/70 hover:text-exo-yellow transition-colors tracking-wider uppercase flex items-center gap-2 cursor-pointer"
class="text-sm text-exo-light-gray 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}
<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>
Downloads
</a>
</nav>
</div>
</header>
+3 -29
View File
@@ -21,7 +21,7 @@
} | null;
nodes?: Record<string, NodeInfo>;
sharding?: "Pipeline" | "Tensor";
runtime?: "MlxRing" | "MlxJaccl";
runtime?: "MlxRing" | "MlxIbv" | "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 === "MlxJaccl");
const isRdma = $derived(runtime === "MlxIbv" || runtime === "MlxJaccl");
// Get interface name for an IP from node data
function getInterfaceForIp(nodeId: string, ip?: string): string | null {
@@ -567,46 +567,20 @@
<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 === "MlxJaccl"
: runtime === "MlxIbv" || 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()}
@@ -6,7 +6,6 @@
capabilities: string[];
sizeRange: { min: number; max: number } | null;
downloadedOnly: boolean;
readyOnly: boolean;
}
type ModelFilterPopoverProps = {
@@ -191,58 +190,34 @@
</div>
</div>
<!-- Availability filters -->
<!-- Downloaded only -->
<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 })}
<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"
>
<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>
<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>
</div>
<!-- Size range -->
@@ -41,7 +41,6 @@
onShowInfo: (group: ModelGroup) => void;
downloadStatusMap?: Map<string, DownloadAvailability>;
launchedAt?: number;
instanceStatuses?: Record<string, { status: string; statusClass: string }>;
};
let {
@@ -57,7 +56,6 @@
onShowInfo,
downloadStatusMap,
launchedAt,
instanceStatuses = {},
}: ModelPickerGroupProps = $props();
// Group-level download status: show if any variant is downloaded
@@ -94,12 +92,6 @@
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) {
@@ -130,35 +122,16 @@
!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
class="border-b border-white/5 last:border-b-0 {!anyVariantFits
? 'opacity-50'
: ''}"
>
<!-- Main row -->
<div
class="flex items-center gap-2 px-3 py-2.5 transition-colors {anyVariantFits ||
anyVariantHasInstance
class="flex items-center gap-2 px-3 py-2.5 transition-colors {anyVariantFits
? 'hover:bg-white/5 cursor-pointer'
: 'cursor-not-allowed'} {isMainSelected
? 'bg-exo-yellow/10 border-l-2 border-exo-yellow'
@@ -168,7 +141,7 @@
onToggleExpand();
} else {
const modelId = group.variants[0]?.id;
if (modelId && (canModelFit(modelId) || instanceStatuses[modelId])) {
if (modelId && canModelFit(modelId)) {
onSelectModel(modelId);
}
}
@@ -182,7 +155,7 @@
onToggleExpand();
} else {
const modelId = group.variants[0]?.id;
if (modelId && (canModelFit(modelId) || instanceStatuses[modelId])) {
if (modelId && canModelFit(modelId)) {
onSelectModel(modelId);
}
}
@@ -373,47 +346,6 @@
</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
@@ -488,30 +420,20 @@
{#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
<button
type="button"
class="w-full flex items-center gap-3 px-3 py-2 pl-10 hover:bg-white/5 transition-colors text-left {!modelCanFit
? '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"
disabled={!modelCanFit}
onclick={() => {
if (modelCanFit || variantHasInstance) {
if (modelCanFit) {
onSelectModel(variant.id);
}
}}
onkeydown={(e) => {
if (e.key === "Enter" || e.key === " ") {
e.preventDefault();
if (modelCanFit) {
onSelectModel(variant.id);
}
}
}}
>
<!-- Quantization badge -->
<span
@@ -556,48 +478,6 @@
{/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
@@ -610,36 +490,7 @@
/>
</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>
</button>
{/each}
</div>
{/if}
@@ -36,7 +36,6 @@
capabilities: string[];
sizeRange: { min: number; max: number } | null;
downloadedOnly: boolean;
readyOnly: boolean;
}
interface HuggingFaceModel {
@@ -50,11 +49,6 @@
type ModelFitStatus = "fits_now" | "fits_cluster_capacity" | "too_large";
export type InstanceStatus = {
status: string;
statusClass: string;
};
type ModelPickerModalProps = {
isOpen: boolean;
models: ModelInfo[];
@@ -81,7 +75,6 @@
macmon_info?: { memory?: { ram_total?: number } };
}
>;
instanceStatuses?: Record<string, InstanceStatus>;
};
let {
@@ -103,7 +96,6 @@
usedMemoryGB,
downloadsData,
topologyNodes,
instanceStatuses = {},
}: ModelPickerModalProps = $props();
// Local state
@@ -115,7 +107,6 @@
capabilities: [],
sizeRange: null,
downloadedOnly: false,
readyOnly: false,
});
let infoGroup = $state<ModelGroup | null>(null);
@@ -449,16 +440,6 @@
);
}
// Filter to ready/running models only
if (filters.readyOnly) {
result = result.filter((g) =>
g.variants.some((v) => {
const s = instanceStatuses[v.id];
return s && s.statusClass === "ready";
}),
);
}
// Sort: fits-now first, then fits-cluster-capacity, then too-large
result.sort((a, b) => {
const getGroupFitRank = (group: ModelGroup): number => {
@@ -531,18 +512,6 @@
);
});
// Split filtered groups into recommended (fits_now) and others for visual separation
const recommendedGroups = $derived(
filteredGroups.filter((g) =>
g.variants.some((v) => getModelFitStatus(v.id) === "fits_now"),
),
);
const otherGroups = $derived(
filteredGroups.filter(
(g) => !g.variants.some((v) => getModelFitStatus(v.id) === "fits_now"),
),
);
function toggleGroupExpanded(groupId: string) {
const next = new Set(expandedGroups);
if (next.has(groupId)) {
@@ -569,19 +538,13 @@
}
function clearFilters() {
filters = {
capabilities: [],
sizeRange: null,
downloadedOnly: false,
readyOnly: false,
};
filters = { capabilities: [], sizeRange: null, downloadedOnly: false };
}
const hasActiveFilters = $derived(
filters.capabilities.length > 0 ||
filters.sizeRange !== null ||
filters.downloadedOnly ||
filters.readyOnly,
filters.downloadedOnly,
);
</script>
@@ -850,7 +813,6 @@
onShowInfo={(g) => (infoGroup = g)}
downloadStatusMap={getVariantDownloadMap(group)}
launchedAt={recentTimestamps.get(group.variants[0]?.id ?? "")}
{instanceStatuses}
/>
{/each}
{/if}
@@ -878,34 +840,7 @@
{/if}
</div>
{:else}
<!-- Recommended for your cluster -->
{#if recommendedGroups.length > 0 && otherGroups.length > 0 && !searchQuery.trim()}
<div
class="sticky top-0 z-10 flex items-center gap-2 px-3 py-2 bg-green-950/60 border-b border-green-500/20 backdrop-blur-sm"
>
<svg
class="w-3.5 h-3.5 text-green-400 flex-shrink-0"
fill="none"
viewBox="0 0 24 24"
stroke="currentColor"
stroke-width="2"
>
<path
stroke-linecap="round"
stroke-linejoin="round"
d="M9 12.75L11.25 15 15 9.75M21 12a9 9 0 11-18 0 9 9 0 0118 0z"
/>
</svg>
<span
class="text-xs font-mono text-green-400 tracking-wider uppercase"
>Recommended for your cluster</span
>
<span class="text-xs font-mono text-green-400/50"
>— fits in available memory</span
>
</div>
{/if}
{#each recommendedGroups as group}
{#each filteredGroups as group}
<ModelPickerGroup
{group}
isExpanded={expandedGroups.has(group.id)}
@@ -918,34 +853,6 @@
{onToggleFavorite}
onShowInfo={(g) => (infoGroup = g)}
downloadStatusMap={getVariantDownloadMap(group)}
{instanceStatuses}
/>
{/each}
<!-- Other models -->
{#if otherGroups.length > 0 && recommendedGroups.length > 0 && !searchQuery.trim()}
<div
class="sticky top-0 z-10 flex items-center gap-2 px-3 py-2 bg-exo-dark-gray/80 border-y border-exo-medium-gray/20 backdrop-blur-sm"
>
<span
class="text-xs font-mono text-white/40 tracking-wider uppercase"
>Other models</span
>
</div>
{/if}
{#each otherGroups as group}
<ModelPickerGroup
{group}
isExpanded={expandedGroups.has(group.id)}
isFavorite={favorites.has(group.id)}
{selectedModelId}
{canModelFit}
{getModelFitStatus}
onToggleExpand={() => toggleGroupExpanded(group.id)}
onSelectModel={handleSelect}
{onToggleFavorite}
onShowInfo={(g) => (infoGroup = g)}
downloadStatusMap={getVariantDownloadMap(group)}
{instanceStatuses}
/>
{/each}
{/if}
@@ -968,11 +875,6 @@
>Downloaded</span
>
{/if}
{#if filters.readyOnly}
<span class="px-1.5 py-0.5 bg-green-500/20 text-green-400 rounded"
>Ready</span
>
{/if}
{#if filters.sizeRange}
<span class="px-1.5 py-0.5 bg-exo-yellow/20 text-exo-yellow rounded">
{filters.sizeRange.min}GB - {filters.sizeRange.max}GB
@@ -1,117 +0,0 @@
<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>
-1
View File
@@ -12,4 +12,3 @@ 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";
+20 -46
View File
@@ -254,6 +254,20 @@ interface RawStateResponse {
string,
{ total: { inBytes: number }; available: { inBytes: number } }
>;
// MetaInstances (declarative instance constraints)
metaInstances?: Record<string, MetaInstanceData>;
}
export interface MetaInstanceData {
metaInstanceId: string;
modelId: string;
sharding: string;
instanceMeta: string;
minNodes: number;
nodeIds: string[] | null;
placementError: string | null;
consecutiveFailures: number;
lastFailureError: string | null;
}
export interface MessageAttachment {
@@ -553,6 +567,7 @@ class AppStore {
previewNodeFilter = $state<Set<string>>(new Set());
lastUpdate = $state<number | null>(null);
nodeIdentities = $state<Record<string, RawNodeIdentity>>({});
metaInstances = $state<Record<string, MetaInstanceData>>({});
thunderboltBridgeCycles = $state<string[][]>([]);
nodeThunderbolt = $state<
Record<
@@ -589,12 +604,6 @@ class AppStore {
// Image editing state
editingImage = $state<EditingImage | null>(null);
/** True when the backend is reachable. */
isConnected = $state<boolean>(true);
/** Number of consecutive fetch failures. */
private consecutiveFailures = 0;
private static readonly CONNECTION_LOST_THRESHOLD = 3;
private fetchInterval: ReturnType<typeof setInterval> | null = null;
private previewsInterval: ReturnType<typeof setInterval> | null = null;
private lastConversationPersistTs = 0;
@@ -847,10 +856,6 @@ class AppStore {
this.thinkingEnabled = conversation.enableThinking ?? true;
this.refreshConversationModelFromInstances();
// Sync global selection to the loaded conversation's model so reactive
// effects in +page.svelte can determine the correct chat launch state.
this.selectedChatModel = conversation.modelId || "";
return true;
}
@@ -1295,24 +1300,14 @@ class AppStore {
this.nodeThunderbolt = data.nodeThunderbolt ?? {};
// RDMA ctl status per node
this.nodeRdmaCtl = data.nodeRdmaCtl ?? {};
// MetaInstances
this.metaInstances = data.metaInstances ?? {};
// Thunderbolt bridge cycles
this.thunderboltBridgeCycles = data.thunderboltBridgeCycles ?? [];
// Thunderbolt bridge status per node
this.nodeThunderboltBridge = data.nodeThunderboltBridge ?? {};
this.lastUpdate = Date.now();
// Connection recovered
if (!this.isConnected) {
this.isConnected = true;
}
this.consecutiveFailures = 0;
} catch (error) {
this.consecutiveFailures++;
if (
this.consecutiveFailures >= AppStore.CONNECTION_LOST_THRESHOLD &&
this.isConnected
) {
this.isConnected = false;
}
console.error("Error fetching state:", error);
}
}
@@ -1853,7 +1848,7 @@ class AppStore {
assistantMessage.id,
(msg) => {
msg.content =
"No model is loaded yet. Select a model from the sidebar to get started — it will download and load automatically.";
"Error: No model available. Please launch an instance first.";
},
);
this.syncActiveMessagesIfNeeded(targetConversationId);
@@ -2324,7 +2319,7 @@ class AppStore {
const modelToUse = this.getModelForRequest();
if (!modelToUse) {
throw new Error(
"No model is loaded yet. Select a model from the sidebar to get started — it will download and load automatically.",
"No model selected and no running instances available. Please launch an instance first.",
);
}
@@ -3158,23 +3153,6 @@ class AppStore {
return (await response.json()) as TraceStatsResponse;
}
/**
* Delete traces by task IDs
*/
async deleteTraces(
taskIds: string[],
): Promise<{ deleted: string[]; notFound: string[] }> {
const response = await fetch("/v1/traces/delete", {
method: "POST",
headers: { "Content-Type": "application/json" },
body: JSON.stringify({ taskIds }),
});
if (!response.ok) {
throw new Error(`Failed to delete traces: ${response.status}`);
}
return await response.json();
}
/**
* Get the URL for the raw trace file (for Perfetto)
*/
@@ -3196,6 +3174,7 @@ export const totalTokens = () => appStore.totalTokens;
export const prefillProgress = () => appStore.prefillProgress;
export const topologyData = () => appStore.topologyData;
export const instances = () => appStore.instances;
export const metaInstances = () => appStore.metaInstances;
export const runners = () => appStore.runners;
export const downloads = () => appStore.downloads;
export const nodeDisk = () => appStore.nodeDisk;
@@ -3284,9 +3263,6 @@ export const setChatSidebarVisible = (visible: boolean) =>
appStore.setChatSidebarVisible(visible);
export const refreshState = () => appStore.fetchState();
// Connection status
export const isConnected = () => appStore.isConnected;
// Node identities (for OS version mismatch detection)
export const nodeIdentities = () => appStore.nodeIdentities;
@@ -3318,5 +3294,3 @@ export const fetchTraceStats = (taskId: string) =>
appStore.fetchTraceStats(taskId);
export const getTraceRawUrl = (taskId: string) =>
appStore.getTraceRawUrl(taskId);
export const deleteTraces = (taskIds: string[]) =>
appStore.deleteTraces(taskIds);
-87
View File
@@ -1,87 +0,0 @@
/**
* Toast notification store - Global notification system for the EXO dashboard.
*
* Usage:
* import { addToast, dismissToast, toasts } from "$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()}-${Math.random().toString(36).slice(2, 8)}`;
}
export function addToast(input: ToastInput): string {
const id = generateId();
const duration = input.persistent
? 0
: (input.duration ?? DEFAULT_DURATIONS[input.type]);
const toast: Toast = {
id,
type: input.type,
message: input.message,
duration,
createdAt: Date.now(),
};
toastList = [...toastList, toast];
if (duration > 0) {
const timer = setTimeout(() => dismissToast(id), duration);
timers.set(id, timer);
}
return id;
}
export function dismissToast(id: string): void {
const timer = timers.get(id);
if (timer) {
clearTimeout(timer);
timers.delete(id);
}
toastList = toastList.filter((t) => t.id !== id);
}
/** Dismiss all toasts matching a message (useful for dedup). */
export function dismissByMessage(message: string): void {
const matching = toastList.filter((t) => t.message === message);
for (const t of matching) {
dismissToast(t.id);
}
}
export function toasts(): Toast[] {
return toastList;
}
-4
View File
@@ -1,7 +1,5 @@
<script lang="ts">
import "../app.css";
import ToastContainer from "$lib/components/ToastContainer.svelte";
import ConnectionBanner from "$lib/components/ConnectionBanner.svelte";
let { children } = $props();
</script>
@@ -12,7 +10,5 @@
</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
+25 -69
View File
@@ -412,7 +412,7 @@
<div>{col.label}</div>
{#if col.diskAvailable != null}
<div
class="text-[9px] text-white/70 normal-case tracking-normal mt-0.5"
class="text-[9px] text-exo-light-gray/60 normal-case tracking-normal mt-0.5"
>
{formatBytes(col.diskAvailable)} free
</div>
@@ -436,7 +436,7 @@
</div>
{#if row.prettyName}
<div
class="text-[10px] text-white/60"
class="text-[10px] text-exo-light-gray/60"
title={row.modelId}
>
{row.modelId}
@@ -450,7 +450,7 @@
title="View model details"
>
<svg
class="w-4 h-4 text-white/60 hover:text-white/80"
class="w-4 h-4 text-white/30 hover:text-white/60"
viewBox="0 0 24 24"
fill="currentColor"
>
@@ -469,11 +469,11 @@
<td class="px-4 py-3 text-center align-middle">
{#if cell.kind === "completed"}
<div
class="flex flex-col items-center gap-1"
class="flex flex-col items-center gap-0.5"
title="Completed ({formatBytes(cell.totalBytes)})"
>
<svg
class="w-7 h-7 text-green-400"
class="w-5 h-5 text-green-400"
viewBox="0 0 20 20"
fill="currentColor"
>
@@ -483,18 +483,18 @@
clip-rule="evenodd"
></path>
</svg>
<span class="text-xs text-white/70"
<span class="text-[10px] text-exo-light-gray/70"
>{formatBytes(cell.totalBytes)}</span
>
<button
type="button"
class="text-white/50 hover:text-red-400 transition-colors mt-0.5 cursor-pointer"
class="text-exo-light-gray/40 hover:text-red-400 transition-colors mt-0.5"
onclick={() =>
deleteDownload(col.nodeId, row.modelId)}
title="Delete from this node"
>
<svg
class="w-5 h-5"
class="w-3.5 h-3.5"
viewBox="0 0 20 20"
fill="none"
stroke="currentColor"
@@ -517,11 +517,11 @@
cell.speed,
)} - ETA {formatEta(cell.etaMs)}"
>
<span class="text-exo-yellow text-sm font-medium"
<span class="text-exo-yellow text-xs font-medium"
>{clampPercent(cell.percentage).toFixed(1)}%</span
>
<div
class="w-16 h-2 bg-exo-black/60 rounded-sm overflow-hidden"
class="w-14 h-1.5 bg-exo-black/60 rounded-sm overflow-hidden"
>
<div
class="h-full bg-gradient-to-r from-exo-yellow to-exo-yellow/70 transition-all duration-300"
@@ -530,25 +530,25 @@
).toFixed(1)}%"
></div>
</div>
<span class="text-[10px] text-white/70"
<span class="text-[9px] text-exo-light-gray/60"
>{formatSpeed(cell.speed)}</span
>
</div>
{:else if cell.kind === "pending"}
<div
class="flex flex-col items-center gap-1"
class="flex flex-col items-center gap-0.5"
title={cell.downloaded > 0
? `${formatBytes(cell.downloaded)} / ${formatBytes(cell.total)} downloaded (paused)`
? `${formatBytes(cell.downloaded)} / ${formatBytes(cell.total)} downloaded`
: "Download pending"}
>
{#if cell.downloaded > 0 && cell.total > 0}
<span class="text-white/70 text-xs"
<span class="text-exo-light-gray/70 text-[10px]"
>{formatBytes(cell.downloaded)} / {formatBytes(
cell.total,
)}</span
>
<div
class="w-full h-1.5 bg-white/10 rounded-full overflow-hidden"
class="w-full h-1 bg-white/10 rounded-full overflow-hidden"
>
<div
class="h-full bg-exo-light-gray/40 rounded-full"
@@ -558,65 +558,21 @@
).toFixed(1)}%"
></div>
</div>
{#if row.shardMetadata}
<button
type="button"
class="text-white/50 hover:text-exo-yellow transition-colors cursor-pointer"
onclick={() =>
startDownload(col.nodeId, row.shardMetadata!)}
title="Resume download on this node"
>
<svg
class="w-5 h-5"
viewBox="0 0 20 20"
fill="none"
stroke="currentColor"
stroke-width="2"
>
<path
d="M10 3v10m0 0l-3-3m3 3l3-3M3 17h14"
stroke-linecap="round"
stroke-linejoin="round"
></path>
</svg>
</button>
{:else}
<span class="text-white/50 text-[10px]">paused</span
>
{/if}
{:else if row.shardMetadata}
<button
type="button"
class="text-white/50 hover:text-exo-yellow transition-colors cursor-pointer"
onclick={() =>
startDownload(col.nodeId, row.shardMetadata!)}
title="Start download on this node"
<span class="text-exo-light-gray/40 text-[9px]"
>paused</span
>
<svg
class="w-6 h-6"
viewBox="0 0 20 20"
fill="none"
stroke="currentColor"
stroke-width="2"
>
<path
d="M10 3v10m0 0l-3-3m3 3l3-3M3 17h14"
stroke-linecap="round"
stroke-linejoin="round"
></path>
</svg>
</button>
{:else}
<span class="text-white/40 text-sm">...</span>
<span class="text-exo-light-gray/50 text-sm">...</span
>
{/if}
</div>
{:else if cell.kind === "failed"}
<div
class="flex flex-col items-center gap-1"
class="flex flex-col items-center gap-0.5"
title="Download failed"
>
<svg
class="w-7 h-7 text-red-400"
class="w-5 h-5 text-red-400"
viewBox="0 0 20 20"
fill="currentColor"
>
@@ -629,13 +585,13 @@
{#if row.shardMetadata}
<button
type="button"
class="text-white/50 hover:text-exo-yellow transition-colors cursor-pointer"
class="text-exo-light-gray/40 hover:text-exo-yellow transition-colors"
onclick={() =>
startDownload(col.nodeId, row.shardMetadata!)}
title="Retry download on this node"
>
<svg
class="w-5 h-5"
class="w-3.5 h-3.5"
viewBox="0 0 20 20"
fill="none"
stroke="currentColor"
@@ -661,13 +617,13 @@
{#if row.shardMetadata}
<button
type="button"
class="text-white/50 hover:text-exo-yellow transition-colors mt-0.5 opacity-0 group-hover:opacity-100 cursor-pointer"
class="text-exo-light-gray/30 hover:text-exo-yellow transition-colors mt-0.5 opacity-0 group-hover:opacity-100"
onclick={() =>
startDownload(col.nodeId, row.shardMetadata!)}
title="Download to this node"
>
<svg
class="w-5 h-5"
class="w-3.5 h-3.5"
viewBox="0 0 20 20"
fill="none"
stroke="currentColor"
+3 -90
View File
@@ -3,7 +3,6 @@
import {
listTraces,
getTraceRawUrl,
deleteTraces,
type TraceListItem,
} from "$lib/stores/app.svelte";
import HeaderNav from "$lib/components/HeaderNav.svelte";
@@ -11,51 +10,6 @@
let traces = $state<TraceListItem[]>([]);
let loading = $state(true);
let error = $state<string | null>(null);
let selectedIds = $state<Set<string>>(new Set());
let deleting = $state(false);
let allSelected = $derived(
traces.length > 0 && selectedIds.size === traces.length,
);
function toggleSelect(taskId: string) {
const next = new Set(selectedIds);
if (next.has(taskId)) {
next.delete(taskId);
} else {
next.add(taskId);
}
selectedIds = next;
}
function toggleSelectAll() {
if (allSelected) {
selectedIds = new Set();
} else {
selectedIds = new Set(traces.map((t) => t.taskId));
}
}
async function handleDelete() {
if (selectedIds.size === 0) return;
const count = selectedIds.size;
if (
!confirm(
`Delete ${count} trace${count === 1 ? "" : "s"}? This cannot be undone.`,
)
)
return;
deleting = true;
try {
await deleteTraces([...selectedIds]);
selectedIds = new Set();
await refresh();
} catch (e) {
error = e instanceof Error ? e.message : "Failed to delete traces";
} finally {
deleting = false;
}
}
function formatBytes(bytes: number): string {
if (!bytes || bytes <= 0) return "0B";
@@ -155,16 +109,6 @@
</h1>
</div>
<div class="flex items-center gap-3">
{#if selectedIds.size > 0}
<button
type="button"
class="text-xs font-mono text-red-400 hover:text-red-300 transition-colors uppercase border border-red-500/40 px-2 py-1 rounded"
onclick={handleDelete}
disabled={deleting}
>
{deleting ? "Deleting..." : `Delete (${selectedIds.size})`}
</button>
{/if}
<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"
@@ -199,41 +143,14 @@
</div>
{:else}
<div class="space-y-3">
<div class="flex items-center gap-2 px-1">
<button
type="button"
class="text-xs font-mono uppercase transition-colors {allSelected
? 'text-exo-yellow'
: 'text-exo-light-gray hover:text-exo-yellow'}"
onclick={toggleSelectAll}
>
{allSelected ? "Deselect all" : "Select all"}
</button>
</div>
{#each traces as trace}
{@const isSelected = selectedIds.has(trace.taskId)}
<!-- svelte-ignore a11y_no_static_element_interactions -->
<div
role="button"
tabindex="0"
class="w-full text-left rounded border-l-2 border-r border-t border-b transition-all p-4 flex items-center justify-between gap-4 cursor-pointer {isSelected
? 'bg-exo-yellow/10 border-l-exo-yellow border-r-exo-medium-gray/30 border-t-exo-medium-gray/30 border-b-exo-medium-gray/30'
: 'bg-exo-black/30 border-l-transparent border-r-exo-medium-gray/30 border-t-exo-medium-gray/30 border-b-exo-medium-gray/30 hover:bg-white/[0.03]'}"
onclick={() => toggleSelect(trace.taskId)}
onkeydown={(e) => {
if (e.key === "Enter" || e.key === " ") {
e.preventDefault();
toggleSelect(trace.taskId);
}
}}
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 transition-colors truncate block {isSelected
? 'text-exo-yellow'
: 'text-white hover:text-exo-yellow'}"
onclick={(e) => e.stopPropagation()}
class="text-sm font-mono text-white hover:text-exo-yellow transition-colors truncate block"
>
{trace.taskId}
</a>
@@ -243,11 +160,7 @@
)}
</div>
</div>
<!-- svelte-ignore a11y_click_events_have_key_events -->
<div
class="flex items-center gap-2 shrink-0"
onclick={(e) => e.stopPropagation()}
>
<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"
-22
View File
@@ -183,27 +183,6 @@ Same schema as `/v1/chat/completions`.
**Response:**
Chat completion plus benchmarking metrics.
### Cancel Command
**POST** `/v1/cancel/{command_id}`
Cancels an active generation command (text or image). Notifies workers and closes the stream.
**Path parameters:**
* `command_id`: string, ID of the command to cancel
**Response (example):**
```json
{
"message": "Command cancelled.",
"command_id": "cmd-abc-123"
}
```
Returns 404 if the command is not found or already completed.
## 5. Image Generation & Editing
### Image Generation
@@ -287,7 +266,6 @@ GET /v1/models
POST /v1/chat/completions
POST /bench/chat/completions
POST /v1/cancel/{command_id}
POST /v1/images/generations
POST /bench/images/generations
-1
View File
@@ -108,7 +108,6 @@
package = pkgsSwift.swiftPackages.swift-format;
};
shfmt.enable = true;
taplo.enable = true;
};
};
+3 -3
View File
@@ -41,7 +41,7 @@ let
mlx = stdenv.mkDerivation rec {
pname = "mlx";
version = let v = "0.30.7.dev20260225+257d5692"; in
version = let v = "0.30.7.dev20260220+13998a05"; 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;
@@ -49,8 +49,8 @@ let
src = fetchFromGitHub {
owner = "rltakashige";
repo = "mlx-jaccl-fix-small-recv";
rev = "257d5692fc7af6bba3b8afaeb63c549b7d1e43d5";
hash = "sha256-GosFIWxIB48Egb1MqJrR3xhsUsQeWdRk5rV93USY6wQ=";
rev = "13998a054715edcdc93618fb1496c79c7c25ff7c";
hash = "sha256-fAqA3hFwNBx7FcoGnhQsIFpAIRbC2EerACm4Fvne0Cc=";
};
patches = [
Binary file not shown.

Before

Width:  |  Height:  |  Size: 6.8 KiB

-112
View File
@@ -1,112 +0,0 @@
#!/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
@@ -1,91 +0,0 @@
#!/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}")
+52 -52
View File
@@ -1,35 +1,35 @@
[project]
name = "exo"
version = "0.3.68"
version = "0.3.0"
description = "Exo"
readme = "README.md"
requires-python = ">=3.13"
dependencies = [
"aiofiles>=24.1.0",
"aiohttp>=3.12.14",
"types-aiofiles>=24.1.0.20250708",
"pydantic>=2.11.7",
"fastapi>=0.116.1",
"filelock>=3.18.0",
"rustworkx>=0.17.1",
"huggingface-hub>=0.33.4",
"psutil>=7.0.0",
"loguru>=0.7.3",
"exo_pyo3_bindings", # rust bindings
"anyio==4.11.0",
"mlx; sys_platform == 'darwin'",
"mlx[cpu]==0.30.6; sys_platform == 'linux'",
"mlx-lm",
"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==0.15.5",
"python-multipart>=0.0.21",
"msgspec>=0.19.0",
"zstandard>=0.23.0",
"aiofiles>=24.1.0",
"aiohttp>=3.12.14",
"types-aiofiles>=24.1.0.20250708",
"pydantic>=2.11.7",
"fastapi>=0.116.1",
"filelock>=3.18.0",
"rustworkx>=0.17.1",
"huggingface-hub>=0.33.4",
"psutil>=7.0.0",
"loguru>=0.7.3",
"exo_pyo3_bindings", # rust bindings
"anyio==4.11.0",
"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==0.15.5",
"python-multipart>=0.0.21",
"msgspec>=0.19.0",
"zstandard>=0.23.0",
]
[project.scripts]
@@ -38,12 +38,12 @@ exo = "exo.main:main"
# dependencies only required for development
[dependency-groups]
dev = [
"basedpyright>=1.29.0",
"pyinstaller>=6.17.0",
"pytest>=8.4.0",
"pytest-asyncio>=1.0.0",
"pytest-env",
"ruff>=0.11.13",
"basedpyright>=1.29.0",
"pyinstaller>=6.17.0",
"pytest>=8.4.0",
"pytest-asyncio>=1.0.0",
"pytest-env",
"ruff>=0.11.13",
]
# mlx[cuda] requires a newer version of mlx. the ideal on linux is: default to mlx[cpu] unless[cuda] specified.
@@ -57,12 +57,15 @@ dev = [
###
[tool.uv.workspace]
members = ["rust/exo_pyo3_bindings", "bench"]
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/ml-explore/mlx-lm", rev = "834fac934c4e04de9b3d723e2b9287a2c60cfd4a" }
#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 }
@@ -92,15 +95,7 @@ reportUnnecessaryTypeIgnoreComment = "error"
pythonVersion = "3.13"
pythonPlatform = "Darwin"
exclude = [
"**/.venv",
"**/venv",
"**/__pycache__",
"**/exo_scripts",
"**/.direnv",
"**/rust",
"**/.github",
]
exclude = ["**/.venv", "**/venv", "**/__pycache__", "**/exo_scripts", "**/.direnv", "**/rust", "**/.github"]
stubPath = ".mlx_typings"
[[tool.basedpyright.executionEnvironments]]
@@ -114,18 +109,17 @@ root = "src"
[tool.uv]
required-version = ">=0.8.6"
prerelease = "allow"
environments = ["sys_platform == 'darwin'", "sys_platform == 'linux'"]
environments = [
"sys_platform == 'darwin'",
"sys_platform == 'linux'",
]
###
# ruff configuration
###
[tool.ruff]
extend-exclude = [
"shared/protobufs/**",
"*mlx_typings/**",
"rust/exo_pyo3_bindings/**",
]
extend-exclude = ["shared/protobufs/**", "*mlx_typings/**", "rust/exo_pyo3_bindings/**"]
[tool.ruff.lint]
extend-select = ["I", "N", "B", "A", "PIE", "SIM"]
@@ -133,7 +127,13 @@ extend-select = ["I", "N", "B", "A", "PIE", "SIM"]
[tool.pytest.ini_options]
pythonpath = "."
asyncio_mode = "auto"
markers = ["slow: marks tests as slow (deselected by default)"]
env = ["EXO_TESTS=1"]
markers = [
"slow: marks tests as slow (deselected by default)"
]
env = [
"EXO_TESTS=1"
]
addopts = "-m 'not slow' --ignore=tests/start_distributed_test.py"
filterwarnings = ["ignore:builtin type Swig:DeprecationWarning"]
filterwarnings = [
"ignore:builtin type Swig:DeprecationWarning",
]
@@ -1,12 +0,0 @@
model_id = "mlx-community/Qwen3.5-122B-A10B-4bit"
n_layers = 48
hidden_size = 3072
supports_tensor = true
tasks = ["TextGeneration"]
family = "qwen"
quantization = "4bit"
base_model = "Qwen3.5 122B A10B"
capabilities = ["text", "thinking", "thinking_toggle"]
[storage_size]
in_bytes = 69593314272
@@ -1,12 +0,0 @@
model_id = "mlx-community/Qwen3.5-122B-A10B-6bit"
n_layers = 48
hidden_size = 3072
supports_tensor = true
tasks = ["TextGeneration"]
family = "qwen"
quantization = "6bit"
base_model = "Qwen3.5 122B A10B"
capabilities = ["text", "thinking", "thinking_toggle"]
[storage_size]
in_bytes = 100120675296
@@ -1,12 +0,0 @@
model_id = "mlx-community/Qwen3.5-122B-A10B-8bit"
n_layers = 48
hidden_size = 3072
supports_tensor = true
tasks = ["TextGeneration"]
family = "qwen"
quantization = "8bit"
base_model = "Qwen3.5 122B A10B"
capabilities = ["text", "thinking", "thinking_toggle"]
[storage_size]
in_bytes = 130648036320
@@ -1,12 +0,0 @@
model_id = "mlx-community/Qwen3.5-122B-A10B-bf16"
n_layers = 48
hidden_size = 3072
supports_tensor = true
tasks = ["TextGeneration"]
family = "qwen"
quantization = "bf16"
base_model = "Qwen3.5 122B A10B"
capabilities = ["text", "thinking", "thinking_toggle"]
[storage_size]
in_bytes = 245125640160
@@ -1,12 +0,0 @@
model_id = "mlx-community/Qwen3.5-27B-4bit"
n_layers = 64
hidden_size = 5120
supports_tensor = true
tasks = ["TextGeneration"]
family = "qwen"
quantization = "4bit"
base_model = "Qwen3.5 27B"
capabilities = ["text", "thinking"]
[storage_size]
in_bytes = 16054266848
@@ -1,12 +0,0 @@
model_id = "mlx-community/Qwen3.5-27B-8bit"
n_layers = 64
hidden_size = 5120
supports_tensor = true
tasks = ["TextGeneration"]
family = "qwen"
quantization = "8bit"
base_model = "Qwen3.5 27B"
capabilities = ["text", "thinking"]
[storage_size]
in_bytes = 29500943328
@@ -1,12 +0,0 @@
model_id = "mlx-community/Qwen3.5-2B-MLX-8bit"
n_layers = 24
hidden_size = 2048
supports_tensor = true
tasks = ["TextGeneration"]
family = "qwen"
quantization = "8bit"
base_model = "Qwen3.5 2B"
capabilities = ["text", "thinking"]
[storage_size]
in_bytes = 2662787264
@@ -1,12 +0,0 @@
model_id = "mlx-community/Qwen3.5-35B-A3B-4bit"
n_layers = 40
hidden_size = 2048
supports_tensor = true
tasks = ["TextGeneration"]
family = "qwen"
quantization = "4bit"
base_model = "Qwen3.5 35B A3B"
capabilities = ["text", "thinking", "thinking_toggle"]
[storage_size]
in_bytes = 20391405152
@@ -1,12 +0,0 @@
model_id = "mlx-community/Qwen3.5-35B-A3B-8bit"
n_layers = 40
hidden_size = 2048
supports_tensor = true
tasks = ["TextGeneration"]
family = "qwen"
quantization = "8bit"
base_model = "Qwen3.5 35B A3B"
capabilities = ["text", "thinking", "thinking_toggle"]
[storage_size]
in_bytes = 37721130592
@@ -1,12 +0,0 @@
model_id = "mlx-community/Qwen3.5-397B-A17B-4bit"
n_layers = 60
hidden_size = 4096
supports_tensor = true
tasks = ["TextGeneration"]
family = "qwen"
quantization = "4bit"
base_model = "Qwen3.5 397B A17B"
capabilities = ["text", "thinking", "thinking_toggle"]
[storage_size]
in_bytes = 223860768352
@@ -1,12 +0,0 @@
model_id = "mlx-community/Qwen3.5-397B-A17B-6bit"
n_layers = 60
hidden_size = 4096
supports_tensor = true
tasks = ["TextGeneration"]
family = "qwen"
quantization = "6bit"
base_model = "Qwen3.5 397B A17B"
capabilities = ["text", "thinking", "thinking_toggle"]
[storage_size]
in_bytes = 322946674272
@@ -1,12 +0,0 @@
model_id = "mlx-community/Qwen3.5-397B-A17B-8bit"
n_layers = 60
hidden_size = 4096
supports_tensor = true
tasks = ["TextGeneration"]
family = "qwen"
quantization = "8bit"
base_model = "Qwen3.5 397B A17B"
capabilities = ["text", "thinking", "thinking_toggle"]
[storage_size]
in_bytes = 422032580192
@@ -1,12 +0,0 @@
model_id = "mlx-community/Qwen3.5-9B-4bit"
n_layers = 32
hidden_size = 4096
supports_tensor = true
tasks = ["TextGeneration"]
family = "qwen"
quantization = "4bit"
base_model = "Qwen3.5 9B"
capabilities = ["text", "thinking"]
[storage_size]
in_bytes = 5950062560
@@ -1,12 +0,0 @@
model_id = "mlx-community/Qwen3.5-9B-8bit"
n_layers = 32
hidden_size = 4096
supports_tensor = true
tasks = ["TextGeneration"]
family = "qwen"
quantization = "8bit"
base_model = "Qwen3.5 9B"
capabilities = ["text", "thinking"]
[storage_size]
in_bytes = 10426433504
+11 -15
View File
@@ -26,24 +26,20 @@ networking = { workspace = true }
# interop
pyo3 = { version = "0.27.2", features = [
# "abi3-py313", # tells pyo3 (and maturin) to build using the stable ABI with minimum Python version 3.13
# "nightly", # enables better-supported GIL integration
"experimental-async", # async support in #[pyfunction] & #[pymethods]
#"experimental-inspect", # inspection of generated binary => easier to automate type-hint generation
#"py-clone", # adding Clone-ing of `Py<T>` without GIL (may cause panics - remove if panics happen)
# "multiple-pymethods", # allows multiple #[pymethods] sections per class
# "abi3-py313", # tells pyo3 (and maturin) to build using the stable ABI with minimum Python version 3.13
# "nightly", # enables better-supported GIL integration
"experimental-async", # async support in #[pyfunction] & #[pymethods]
#"experimental-inspect", # inspection of generated binary => easier to automate type-hint generation
#"py-clone", # adding Clone-ing of `Py<T>` without GIL (may cause panics - remove if panics happen)
# "multiple-pymethods", # allows multiple #[pymethods] sections per class
# integrations with other libraries
# "arc_lock", "bigdecimal", "either", "hashbrown", "indexmap", "num-bigint", "num-complex", "num-rational",
# "ordered-float", "rust_decimal", "smallvec",
# "anyhow", "chrono", "chrono-local", "chrono-tz", "eyre", "jiff-02", "lock_api", "parking-lot", "time", "serde",
# integrations with other libraries
# "arc_lock", "bigdecimal", "either", "hashbrown", "indexmap", "num-bigint", "num-complex", "num-rational",
# "ordered-float", "rust_decimal", "smallvec",
# "anyhow", "chrono", "chrono-local", "chrono-tz", "eyre", "jiff-02", "lock_api", "parking-lot", "time", "serde",
] }
pyo3-stub-gen = { version = "0.17.2" }
pyo3-async-runtimes = { version = "0.27.0", features = [
"attributes",
"tokio-runtime",
"testing",
] }
pyo3-async-runtimes = { version = "0.27.0", features = ["attributes", "tokio-runtime", "testing"] }
pyo3-log = "0.13.2"
# macro dependencies
+55 -29
View File
@@ -2,6 +2,7 @@
# ruff: noqa: E501, F401
import builtins
import enum
import typing
@typing.final
@@ -10,6 +11,29 @@ class AllQueuesFullError(builtins.Exception):
def __repr__(self) -> builtins.str: ...
def __str__(self) -> builtins.str: ...
@typing.final
class ConnectionUpdate:
@property
def update_type(self) -> ConnectionUpdateType:
r"""
Whether this is a connection or disconnection event
"""
@property
def peer_id(self) -> builtins.str:
r"""
Identity of the peer that we have connected to or disconnected from.
"""
@property
def remote_ipv4(self) -> builtins.str:
r"""
Remote connection's IPv4 address.
"""
@property
def remote_tcp_port(self) -> builtins.int:
r"""
Remote connection's TCP port.
"""
@typing.final
class Keypair:
r"""
@@ -34,15 +58,21 @@ class Keypair:
Convert the `Keypair` into the corresponding `PeerId` string, which we use as our `NodeId`.
"""
@typing.final
class MessageTooLargeError(builtins.Exception):
def __new__(cls, *args: typing.Any) -> MessageTooLargeError: ...
def __repr__(self) -> builtins.str: ...
def __str__(self) -> builtins.str: ...
@typing.final
class NetworkingHandle:
def __new__(cls, identity: Keypair) -> NetworkingHandle: ...
async def connection_update_recv(self) -> ConnectionUpdate:
r"""
Receives the next `ConnectionUpdate` from networking.
"""
async def connection_update_recv_many(self, limit: builtins.int) -> builtins.list[ConnectionUpdate]:
r"""
Receives at most `limit` `ConnectionUpdate`s from networking and returns them.
For `limit = 0`, an empty collection of `ConnectionUpdate`s will be returned immediately.
For `limit > 0`, if there are no `ConnectionUpdate`s in the channel's queue this method
will sleep until a `ConnectionUpdate`s is sent.
"""
async def gossipsub_subscribe(self, topic: builtins.str) -> builtins.bool:
r"""
Subscribe to a `GossipSub` topic.
@@ -61,7 +91,18 @@ class NetworkingHandle:
If no peers are found that subscribe to this topic, throws `NoPeersSubscribedToTopicError` exception.
"""
async def recv(self) -> PyFromSwarm: ...
async def gossipsub_recv(self) -> tuple[builtins.str, bytes]:
r"""
Receives the next message from the `GossipSub` network.
"""
async def gossipsub_recv_many(self, limit: builtins.int) -> builtins.list[tuple[builtins.str, bytes]]:
r"""
Receives at most `limit` messages from the `GossipSub` network and returns them.
For `limit = 0`, an empty collection of messages will be returned immediately.
For `limit > 0`, if there are no messages in the channel's queue this method
will sleep until a message is sent.
"""
@typing.final
class NoPeersSubscribedToTopicError(builtins.Exception):
@@ -69,26 +110,11 @@ class NoPeersSubscribedToTopicError(builtins.Exception):
def __repr__(self) -> builtins.str: ...
def __str__(self) -> builtins.str: ...
class PyFromSwarm:
@typing.final
class Connection(PyFromSwarm):
__match_args__ = ("peer_id", "connected",)
@property
def peer_id(self) -> builtins.str: ...
@property
def connected(self) -> builtins.bool: ...
def __new__(cls, peer_id: builtins.str, connected: builtins.bool) -> PyFromSwarm.Connection: ...
@typing.final
class Message(PyFromSwarm):
__match_args__ = ("origin", "topic", "data",)
@property
def origin(self) -> builtins.str: ...
@property
def topic(self) -> builtins.str: ...
@property
def data(self) -> bytes: ...
def __new__(cls, origin: builtins.str, topic: builtins.str, data: bytes) -> PyFromSwarm.Message: ...
...
@typing.final
class ConnectionUpdateType(enum.Enum):
r"""
Connection or disconnection event discriminant type.
"""
Connected = ...
Disconnected = ...
+7 -4
View File
@@ -4,18 +4,21 @@ build-backend = "maturin"
[project]
name = "exo_pyo3_bindings"
version = "0.2.0"
version = "0.1.0"
description = "Add your description here"
readme = "README.md"
authors = [
{ name = "Andrei Cravtov", email = "the.andrei.cravtov@gmail.com" },
{ name = "Evan Quiney", email = "evanev7@gmail.com" },
{ name = "Andrei Cravtov", email = "the.andrei.cravtov@gmail.com" }
]
requires-python = ">=3.13"
dependencies = []
[dependency-groups]
dev = ["exo_pyo3_bindings", "pytest>=8.4.0", "pytest-asyncio>=1.0.0"]
dev = [
"exo_pyo3_bindings",
"pytest>=8.4.0",
"pytest-asyncio>=1.0.0",
]
[tool.maturin]
#purelib = true
-3
View File
@@ -155,9 +155,6 @@ pub(crate) mod ext {
fn main_module(m: &Bound<'_, PyModule>) -> PyResult<()> {
// install logger
pyo3_log::init();
let mut builder = tokio::runtime::Builder::new_multi_thread();
builder.enable_all();
pyo3_async_runtimes::tokio::init(builder);
// TODO: for now this is all NOT a submodule, but figure out how to make the submodule system
// work with maturin, where the types generate correctly, in the right folder, without
+375 -115
View File
@@ -1,24 +1,26 @@
use std::pin::Pin;
use std::sync::Arc;
#![allow(
clippy::multiple_inherent_impl,
clippy::unnecessary_wraps,
clippy::unused_self,
clippy::needless_pass_by_value
)]
use crate::r#const::MPSC_CHANNEL_SIZE;
use crate::ext::{ByteArrayExt as _, FutureExt, PyErrExt as _};
use crate::ext::{ResultExt as _, TokioMpscSenderExt as _};
use crate::ext::{ResultExt as _, TokioMpscReceiverExt as _, TokioMpscSenderExt as _};
use crate::ident::PyKeypair;
use crate::networking::exception::{
PyAllQueuesFullError, PyMessageTooLargeError, PyNoPeersSubscribedToTopicError,
};
use crate::pyclass;
use futures_lite::{Stream, StreamExt as _};
use libp2p::gossipsub::PublishError;
use networking::swarm::{FromSwarm, ToSwarm, create_swarm};
use pyo3::exceptions::PyRuntimeError;
use libp2p::futures::StreamExt as _;
use libp2p::gossipsub;
use libp2p::gossipsub::{IdentTopic, Message, MessageId, PublishError};
use libp2p::swarm::SwarmEvent;
use networking::discovery;
use networking::swarm::create_swarm;
use pyo3::prelude::{PyModule, PyModuleMethods as _};
use pyo3::types::PyBytes;
use pyo3::{Bound, Py, PyAny, PyErr, PyResult, Python, pymethods};
use pyo3_stub_gen::derive::{
gen_methods_from_python, gen_stub_pyclass, gen_stub_pyclass_complex_enum, gen_stub_pymethods,
};
use pyo3::{Bound, Py, PyErr, PyResult, PyTraverseError, PyVisit, Python, pymethods};
use pyo3_stub_gen::derive::{gen_stub_pyclass, gen_stub_pyclass_enum, gen_stub_pymethods};
use std::net::IpAddr;
use tokio::sync::{Mutex, mpsc, oneshot};
mod exception {
@@ -96,78 +98,237 @@ mod exception {
Self::MSG.to_string()
}
}
}
#[gen_stub_pyclass]
#[pyclass(frozen, extends=PyException, name="MessageTooLargeError")]
pub struct PyMessageTooLargeError {}
/// Connection or disconnection event discriminant type.
#[gen_stub_pyclass_enum]
#[pyclass(eq, eq_int, name = "ConnectionUpdateType")]
#[derive(Debug, Clone, PartialEq)]
enum PyConnectionUpdateType {
Connected = 0,
Disconnected,
}
impl PyMessageTooLargeError {
const MSG: &'static str = "Gossipsub message exceeds max_transmit_size. Reduce prompt length or increase the limit.";
#[gen_stub_pyclass]
#[pyclass(frozen, name = "ConnectionUpdate")]
#[derive(Debug, Clone)]
struct PyConnectionUpdate {
/// Whether this is a connection or disconnection event
#[pyo3(get)]
update_type: PyConnectionUpdateType,
pub(crate) fn new_err() -> PyErr {
PyErr::new::<Self, _>(())
/// Identity of the peer that we have connected to or disconnected from.
#[pyo3(get)]
peer_id: String,
/// Remote connection's IPv4 address.
#[pyo3(get)]
remote_ipv4: String,
/// Remote connection's TCP port.
#[pyo3(get)]
remote_tcp_port: u16,
}
enum ToTask {
GossipsubSubscribe {
topic: String,
result_tx: oneshot::Sender<PyResult<bool>>,
},
GossipsubUnsubscribe {
topic: String,
result_tx: oneshot::Sender<bool>,
},
GossipsubPublish {
topic: String,
data: Vec<u8>,
result_tx: oneshot::Sender<PyResult<MessageId>>,
},
}
#[allow(clippy::enum_glob_use)]
async fn networking_task(
mut swarm: networking::swarm::Swarm,
mut to_task_rx: mpsc::Receiver<ToTask>,
connection_update_tx: mpsc::Sender<PyConnectionUpdate>,
gossipsub_message_tx: mpsc::Sender<(String, Vec<u8>)>,
) {
use SwarmEvent::*;
use ToTask::*;
use networking::swarm::BehaviourEvent::*;
log::info!("RUST: networking task started");
loop {
tokio::select! {
message = to_task_rx.recv() => {
// handle closed channel
let Some(message) = message else {
log::info!("RUST: channel closed");
break;
};
// dispatch incoming messages
match message {
GossipsubSubscribe { topic, result_tx } => {
// try to subscribe
let result = swarm.behaviour_mut()
.gossipsub.subscribe(&IdentTopic::new(topic));
// send response oneshot
if let Err(e) = result_tx.send(result.pyerr()) {
log::error!("RUST: could not subscribe to gossipsub topic since channel already closed: {e:?}");
continue;
}
}
GossipsubUnsubscribe { topic, result_tx } => {
// try to unsubscribe from the topic
let result = swarm.behaviour_mut()
.gossipsub.unsubscribe(&IdentTopic::new(topic));
// send response oneshot (or exit if connection closed)
if let Err(e) = result_tx.send(result) {
log::error!("RUST: could not unsubscribe from gossipsub topic since channel already closed: {e:?}");
continue;
}
}
GossipsubPublish { topic, data, result_tx } => {
// try to publish the data -> catch NoPeersSubscribedToTopic error & convert to correct exception
let result = swarm.behaviour_mut().gossipsub.publish(
IdentTopic::new(topic), data);
let pyresult: PyResult<MessageId> = if let Err(PublishError::NoPeersSubscribedToTopic) = result {
Err(exception::PyNoPeersSubscribedToTopicError::new_err())
} else if let Err(PublishError::AllQueuesFull(_)) = result {
Err(exception::PyAllQueuesFullError::new_err())
} else {
result.pyerr()
};
// send response oneshot (or exit if connection closed)
if let Err(e) = result_tx.send(pyresult) {
log::error!("RUST: could not publish gossipsub message since channel already closed: {e:?}");
continue;
}
}
}
}
// architectural solution to this problem:
// create keep_alive behavior who's job it is to dial peers discovered by mDNS (and drop when expired)
// -> it will emmit TRUE connected/disconnected events consumable elsewhere
//
// gossipsub will feed off-of dial attempts created by networking, and that will bootstrap its' peers list
// then for actual communication it will dial those peers if need-be
swarm_event = swarm.select_next_some() => {
match swarm_event {
Behaviour(Gossipsub(gossipsub::Event::Message {
message: Message {
topic,
data,
..
},
..
})) => {
// topic-ID is just the topic hash!!! (since we used identity hasher)
let message = (topic.into_string(), data);
// send incoming message to channel (or exit if connection closed)
if let Err(e) = gossipsub_message_tx.send(message).await {
log::error!("RUST: could not send incoming gossipsub message since channel already closed: {e}");
continue;
}
},
Behaviour(Discovery(discovery::Event::ConnectionEstablished { peer_id, remote_ip, remote_tcp_port, .. })) => {
// grab IPv4 string
let remote_ipv4 = match remote_ip {
IpAddr::V4(ip) => ip.to_string(),
IpAddr::V6(ip) => {
log::warn!("RUST: ignoring connection to IPv6 address: {ip}");
continue;
}
};
// send connection event to channel (or exit if connection closed)
if let Err(e) = connection_update_tx.send(PyConnectionUpdate {
update_type: PyConnectionUpdateType::Connected,
peer_id: peer_id.to_base58(),
remote_ipv4,
remote_tcp_port,
}).await {
log::error!("RUST: could not send connection update since channel already closed: {e}");
continue;
}
},
Behaviour(Discovery(discovery::Event::ConnectionClosed { peer_id, remote_ip, remote_tcp_port, .. })) => {
// grab IPv4 string
let remote_ipv4 = match remote_ip {
IpAddr::V4(ip) => ip.to_string(),
IpAddr::V6(ip) => {
log::warn!("RUST: ignoring disconnection from IPv6 address: {ip}");
continue;
}
};
// send disconnection event to channel (or exit if connection closed)
if let Err(e) = connection_update_tx.send(PyConnectionUpdate {
update_type: PyConnectionUpdateType::Disconnected,
peer_id: peer_id.to_base58(),
remote_ipv4,
remote_tcp_port,
}).await {
log::error!("RUST: could not send connection update since channel already closed: {e}");
continue;
}
},
e => {
log::info!("RUST: other event {e:?}");
}
}
}
}
}
#[gen_stub_pymethods]
#[pymethods]
impl PyMessageTooLargeError {
#[new]
#[pyo3(signature = (*args))]
#[allow(unused_variables)]
pub(crate) fn new(args: &Bound<'_, PyTuple>) -> Self {
Self {}
}
fn __repr__(&self) -> String {
format!("MessageTooLargeError(\"{}\")", Self::MSG)
}
fn __str__(&self) -> String {
Self::MSG.to_string()
}
}
log::info!("RUST: networking task stopped");
}
#[gen_stub_pyclass]
#[pyclass(name = "NetworkingHandle")]
#[derive(Debug)]
struct PyNetworkingHandle {
// channels
pub to_swarm: mpsc::Sender<ToSwarm>,
pub swarm: Arc<Mutex<Pin<Box<dyn Stream<Item = FromSwarm> + Send>>>>,
to_task_tx: Option<mpsc::Sender<ToTask>>,
connection_update_rx: Mutex<mpsc::Receiver<PyConnectionUpdate>>,
gossipsub_message_rx: Mutex<mpsc::Receiver<(String, Vec<u8>)>>,
}
#[gen_stub_pyclass_complex_enum]
#[pyclass]
enum PyFromSwarm {
Connection {
peer_id: String,
connected: bool,
},
Message {
origin: String,
topic: String,
data: Py<PyBytes>,
},
impl Drop for PyNetworkingHandle {
fn drop(&mut self) {
// TODO: may or may not need to await a "kill-signal" oneshot channel message,
// to ensure that the networking task is done BEFORE exiting the clear function...
// but this may require GIL?? and it may not be safe to call GIL here??
self.to_task_tx = None; // Using Option<T> as a trick to force channel to be dropped
}
}
impl From<FromSwarm> for PyFromSwarm {
fn from(value: FromSwarm) -> Self {
match value {
FromSwarm::Discovered { peer_id } => Self::Connection {
peer_id: peer_id.to_base58(),
connected: true,
},
FromSwarm::Expired { peer_id } => Self::Connection {
peer_id: peer_id.to_base58(),
connected: false,
},
FromSwarm::Message { from, topic, data } => Self::Message {
origin: from.to_base58(),
topic: topic,
data: data.pybytes(),
},
#[allow(clippy::expect_used)]
impl PyNetworkingHandle {
fn new(
to_task_tx: mpsc::Sender<ToTask>,
connection_update_rx: mpsc::Receiver<PyConnectionUpdate>,
gossipsub_message_rx: mpsc::Receiver<(String, Vec<u8>)>,
) -> Self {
Self {
to_task_tx: Some(to_task_tx),
connection_update_rx: Mutex::new(connection_update_rx),
gossipsub_message_rx: Mutex::new(gossipsub_message_rx),
}
}
const fn to_task_tx(&self) -> &mpsc::Sender<ToTask> {
self.to_task_tx
.as_ref()
.expect("The sender should only be None after de-initialization.")
}
}
#[gen_stub_pymethods]
@@ -181,36 +342,97 @@ impl PyNetworkingHandle {
#[new]
fn py_new(identity: Bound<'_, PyKeypair>) -> PyResult<Self> {
use pyo3_async_runtimes::tokio::get_runtime;
// create communication channels
let (to_swarm, from_client) = mpsc::channel(MPSC_CHANNEL_SIZE);
let (to_task_tx, to_task_rx) = mpsc::channel(MPSC_CHANNEL_SIZE);
let (connection_update_tx, connection_update_rx) = mpsc::channel(MPSC_CHANNEL_SIZE);
let (gossipsub_message_tx, gossipsub_message_rx) = mpsc::channel(MPSC_CHANNEL_SIZE);
// get identity
let identity = identity.borrow().0.clone();
// create networking swarm (within tokio context!! or it crashes)
let _guard = pyo3_async_runtimes::tokio::get_runtime().enter();
let swarm = create_swarm(identity, from_client).pyerr()?.into_stream();
let swarm = get_runtime()
.block_on(async { create_swarm(identity) })
.pyerr()?;
Ok(Self {
swarm: Arc::new(Mutex::new(swarm)),
to_swarm,
})
// spawn tokio task running the networking logic
get_runtime().spawn(async move {
networking_task(
swarm,
to_task_rx,
connection_update_tx,
gossipsub_message_tx,
)
.await;
});
Ok(Self::new(
to_task_tx,
connection_update_rx,
gossipsub_message_rx,
))
}
#[gen_stub(skip)]
fn recv<'py>(&'py self, py: Python<'py>) -> PyResult<Bound<'py, PyAny>> {
let swarm = Arc::clone(&self.swarm);
pyo3_async_runtimes::tokio::future_into_py(py, async move {
swarm
.try_lock()
.map_err(|_| PyRuntimeError::new_err("called recv twice concurrently"))?
.next()
.await
.ok_or(PyErr::receiver_channel_closed())
.map(PyFromSwarm::from)
})
const fn __traverse__(&self, _visit: PyVisit<'_>) -> Result<(), PyTraverseError> {
Ok(()) // This is needed purely so `__clear__` can work
}
#[gen_stub(skip)]
fn __clear__(&mut self) {
// TODO: may or may not need to await a "kill-signal" oneshot channel message,
// to ensure that the networking task is done BEFORE exiting the clear function...
// but this may require GIL?? and it may not be safe to call GIL here??
self.to_task_tx = None; // Using Option<T> as a trick to force channel to be dropped
}
// ---- Connection update receiver methods ----
/// Receives the next `ConnectionUpdate` from networking.
async fn connection_update_recv(&self) -> PyResult<PyConnectionUpdate> {
self.connection_update_rx
.lock()
.allow_threads_py() // allow-threads-aware async call
.await
.recv_py()
.allow_threads_py() // allow-threads-aware async call
.await
}
/// Receives at most `limit` `ConnectionUpdate`s from networking and returns them.
///
/// For `limit = 0`, an empty collection of `ConnectionUpdate`s will be returned immediately.
/// For `limit > 0`, if there are no `ConnectionUpdate`s in the channel's queue this method
/// will sleep until a `ConnectionUpdate`s is sent.
async fn connection_update_recv_many(&self, limit: usize) -> PyResult<Vec<PyConnectionUpdate>> {
self.connection_update_rx
.lock()
.allow_threads_py() // allow-threads-aware async call
.await
.recv_many_py(limit)
.allow_threads_py() // allow-threads-aware async call
.await
}
// TODO: rn this blocks main thread if anything else is awaiting the channel (bc its a mutex)
// so its too dangerous to expose just yet. figure out a better semantics for handling this,
// so things don't randomly block
// /// Tries to receive the next `ConnectionUpdate` from networking.
// fn connection_update_try_recv(&self) -> PyResult<Option<PyConnectionUpdate>> {
// self.connection_update_rx.blocking_lock().try_recv_py()
// }
//
// /// Checks if the `ConnectionUpdate` channel is empty.
// fn connection_update_is_empty(&self) -> bool {
// self.connection_update_rx.blocking_lock().is_empty()
// }
//
// /// Returns the number of `ConnectionUpdate`s in the channel.
// fn connection_update_len(&self) -> usize {
// self.connection_update_rx.blocking_lock().len()
// }
// ---- Gossipsub management methods ----
/// Subscribe to a `GossipSub` topic.
@@ -220,10 +442,10 @@ impl PyNetworkingHandle {
let (tx, rx) = oneshot::channel();
// send off request to subscribe
self.to_swarm
.send_py(ToSwarm::Subscribe {
self.to_task_tx()
.send_py(ToTask::GossipsubSubscribe {
topic,
result_sender: tx,
result_tx: tx,
})
.allow_threads_py() // allow-threads-aware async call
.await?;
@@ -232,7 +454,6 @@ impl PyNetworkingHandle {
rx.allow_threads_py() // allow-threads-aware async call
.await
.map_err(|_| PyErr::receiver_channel_closed())?
.pyerr()
}
/// Unsubscribes from a `GossipSub` topic.
@@ -242,10 +463,10 @@ impl PyNetworkingHandle {
let (tx, rx) = oneshot::channel();
// send off request to unsubscribe
self.to_swarm
.send_py(ToSwarm::Unsubscribe {
self.to_task_tx()
.send_py(ToTask::GossipsubUnsubscribe {
topic,
result_sender: tx,
result_tx: tx,
})
.allow_threads_py() // allow-threads-aware async call
.await?;
@@ -264,11 +485,11 @@ impl PyNetworkingHandle {
// send off request to subscribe
let data = Python::attach(|py| Vec::from(data.as_bytes(py)));
self.to_swarm
.send_py(ToSwarm::Publish {
self.to_task_tx()
.send_py(ToTask::GossipsubPublish {
topic,
data,
result_sender: tx,
result_tx: tx,
})
.allow_threads_py() // allow-threads-aware async call
.await?;
@@ -277,35 +498,74 @@ impl PyNetworkingHandle {
let _ = rx
.allow_threads_py() // allow-threads-aware async call
.await
.map_err(|_| PyErr::receiver_channel_closed())?
.map_err(|e| match e {
PublishError::AllQueuesFull(_) => PyAllQueuesFullError::new_err(),
PublishError::MessageTooLarge => PyMessageTooLargeError::new_err(),
PublishError::NoPeersSubscribedToTopic => {
PyNoPeersSubscribedToTopicError::new_err()
}
e => PyRuntimeError::new_err(e.to_string()),
})?;
.map_err(|_| PyErr::receiver_channel_closed())??;
Ok(())
}
}
pyo3_stub_gen::inventory::submit! {
gen_methods_from_python! {
r#"
class PyNetworkingHandle:
async def recv() -> PyFromSwarm: ...
"#
// ---- Gossipsub message receiver methods ----
/// Receives the next message from the `GossipSub` network.
async fn gossipsub_recv(&self) -> PyResult<(String, Py<PyBytes>)> {
self.gossipsub_message_rx
.lock()
.allow_threads_py() // allow-threads-aware async call
.await
.recv_py()
.allow_threads_py() // allow-threads-aware async call
.await
.map(|(t, d)| (t, d.pybytes()))
}
/// Receives at most `limit` messages from the `GossipSub` network and returns them.
///
/// For `limit = 0`, an empty collection of messages will be returned immediately.
/// For `limit > 0`, if there are no messages in the channel's queue this method
/// will sleep until a message is sent.
async fn gossipsub_recv_many(&self, limit: usize) -> PyResult<Vec<(String, Py<PyBytes>)>> {
Ok(self
.gossipsub_message_rx
.lock()
.allow_threads_py() // allow-threads-aware async call
.await
.recv_many_py(limit)
.allow_threads_py() // allow-threads-aware async call
.await?
.into_iter()
.map(|(t, d)| (t, d.pybytes()))
.collect())
}
// TODO: rn this blocks main thread if anything else is awaiting the channel (bc its a mutex)
// so its too dangerous to expose just yet. figure out a better semantics for handling this,
// so things don't randomly block
// /// Tries to receive the next message from the `GossipSub` network.
// fn gossipsub_try_recv(&self) -> PyResult<Option<(String, Py<PyBytes>)>> {
// Ok(self
// .gossipsub_message_rx
// .blocking_lock()
// .try_recv_py()?
// .map(|(t, d)| (t, d.pybytes())))
// }
//
// /// Checks if the `GossipSub` message channel is empty.
// fn gossipsub_is_empty(&self) -> bool {
// self.gossipsub_message_rx.blocking_lock().is_empty()
// }
//
// /// Returns the number of `GossipSub` messages in the channel.
// fn gossipsub_len(&self) -> usize {
// self.gossipsub_message_rx.blocking_lock().len()
// }
}
pub fn networking_submodule(m: &Bound<'_, PyModule>) -> PyResult<()> {
m.add_class::<exception::PyNoPeersSubscribedToTopicError>()?;
m.add_class::<exception::PyAllQueuesFullError>()?;
m.add_class::<exception::PyMessageTooLargeError>()?;
m.add_class::<PyConnectionUpdateType>()?;
m.add_class::<PyConnectionUpdate>()?;
m.add_class::<PyConnectionUpdateType>()?;
m.add_class::<PyNetworkingHandle>()?;
m.add_class::<PyFromSwarm>()?;
Ok(())
}
+13 -15
View File
@@ -1,20 +1,16 @@
import asyncio
import pytest
from exo_pyo3_bindings import (
Keypair,
NetworkingHandle,
NoPeersSubscribedToTopicError,
PyFromSwarm,
)
from exo_pyo3_bindings import Keypair, NetworkingHandle, NoPeersSubscribedToTopicError
@pytest.mark.asyncio
async def test_sleep_on_multiple_items() -> None:
print("PYTHON: starting handle")
h = NetworkingHandle(Keypair.generate())
h = NetworkingHandle(Keypair.generate_ed25519())
rt = asyncio.create_task(_await_recv(h))
ct = asyncio.create_task(_await_cons(h))
mt = asyncio.create_task(_await_msg(h))
# sleep for 4 ticks
for i in range(4):
@@ -26,11 +22,13 @@ async def test_sleep_on_multiple_items() -> None:
print("caught it", e)
async def _await_recv(h: NetworkingHandle):
async def _await_cons(h: NetworkingHandle):
while True:
event = await h.recv()
match event:
case PyFromSwarm.Connection() as c:
print(f"PYTHON: connection update: {c}")
case PyFromSwarm.Message() as m:
print(f"PYTHON: message: {m}")
c = await h.connection_update_recv()
print(f"PYTHON: connection update: {c}")
async def _await_msg(h: NetworkingHandle):
while True:
m = await h.gossipsub_recv()
print(f"PYTHON: message: {m}")
+2 -7
View File
@@ -21,17 +21,13 @@ extend = { workspace = true }
delegate = { workspace = true }
# async
async-stream = { workspace = true }
tokio = { workspace = true, features = ["full"] }
futures-lite = { workspace = true }
futures-timer = { workspace = true }
tokio = { workspace = true, features = ["full"] }
# utility dependencies
util = { workspace = true }
tracing-subscriber = { version = "0.3.19", features = [
"default",
"env-filter",
] }
tracing-subscriber = { version = "0.3.19", features = ["default", "env-filter"] }
keccak-const = { workspace = true }
# tracing/logging
@@ -39,4 +35,3 @@ log = { workspace = true }
# networking
libp2p = { workspace = true, features = ["full"] }
pin-project = "1.1.10"
+49 -56
View File
@@ -1,9 +1,7 @@
use futures_lite::StreamExt;
use libp2p::identity;
use networking::swarm;
use networking::swarm::{FromSwarm, ToSwarm};
use tokio::sync::{mpsc, oneshot};
use tokio::{io, io::AsyncBufReadExt as _};
use libp2p::{gossipsub, identity, swarm::SwarmEvent};
use networking::{discovery, swarm};
use tokio::{io, io::AsyncBufReadExt as _, select};
use tracing_subscriber::EnvFilter;
use tracing_subscriber::filter::LevelFilter;
@@ -13,69 +11,64 @@ async fn main() {
.with_env_filter(EnvFilter::from_default_env().add_directive(LevelFilter::INFO.into()))
.try_init();
let (to_swarm, from_client) = mpsc::channel(20);
// Configure swarm
let mut swarm = swarm::create_swarm(identity::Keypair::generate_ed25519(), from_client)
.expect("Swarm creation failed")
.into_stream();
let mut swarm =
swarm::create_swarm(identity::Keypair::generate_ed25519()).expect("Swarm creation failed");
// Create a Gossipsub topic & subscribe
let (tx, rx) = oneshot::channel();
_ = to_swarm
.send(ToSwarm::Subscribe {
topic: "test-net".to_string(),
result_sender: tx,
})
.await
.expect("should send");
let topic = gossipsub::IdentTopic::new("test-net");
swarm
.behaviour_mut()
.gossipsub
.subscribe(&topic)
.expect("Subscribing to topic failed");
// Read full lines from stdin
let mut stdin = io::BufReader::new(io::stdin()).lines();
println!("Enter messages via STDIN and they will be sent to connected peers using Gossipsub");
tokio::task::spawn(async move {
rx.await
.expect("tx not dropped")
.expect("subscribe shouldn't fail");
loop {
if let Ok(Some(line)) = stdin.next_line().await {
let (tx, rx) = oneshot::channel();
if let Err(e) = to_swarm
.send(swarm::ToSwarm::Publish {
topic: "test-net".to_string(),
data: line.as_bytes().to_vec(),
result_sender: tx,
})
.await
{
println!("Send error: {e:?}");
return;
};
match rx.await {
Ok(Err(e)) => println!("Publish error: {e:?}"),
Err(e) => println!("Publish error: {e:?}"),
Ok(_) => {}
}
}
}
});
// Kick it off
loop {
// on gossipsub outgoing
match swarm.next().await {
// on gossipsub incoming
Some(FromSwarm::Discovered { peer_id }) => {
println!("\n\nconnected to {peer_id}\n\n")
select! {
// on gossipsub outgoing
Ok(Some(line)) = stdin.next_line() => {
if let Err(e) = swarm
.behaviour_mut().gossipsub
.publish(topic.clone(), line.as_bytes()) {
println!("Publish error: {e:?}");
}
}
Some(FromSwarm::Expired { peer_id }) => {
println!("\n\ndisconnected from {peer_id}\n\n")
event = swarm.next() => match event {
// on gossipsub incoming
Some(SwarmEvent::Behaviour(swarm::BehaviourEvent::Gossipsub(gossipsub::Event::Message {
propagation_source: peer_id,
message_id: id,
message,
}))) => println!(
"\n\nGot message: '{}' with id: {id} from peer: {peer_id}\n\n",
String::from_utf8_lossy(&message.data),
),
// on discovery
Some(SwarmEvent::Behaviour(swarm::BehaviourEvent::Discovery(e)) )=> match e {
discovery::Event::ConnectionEstablished {
peer_id, connection_id, remote_ip, remote_tcp_port
} => {
println!("\n\nConnected to: {peer_id}; connection ID: {connection_id}; remote IP: {remote_ip}; remote TCP port: {remote_tcp_port}\n\n");
}
discovery::Event::ConnectionClosed {
peer_id, connection_id, remote_ip, remote_tcp_port
} => {
eprintln!("\n\nDisconnected from: {peer_id}; connection ID: {connection_id}; remote IP: {remote_ip}; remote TCP port: {remote_tcp_port}\n\n");
}
}
// ignore outgoing errors: those are normal
e@Some(SwarmEvent::OutgoingConnectionError { .. }) => { log::debug!("Outgoing connection error: {e:?}"); }
// otherwise log any other event
e => { log::info!("Other event {e:?}"); }
}
Some(FromSwarm::Message { from, topic, data }) => {
println!("{topic}/{from}:\n{}", String::from_utf8_lossy(&data))
}
None => {}
}
}
}
+6 -136
View File
@@ -1,11 +1,9 @@
use std::pin::Pin;
use crate::alias;
use crate::swarm::transport::tcp_transport;
use crate::{alias, discovery};
pub use behaviour::{Behaviour, BehaviourEvent};
use futures_lite::{Stream, StreamExt};
use libp2p::{PeerId, SwarmBuilder, gossipsub, identity, swarm::SwarmEvent};
use tokio::sync::{mpsc, oneshot};
use libp2p::{SwarmBuilder, identity};
pub type Swarm = libp2p::Swarm<Behaviour>;
/// The current version of the network: this prevents devices running different versions of the
/// software from interacting with each other.
@@ -17,136 +15,8 @@ use tokio::sync::{mpsc, oneshot};
pub const NETWORK_VERSION: &[u8] = b"v0.0.1";
pub const OVERRIDE_VERSION_ENV_VAR: &str = "EXO_LIBP2P_NAMESPACE";
// Uses oneshot senders to emulate function calling apis while avoiding requiring unique ownership
// of the Swarm.
pub enum ToSwarm {
Unsubscribe {
topic: String,
result_sender: oneshot::Sender<bool>,
},
Subscribe {
topic: String,
result_sender: oneshot::Sender<Result<bool, gossipsub::SubscriptionError>>,
},
Publish {
topic: String,
data: Vec<u8>,
result_sender: oneshot::Sender<Result<gossipsub::MessageId, gossipsub::PublishError>>,
},
}
pub enum FromSwarm {
Message {
from: PeerId,
topic: String,
data: Vec<u8>,
},
Discovered {
peer_id: PeerId,
},
Expired {
peer_id: PeerId,
},
}
pub struct Swarm {
swarm: libp2p::Swarm<Behaviour>,
from_client: mpsc::Receiver<ToSwarm>,
}
impl Swarm {
pub fn into_stream(self) -> Pin<Box<dyn Stream<Item = FromSwarm> + Send>> {
let Swarm {
mut swarm,
mut from_client,
} = self;
let stream = async_stream::stream! {
loop {
tokio::select! {
msg = from_client.recv() => {
let Some(msg) = msg else { break };
on_message(&mut swarm, msg);
}
event = swarm.next() => {
let Some(event) = event else { break };
if let Some(item) = filter_swarm_event(event) {
yield item;
}
}
}
}
};
Box::pin(stream)
}
}
fn on_message(swarm: &mut libp2p::Swarm<Behaviour>, message: ToSwarm) {
match message {
ToSwarm::Subscribe {
topic,
result_sender,
} => {
let result = swarm
.behaviour_mut()
.gossipsub
.subscribe(&gossipsub::IdentTopic::new(topic));
_ = result_sender.send(result);
}
ToSwarm::Unsubscribe {
topic,
result_sender,
} => {
let result = swarm
.behaviour_mut()
.gossipsub
.unsubscribe(&gossipsub::IdentTopic::new(topic));
_ = result_sender.send(result);
}
ToSwarm::Publish {
topic,
data,
result_sender,
} => {
let result = swarm
.behaviour_mut()
.gossipsub
.publish(gossipsub::IdentTopic::new(topic), data);
_ = result_sender.send(result);
}
}
}
fn filter_swarm_event(event: SwarmEvent<BehaviourEvent>) -> Option<FromSwarm> {
match event {
SwarmEvent::Behaviour(BehaviourEvent::Gossipsub(gossipsub::Event::Message {
message:
gossipsub::Message {
source: Some(peer_id),
topic,
data,
..
},
..
})) => Some(FromSwarm::Message {
from: peer_id,
topic: topic.into_string(),
data,
}),
SwarmEvent::Behaviour(BehaviourEvent::Discovery(
discovery::Event::ConnectionEstablished { peer_id, .. },
)) => Some(FromSwarm::Discovered { peer_id }),
SwarmEvent::Behaviour(BehaviourEvent::Discovery(discovery::Event::ConnectionClosed {
peer_id,
..
})) => Some(FromSwarm::Expired { peer_id }),
_ => None,
}
}
/// Create and configure a swarm which listens to all ports on OS
pub fn create_swarm(
keypair: identity::Keypair,
from_client: mpsc::Receiver<ToSwarm>,
) -> alias::AnyResult<Swarm> {
pub fn create_swarm(keypair: identity::Keypair) -> alias::AnyResult<Swarm> {
let mut swarm = SwarmBuilder::with_existing_identity(keypair)
.with_tokio()
.with_other_transport(tcp_transport)?
@@ -155,7 +25,7 @@ pub fn create_swarm(
// Listen on all interfaces and whatever port the OS assigns
swarm.listen_on("/ip4/0.0.0.0/tcp/0".parse()?)?;
Ok(Swarm { swarm, from_client })
Ok(swarm)
}
mod transport {
+92 -6
View File
@@ -1,8 +1,11 @@
import asyncio
import socket
from dataclasses import dataclass, field
from random import random
import anyio
from anyio import current_time
from anyio.abc import TaskGroup
from loguru import logger
from exo.download.download_utils import (
@@ -20,9 +23,13 @@ from exo.shared.types.commands import (
ForwarderDownloadCommand,
StartDownload,
)
from exo.shared.types.common import NodeId
from exo.shared.types.common import NodeId, SessionId, SystemId
from exo.shared.types.events import (
Event,
EventId,
# TODO(evan): just for acks, should delete this ASAP
GlobalForwarderEvent,
LocalForwarderEvent,
NodeDownloadProgress,
)
from exo.shared.types.worker.downloads import (
@@ -33,28 +40,41 @@ from exo.shared.types.worker.downloads import (
DownloadProgress,
)
from exo.shared.types.worker.shards import PipelineShardMetadata, ShardMetadata
from exo.utils.channels import Receiver, Sender
from exo.utils.task_group import TaskGroup
from exo.utils.channels import Receiver, Sender, channel
@dataclass
class DownloadCoordinator:
node_id: NodeId
session_id: SessionId
shard_downloader: ShardDownloader
download_command_receiver: Receiver[ForwarderDownloadCommand]
event_sender: Sender[Event]
local_event_sender: Sender[LocalForwarderEvent]
# ack stuff
_global_event_receiver: Receiver[GlobalForwarderEvent]
_out_for_delivery: dict[EventId, LocalForwarderEvent] = field(default_factory=dict)
offline: bool = False
_system_id: SystemId = field(default_factory=SystemId)
# Local state
download_status: dict[ModelId, DownloadProgress] = field(default_factory=dict)
active_downloads: dict[ModelId, asyncio.Task[None]] = field(default_factory=dict)
_tg: TaskGroup = field(init=False, default_factory=TaskGroup)
# Internal event channel for forwarding (initialized in __post_init__)
event_sender: Sender[Event] = field(init=False)
event_receiver: Receiver[Event] = field(init=False)
_tg: TaskGroup = field(init=False, default_factory=anyio.create_task_group)
# Per-model throttle for download progress events
_last_progress_time: dict[ModelId, float] = field(default_factory=dict)
def __post_init__(self) -> None:
self.event_sender, self.event_receiver = channel[Event]()
if self.offline:
self.shard_downloader.set_internet_connection(False)
self.shard_downloader.on_progress(self._download_progress_callback)
def _model_dir(self, model_id: ModelId) -> str:
@@ -103,16 +123,65 @@ class DownloadCoordinator:
logger.info(
f"Starting DownloadCoordinator{' (offline mode)' if self.offline else ''}"
)
if not self.offline:
self._test_internet_connection()
try:
async with self._tg as tg:
tg.start_soon(self._command_processor)
tg.start_soon(self._forward_events)
tg.start_soon(self._emit_existing_download_progress)
tg.start_soon(self._resend_out_for_delivery)
tg.start_soon(self._clear_ofd)
if not self.offline:
tg.start_soon(self._check_internet_connection)
finally:
for task in self.active_downloads.values():
task.cancel()
def _test_internet_connection(self) -> None:
# Try multiple endpoints since some ISPs/networks block specific IPs
for host in ("1.1.1.1", "8.8.8.8", "1.0.0.1"):
try:
socket.create_connection((host, 443), timeout=3).close()
self.shard_downloader.set_internet_connection(True)
logger.debug(f"Internet connectivity: True (via {host})")
return
except OSError:
continue
self.shard_downloader.set_internet_connection(False)
logger.debug("Internet connectivity: False")
async def _check_internet_connection(self) -> None:
first_connection = True
while True:
await asyncio.sleep(10)
# Assume that internet connection is set to False on 443 errors.
if self.shard_downloader.internet_connection:
continue
self._test_internet_connection()
if first_connection and self.shard_downloader.internet_connection:
first_connection = False
self._tg.start_soon(self._emit_existing_download_progress)
def shutdown(self) -> None:
self._tg.cancel_tasks()
self._tg.cancel_scope.cancel()
# directly copied from worker
async def _resend_out_for_delivery(self) -> None:
# This can also be massively tightened, we should check events are at least a certain age before resending.
# Exponential backoff would also certainly help here.
while True:
await anyio.sleep(1 + random())
for event in self._out_for_delivery.copy().values():
await self.local_event_sender.send(event)
async def _clear_ofd(self) -> None:
with self._global_event_receiver as events:
async for event in events:
self._out_for_delivery.pop(event.event.event_id, None)
async def _command_processor(self) -> None:
with self.download_command_receiver as commands:
@@ -286,6 +355,23 @@ class DownloadCoordinator:
)
del self.download_status[model_id]
async def _forward_events(self) -> None:
idx = 0
with self.event_receiver as events:
async for event in events:
fe = LocalForwarderEvent(
origin_idx=idx,
origin=self._system_id,
session=self.session_id,
event=event,
)
idx += 1
logger.debug(
f"DownloadCoordinator published event {idx}: {str(event)[:100]}"
)
await self.local_event_sender.send(fe)
self._out_for_delivery[event.event_id] = fe
async def _emit_existing_download_progress(self) -> None:
try:
while True:
+2 -9
View File
@@ -314,13 +314,9 @@ async def fetch_file_list_with_cache(
_fetched_file_lists_this_session.add(cache_key)
return file_list
except Exception as e:
logger.opt(exception=e).warning(
"Ran into exception when fetching file list from HF."
)
if await aios.path.exists(cache_file):
logger.warning(
f"No cached file list for {model_id} - using local file list"
f"No internet and no cached file list for {model_id} - using local file list"
)
async with aiofiles.open(cache_file, "r") as f:
return TypeAdapter(list[FileListEntry]).validate_json(await f.read())
@@ -827,7 +823,6 @@ async def download_shard(
for file in filtered_file_list:
downloaded_bytes = await get_downloaded_size(target_dir / file.path)
final_file_exists = await aios.path.exists(target_dir / file.path)
file_progress[file.path] = RepoFileDownloadProgress(
repo_id=shard.model_card.model_id,
repo_revision=revision,
@@ -837,9 +832,7 @@ async def download_shard(
total=Memory.from_bytes(file.size or 0),
speed=0,
eta=timedelta(0),
status="complete"
if final_file_exists and downloaded_bytes == file.size
else "not_started",
status="complete" if downloaded_bytes == file.size else "not_started",
start_time=time.time(),
)
+17 -11
View File
@@ -15,13 +15,9 @@ from exo.shared.types.worker.shards import (
)
def exo_shard_downloader(
max_parallel_downloads: int = 8, offline: bool = False
) -> ShardDownloader:
def exo_shard_downloader(max_parallel_downloads: int = 8) -> ShardDownloader:
return SingletonShardDownloader(
CachedShardDownloader(
ResumableShardDownloader(max_parallel_downloads, offline=offline)
)
CachedShardDownloader(ResumableShardDownloader(max_parallel_downloads))
)
@@ -54,6 +50,10 @@ class SingletonShardDownloader(ShardDownloader):
self.shard_downloader = shard_downloader
self.active_downloads: dict[ShardMetadata, asyncio.Task[Path]] = {}
def set_internet_connection(self, value: bool) -> None:
self.internet_connection = value
self.shard_downloader.set_internet_connection(value)
def on_progress(
self,
callback: Callable[[ShardMetadata, RepoDownloadProgress], Awaitable[None]],
@@ -90,6 +90,10 @@ class CachedShardDownloader(ShardDownloader):
self.shard_downloader = shard_downloader
self.cache: dict[tuple[str, ShardMetadata], Path] = {}
def set_internet_connection(self, value: bool) -> None:
self.internet_connection = value
self.shard_downloader.set_internet_connection(value)
def on_progress(
self,
callback: Callable[[ShardMetadata, RepoDownloadProgress], Awaitable[None]],
@@ -119,9 +123,8 @@ class CachedShardDownloader(ShardDownloader):
class ResumableShardDownloader(ShardDownloader):
def __init__(self, max_parallel_downloads: int = 8, offline: bool = False):
def __init__(self, max_parallel_downloads: int = 8):
self.max_parallel_downloads = max_parallel_downloads
self.offline = offline
self.on_progress_callbacks: list[
Callable[[ShardMetadata, RepoDownloadProgress], Awaitable[None]]
] = []
@@ -148,7 +151,8 @@ class ResumableShardDownloader(ShardDownloader):
self.on_progress_wrapper,
max_parallel_downloads=self.max_parallel_downloads,
allow_patterns=allow_patterns,
skip_internet=self.offline,
skip_internet=not self.internet_connection,
on_connection_lost=lambda: self.set_internet_connection(False),
)
return target_dir
@@ -164,7 +168,8 @@ class ResumableShardDownloader(ShardDownloader):
shard,
self.on_progress_wrapper,
skip_download=True,
skip_internet=self.offline,
skip_internet=not self.internet_connection,
on_connection_lost=lambda: self.set_internet_connection(False),
)
semaphore = asyncio.Semaphore(self.max_parallel_downloads)
@@ -193,6 +198,7 @@ class ResumableShardDownloader(ShardDownloader):
shard,
self.on_progress_wrapper,
skip_download=True,
skip_internet=self.offline,
skip_internet=not self.internet_connection,
on_connection_lost=lambda: self.set_internet_connection(False),
)
return progress
+5
View File
@@ -16,6 +16,11 @@ from exo.shared.types.worker.shards import (
# TODO: the PipelineShardMetadata getting reinstantiated is a bit messy. Should this be a classmethod?
class ShardDownloader(ABC):
internet_connection: bool = False
def set_internet_connection(self, value: bool) -> None:
self.internet_connection = value
@abstractmethod
async def ensure_shard(
self, shard: ShardMetadata, config_only: bool = False
@@ -0,0 +1,98 @@
from typing import Any
import anyio
import pytest
from exo.download.coordinator import DownloadCoordinator
from exo.download.shard_downloader import NoopShardDownloader
from exo.shared.models.model_cards import ModelCard, ModelTask
from exo.shared.types.common import ModelId, NodeId, SessionId
from exo.shared.types.events import (
GlobalForwarderEvent,
LocalForwarderEvent,
NodeDownloadProgress,
)
from exo.shared.types.memory import Memory
from exo.shared.types.worker.downloads import (
DownloadPending,
)
from exo.shared.types.worker.shards import PipelineShardMetadata
from exo.utils.channels import channel
# Use the builtin NoopShardDownloader directly it already implements the required abstract interface.
# No additional subclass is needed for this test.
@pytest.mark.anyio
async def test_ack_behaviour():
# Create channels (type Any for simplicity)
_, command_receiver = channel[Any]()
local_sender, _ = channel[Any]()
global_sender, global_receiver = channel[Any]()
# Minimal identifiers
node_id = NodeId()
session_id = SessionId(master_node_id=node_id, election_clock=0)
# Create a dummy model card and shard metadata
model_id = ModelId("test/model")
model_card = ModelCard(
model_id=model_id,
storage_size=Memory.from_bytes(0),
n_layers=1,
hidden_size=1,
supports_tensor=True,
tasks=[ModelTask.TextGeneration],
)
shard = PipelineShardMetadata(
model_card=model_card,
device_rank=0,
world_size=1,
start_layer=0,
end_layer=1,
n_layers=1,
)
# Instantiate the coordinator with the dummy downloader
coord = DownloadCoordinator(
node_id=node_id,
session_id=session_id,
shard_downloader=NoopShardDownloader(),
download_command_receiver=command_receiver,
local_event_sender=local_sender,
_global_event_receiver=global_receiver,
)
async with anyio.create_task_group() as tg:
# Start the forwarding and ackclearing loops
tg.start_soon(coord._forward_events) # pyright: ignore[reportPrivateUsage]
tg.start_soon(coord._clear_ofd) # pyright: ignore[reportPrivateUsage]
# Send a pending download progress event via the internal event sender
pending = DownloadPending(
node_id=node_id,
shard_metadata=shard,
model_directory="/tmp/model",
)
await coord.event_sender.send(NodeDownloadProgress(download_progress=pending))
# Allow the forwarder to process the event
await anyio.sleep(0.1)
# There should be exactly one entry awaiting ACK
assert len(coord._out_for_delivery) == 1 # pyright: ignore[reportPrivateUsage]
# Retrieve the stored LocalForwarderEvent
stored_fe: LocalForwarderEvent = next(iter(coord._out_for_delivery.values())) # pyright: ignore[reportPrivateUsage]
# Simulate receiving a global ack for this event
ack = GlobalForwarderEvent(
origin_idx=0,
origin=node_id,
session=session_id,
event=stored_fe.event,
)
await global_sender.send(ack)
# Give the clearofd task a moment to process the ack
await anyio.sleep(0.1)
# The outfordelivery map should now be empty
assert len(coord._out_for_delivery) == 0 # pyright: ignore[reportPrivateUsage]
# Cancel background tasks
tg.cancel_scope.cancel()
+33 -50
View File
@@ -7,6 +7,7 @@ from dataclasses import dataclass, field
from typing import Self
import anyio
from anyio.abc import TaskGroup
from loguru import logger
from pydantic import PositiveInt
@@ -15,7 +16,6 @@ from exo.download.coordinator import DownloadCoordinator
from exo.download.impl_shard_downloader import exo_shard_downloader
from exo.master.api import API # TODO: should API be in master?
from exo.master.main import Master
from exo.routing.event_router import EventRouter
from exo.routing.router import Router, get_node_id_keypair
from exo.shared.constants import EXO_LOG
from exo.shared.election import Election, ElectionResult
@@ -23,14 +23,12 @@ from exo.shared.logging import logger_cleanup, logger_setup
from exo.shared.types.common import NodeId, SessionId
from exo.utils.channels import Receiver, channel
from exo.utils.pydantic_ext import CamelCaseModel
from exo.utils.task_group import TaskGroup
from exo.worker.main import Worker
@dataclass
class Node:
router: Router
event_router: EventRouter
download_coordinator: DownloadCoordinator | None
worker: Worker | None
election: Election # Every node participates in election, as we do want a node to become master even if it isn't a master candidate if no master candidates are present.
@@ -40,7 +38,7 @@ class Node:
node_id: NodeId
offline: bool
_tg: TaskGroup = field(init=False, default_factory=TaskGroup)
_tg: TaskGroup = field(init=False, default_factory=anyio.create_task_group)
@classmethod
async def create(cls, args: "Args") -> Self:
@@ -54,12 +52,6 @@ class Node:
await router.register_topic(topics.ELECTION_MESSAGES)
await router.register_topic(topics.CONNECTION_MESSAGES)
await router.register_topic(topics.DOWNLOAD_COMMANDS)
event_router = EventRouter(
session_id,
command_sender=router.sender(topics.COMMANDS),
external_outbound=router.sender(topics.LOCAL_EVENTS),
external_inbound=router.receiver(topics.GLOBAL_EVENTS),
)
logger.info(f"Starting node {node_id}")
@@ -67,10 +59,13 @@ class Node:
if not args.no_downloads:
download_coordinator = DownloadCoordinator(
node_id,
exo_shard_downloader(offline=args.offline),
event_sender=event_router.sender(),
session_id,
exo_shard_downloader(),
download_command_receiver=router.receiver(topics.DOWNLOAD_COMMANDS),
local_event_sender=router.sender(topics.LOCAL_EVENTS),
offline=args.offline,
# TODO(evan): remove
_global_event_receiver=router.receiver(topics.GLOBAL_EVENTS),
)
else:
download_coordinator = None
@@ -78,8 +73,9 @@ class Node:
if args.spawn_api:
api = API(
node_id,
session_id,
port=args.api_port,
event_receiver=event_router.receiver(),
global_event_receiver=router.receiver(topics.GLOBAL_EVENTS),
command_sender=router.sender(topics.COMMANDS),
download_command_sender=router.sender(topics.DOWNLOAD_COMMANDS),
election_receiver=router.receiver(topics.ELECTION_MESSAGES),
@@ -90,8 +86,9 @@ class Node:
if not args.no_worker:
worker = Worker(
node_id,
event_receiver=event_router.receiver(),
event_sender=event_router.sender(),
session_id,
global_event_receiver=router.receiver(topics.GLOBAL_EVENTS),
local_event_sender=router.sender(topics.LOCAL_EVENTS),
command_sender=router.sender(topics.COMMANDS),
download_command_sender=router.sender(topics.DOWNLOAD_COMMANDS),
)
@@ -102,7 +99,6 @@ class Node:
master = Master(
node_id,
session_id,
event_sender=event_router.sender(),
global_event_sender=router.sender(topics.GLOBAL_EVENTS),
local_event_receiver=router.receiver(topics.LOCAL_EVENTS),
command_receiver=router.receiver(topics.COMMANDS),
@@ -125,7 +121,6 @@ class Node:
return cls(
router,
event_router,
download_coordinator,
worker,
election,
@@ -141,7 +136,6 @@ class Node:
signal.signal(signal.SIGINT, lambda _, __: self.shutdown())
signal.signal(signal.SIGTERM, lambda _, __: self.shutdown())
tg.start_soon(self.router.run)
tg.start_soon(self.event_router.run)
tg.start_soon(self.election.run)
if self.download_coordinator:
tg.start_soon(self.download_coordinator.run)
@@ -155,11 +149,11 @@ class Node:
def shutdown(self):
# if this is our second call to shutdown, just sys.exit
if self._tg.cancel_called():
if self._tg.cancel_scope.cancel_called:
import sys
sys.exit(1)
self._tg.cancel_tasks()
self._tg.cancel_scope.cancel()
async def _elect_loop(self):
with self.election_result_receiver as results:
@@ -176,17 +170,6 @@ class Node:
# - Shutdown and re-create the worker
# - Shut down and re-create the API
if result.is_new_master:
await anyio.sleep(0)
self.event_router.shutdown()
self.event_router = EventRouter(
result.session_id,
self.router.sender(topics.COMMANDS),
self.router.receiver(topics.GLOBAL_EVENTS),
self.router.sender(topics.LOCAL_EVENTS),
)
self._tg.start_soon(self.event_router.run)
if (
result.session_id.master_node_id == self.node_id
and self.master is not None
@@ -200,7 +183,6 @@ class Node:
self.master = Master(
self.node_id,
result.session_id,
event_sender=self.event_router.sender(),
global_event_sender=self.router.sender(topics.GLOBAL_EVENTS),
local_event_receiver=self.router.receiver(topics.LOCAL_EVENTS),
command_receiver=self.router.receiver(topics.COMMANDS),
@@ -223,16 +205,22 @@ class Node:
f"Node {result.session_id.master_node_id} elected master"
)
if result.is_new_master:
await anyio.sleep(0)
if self.download_coordinator:
self.download_coordinator.shutdown()
self.download_coordinator = DownloadCoordinator(
self.node_id,
exo_shard_downloader(offline=self.offline),
event_sender=self.event_router.sender(),
result.session_id,
exo_shard_downloader(),
download_command_receiver=self.router.receiver(
topics.DOWNLOAD_COMMANDS
),
local_event_sender=self.router.sender(topics.LOCAL_EVENTS),
offline=self.offline,
# TODO(evan): remove
_global_event_receiver=self.router.receiver(
topics.GLOBAL_EVENTS
),
)
self._tg.start_soon(self.download_coordinator.run)
if self.worker:
@@ -240,8 +228,11 @@ class Node:
# TODO: add profiling etc to resource monitor
self.worker = Worker(
self.node_id,
event_receiver=self.event_router.receiver(),
event_sender=self.event_router.sender(),
result.session_id,
global_event_receiver=self.router.receiver(
topics.GLOBAL_EVENTS
),
local_event_sender=self.router.sender(topics.LOCAL_EVENTS),
command_sender=self.router.sender(topics.COMMANDS),
download_command_sender=self.router.sender(
topics.DOWNLOAD_COMMANDS
@@ -249,7 +240,7 @@ class Node:
)
self._tg.start_soon(self.worker.run)
if self.api:
self.api.reset(result.won_clock, self.event_router.receiver())
self.api.reset(result.session_id, result.won_clock)
else:
if self.api:
self.api.unpause(result.won_clock)
@@ -261,7 +252,7 @@ def main():
target = min(max(soft, 65535), hard)
resource.setrlimit(resource.RLIMIT_NOFILE, (target, hard))
mp.set_start_method("spawn", force=True)
mp.set_start_method("spawn")
# TODO: Refactor the current verbosity system
logger_setup(EXO_LOG, args.verbosity)
logger.info("Starting EXO")
@@ -279,16 +270,9 @@ def main():
logger.info("FAST_SYNCH forced OFF")
node = anyio.run(Node.create, args)
try:
anyio.run(node.run)
except BaseException as exception:
logger.opt(exception=exception).critical(
"EXO terminated due to unhandled exception"
)
raise
finally:
logger.info("EXO Shutdown complete")
logger_cleanup()
anyio.run(node.run)
logger.info("EXO Shutdown complete")
logger_cleanup()
class Args(CamelCaseModel):
@@ -299,7 +283,7 @@ class Args(CamelCaseModel):
tb_only: bool = False
no_worker: bool = False
no_downloads: bool = False
offline: bool = os.getenv("EXO_OFFLINE", "false").lower() == "true"
offline: bool = False
fast_synch: bool | None = None # None = auto, True = force on, False = force off
@classmethod
@@ -350,7 +334,6 @@ class Args(CamelCaseModel):
parser.add_argument(
"--offline",
action="store_true",
default=os.getenv("EXO_OFFLINE", "false").lower() == "true",
help="Run in offline/air-gapped mode: skip internet checks, use only pre-staged local models",
)
fast_synch_group = parser.add_mutually_exclusive_group()
+116 -87
View File
@@ -11,7 +11,8 @@ from typing import Annotated, Literal, cast
from uuid import uuid4
import anyio
from anyio import BrokenResourceError
from anyio import BrokenResourceError, create_task_group
from anyio.abc import TaskGroup
from fastapi import FastAPI, File, Form, HTTPException, Query, Request, UploadFile
from fastapi.middleware.cors import CORSMiddleware
from fastapi.responses import FileResponse, JSONResponse, StreamingResponse
@@ -50,7 +51,6 @@ from exo.master.placement import place_instance as get_instance_placements
from exo.shared.apply import apply
from exo.shared.constants import (
DASHBOARD_DIR,
EXO_CACHE_HOME,
EXO_EVENT_LOG_DIR,
EXO_IMAGE_CACHE_DIR,
EXO_MAX_CHUNK_SIZE,
@@ -73,17 +73,17 @@ from exo.shared.types.api import (
BenchChatCompletionResponse,
BenchImageGenerationResponse,
BenchImageGenerationTaskParams,
CancelCommandResponse,
ChatCompletionChoice,
ChatCompletionMessage,
ChatCompletionRequest,
ChatCompletionResponse,
CreateInstanceParams,
CreateInstanceResponse,
CreateMetaInstanceParams,
CreateMetaInstanceResponse,
DeleteDownloadResponse,
DeleteInstanceResponse,
DeleteTracesRequest,
DeleteTracesResponse,
DeleteMetaInstanceResponse,
ErrorInfo,
ErrorResponse,
FinishReason,
@@ -129,8 +129,10 @@ from exo.shared.types.claude_api import (
from exo.shared.types.commands import (
Command,
CreateInstance,
CreateMetaInstance,
DeleteDownload,
DeleteInstance,
DeleteMetaInstance,
DownloadCommand,
ForwarderCommand,
ForwarderDownloadCommand,
@@ -143,14 +145,23 @@ from exo.shared.types.commands import (
TaskFinished,
TextGeneration,
)
from exo.shared.types.common import CommandId, Id, NodeId, SystemId
from exo.shared.types.common import (
CommandId,
Id,
MetaInstanceId,
NodeId,
SessionId,
SystemId,
)
from exo.shared.types.events import (
ChunkGenerated,
Event,
GlobalForwarderEvent,
IndexedEvent,
TracesMerged,
)
from exo.shared.types.memory import Memory
from exo.shared.types.meta_instance import MetaInstance
from exo.shared.types.ollama_api import (
OllamaChatRequest,
OllamaChatResponse,
@@ -174,10 +185,9 @@ from exo.shared.types.worker.instances import Instance, InstanceId, InstanceMeta
from exo.shared.types.worker.shards import Sharding
from exo.utils.banner import print_startup_banner
from exo.utils.channels import Receiver, Sender, channel
from exo.utils.task_group import TaskGroup
from exo.utils.event_buffer import OrderedBuffer
_API_EVENT_LOG_DIR = EXO_EVENT_LOG_DIR / "api"
ONBOARDING_COMPLETE_FILE = EXO_CACHE_HOME / "onboarding_complete"
def _format_to_content_type(image_format: Literal["png", "jpeg", "webp"] | None) -> str:
@@ -197,9 +207,10 @@ class API:
def __init__(
self,
node_id: NodeId,
session_id: SessionId,
*,
port: int,
event_receiver: Receiver[IndexedEvent],
global_event_receiver: Receiver[GlobalForwarderEvent],
command_sender: Sender[ForwarderCommand],
download_command_sender: Sender[ForwarderDownloadCommand],
# This lets us pause the API if an election is running
@@ -210,9 +221,11 @@ class API:
self._system_id = SystemId()
self.command_sender = command_sender
self.download_command_sender = download_command_sender
self.event_receiver = event_receiver
self.global_event_receiver = global_event_receiver
self.election_receiver = election_receiver
self.event_buffer: OrderedBuffer[Event] = OrderedBuffer[Event]()
self.node_id: NodeId = node_id
self.session_id: SessionId = session_id
self.last_completed_election: int = 0
self.port = port
@@ -250,20 +263,19 @@ class API:
CommandId, Sender[ImageChunk | ErrorChunk]
] = {}
self._image_store = ImageStore(EXO_IMAGE_CACHE_DIR)
self._tg: TaskGroup = TaskGroup()
self._tg: TaskGroup | None = None
def reset(self, result_clock: int, event_receiver: Receiver[IndexedEvent]):
def reset(self, new_session_id: SessionId, result_clock: int):
logger.info("Resetting API State")
self._event_log.close()
self._event_log = DiskEventLog(_API_EVENT_LOG_DIR)
self.state = State()
self._system_id = SystemId()
self.session_id = new_session_id
self.event_buffer = OrderedBuffer[Event]()
self._text_generation_queues = {}
self._image_generation_queues = {}
self.unpause(result_clock)
self.event_receiver.close()
self.event_receiver = event_receiver
self._tg.start_soon(self._apply_state)
def unpause(self, result_clock: int):
logger.info("Unpausing API")
@@ -304,6 +316,9 @@ class API:
self.app.get("/instance/previews")(self.get_placement_previews)
self.app.get("/instance/{instance_id}")(self.get_instance)
self.app.delete("/instance/{instance_id}")(self.delete_instance)
self.app.get("/meta_instances")(self.list_meta_instances)
self.app.post("/meta_instance")(self.create_meta_instance)
self.app.delete("/meta_instance/{meta_instance_id}")(self.delete_meta_instance)
self.app.get("/models")(self.get_models)
self.app.get("/v1/models")(self.get_models)
self.app.post("/models/add")(self.add_custom_model)
@@ -323,7 +338,6 @@ class API:
self.app.get("/images/{image_id}")(self.get_image)
self.app.post("/v1/messages", response_model=None)(self.claude_messages)
self.app.post("/v1/responses", response_model=None)(self.openai_responses)
self.app.post("/v1/cancel/{command_id}")(self.cancel_command)
# Ollama API
self.app.head("/ollama/")(self.ollama_version)
@@ -344,20 +358,32 @@ class API:
self.app.post("/download/start")(self.start_download)
self.app.delete("/download/{node_id}/{model_id:path}")(self.delete_download)
self.app.get("/v1/traces")(self.list_traces)
self.app.post("/v1/traces/delete")(self.delete_traces)
self.app.get("/v1/traces/{task_id}")(self.get_trace)
self.app.get("/v1/traces/{task_id}/stats")(self.get_trace_stats)
self.app.get("/v1/traces/{task_id}/raw")(self.get_trace_raw)
self.app.get("/onboarding")(self.get_onboarding)
self.app.post("/onboarding")(self.complete_onboarding)
async def place_instance(self, payload: PlaceInstanceParams):
model_card = await ModelCard.load(payload.model_id)
command = PlaceInstance(
model_card=await ModelCard.load(payload.model_id),
model_card=model_card,
sharding=payload.sharding,
instance_meta=payload.instance_meta,
min_nodes=payload.min_nodes,
)
# Validate placement before sending — fail fast with a clear error
# instead of silently dropping the command in the master.
try:
get_instance_placements(
command,
topology=self.state.topology,
current_instances=self.state.instances,
node_memory=self.state.node_memory,
node_network=self.state.node_network,
)
except ValueError as exc:
raise HTTPException(status_code=400, detail=str(exc)) from exc
await self._send(command)
return CreateInstanceResponse(
@@ -569,23 +595,42 @@ class API:
instance_id=instance_id,
)
async def cancel_command(self, command_id: CommandId) -> CancelCommandResponse:
"""Cancel an active command by closing its stream and notifying workers."""
sender = self._text_generation_queues.get(
command_id
) or self._image_generation_queues.get(command_id)
if sender is None:
raise HTTPException(
status_code=404,
detail="Command not found or already completed",
)
def list_meta_instances(self) -> dict[MetaInstanceId, MetaInstance]:
return dict(self.state.meta_instances)
await self._send(TaskCancelled(cancelled_command_id=command_id))
sender.close()
async def create_meta_instance(
self, payload: CreateMetaInstanceParams
) -> CreateMetaInstanceResponse:
meta_instance = MetaInstance(
model_id=payload.model_id,
sharding=payload.sharding,
instance_meta=payload.instance_meta,
min_nodes=payload.min_nodes,
node_ids=payload.node_ids,
)
command = CreateMetaInstance(meta_instance=meta_instance)
await self._send(command)
return CreateMetaInstanceResponse(
message="Command received.",
command_id=command.command_id,
meta_instance_id=meta_instance.meta_instance_id,
)
return CancelCommandResponse(
message="Command cancelled.",
command_id=command_id,
async def delete_meta_instance(
self, meta_instance_id: MetaInstanceId
) -> DeleteMetaInstanceResponse:
meta = self.state.meta_instances.get(meta_instance_id)
if not meta:
raise HTTPException(status_code=404, detail="MetaInstance not found")
# Command processor handles cascade-deleting backing instances
command = DeleteMetaInstance(meta_instance_id=meta_instance_id)
await self._send(command)
return DeleteMetaInstanceResponse(
message="Command received.",
command_id=command.command_id,
meta_instance_id=meta_instance_id,
)
async def _token_chunk_stream(
@@ -1611,7 +1656,8 @@ class API:
shutdown_ev = anyio.Event()
try:
async with self._tg as tg:
async with create_task_group() as tg:
self._tg = tg
logger.info("Starting API")
tg.start_soon(self._apply_state)
tg.start_soon(self._pause_on_new_election)
@@ -1626,7 +1672,7 @@ class API:
finally:
self._event_log.close()
self.command_sender.close()
self.event_receiver.close()
self.global_event_receiver.close()
async def run_api(self, ev: anyio.Event):
cfg = Config()
@@ -1643,31 +1689,38 @@ class API:
)
async def _apply_state(self):
with self.event_receiver as events:
async for i_event in events:
self._event_log.append(i_event.event)
self.state = apply(self.state, i_event)
event = i_event.event
with self.global_event_receiver as events:
async for f_event in events:
if f_event.session != self.session_id:
continue
if f_event.origin != self.session_id.master_node_id:
continue
self.event_buffer.ingest(f_event.origin_idx, f_event.event)
for idx, event in self.event_buffer.drain_indexed():
self._event_log.append(event)
self.state = apply(self.state, IndexedEvent(event=event, idx=idx))
if isinstance(event, ChunkGenerated):
if queue := self._image_generation_queues.get(
event.command_id, None
):
assert isinstance(event.chunk, ImageChunk)
try:
await queue.send(event.chunk)
except BrokenResourceError:
self._image_generation_queues.pop(event.command_id, None)
if queue := self._text_generation_queues.get(
event.command_id, None
):
assert not isinstance(event.chunk, ImageChunk)
try:
await queue.send(event.chunk)
except BrokenResourceError:
self._text_generation_queues.pop(event.command_id, None)
if isinstance(event, TracesMerged):
self._save_merged_trace(event)
if isinstance(event, ChunkGenerated):
if queue := self._image_generation_queues.get(
event.command_id, None
):
assert isinstance(event.chunk, ImageChunk)
try:
await queue.send(event.chunk)
except BrokenResourceError:
self._image_generation_queues.pop(
event.command_id, None
)
if queue := self._text_generation_queues.get(
event.command_id, None
):
assert not isinstance(event.chunk, ImageChunk)
try:
await queue.send(event.chunk)
except BrokenResourceError:
self._text_generation_queues.pop(event.command_id, None)
if isinstance(event, TracesMerged):
self._save_merged_trace(event)
def _save_merged_trace(self, event: TracesMerged) -> None:
traces = [
@@ -1731,12 +1784,8 @@ class API:
await self._send_download(command)
return DeleteDownloadResponse(command_id=command.command_id)
@staticmethod
def _get_trace_path(task_id: str) -> Path:
trace_path = EXO_TRACING_CACHE_DIR / f"trace_{task_id}.json"
if not trace_path.resolve().is_relative_to(EXO_TRACING_CACHE_DIR.resolve()):
raise HTTPException(status_code=400, detail=f"Invalid task ID: {task_id}")
return trace_path
def _get_trace_path(self, task_id: str) -> Path:
return EXO_TRACING_CACHE_DIR / f"trace_{task_id}.json"
async def list_traces(self) -> TraceListResponse:
traces: list[TraceListItem] = []
@@ -1834,23 +1883,3 @@ class API:
media_type="application/json",
filename=f"trace_{task_id}.json",
)
async def delete_traces(self, request: DeleteTracesRequest) -> DeleteTracesResponse:
deleted: list[str] = []
not_found: list[str] = []
for task_id in request.task_ids:
trace_path = self._get_trace_path(task_id)
if trace_path.exists():
trace_path.unlink()
deleted.append(task_id)
else:
not_found.append(task_id)
return DeleteTracesResponse(deleted=deleted, not_found=not_found)
async def get_onboarding(self) -> JSONResponse:
return JSONResponse({"completed": ONBOARDING_COMPLETE_FILE.exists()})
async def complete_onboarding(self) -> JSONResponse:
ONBOARDING_COMPLETE_FILE.parent.mkdir(parents=True, exist_ok=True)
ONBOARDING_COMPLETE_FILE.write_text("true")
return JSONResponse({"completed": True})
+139 -52
View File
@@ -1,6 +1,8 @@
from datetime import datetime, timedelta, timezone
from collections.abc import Sequence
from datetime import datetime, timezone
import anyio
from anyio.abc import TaskGroup
from loguru import logger
from exo.master.event_log import DiskEventLog
@@ -11,11 +13,22 @@ from exo.master.placement import (
get_transition_events,
place_instance,
)
from exo.master.process_managers import ProcessManager
from exo.master.process_managers.instance_health import InstanceHealthReconciler
from exo.master.process_managers.meta_instance import MetaInstanceReconciler
from exo.master.process_managers.node_timeout import NodeTimeoutReconciler
from exo.master.reconcile import (
find_unsatisfied_meta_instances,
try_place_for_meta_instance,
)
from exo.shared.apply import apply
from exo.shared.constants import EXO_EVENT_LOG_DIR, EXO_TRACING_ENABLED
from exo.shared.models.model_cards import ModelCard
from exo.shared.types.commands import (
CreateInstance,
CreateMetaInstance,
DeleteInstance,
DeleteMetaInstance,
ForwarderCommand,
ForwarderDownloadCommand,
ImageEdits,
@@ -36,8 +49,10 @@ from exo.shared.types.events import (
InputChunkReceived,
InstanceDeleted,
LocalForwarderEvent,
MetaInstanceCreated,
MetaInstanceDeleted,
MetaInstancePlacementFailed,
NodeGatheredInfo,
NodeTimedOut,
TaskCreated,
TaskDeleted,
TaskStatusUpdated,
@@ -62,7 +77,6 @@ from exo.shared.types.tasks import (
from exo.shared.types.worker.instances import InstanceId
from exo.utils.channels import Receiver, Sender
from exo.utils.event_buffer import MultiSourceBuffer
from exo.utils.task_group import TaskGroup
class Master:
@@ -72,26 +86,29 @@ class Master:
session_id: SessionId,
*,
command_receiver: Receiver[ForwarderCommand],
event_sender: Sender[Event],
local_event_receiver: Receiver[LocalForwarderEvent],
global_event_sender: Sender[GlobalForwarderEvent],
download_command_sender: Sender[ForwarderDownloadCommand],
):
self.state = State()
self._tg: TaskGroup = anyio.create_task_group()
self.node_id = node_id
self.session_id = session_id
self.state = State()
self._tg: TaskGroup = TaskGroup()
self.command_task_mapping: dict[CommandId, TaskId] = {}
self.command_receiver = command_receiver
self.local_event_receiver = local_event_receiver
self.global_event_sender = global_event_sender
self.download_command_sender = download_command_sender
self.event_sender = event_sender
self._system_id = SystemId()
self._multi_buffer = MultiSourceBuffer[SystemId, Event]()
self._event_log = DiskEventLog(EXO_EVENT_LOG_DIR / "master")
self._pending_traces: dict[TaskId, dict[int, list[TraceEventData]]] = {}
self._expected_ranks: dict[TaskId, set[int]] = {}
self._process_managers: Sequence[ProcessManager] = [
InstanceHealthReconciler(),
NodeTimeoutReconciler(),
MetaInstanceReconciler(),
]
async def run(self):
logger.info("Starting Master")
@@ -100,7 +117,7 @@ class Master:
async with self._tg as tg:
tg.start_soon(self._event_processor)
tg.start_soon(self._command_processor)
tg.start_soon(self._plan)
tg.start_soon(self._reconcile)
finally:
self._event_log.close()
self.global_event_sender.close()
@@ -109,7 +126,7 @@ class Master:
async def shutdown(self):
logger.info("Stopping Master")
self._tg.cancel_tasks()
self._tg.cancel_scope.cancel()
async def _command_processor(self) -> None:
with self.command_receiver as commands:
@@ -287,6 +304,86 @@ class Master:
)
)
generated_events.extend(transition_events)
case CreateMetaInstance():
logger.info(
f"Creating MetaInstance for {command.meta_instance.model_id}"
f" (min_nodes={command.meta_instance.min_nodes},"
f" sharding={command.meta_instance.sharding})"
)
# Apply immediately so self.state is fresh across
# the await below and the reconciler won't race.
await self._apply_and_broadcast(
MetaInstanceCreated(meta_instance=command.meta_instance)
)
# Immediate placement attempt for responsiveness
model_card = await ModelCard.load(
command.meta_instance.model_id
)
# Re-check: reconciler may have satisfied it during the await
meta_id = command.meta_instance.meta_instance_id
still_unsatisfied = any(
m.meta_instance_id == meta_id
for m in find_unsatisfied_meta_instances(
self.state.meta_instances,
self.state.instances,
self.state.topology,
)
)
if still_unsatisfied:
result = try_place_for_meta_instance(
command.meta_instance,
model_card,
self.state.topology,
self.state.instances,
self.state.node_memory,
self.state.node_network,
self.state.tasks,
)
generated_events.extend(result.events)
if result.error is not None:
generated_events.append(
MetaInstancePlacementFailed(
meta_instance_id=meta_id,
reason=result.error,
)
)
case DeleteMetaInstance():
backing_count = sum(
1
for inst in self.state.instances.values()
if inst.meta_instance_id == command.meta_instance_id
)
logger.info(
f"Deleting MetaInstance {command.meta_instance_id}"
f" (cascade-deleting {backing_count} backing instance(s))"
)
generated_events.append(
MetaInstanceDeleted(
meta_instance_id=command.meta_instance_id
)
)
# Cascade-delete backing instances atomically,
# cancelling any active tasks first.
for iid, inst in self.state.instances.items():
if inst.meta_instance_id == command.meta_instance_id:
for task in self.state.tasks.values():
if (
task.instance_id == iid
and task.task_status
in (
TaskStatus.Pending,
TaskStatus.Running,
)
):
generated_events.append(
TaskStatusUpdated(
task_status=TaskStatus.Cancelled,
task_id=task.task_id,
)
)
generated_events.append(
InstanceDeleted(instance_id=iid)
)
case PlaceInstance():
placement = place_instance(
command,
@@ -328,22 +425,17 @@ class Master:
task_id=task_id,
)
)
else:
logger.warning(
f"Nonexistent command {command.cancelled_command_id} cancelled"
)
case TaskFinished():
if (
task_id := self.command_task_mapping.pop(
command.finished_command_id, None
generated_events.append(
TaskDeleted(
task_id=self.command_task_mapping[
command.finished_command_id
]
)
) is not None:
generated_events.append(TaskDeleted(task_id=task_id))
else:
logger.warning(
f"Finished command {command.finished_command_id} finished"
)
)
self.command_task_mapping.pop(
command.finished_command_id, None
)
case RequestEventLog():
# We should just be able to send everything, since other buffers will ignore old messages
# rate limit to 1000 at a time
@@ -354,31 +446,32 @@ class Master:
):
await self._send_event(IndexedEvent(idx=i, event=event))
for event in generated_events:
await self.event_sender.send(event)
await self._apply_and_broadcast(event)
except ValueError as e:
logger.opt(exception=e).warning("Error in command processor")
# These plan loops are the cracks showing in our event sourcing architecture - more things could be commands
async def _plan(self) -> None:
async def _apply_and_broadcast(self, event: Event) -> None:
"""Apply event to state, persist to disk, and broadcast to workers.
State is updated synchronously (before any await), so callers can
rely on ``self.state`` reflecting this event immediately after the
call. Python's cooperative scheduling guarantees no interleaving
between the state read and write.
"""
logger.debug(f"Master indexing event: {str(event)[:100]}")
indexed = IndexedEvent(event=event, idx=len(self._event_log))
self.state = apply(self.state, indexed)
event._master_time_stamp = datetime.now(tz=timezone.utc) # pyright: ignore[reportPrivateUsage]
self._event_log.append(event)
await self._send_event(indexed)
async def _reconcile(self) -> None:
while True:
# kill broken instances
connected_node_ids = set(self.state.topology.list_nodes())
for instance_id, instance in self.state.instances.items():
for node_id in instance.shard_assignments.node_to_runner:
if node_id not in connected_node_ids:
await self.event_sender.send(
InstanceDeleted(instance_id=instance_id)
)
break
# time out dead nodes
for node_id, time in self.state.last_seen.items():
now = datetime.now(tz=timezone.utc)
if now - time > timedelta(seconds=30):
logger.info(f"Manually removing node {node_id} due to inactivity")
await self.event_sender.send(NodeTimedOut(node_id=node_id))
await anyio.sleep(10)
for pm in self._process_managers:
events = await pm.reconcile(self.state)
for event in events:
await self._apply_and_broadcast(event)
await anyio.sleep(1)
async def _event_processor(self) -> None:
with self.local_event_receiver as local_events:
@@ -396,16 +489,10 @@ class Master:
await self._handle_traces_collected(event)
continue
logger.debug(f"Master indexing event: {str(event)[:100]}")
indexed = IndexedEvent(event=event, idx=len(self._event_log))
self.state = apply(self.state, indexed)
event._master_time_stamp = datetime.now(tz=timezone.utc) # pyright: ignore[reportPrivateUsage]
if isinstance(event, NodeGatheredInfo):
event.when = str(datetime.now(tz=timezone.utc))
self._event_log.append(event)
await self._send_event(indexed)
await self._apply_and_broadcast(event)
# This function is re-entrant, take care!
async def _send_event(self, event: IndexedEvent):
@@ -437,7 +524,7 @@ class Master:
for trace_data in self._pending_traces[task_id].values():
all_trace_data.extend(trace_data)
await self.event_sender.send(
await self._apply_and_broadcast(
TracesMerged(task_id=task_id, traces=all_trace_data)
)
+6 -22
View File
@@ -69,17 +69,7 @@ def place_instance(
required_nodes: set[NodeId] | None = None,
) -> dict[InstanceId, Instance]:
cycles = topology.get_cycles()
candidate_cycles = filter(lambda it: len(it) >= command.min_nodes, cycles)
nodes_with_exo = set[NodeId]()
for instance in current_instances.values():
nodes_with_exo |= instance.shard_assignments.node_to_runner.keys()
candidate_cycles = [
Cycle(
cycle.node_ids,
has_other_placement=any(node in nodes_with_exo for node in cycle),
)
for cycle in candidate_cycles
]
candidate_cycles = list(filter(lambda it: len(it) >= command.min_nodes, cycles))
# Filter to cycles containing all required nodes (subset matching)
if required_nodes:
@@ -89,7 +79,7 @@ def place_instance(
if required_nodes.issubset(cycle.node_ids)
]
cycles_with_sufficient_memory = filter_cycles_by_memory(
candidate_cycles, node_memory, command.model_card.storage_size * 1.1
candidate_cycles, node_memory, command.model_card.storage_size
)
if len(cycles_with_sufficient_memory) == 0:
raise ValueError("No cycles found with sufficient memory")
@@ -122,11 +112,7 @@ def place_instance(
cycle for cycle in smallest_cycles if topology.is_rdma_cycle(cycle)
]
if command.instance_meta == InstanceMeta.MlxJaccl:
if not smallest_rdma_cycles:
raise ValueError(
"Requested RDMA (MlxJaccl) but no RDMA-connected cycles available"
)
if command.instance_meta == InstanceMeta.MlxJaccl and smallest_rdma_cycles != []:
smallest_cycles = smallest_rdma_cycles
cycles_with_leaf_nodes: list[Cycle] = [
@@ -143,11 +129,6 @@ def place_instance(
),
)
# Single-node: force Pipeline/Ring (Tensor and Jaccl require multi-node)
if len(selected_cycle) == 1:
command.instance_meta = InstanceMeta.MlxRing
command.sharding = Sharding.Pipeline
shard_assignments = get_shard_assignments(
command.model_card, selected_cycle, command.sharding, node_memory
)
@@ -157,6 +138,9 @@ def place_instance(
instance_id = InstanceId()
target_instances = dict(deepcopy(current_instances))
if len(selected_cycle) == 1:
command.instance_meta = InstanceMeta.MlxRing
match command.instance_meta:
case InstanceMeta.MlxJaccl:
# TODO(evan): shard assignments should contain information about ranks, this is ugly
+1 -6
View File
@@ -29,12 +29,7 @@ def filter_cycles_by_memory(
continue
total_mem = sum(
(
node_memory[node_id].ram_available
if cycle.has_other_placement
else node_memory[node_id].ram_total
for node_id in cycle.node_ids
),
(node_memory[node_id].ram_available for node_id in cycle.node_ids),
start=Memory(),
)
if total_mem >= required_memory:
@@ -0,0 +1,12 @@
from collections.abc import Sequence
from typing import Protocol, runtime_checkable
from exo.shared.types.events import Event
from exo.shared.types.state import State
@runtime_checkable
class ProcessManager(Protocol):
"""A reconciliation step that examines state and returns corrective events."""
async def reconcile(self, state: State) -> Sequence[Event]: ...
@@ -0,0 +1,62 @@
from collections.abc import Sequence
from typing import final
from loguru import logger
from exo.master.reconcile import instance_connections_healthy, instance_runners_failed
from exo.shared.types.events import Event, InstanceDeleted, InstanceRetrying
from exo.shared.types.state import State
MAX_INSTANCE_RETRIES = 3
@final
class InstanceHealthReconciler:
"""Delete instances whose network connections are broken or whose runners have all failed."""
async def reconcile(self, state: State) -> Sequence[Event]:
events: list[Event] = []
for instance_id, instance in state.instances.items():
if not instance_connections_healthy(instance, state.topology):
events.append(
InstanceDeleted(
instance_id=instance_id,
failure_error="Network connection lost",
)
)
continue
is_failed, error_message = instance_runners_failed(
instance, state.runners, state.node_identities
)
if is_failed:
# Retry within the same instance if backed by a MetaInstance
mid = instance.meta_instance_id
mi = state.meta_instances.get(mid) if mid else None
if mid and mi and mi.consecutive_failures < MAX_INSTANCE_RETRIES:
logger.info(
f"Instance {instance_id} failed (attempt"
f" {mi.consecutive_failures + 1}/{MAX_INSTANCE_RETRIES}),"
f" retrying: {error_message}"
)
events.append(
InstanceRetrying(
instance_id=instance_id,
meta_instance_id=mid,
failure_error=error_message or "Runner failed",
)
)
else:
if mid and mi:
logger.warning(
f"Instance {instance_id} exceeded retry limit"
f" ({MAX_INSTANCE_RETRIES}), deleting:"
f" {error_message}"
)
events.append(
InstanceDeleted(
instance_id=instance_id,
failure_error=error_message,
)
)
return events
@@ -0,0 +1,92 @@
from collections.abc import Sequence
from typing import final
import anyio
from loguru import logger
from exo.master.reconcile import (
find_unsatisfied_meta_instances,
try_place_for_meta_instance,
)
from exo.shared.models.model_cards import ModelCard
from exo.shared.types.events import Event, InstanceCreated, MetaInstancePlacementFailed
from exo.shared.types.state import State
from exo.shared.types.worker.instances import Instance, InstanceId
MODEL_CARD_LOAD_TIMEOUT_SECONDS = 10
@final
class MetaInstanceReconciler:
"""Place instances for unsatisfied MetaInstances."""
async def reconcile(self, state: State) -> Sequence[Event]:
all_events: list[Event] = []
# Local copy for intermediate tracking — so placement of B
# sees A's instance and doesn't double-place on same resources.
current_instances: dict[InstanceId, Instance] = dict(state.instances)
unsatisfied = find_unsatisfied_meta_instances(
state.meta_instances,
current_instances,
state.topology,
)
for meta_instance in unsatisfied:
try:
with anyio.fail_after(MODEL_CARD_LOAD_TIMEOUT_SECONDS):
model_card = await ModelCard.load(meta_instance.model_id)
except TimeoutError:
logger.warning(
f"ModelCard.load timed out for {meta_instance.model_id}, skipping this cycle"
)
continue
except Exception as exc:
logger.warning(
f"ModelCard.load failed for {meta_instance.model_id}: {exc}"
)
error = f"Failed to load model card: {exc}"
if meta_instance.placement_error != error:
all_events.append(
MetaInstancePlacementFailed(
meta_instance_id=meta_instance.meta_instance_id,
reason=error,
)
)
continue
result = try_place_for_meta_instance(
meta_instance,
model_card,
state.topology,
current_instances,
state.node_memory,
state.node_network,
state.tasks,
)
# Update local instance map so next placement sees this one
for event in result.events:
if isinstance(event, InstanceCreated):
logger.info(
f"MetaInstance reconciler placed instance"
f" {event.instance.instance_id} for"
f" {meta_instance.model_id}"
)
current_instances[event.instance.instance_id] = event.instance
all_events.extend(result.events)
# Emit placement failure if error differs from what's already in state
if (
result.error is not None
and meta_instance.placement_error != result.error
):
logger.warning(
f"MetaInstance placement failed for"
f" {meta_instance.model_id}: {result.error}"
)
all_events.append(
MetaInstancePlacementFailed(
meta_instance_id=meta_instance.meta_instance_id,
reason=result.error,
)
)
return all_events
@@ -0,0 +1,27 @@
from collections.abc import Sequence
from datetime import datetime, timedelta, timezone
from typing import final
from loguru import logger
from exo.shared.types.events import Event, NodeTimedOut
from exo.shared.types.state import State
_DEFAULT_TIMEOUT = timedelta(seconds=30)
@final
class NodeTimeoutReconciler:
"""Time out nodes that haven't been seen recently."""
def __init__(self, timeout: timedelta = _DEFAULT_TIMEOUT) -> None:
self.timeout = timeout
async def reconcile(self, state: State) -> Sequence[Event]:
now = datetime.now(tz=timezone.utc)
events: list[Event] = []
for node_id, last_seen in state.last_seen.items():
if now - last_seen > self.timeout:
logger.info(f"Removing node {node_id} due to inactivity")
events.append(NodeTimedOut(node_id=node_id))
return events
+244
View File
@@ -0,0 +1,244 @@
from collections.abc import Mapping, Sequence
from typing import NamedTuple
from loguru import logger
from exo.master.placement import get_transition_events, place_instance
from exo.shared.models.model_cards import ModelCard
from exo.shared.topology import Topology
from exo.shared.types.commands import PlaceInstance
from exo.shared.types.common import MetaInstanceId, NodeId
from exo.shared.types.events import Event
from exo.shared.types.meta_instance import MetaInstance
from exo.shared.types.profiling import MemoryUsage, NodeIdentity, NodeNetworkInfo
from exo.shared.types.tasks import Task, TaskId
from exo.shared.types.topology import RDMAConnection, SocketConnection
from exo.shared.types.worker.instances import (
BaseInstance,
Instance,
InstanceId,
MlxJacclInstance,
MlxRingInstance,
)
from exo.shared.types.worker.runners import (
RunnerFailed,
RunnerId,
RunnerShutdown,
RunnerStatus,
)
class PlacementResult(NamedTuple):
"""Result of a placement attempt: events to apply and optional error reason."""
events: Sequence[Event]
error: str | None
def _get_ring_order(instance: BaseInstance) -> list[NodeId]:
"""Reconstruct ring order from shard device_rank."""
node_ranks: list[tuple[NodeId, int]] = []
for node_id, runner_id in instance.shard_assignments.node_to_runner.items():
shard = instance.shard_assignments.runner_to_shard[runner_id]
node_ranks.append((node_id, shard.device_rank))
node_ranks.sort(key=lambda x: x[1])
return [node_id for node_id, _ in node_ranks]
def _ring_connections_healthy(instance: MlxRingInstance, topology: Topology) -> bool:
"""Check that the specific IPs used by a ring instance still exist in the topology."""
ring = _get_ring_order(instance)
n = len(ring)
for node in ring:
hosts = instance.hosts_by_node[node]
for idx in range(n):
host = hosts[idx]
if host.ip in ("0.0.0.0", "198.51.100.1"):
continue # self or placeholder
# Real connection: node → ring[idx]. Check specific IP.
connections = topology.get_all_connections_between(node, ring[idx])
if not any(
isinstance(c, SocketConnection)
and c.sink_multiaddr.ip_address == host.ip
for c in connections
):
return False
return True
def _jaccl_connections_healthy(instance: MlxJacclInstance, topology: Topology) -> bool:
"""Check that the specific RDMA interfaces used by a JACCL instance still exist."""
ring = _get_ring_order(instance)
n = len(ring)
for i in range(n):
for j in range(n):
iface = instance.jaccl_devices[i][j]
if iface is None:
continue
connections = topology.get_all_connections_between(ring[i], ring[j])
if not any(
isinstance(c, RDMAConnection) and c.source_rdma_iface == iface
for c in connections
):
return False
return True
def instance_connections_healthy(instance: Instance, topology: Topology) -> bool:
"""Check that an instance's nodes and specific connections are still in the topology."""
instance_nodes = set(instance.shard_assignments.node_to_runner.keys())
if not all(topology.contains_node(n) for n in instance_nodes):
return False
if len(instance_nodes) <= 1:
return True
match instance:
case MlxRingInstance():
return _ring_connections_healthy(instance, topology)
case MlxJacclInstance():
return _jaccl_connections_healthy(instance, topology)
def instance_runners_failed(
instance: Instance,
runners: Mapping[RunnerId, RunnerStatus],
node_identities: Mapping[NodeId, NodeIdentity],
) -> tuple[bool, str | None]:
"""Check if an instance's runners have all reached terminal failure states.
Returns ``(True, error_message)`` when ALL runners are terminal
(``RunnerFailed`` or ``RunnerShutdown``) and at least one is ``RunnerFailed``.
Returns ``(False, None)`` when runners are still active, haven't reported
yet, or all gracefully shut down (no ``RunnerFailed``).
"""
instance_runner_ids = set(instance.shard_assignments.node_to_runner.values())
if not instance_runner_ids:
return False, None
# Build reverse mapping: runner_id -> node_id
runner_to_node: dict[RunnerId, NodeId] = {
runner_id: node_id
for node_id, runner_id in instance.shard_assignments.node_to_runner.items()
}
has_any_failed = False
error_messages: list[str] = []
for runner_id in instance_runner_ids:
status = runners.get(runner_id)
if status is None:
# Runner hasn't reported yet — instance is still starting
return False, None
if isinstance(status, RunnerFailed):
has_any_failed = True
if status.error_message:
node_id = runner_to_node.get(runner_id)
name = (
node_identities[node_id].friendly_name
if node_id and node_id in node_identities
else node_id or "unknown"
)
error_messages.append(f"{name}: {status.error_message}")
elif isinstance(status, RunnerShutdown):
pass # Terminal but not a failure indicator on its own
else:
# Runner is still active (connecting, loading, running, etc.)
return False, None
if has_any_failed:
return True, "; ".join(error_messages) if error_messages else "Runner failed"
# All runners are Shutdown but none Failed — graceful shutdown, not a failure
return False, None
def instance_satisfies_meta_instance(
meta_instance: MetaInstance,
instance: Instance,
) -> bool:
"""Check if a single instance satisfies a meta-instance's constraints.
This is a pure constraint check (model, min_nodes, node_ids).
Use ``instance_connections_healthy`` separately for topology health.
"""
if instance.shard_assignments.model_id != meta_instance.model_id:
return False
instance_nodes = set(instance.shard_assignments.node_to_runner.keys())
if len(instance_nodes) < meta_instance.min_nodes:
return False
return meta_instance.node_ids is None or set(meta_instance.node_ids).issubset(
instance_nodes
)
def find_unsatisfied_meta_instances(
meta_instances: Mapping[MetaInstanceId, MetaInstance],
instances: Mapping[InstanceId, Instance],
topology: Topology,
) -> Sequence[MetaInstance]:
"""Return meta-instances that have no healthy backing instance."""
unsatisfied: list[MetaInstance] = []
for meta_id, meta_instance in meta_instances.items():
has_healthy_backing = any(
instance.meta_instance_id == meta_id
and instance_connections_healthy(instance, topology)
for instance in instances.values()
)
if not has_healthy_backing:
unsatisfied.append(meta_instance)
return unsatisfied
def try_place_for_meta_instance(
meta_instance: MetaInstance,
model_card: ModelCard,
topology: Topology,
current_instances: Mapping[InstanceId, Instance],
node_memory: Mapping[NodeId, MemoryUsage],
node_network: Mapping[NodeId, NodeNetworkInfo],
tasks: Mapping[TaskId, Task] | None = None,
) -> PlacementResult:
"""Try to place an instance satisfying the meta-instance constraints.
Returns a :class:`PlacementResult` with events on success, or an error
reason on failure.
"""
command = PlaceInstance(
model_card=model_card,
sharding=meta_instance.sharding,
instance_meta=meta_instance.instance_meta,
min_nodes=meta_instance.min_nodes,
)
try:
target_instances = place_instance(
command,
topology,
current_instances,
node_memory,
node_network,
required_nodes=(
set(meta_instance.node_ids) if meta_instance.node_ids else None
),
)
# Tag the new instance with meta_instance_id
new_instance_ids = set(target_instances.keys()) - set(current_instances.keys())
if new_instance_ids:
new_id = next(iter(new_instance_ids))
target_instances[new_id] = target_instances[new_id].model_copy(
update={"meta_instance_id": meta_instance.meta_instance_id}
)
return PlacementResult(
events=list(
get_transition_events(current_instances, target_instances, tasks or {})
),
error=None,
)
except ValueError as e:
logger.debug(
f"MetaInstance placement not possible for {meta_instance.model_id}: {e}"
)
return PlacementResult(events=[], error=str(e))
@@ -1,77 +0,0 @@
# pyright: reportUnusedFunction=false, reportAny=false
from typing import Any
from unittest.mock import AsyncMock, MagicMock
from fastapi import FastAPI
from fastapi.testclient import TestClient
from exo.shared.types.common import CommandId
def _make_api() -> Any:
"""Create a minimal API instance with cancel route and error handler."""
from exo.master.api import API
app = FastAPI()
api = object.__new__(API)
api.app = app
api._text_generation_queues = {} # pyright: ignore[reportPrivateUsage]
api._image_generation_queues = {} # pyright: ignore[reportPrivateUsage]
api._send = AsyncMock() # pyright: ignore[reportPrivateUsage]
api._setup_exception_handlers() # pyright: ignore[reportPrivateUsage]
app.post("/v1/cancel/{command_id}")(api.cancel_command)
return api
def test_cancel_nonexistent_command_returns_404() -> None:
"""Cancel for an unknown command_id returns 404 in OpenAI error format."""
api = _make_api()
client = TestClient(api.app)
response = client.post("/v1/cancel/nonexistent-id")
assert response.status_code == 404
data: dict[str, Any] = response.json()
assert "error" in data
assert data["error"]["message"] == "Command not found or already completed"
assert data["error"]["type"] == "Not Found"
assert data["error"]["code"] == 404
def test_cancel_active_text_generation() -> None:
"""Cancel an active text generation command: returns 200, sender.close() called."""
api = _make_api()
client = TestClient(api.app)
cid = CommandId("text-cmd-123")
sender = MagicMock()
api._text_generation_queues[cid] = sender
response = client.post(f"/v1/cancel/{cid}")
assert response.status_code == 200
data: dict[str, Any] = response.json()
assert data["message"] == "Command cancelled."
assert data["command_id"] == str(cid)
sender.close.assert_called_once()
api._send.assert_called_once()
task_cancelled = api._send.call_args[0][0]
assert task_cancelled.cancelled_command_id == cid
def test_cancel_active_image_generation() -> None:
"""Cancel an active image generation command: returns 200, sender.close() called."""
api = _make_api()
client = TestClient(api.app)
cid = CommandId("img-cmd-456")
sender = MagicMock()
api._image_generation_queues[cid] = sender
response = client.post(f"/v1/cancel/{cid}")
assert response.status_code == 200
data: dict[str, Any] = response.json()
assert data["message"] == "Command cancelled."
assert data["command_id"] == str(cid)
sender.close.assert_called_once()
api._send.assert_called_once()
task_cancelled = api._send.call_args[0][0]
assert task_cancelled.cancelled_command_id == cid
-20
View File
@@ -17,7 +17,6 @@ from exo.shared.types.commands import (
)
from exo.shared.types.common import ModelId, NodeId, SessionId, SystemId
from exo.shared.types.events import (
Event,
GlobalForwarderEvent,
IndexedEvent,
InstanceCreated,
@@ -51,22 +50,6 @@ async def test_master():
command_sender, co_receiver = channel[ForwarderCommand]()
local_event_sender, le_receiver = channel[LocalForwarderEvent]()
fcds, _fcdr = channel[ForwarderDownloadCommand]()
ev_send, ev_recv = channel[Event]()
async def mock_event_router():
idx = 0
sid = SystemId()
with ev_recv as master_events:
async for event in master_events:
await local_event_sender.send(
LocalForwarderEvent(
origin=sid,
origin_idx=idx,
session=session_id,
event=event,
)
)
idx += 1
all_events: list[IndexedEvent] = []
@@ -84,7 +67,6 @@ async def test_master():
master = Master(
node_id,
session_id,
event_sender=ev_send,
global_event_sender=ge_sender,
local_event_receiver=le_receiver,
command_receiver=co_receiver,
@@ -93,7 +75,6 @@ async def test_master():
logger.info("run the master")
async with anyio.create_task_group() as tg:
tg.start_soon(master.run)
tg.start_soon(mock_event_router)
# inject a NodeGatheredInfo event
logger.info("inject a NodeGatheredInfo event")
@@ -216,5 +197,4 @@ async def test_master():
input=[InputMessage(role="user", content="Hello, how are you?")],
)
ev_send.close()
await master.shutdown()
@@ -0,0 +1,778 @@
"""Edge-case and regression tests for MetaInstance lifecycle, concurrent operations, and error handling."""
import pytest
from exo.master.process_managers.instance_health import (
MAX_INSTANCE_RETRIES,
InstanceHealthReconciler,
)
from exo.master.process_managers.meta_instance import MetaInstanceReconciler
from exo.master.reconcile import (
find_unsatisfied_meta_instances,
instance_connections_healthy,
instance_runners_failed,
instance_satisfies_meta_instance,
)
from exo.shared.apply import apply
from exo.shared.models.model_cards import ModelCard, ModelId, ModelTask
from exo.shared.topology import Topology
from exo.shared.types.common import Host, MetaInstanceId, NodeId
from exo.shared.types.events import (
IndexedEvent,
InstanceCreated,
InstanceDeleted,
InstanceRetrying,
MetaInstanceCreated,
MetaInstanceDeleted,
MetaInstancePlacementFailed,
TaskStatusUpdated,
)
from exo.shared.types.memory import Memory
from exo.shared.types.meta_instance import MetaInstance
from exo.shared.types.multiaddr import Multiaddr
from exo.shared.types.profiling import NodeIdentity
from exo.shared.types.state import State
from exo.shared.types.tasks import LoadModel, TaskId, TaskStatus
from exo.shared.types.topology import Connection, SocketConnection
from exo.shared.types.worker.instances import (
InstanceId,
MlxRingInstance,
)
from exo.shared.types.worker.runners import (
RunnerFailed,
RunnerId,
RunnerReady,
ShardAssignments,
)
from exo.shared.types.worker.shards import PipelineShardMetadata
# --- Helpers (copied from test_reconcile.py for independence) ---
def _model_card(model_id: str = "test-org/test-model") -> ModelCard:
return ModelCard(
model_id=ModelId(model_id),
storage_size=Memory.from_kb(1000),
n_layers=10,
hidden_size=30,
supports_tensor=True,
tasks=[ModelTask.TextGeneration],
)
def _topology(*node_ids: str, connect: bool = True) -> Topology:
t = Topology()
nodes = [NodeId(n) for n in node_ids]
for n in nodes:
t.add_node(n)
if connect and len(nodes) > 1:
for i in range(len(nodes)):
j = (i + 1) % len(nodes)
t.add_connection(
Connection(
source=nodes[i],
sink=nodes[j],
edge=SocketConnection(
sink_multiaddr=Multiaddr(
address=f"/ip4/10.0.0.{j + 1}/tcp/50000"
)
),
)
)
t.add_connection(
Connection(
source=nodes[j],
sink=nodes[i],
edge=SocketConnection(
sink_multiaddr=Multiaddr(
address=f"/ip4/10.0.0.{i + 1}/tcp/50000"
)
),
)
)
return t
def _meta_instance(
model_id: str = "test-org/test-model",
*,
min_nodes: int = 1,
node_ids: list[NodeId] | None = None,
meta_instance_id: MetaInstanceId | None = None,
consecutive_failures: int = 0,
last_failure_error: str | None = None,
placement_error: str | None = None,
) -> MetaInstance:
return MetaInstance(
meta_instance_id=meta_instance_id or MetaInstanceId(),
model_id=ModelId(model_id),
min_nodes=min_nodes,
node_ids=node_ids,
consecutive_failures=consecutive_failures,
last_failure_error=last_failure_error,
placement_error=placement_error,
)
def _instance(
model_id: str = "test-org/test-model",
node_ids: list[str] | None = None,
instance_id: InstanceId | None = None,
meta_instance_id: MetaInstanceId | None = None,
) -> tuple[InstanceId, MlxRingInstance]:
iid = instance_id or InstanceId()
nodes = node_ids or ["node-a"]
n = len(nodes)
mc = _model_card(model_id)
ephemeral_port = 50000
node_to_runner = {NodeId(nd): RunnerId() for nd in nodes}
runner_to_shard = {
runner_id: PipelineShardMetadata(
model_card=mc,
device_rank=i,
world_size=n,
start_layer=0,
end_layer=mc.n_layers,
n_layers=mc.n_layers,
)
for i, runner_id in enumerate(node_to_runner.values())
}
hosts_by_node: dict[NodeId, list[Host]] = {}
for r, node_str in enumerate(nodes):
hosts: list[Host] = []
for idx in range(n):
if idx == r:
hosts.append(Host(ip="0.0.0.0", port=ephemeral_port))
elif n > 1 and idx in ((r - 1) % n, (r + 1) % n):
hosts.append(Host(ip=f"10.0.0.{idx + 1}", port=ephemeral_port))
else:
hosts.append(Host(ip="198.51.100.1", port=0))
hosts_by_node[NodeId(node_str)] = hosts
return iid, MlxRingInstance(
instance_id=iid,
shard_assignments=ShardAssignments(
model_id=ModelId(model_id),
runner_to_shard=runner_to_shard,
node_to_runner=node_to_runner,
),
hosts_by_node=hosts_by_node,
ephemeral_port=ephemeral_port,
meta_instance_id=meta_instance_id,
)
# =============================================================================
# 1. MetaInstance lifecycle edge cases
# =============================================================================
def test_meta_instance_model_is_frozen():
"""MetaInstance should be immutable (frozen model)."""
meta = _meta_instance()
try:
meta.model_id = ModelId("something-else")
raise AssertionError("Should have raised")
except Exception:
pass # Expected — frozen model
def test_meta_instance_created_then_deleted_roundtrip():
"""Create and delete a MetaInstance through apply — state should be clean."""
state = State()
meta = _meta_instance()
state = apply(
state, IndexedEvent(idx=0, event=MetaInstanceCreated(meta_instance=meta))
)
assert meta.meta_instance_id in state.meta_instances
state = apply(
state,
IndexedEvent(
idx=1, event=MetaInstanceDeleted(meta_instance_id=meta.meta_instance_id)
),
)
assert meta.meta_instance_id not in state.meta_instances
assert len(state.meta_instances) == 0
def test_delete_nonexistent_meta_instance_is_safe():
"""Deleting a MetaInstance that doesn't exist should not crash."""
state = State()
event = MetaInstanceDeleted(meta_instance_id=MetaInstanceId("nonexistent"))
new_state = apply(state, IndexedEvent(idx=0, event=event))
assert len(new_state.meta_instances) == 0
def test_placement_failed_for_nonexistent_meta_instance_is_safe():
"""MetaInstancePlacementFailed for unknown ID should not crash."""
state = State()
event = MetaInstancePlacementFailed(
meta_instance_id=MetaInstanceId("nonexistent"),
reason="test",
)
new_state = apply(state, IndexedEvent(idx=0, event=event))
assert len(new_state.meta_instances) == 0
def test_multiple_meta_instances_for_same_model():
"""Multiple MetaInstances for the same model are tracked independently."""
state = State()
meta_a = _meta_instance("test-org/model-x")
meta_b = _meta_instance("test-org/model-x")
state = apply(
state, IndexedEvent(idx=0, event=MetaInstanceCreated(meta_instance=meta_a))
)
state = apply(
state, IndexedEvent(idx=1, event=MetaInstanceCreated(meta_instance=meta_b))
)
assert len(state.meta_instances) == 2
assert meta_a.meta_instance_id in state.meta_instances
assert meta_b.meta_instance_id in state.meta_instances
# =============================================================================
# 2. Retry logic edge cases
# =============================================================================
def test_retry_counter_resets_on_successful_instance_creation():
"""When a new instance is created for a meta-instance, failures should reset."""
meta = _meta_instance(consecutive_failures=2, last_failure_error="old")
_, inst = _instance(node_ids=["node-a"], meta_instance_id=meta.meta_instance_id)
state = State(meta_instances={meta.meta_instance_id: meta})
state = apply(state, IndexedEvent(idx=0, event=InstanceCreated(instance=inst)))
mi = state.meta_instances[meta.meta_instance_id]
assert mi.consecutive_failures == 0
# last_failure_error is preserved (for UI display)
assert mi.last_failure_error == "old"
async def test_retry_count_increments_through_full_cycle():
"""Walk through MAX_INSTANCE_RETRIES worth of retries, then verify delete."""
meta = _meta_instance()
iid, inst = _instance(node_ids=["node-a"], meta_instance_id=meta.meta_instance_id)
topology = _topology("node-a")
state = State(
meta_instances={meta.meta_instance_id: meta},
instances={iid: inst},
topology=topology,
)
runner_ids = list(inst.shard_assignments.node_to_runner.values())
for idx, i in enumerate(range(MAX_INSTANCE_RETRIES)):
# Simulate runners failing
state_with_runners = state.model_copy(
update={"runners": {runner_ids[0]: RunnerFailed(error_message=f"fail-{i}")}}
)
reconciler = InstanceHealthReconciler()
events = await reconciler.reconcile(state_with_runners)
assert len(events) == 1
assert isinstance(events[0], InstanceRetrying), f"iteration {i}"
state = apply(state, IndexedEvent(idx=idx, event=events[0]))
# After MAX_INSTANCE_RETRIES retries, failure counter should be at max
mi = state.meta_instances[meta.meta_instance_id]
assert mi.consecutive_failures == MAX_INSTANCE_RETRIES
# Next failure should result in deletion
state_with_runners = state.model_copy(
update={"runners": {runner_ids[0]: RunnerFailed(error_message="final")}}
)
reconciler = InstanceHealthReconciler()
events = await reconciler.reconcile(state_with_runners)
assert len(events) == 1
assert isinstance(events[0], InstanceDeleted)
async def test_health_reconciler_respects_exact_limit():
"""At exactly MAX_INSTANCE_RETRIES, reconciler should delete, not retry."""
meta = _meta_instance(consecutive_failures=MAX_INSTANCE_RETRIES)
iid, inst = _instance(node_ids=["node-a"], meta_instance_id=meta.meta_instance_id)
runner_ids = list(inst.shard_assignments.node_to_runner.values())
state = State(
meta_instances={meta.meta_instance_id: meta},
instances={iid: inst},
runners={runner_ids[0]: RunnerFailed(error_message="OOM")},
topology=_topology("node-a"),
)
reconciler = InstanceHealthReconciler()
events = await reconciler.reconcile(state)
assert len(events) == 1
assert isinstance(events[0], InstanceDeleted)
async def test_health_reconciler_at_limit_minus_one_retries():
"""At MAX_INSTANCE_RETRIES - 1, reconciler should still retry."""
meta = _meta_instance(consecutive_failures=MAX_INSTANCE_RETRIES - 1)
iid, inst = _instance(node_ids=["node-a"], meta_instance_id=meta.meta_instance_id)
runner_ids = list(inst.shard_assignments.node_to_runner.values())
state = State(
meta_instances={meta.meta_instance_id: meta},
instances={iid: inst},
runners={runner_ids[0]: RunnerFailed(error_message="OOM")},
topology=_topology("node-a"),
)
reconciler = InstanceHealthReconciler()
events = await reconciler.reconcile(state)
assert len(events) == 1
assert isinstance(events[0], InstanceRetrying)
# =============================================================================
# 3. Error handling edge cases
# =============================================================================
def test_runners_failed_with_empty_error_message():
"""RunnerFailed with empty error_message should still report as failed."""
_, inst = _instance(node_ids=["node-a"])
runners = {
rid: RunnerFailed(error_message="")
for rid in inst.shard_assignments.node_to_runner.values()
}
is_failed, error = instance_runners_failed(inst, runners, {})
assert is_failed is True
# Empty error message means we get the fallback
assert error == "Runner failed"
def test_runners_failed_with_none_error_message():
"""RunnerFailed with None error_message should still report as failed."""
_, inst = _instance(node_ids=["node-a"])
runners = {
rid: RunnerFailed(error_message=None)
for rid in inst.shard_assignments.node_to_runner.values()
}
is_failed, error = instance_runners_failed(inst, runners, {})
assert is_failed is True
assert error == "Runner failed"
def test_runners_failed_collects_all_error_messages():
"""With multiple failed runners, all error messages should be collected."""
_, inst = _instance(node_ids=["node-a", "node-b", "node-c"])
runner_ids = list(inst.shard_assignments.node_to_runner.values())
runners = {
runner_ids[0]: RunnerFailed(error_message="OOM on GPU 0"),
runner_ids[1]: RunnerFailed(error_message="OOM on GPU 1"),
runner_ids[2]: RunnerFailed(error_message="OOM on GPU 2"),
}
is_failed, error = instance_runners_failed(inst, runners, {})
assert is_failed is True
assert error is not None
assert "OOM on GPU 0" in error
assert "OOM on GPU 1" in error
assert "OOM on GPU 2" in error
def test_runners_failed_includes_friendly_name():
"""Error messages should include node friendly names when available."""
_, inst = _instance(node_ids=["node-a"])
node_id = NodeId("node-a")
runner_ids = list(inst.shard_assignments.node_to_runner.values())
runners = {runner_ids[0]: RunnerFailed(error_message="OOM")}
identities = {node_id: NodeIdentity(friendly_name="My Mac Studio")}
is_failed, error = instance_runners_failed(inst, runners, identities)
assert is_failed is True
assert error is not None
assert "My Mac Studio" in error
def test_instance_retrying_for_missing_instance_is_safe():
"""InstanceRetrying for an instance not in state should not crash.
NOTE: When the instance is missing, the handler returns early WITHOUT
incrementing the MetaInstance failure counter. This means stale retry
events for already-deleted instances are silently dropped. This is
acceptable since the InstanceDeleted handler already increments failures.
"""
meta = _meta_instance()
state = State(meta_instances={meta.meta_instance_id: meta})
event = InstanceRetrying(
instance_id=InstanceId("nonexistent"),
meta_instance_id=meta.meta_instance_id,
failure_error="crash",
)
new_state = apply(state, IndexedEvent(idx=0, event=event))
# Does not crash, but failure count is NOT incremented (early return)
mi = new_state.meta_instances[meta.meta_instance_id]
assert mi.consecutive_failures == 0
# =============================================================================
# 4. Backward compatibility
# =============================================================================
def test_instance_without_meta_instance_id_works():
"""Instances created without meta_instance_id should still function normally."""
_, inst = _instance(node_ids=["node-a"])
assert inst.meta_instance_id is None
topology = _topology("node-a")
assert instance_connections_healthy(inst, topology) is True
def test_instance_deleted_without_meta_does_not_affect_meta_instances():
"""Deleting an instance without meta_instance_id should not affect meta_instances."""
meta = _meta_instance()
iid, inst = _instance(node_ids=["node-a"]) # no meta_instance_id
state = State(
meta_instances={meta.meta_instance_id: meta},
instances={iid: inst},
)
event = InstanceDeleted(instance_id=iid, failure_error="crash")
new_state = apply(state, IndexedEvent(idx=0, event=event))
mi = new_state.meta_instances[meta.meta_instance_id]
assert mi.consecutive_failures == 0 # unchanged
def test_satisfies_ignores_meta_instance_id_binding():
"""instance_satisfies_meta_instance checks constraints only, not binding."""
meta = _meta_instance()
_, inst = _instance(node_ids=["node-a"]) # no meta_instance_id set
# Should match on constraints (model, min_nodes) regardless of binding
assert instance_satisfies_meta_instance(meta, inst) is True
def test_find_unsatisfied_uses_binding_not_constraints():
"""find_unsatisfied checks meta_instance_id binding, not just constraint matching."""
meta = _meta_instance()
# Instance matches constraints but is NOT bound to this meta_instance
iid, inst = _instance(node_ids=["node-a"])
topology = _topology("node-a")
result = find_unsatisfied_meta_instances(
{meta.meta_instance_id: meta}, {iid: inst}, topology
)
# Should be unsatisfied because instance.meta_instance_id != meta.meta_instance_id
assert list(result) == [meta]
# =============================================================================
# 5. Concurrent / multi-instance scenarios
# =============================================================================
async def test_health_reconciler_handles_multiple_failing_instances():
"""Multiple instances failing simultaneously should each get their own event."""
meta_a = _meta_instance()
meta_b = _meta_instance()
iid_a, inst_a = _instance(
node_ids=["node-a"], meta_instance_id=meta_a.meta_instance_id
)
iid_b, inst_b = _instance(
node_ids=["node-b"], meta_instance_id=meta_b.meta_instance_id
)
runner_ids_a = list(inst_a.shard_assignments.node_to_runner.values())
runner_ids_b = list(inst_b.shard_assignments.node_to_runner.values())
state = State(
meta_instances={
meta_a.meta_instance_id: meta_a,
meta_b.meta_instance_id: meta_b,
},
instances={iid_a: inst_a, iid_b: inst_b},
runners={
runner_ids_a[0]: RunnerFailed(error_message="OOM"),
runner_ids_b[0]: RunnerFailed(error_message="OOM"),
},
topology=_topology("node-a", "node-b"),
)
reconciler = InstanceHealthReconciler()
events = await reconciler.reconcile(state)
assert len(events) == 2
# Both should be InstanceRetrying since failures < MAX
assert all(isinstance(e, InstanceRetrying) for e in events)
instance_ids = {e.instance_id for e in events} # type: ignore[union-attr]
assert instance_ids == {iid_a, iid_b}
async def test_health_reconciler_mixed_healthy_and_failing():
"""Only failing instances should produce events; healthy ones should not."""
meta_healthy = _meta_instance()
meta_failing = _meta_instance()
iid_h, inst_h = _instance(
node_ids=["node-a"], meta_instance_id=meta_healthy.meta_instance_id
)
iid_f, inst_f = _instance(
node_ids=["node-b"], meta_instance_id=meta_failing.meta_instance_id
)
runner_ids_h = list(inst_h.shard_assignments.node_to_runner.values())
runner_ids_f = list(inst_f.shard_assignments.node_to_runner.values())
state = State(
meta_instances={
meta_healthy.meta_instance_id: meta_healthy,
meta_failing.meta_instance_id: meta_failing,
},
instances={iid_h: inst_h, iid_f: inst_f},
runners={
runner_ids_h[0]: RunnerReady(),
runner_ids_f[0]: RunnerFailed(error_message="crash"),
},
topology=_topology("node-a", "node-b"),
)
reconciler = InstanceHealthReconciler()
events = await reconciler.reconcile(state)
assert len(events) == 1
assert isinstance(events[0], InstanceRetrying)
assert events[0].instance_id == iid_f
async def test_meta_instance_reconciler_empty_state():
"""MetaInstanceReconciler with no meta_instances should produce no events."""
state = State()
reconciler = MetaInstanceReconciler()
events = await reconciler.reconcile(state)
assert len(events) == 0
# =============================================================================
# 6. Placement error tracking
# =============================================================================
def test_placement_failed_sets_error():
"""MetaInstancePlacementFailed should set placement_error on the MetaInstance."""
meta = _meta_instance()
state = State(meta_instances={meta.meta_instance_id: meta})
event = MetaInstancePlacementFailed(
meta_instance_id=meta.meta_instance_id,
reason="Not enough memory",
)
new_state = apply(state, IndexedEvent(idx=0, event=event))
mi = new_state.meta_instances[meta.meta_instance_id]
assert mi.placement_error == "Not enough memory"
def test_instance_created_clears_placement_error():
"""InstanceCreated should clear placement_error on the MetaInstance."""
meta = _meta_instance(placement_error="Not enough memory")
_, inst = _instance(node_ids=["node-a"], meta_instance_id=meta.meta_instance_id)
state = State(meta_instances={meta.meta_instance_id: meta})
state = apply(state, IndexedEvent(idx=0, event=InstanceCreated(instance=inst)))
mi = state.meta_instances[meta.meta_instance_id]
assert mi.placement_error is None
def test_placement_error_does_not_increment_failures():
"""Placement failures should only set placement_error, not increment consecutive_failures."""
meta = _meta_instance()
state = State(meta_instances={meta.meta_instance_id: meta})
event = MetaInstancePlacementFailed(
meta_instance_id=meta.meta_instance_id,
reason="No resources",
)
new_state = apply(state, IndexedEvent(idx=0, event=event))
mi = new_state.meta_instances[meta.meta_instance_id]
assert mi.consecutive_failures == 0
assert mi.placement_error == "No resources"
# =============================================================================
# 7. State serialization roundtrip
# =============================================================================
def test_state_with_meta_instances_serializes():
"""State with meta_instances should serialize and deserialize correctly."""
meta = _meta_instance(consecutive_failures=2, last_failure_error="test")
iid, inst = _instance(node_ids=["node-a"], meta_instance_id=meta.meta_instance_id)
state = State(
meta_instances={meta.meta_instance_id: meta},
instances={iid: inst},
)
json_str = state.model_dump_json()
restored = State.model_validate_json(json_str)
assert meta.meta_instance_id in restored.meta_instances
mi = restored.meta_instances[meta.meta_instance_id]
assert mi.model_id == meta.model_id
assert mi.consecutive_failures == 2
assert mi.last_failure_error == "test"
assert iid in restored.instances
assert restored.instances[iid].meta_instance_id == meta.meta_instance_id
# =============================================================================
# 8. MetaInstanceReconciler error handling
# =============================================================================
async def test_meta_instance_reconciler_model_load_error_emits_placement_failed(
monkeypatch: "pytest.MonkeyPatch",
):
"""When ModelCard.load raises, reconciler emits MetaInstancePlacementFailed."""
import exo.master.process_managers.meta_instance as mi_mod
meta = _meta_instance()
topo = _topology("node-a")
state = State(
meta_instances={meta.meta_instance_id: meta},
topology=topo,
)
async def _failing_load(_model_id: ModelId) -> ModelCard:
raise RuntimeError("Network error")
monkeypatch.setattr(
mi_mod, "ModelCard", type("MC", (), {"load": staticmethod(_failing_load)})
)
reconciler = MetaInstanceReconciler()
events = await reconciler.reconcile(state)
placement_failed = [e for e in events if isinstance(e, MetaInstancePlacementFailed)]
assert len(placement_failed) == 1
assert "Failed to load model card" in placement_failed[0].reason
assert meta.meta_instance_id == placement_failed[0].meta_instance_id
async def test_meta_instance_reconciler_model_load_error_skips_dedup(
monkeypatch: "pytest.MonkeyPatch",
):
"""When ModelCard.load error matches existing placement_error, no duplicate event."""
import exo.master.process_managers.meta_instance as mi_mod
meta = _meta_instance(placement_error="Failed to load model card: Network error")
topo = _topology("node-a")
state = State(
meta_instances={meta.meta_instance_id: meta},
topology=topo,
)
async def _failing_load(_model_id: ModelId) -> ModelCard:
raise RuntimeError("Network error")
monkeypatch.setattr(
mi_mod, "ModelCard", type("MC", (), {"load": staticmethod(_failing_load)})
)
reconciler = MetaInstanceReconciler()
events = await reconciler.reconcile(state)
# Error matches existing placement_error, so no duplicate event emitted
assert len(events) == 0
async def test_meta_instance_reconciler_continues_after_error(
monkeypatch: "pytest.MonkeyPatch",
):
"""Reconciler should continue to next meta-instance after one fails to load."""
import exo.master.process_managers.meta_instance as mi_mod
meta_a = _meta_instance(model_id="org/model-a")
meta_b = _meta_instance(model_id="org/model-b")
topo = _topology("node-a")
state = State(
meta_instances={
meta_a.meta_instance_id: meta_a,
meta_b.meta_instance_id: meta_b,
},
topology=topo,
)
call_count = 0
async def _load_second_fails(model_id: ModelId) -> ModelCard:
nonlocal call_count
call_count += 1
raise RuntimeError(f"Cannot load {model_id}")
monkeypatch.setattr(
mi_mod, "ModelCard", type("MC", (), {"load": staticmethod(_load_second_fails)})
)
reconciler = MetaInstanceReconciler()
events = await reconciler.reconcile(state)
# Both meta-instances should have been attempted (not short-circuited)
assert call_count == 2
# Both should have placement failed events
placement_failed = [e for e in events if isinstance(e, MetaInstancePlacementFailed)]
assert len(placement_failed) == 2
# =============================================================================
# 8. Cascade delete with task cancellation
# =============================================================================
def test_cascade_delete_cancels_active_tasks():
"""Deleting a MetaInstance should cancel tasks on backing instances.
Regression test: previously, cascade-deleting backing instances via
DeleteMetaInstance did not emit TaskStatusUpdated(Cancelled) for active
tasks, leaving orphaned task references in state.
"""
meta = _meta_instance()
iid, inst = _instance(node_ids=["node-a"], meta_instance_id=meta.meta_instance_id)
task_id = TaskId()
task = LoadModel(task_id=task_id, instance_id=iid, task_status=TaskStatus.Running)
# Build state with meta-instance, backing instance, and active task
state = State(
meta_instances={meta.meta_instance_id: meta},
instances={iid: inst},
tasks={task_id: task},
topology=_topology("node-a"),
)
# Simulate the cascade-delete event sequence produced by main.py:
# 1. MetaInstanceDeleted
# 2. TaskStatusUpdated(Cancelled) for active tasks
# 3. InstanceDeleted
idx = 0
state = apply(
state,
IndexedEvent(
idx=idx,
event=MetaInstanceDeleted(meta_instance_id=meta.meta_instance_id),
),
)
idx += 1
state = apply(
state,
IndexedEvent(
idx=idx,
event=TaskStatusUpdated(task_id=task_id, task_status=TaskStatus.Cancelled),
),
)
idx += 1
state = apply(
state,
IndexedEvent(idx=idx, event=InstanceDeleted(instance_id=iid)),
)
# Verify everything is cleaned up
assert len(state.meta_instances) == 0
assert len(state.instances) == 0
assert state.tasks[task_id].task_status == TaskStatus.Cancelled
def test_cascade_delete_skips_completed_tasks():
"""Cascade delete should only cancel Pending/Running tasks, not completed ones."""
meta = _meta_instance()
iid, inst = _instance(node_ids=["node-a"], meta_instance_id=meta.meta_instance_id)
running_task_id = TaskId()
completed_task_id = TaskId()
running_task = LoadModel(
task_id=running_task_id, instance_id=iid, task_status=TaskStatus.Running
)
completed_task = LoadModel(
task_id=completed_task_id, instance_id=iid, task_status=TaskStatus.Complete
)
state = State(
meta_instances={meta.meta_instance_id: meta},
instances={iid: inst},
tasks={running_task_id: running_task, completed_task_id: completed_task},
topology=_topology("node-a"),
)
# Only the running task should be cancelled — we verify the logic pattern
# by checking which tasks are Pending or Running
active_tasks = [
t
for t in state.tasks.values()
if t.instance_id == iid
and t.task_status in (TaskStatus.Pending, TaskStatus.Running)
]
assert len(active_tasks) == 1
assert active_tasks[0].task_id == running_task_id
+1 -117
View File
@@ -14,12 +14,10 @@ from exo.shared.models.model_cards import ModelCard, ModelId, ModelTask
from exo.shared.topology import Topology
from exo.shared.types.commands import PlaceInstance
from exo.shared.types.common import CommandId, NodeId
from exo.shared.types.events import InstanceCreated, InstanceDeleted, TaskStatusUpdated
from exo.shared.types.events import InstanceCreated, InstanceDeleted
from exo.shared.types.memory import Memory
from exo.shared.types.multiaddr import Multiaddr
from exo.shared.types.profiling import NetworkInterfaceInfo, NodeNetworkInfo
from exo.shared.types.tasks import TaskId, TaskStatus, TextGeneration
from exo.shared.types.text_generation import InputMessage, TextGenerationTaskParams
from exo.shared.types.topology import Connection, SocketConnection
from exo.shared.types.worker.instances import (
Instance,
@@ -458,117 +456,3 @@ def test_tensor_rdma_backend_connectivity_matrix(
else:
ip_part = coordinator.split(":")[0]
assert len(ip_part.split(".")) == 4
def _make_task(
instance_id: InstanceId,
status: TaskStatus = TaskStatus.Running,
) -> TextGeneration:
return TextGeneration(
task_id=TaskId(),
task_status=status,
instance_id=instance_id,
command_id=CommandId(),
task_params=TextGenerationTaskParams(
model=ModelId("test-model"),
input=[InputMessage(role="user", content="hello")],
),
)
def test_get_transition_events_delete_instance_cancels_running_tasks(
instance: Instance,
):
# arrange
instance_id = InstanceId()
current_instances: dict[InstanceId, Instance] = {instance_id: instance}
target_instances: dict[InstanceId, Instance] = {}
task = _make_task(instance_id, TaskStatus.Running)
tasks = {task.task_id: task}
# act
events = get_transition_events(current_instances, target_instances, tasks)
# assert cancellation event should come before the deletion event
assert len(events) == 2
assert isinstance(events[0], TaskStatusUpdated)
assert events[0].task_id == task.task_id
assert events[0].task_status == TaskStatus.Cancelled
assert isinstance(events[1], InstanceDeleted)
assert events[1].instance_id == instance_id
def test_get_transition_events_delete_instance_cancels_pending_tasks(
instance: Instance,
):
# arrange
instance_id = InstanceId()
current_instances: dict[InstanceId, Instance] = {instance_id: instance}
target_instances: dict[InstanceId, Instance] = {}
task = _make_task(instance_id, TaskStatus.Pending)
tasks = {task.task_id: task}
# act
events = get_transition_events(current_instances, target_instances, tasks)
# assert
assert len(events) == 2
assert isinstance(events[0], TaskStatusUpdated)
assert events[0].task_id == task.task_id
assert events[0].task_status == TaskStatus.Cancelled
assert isinstance(events[1], InstanceDeleted)
def test_get_transition_events_delete_instance_ignores_completed_tasks(
instance: Instance,
):
# arrange
instance_id = InstanceId()
current_instances: dict[InstanceId, Instance] = {instance_id: instance}
target_instances: dict[InstanceId, Instance] = {}
tasks = {
t.task_id: t
for t in [
_make_task(instance_id, TaskStatus.Complete),
_make_task(instance_id, TaskStatus.Failed),
_make_task(instance_id, TaskStatus.TimedOut),
_make_task(instance_id, TaskStatus.Cancelled),
]
}
# act
events = get_transition_events(current_instances, target_instances, tasks)
# assert only the InstanceDeleted event, no cancellations
assert len(events) == 1
assert isinstance(events[0], InstanceDeleted)
def test_get_transition_events_delete_instance_cancels_only_matching_tasks(
instance: Instance,
):
# arrange
instance_id_a = InstanceId()
instance_id_b = InstanceId()
current_instances: dict[InstanceId, Instance] = {
instance_id_a: instance,
instance_id_b: instance,
}
# only delete instance A, keep instance B
target_instances: dict[InstanceId, Instance] = {instance_id_b: instance}
task_a = _make_task(instance_id_a, TaskStatus.Running)
task_b = _make_task(instance_id_b, TaskStatus.Running)
tasks = {task_a.task_id: task_a, task_b.task_id: task_b}
# act
events = get_transition_events(current_instances, target_instances, tasks)
# assert only task_a should be cancelled
cancel_events = [e for e in events if isinstance(e, TaskStatusUpdated)]
delete_events = [e for e in events if isinstance(e, InstanceDeleted)]
assert len(cancel_events) == 1
assert cancel_events[0].task_id == task_a.task_id
assert cancel_events[0].task_status == TaskStatus.Cancelled
assert len(delete_events) == 1
assert delete_events[0].instance_id == instance_id_a
+742
View File
@@ -0,0 +1,742 @@
from exo.master.process_managers.instance_health import InstanceHealthReconciler
from exo.master.reconcile import (
find_unsatisfied_meta_instances,
instance_connections_healthy,
instance_runners_failed,
instance_satisfies_meta_instance,
)
from exo.shared.apply import apply
from exo.shared.models.model_cards import ModelCard, ModelId, ModelTask
from exo.shared.topology import Topology
from exo.shared.types.common import Host, MetaInstanceId, NodeId
from exo.shared.types.events import (
IndexedEvent,
InstanceCreated,
InstanceDeleted,
InstanceRetrying,
MetaInstanceCreated,
MetaInstanceDeleted,
)
from exo.shared.types.memory import Memory
from exo.shared.types.meta_instance import MetaInstance
from exo.shared.types.multiaddr import Multiaddr
from exo.shared.types.state import State
from exo.shared.types.topology import Connection, SocketConnection
from exo.shared.types.worker.instances import (
InstanceId,
MlxRingInstance,
)
from exo.shared.types.worker.runners import (
RunnerFailed,
RunnerId,
RunnerLoading,
RunnerReady,
RunnerShutdown,
ShardAssignments,
)
from exo.shared.types.worker.shards import PipelineShardMetadata
def _model_card(model_id: str = "test-org/test-model") -> ModelCard:
return ModelCard(
model_id=ModelId(model_id),
storage_size=Memory.from_kb(1000),
n_layers=10,
hidden_size=30,
supports_tensor=True,
tasks=[ModelTask.TextGeneration],
)
def _topology(*node_ids: str, connect: bool = True) -> Topology:
"""Build a topology with nodes connected in a bidirectional ring with unique IPs.
Node at index ``i`` gets IP ``10.0.0.{i+1}``. Edges go in both directions
between consecutive nodes (including wrap-around).
"""
t = Topology()
nodes = [NodeId(n) for n in node_ids]
for n in nodes:
t.add_node(n)
if connect and len(nodes) > 1:
for i in range(len(nodes)):
j = (i + 1) % len(nodes)
t.add_connection(
Connection(
source=nodes[i],
sink=nodes[j],
edge=SocketConnection(
sink_multiaddr=Multiaddr(
address=f"/ip4/10.0.0.{j + 1}/tcp/50000"
)
),
)
)
t.add_connection(
Connection(
source=nodes[j],
sink=nodes[i],
edge=SocketConnection(
sink_multiaddr=Multiaddr(
address=f"/ip4/10.0.0.{i + 1}/tcp/50000"
)
),
)
)
return t
def _meta_instance(
model_id: str = "test-org/test-model",
*,
min_nodes: int = 1,
node_ids: list[NodeId] | None = None,
meta_instance_id: MetaInstanceId | None = None,
) -> MetaInstance:
return MetaInstance(
meta_instance_id=meta_instance_id or MetaInstanceId(),
model_id=ModelId(model_id),
min_nodes=min_nodes,
node_ids=node_ids,
)
def _instance(
model_id: str = "test-org/test-model",
node_ids: list[str] | None = None,
instance_id: InstanceId | None = None,
meta_instance_id: MetaInstanceId | None = None,
) -> tuple[InstanceId, MlxRingInstance]:
"""Create a test instance with hosts_by_node matching ``_topology()`` IPs."""
iid = instance_id or InstanceId()
nodes = node_ids or ["node-a"]
n = len(nodes)
mc = _model_card(model_id)
ephemeral_port = 50000
node_to_runner = {NodeId(nd): RunnerId() for nd in nodes}
runner_to_shard = {
runner_id: PipelineShardMetadata(
model_card=mc,
device_rank=i,
world_size=n,
start_layer=0,
end_layer=mc.n_layers,
n_layers=mc.n_layers,
)
for i, runner_id in enumerate(node_to_runner.values())
}
# Build hosts_by_node with IPs matching _topology() convention:
# node at index idx has IP 10.0.0.{idx+1}
hosts_by_node: dict[NodeId, list[Host]] = {}
for r, node_str in enumerate(nodes):
hosts: list[Host] = []
for idx in range(n):
if idx == r:
hosts.append(Host(ip="0.0.0.0", port=ephemeral_port))
elif n > 1 and idx in ((r - 1) % n, (r + 1) % n):
hosts.append(Host(ip=f"10.0.0.{idx + 1}", port=ephemeral_port))
else:
hosts.append(Host(ip="198.51.100.1", port=0))
hosts_by_node[NodeId(node_str)] = hosts
return iid, MlxRingInstance(
instance_id=iid,
shard_assignments=ShardAssignments(
model_id=ModelId(model_id),
runner_to_shard=runner_to_shard,
node_to_runner=node_to_runner,
),
hosts_by_node=hosts_by_node,
ephemeral_port=ephemeral_port,
meta_instance_id=meta_instance_id,
)
# --- instance_satisfies_meta_instance (pure constraint matching) ---
def test_satisfies_matching_model():
meta = _meta_instance()
_, inst = _instance(node_ids=["node-a"])
assert instance_satisfies_meta_instance(meta, inst) is True
def test_not_satisfies_wrong_model():
meta = _meta_instance("test-org/model-a")
_, inst = _instance("test-org/model-b")
assert instance_satisfies_meta_instance(meta, inst) is False
def test_not_satisfies_missing_required_node():
meta = _meta_instance(node_ids=[NodeId("node-c")])
_, inst = _instance(node_ids=["node-a", "node-b"])
assert instance_satisfies_meta_instance(meta, inst) is False
def test_not_satisfies_fewer_than_min_nodes():
meta = _meta_instance(min_nodes=3)
_, inst = _instance(node_ids=["node-a", "node-b"])
assert instance_satisfies_meta_instance(meta, inst) is False
def test_satisfies_with_node_ids_specified():
meta = _meta_instance(node_ids=[NodeId("node-a"), NodeId("node-b")], min_nodes=2)
_, inst = _instance(node_ids=["node-a", "node-b", "node-c"])
assert instance_satisfies_meta_instance(meta, inst) is True
# --- instance_connections_healthy ---
def test_healthy_single_node_present():
_, inst = _instance(node_ids=["node-a"])
topology = _topology("node-a")
assert instance_connections_healthy(inst, topology) is True
def test_unhealthy_single_node_missing():
_, inst = _instance(node_ids=["node-a"])
topology = Topology() # empty
assert instance_connections_healthy(inst, topology) is False
def test_healthy_two_node_ring():
_, inst = _instance(node_ids=["node-a", "node-b"])
topology = _topology("node-a", "node-b")
assert instance_connections_healthy(inst, topology) is True
def test_unhealthy_two_node_edge_removed():
"""Nodes present but edge removed — ring broken."""
_, inst = _instance(node_ids=["node-a", "node-b"])
topology = _topology("node-a", "node-b", connect=False)
assert instance_connections_healthy(inst, topology) is False
def test_unhealthy_two_node_ip_changed():
"""Edge exists but with a different IP than instance was configured with."""
_, inst = _instance(node_ids=["node-a", "node-b"])
# Build topology with different IPs than _instance() expects
topology = Topology()
topology.add_node(NodeId("node-a"))
topology.add_node(NodeId("node-b"))
topology.add_connection(
Connection(
source=NodeId("node-a"),
sink=NodeId("node-b"),
edge=SocketConnection(
sink_multiaddr=Multiaddr(address="/ip4/192.168.99.99/tcp/50000")
),
)
)
topology.add_connection(
Connection(
source=NodeId("node-b"),
sink=NodeId("node-a"),
edge=SocketConnection(
sink_multiaddr=Multiaddr(address="/ip4/192.168.99.98/tcp/50000")
),
)
)
assert instance_connections_healthy(inst, topology) is False
def test_healthy_three_node_ring():
_, inst = _instance(node_ids=["node-a", "node-b", "node-c"])
topology = _topology("node-a", "node-b", "node-c")
assert instance_connections_healthy(inst, topology) is True
def test_unhealthy_three_node_one_edge_removed():
"""Remove one edge from a three-node ring — instance unhealthy."""
_, inst = _instance(node_ids=["node-a", "node-b", "node-c"])
# Build topology with one direction of one edge missing
topology = Topology()
nodes = [NodeId("node-a"), NodeId("node-b"), NodeId("node-c")]
for n in nodes:
topology.add_node(n)
# Add all edges except node-a → node-b
topology.add_connection(
Connection(
source=nodes[1],
sink=nodes[0],
edge=SocketConnection(
sink_multiaddr=Multiaddr(address="/ip4/10.0.0.1/tcp/50000")
),
)
)
topology.add_connection(
Connection(
source=nodes[1],
sink=nodes[2],
edge=SocketConnection(
sink_multiaddr=Multiaddr(address="/ip4/10.0.0.3/tcp/50000")
),
)
)
topology.add_connection(
Connection(
source=nodes[2],
sink=nodes[1],
edge=SocketConnection(
sink_multiaddr=Multiaddr(address="/ip4/10.0.0.2/tcp/50000")
),
)
)
topology.add_connection(
Connection(
source=nodes[2],
sink=nodes[0],
edge=SocketConnection(
sink_multiaddr=Multiaddr(address="/ip4/10.0.0.1/tcp/50000")
),
)
)
topology.add_connection(
Connection(
source=nodes[0],
sink=nodes[2],
edge=SocketConnection(
sink_multiaddr=Multiaddr(address="/ip4/10.0.0.3/tcp/50000")
),
)
)
# Missing: node-a → node-b (ip 10.0.0.2)
assert instance_connections_healthy(inst, topology) is False
def test_unhealthy_node_missing_from_topology():
"""Instance has a node that's not in the topology at all."""
_, inst = _instance(node_ids=["node-a", "node-b"])
topology = _topology("node-a") # node-b not present
assert instance_connections_healthy(inst, topology) is False
def test_healthy_extra_nodes_in_topology():
"""Extra nodes in topology don't affect instance health."""
_, inst = _instance(node_ids=["node-a", "node-b"])
topology = _topology("node-a", "node-b", "node-c")
assert instance_connections_healthy(inst, topology) is True
# --- find_unsatisfied_meta_instances ---
def test_unsatisfied_no_meta_instances():
result = find_unsatisfied_meta_instances({}, {}, Topology())
assert list(result) == []
def test_unsatisfied_one_satisfied():
meta = _meta_instance()
id_a, inst_a = _instance(meta_instance_id=meta.meta_instance_id)
topology = _topology("node-a")
result = find_unsatisfied_meta_instances(
{meta.meta_instance_id: meta},
{id_a: inst_a},
topology,
)
assert list(result) == []
def test_unsatisfied_one_not_satisfied():
meta = _meta_instance("test-org/model-x")
id_a, inst_a = _instance("test-org/model-y")
topology = _topology("node-a")
result = find_unsatisfied_meta_instances(
{meta.meta_instance_id: meta}, {id_a: inst_a}, topology
)
assert list(result) == [meta]
def test_unsatisfied_mix():
meta_satisfied = _meta_instance("test-org/model-a")
meta_unsatisfied = _meta_instance("test-org/model-b")
id_a, inst_a = _instance(
"test-org/model-a", meta_instance_id=meta_satisfied.meta_instance_id
)
topology = _topology("node-a")
result = find_unsatisfied_meta_instances(
{
meta_satisfied.meta_instance_id: meta_satisfied,
meta_unsatisfied.meta_instance_id: meta_unsatisfied,
},
{id_a: inst_a},
topology,
)
assert list(result) == [meta_unsatisfied]
def test_unsatisfied_node_disconnect():
meta = _meta_instance()
id_a, inst_a = _instance(
node_ids=["node-a", "node-b"], meta_instance_id=meta.meta_instance_id
)
topology = _topology("node-a") # node-b disconnected
result = find_unsatisfied_meta_instances(
{meta.meta_instance_id: meta},
{id_a: inst_a},
topology,
)
assert list(result) == [meta]
def test_unsatisfied_edge_break():
"""Instance exists but its connections broke — meta-instance becomes unsatisfied."""
meta = _meta_instance()
id_a, inst_a = _instance(
node_ids=["node-a", "node-b"], meta_instance_id=meta.meta_instance_id
)
topology = _topology("node-a", "node-b", connect=False) # nodes present, no edges
result = find_unsatisfied_meta_instances(
{meta.meta_instance_id: meta},
{id_a: inst_a},
topology,
)
assert list(result) == [meta]
def test_unsatisfied_idempotent():
meta = _meta_instance("test-org/model-x")
topology = _topology("node-a")
meta_instances = {meta.meta_instance_id: meta}
instances: dict[InstanceId, MlxRingInstance] = {}
result_1 = list(
find_unsatisfied_meta_instances(meta_instances, instances, topology)
)
result_2 = list(
find_unsatisfied_meta_instances(meta_instances, instances, topology)
)
assert result_1 == result_2
def test_unsatisfied_exclusive_binding():
"""Two MetaInstances for the same model: one is bound via meta_instance_id, the other is unsatisfied."""
meta_a = _meta_instance("test-org/model-x")
meta_b = _meta_instance("test-org/model-x")
id_inst, inst = _instance(
"test-org/model-x", meta_instance_id=meta_a.meta_instance_id
)
topology = _topology("node-a")
result = find_unsatisfied_meta_instances(
{
meta_a.meta_instance_id: meta_a,
meta_b.meta_instance_id: meta_b,
},
{id_inst: inst},
topology,
)
assert list(result) == [meta_b]
# --- apply handlers ---
def test_apply_meta_instance_created():
state = State()
meta = _meta_instance()
event = MetaInstanceCreated(meta_instance=meta)
new_state = apply(state, IndexedEvent(idx=0, event=event))
assert meta.meta_instance_id in new_state.meta_instances
assert new_state.meta_instances[meta.meta_instance_id] == meta
def test_apply_meta_instance_deleted():
meta = _meta_instance()
state = State(meta_instances={meta.meta_instance_id: meta})
event = MetaInstanceDeleted(meta_instance_id=meta.meta_instance_id)
new_state = apply(state, IndexedEvent(idx=0, event=event))
assert meta.meta_instance_id not in new_state.meta_instances
def test_apply_meta_instance_deleted_clears_failure_info():
meta = _meta_instance().model_copy(
update={"consecutive_failures": 2, "last_failure_error": "OOM"}
)
state = State(meta_instances={meta.meta_instance_id: meta})
event = MetaInstanceDeleted(meta_instance_id=meta.meta_instance_id)
new_state = apply(state, IndexedEvent(idx=0, event=event))
assert meta.meta_instance_id not in new_state.meta_instances
# --- instance_runners_failed ---
def test_runners_failed_all_failed():
"""All runners in RunnerFailed -> instance is failed."""
_, inst = _instance(node_ids=["node-a", "node-b"])
runners = {
rid: RunnerFailed(error_message="OOM")
for rid in inst.shard_assignments.node_to_runner.values()
}
is_failed, error = instance_runners_failed(inst, runners, {})
assert is_failed is True
assert error is not None
assert "OOM" in error
def test_runners_failed_mixed_failed_shutdown():
"""One Failed + one Shutdown = failed."""
_, inst = _instance(node_ids=["node-a", "node-b"])
runner_ids = list(inst.shard_assignments.node_to_runner.values())
runners = {
runner_ids[0]: RunnerFailed(error_message="crash"),
runner_ids[1]: RunnerShutdown(),
}
is_failed, error = instance_runners_failed(inst, runners, {})
assert is_failed is True
assert error is not None
assert "crash" in error
def test_runners_not_failed_all_shutdown():
"""All Shutdown (graceful) = not a failure."""
_, inst = _instance(node_ids=["node-a"])
runners = {
rid: RunnerShutdown() for rid in inst.shard_assignments.node_to_runner.values()
}
is_failed, _ = instance_runners_failed(inst, runners, {})
assert is_failed is False
def test_runners_not_failed_still_active():
"""Some runners still active = not failed yet."""
_, inst = _instance(node_ids=["node-a", "node-b"])
runner_ids = list(inst.shard_assignments.node_to_runner.values())
runners = {
runner_ids[0]: RunnerFailed(error_message="OOM"),
runner_ids[1]: RunnerLoading(),
}
is_failed, _ = instance_runners_failed(inst, runners, {})
assert is_failed is False
def test_runners_not_failed_no_status():
"""Runner not yet reported = not failed."""
_, inst = _instance(node_ids=["node-a"])
is_failed, _ = instance_runners_failed(inst, {}, {})
assert is_failed is False
def test_runners_not_failed_healthy():
"""Runners in Ready state = not failed."""
_, inst = _instance(node_ids=["node-a"])
runners = {
rid: RunnerReady() for rid in inst.shard_assignments.node_to_runner.values()
}
is_failed, _ = instance_runners_failed(inst, runners, {})
assert is_failed is False
# --- failure tracking in apply_instance_deleted ---
def test_apply_instance_deleted_tracks_failure():
"""InstanceDeleted with failure_error increments meta instance failure count."""
meta = _meta_instance()
iid, inst = _instance(node_ids=["node-a"], meta_instance_id=meta.meta_instance_id)
state = State(
meta_instances={meta.meta_instance_id: meta},
instances={iid: inst},
)
event = InstanceDeleted(instance_id=iid, failure_error="Runner OOM")
new_state = apply(state, IndexedEvent(idx=0, event=event))
mi = new_state.meta_instances[meta.meta_instance_id]
assert mi.consecutive_failures == 1
assert mi.last_failure_error == "Runner OOM"
def test_apply_instance_deleted_increments_failure():
"""Subsequent failures increment the counter."""
meta = _meta_instance().model_copy(
update={"consecutive_failures": 2, "last_failure_error": "previous error"}
)
iid, inst = _instance(node_ids=["node-a"], meta_instance_id=meta.meta_instance_id)
state = State(
meta_instances={meta.meta_instance_id: meta},
instances={iid: inst},
)
event = InstanceDeleted(instance_id=iid, failure_error="new error")
new_state = apply(state, IndexedEvent(idx=0, event=event))
mi = new_state.meta_instances[meta.meta_instance_id]
assert mi.consecutive_failures == 3
assert mi.last_failure_error == "new error"
def test_apply_instance_deleted_no_failure_no_tracking():
"""InstanceDeleted without failure_error does not track."""
meta = _meta_instance()
iid, inst = _instance(node_ids=["node-a"], meta_instance_id=meta.meta_instance_id)
state = State(
meta_instances={meta.meta_instance_id: meta},
instances={iid: inst},
)
event = InstanceDeleted(instance_id=iid)
new_state = apply(state, IndexedEvent(idx=0, event=event))
mi = new_state.meta_instances[meta.meta_instance_id]
assert mi.consecutive_failures == 0
def test_apply_instance_deleted_orphan_no_tracking():
"""InstanceDeleted for orphan instance (no meta_instance_id) does not track."""
iid, inst = _instance(node_ids=["node-a"])
state = State(instances={iid: inst})
event = InstanceDeleted(instance_id=iid, failure_error="crash")
new_state = apply(state, IndexedEvent(idx=0, event=event))
assert len(new_state.meta_instances) == 0
# --- InstanceRetrying ---
def test_apply_instance_retrying_removes_runners():
"""InstanceRetrying removes the instance's runners from state but keeps the instance."""
meta = _meta_instance()
iid, inst = _instance(
node_ids=["node-a", "node-b"], meta_instance_id=meta.meta_instance_id
)
runner_ids = list(inst.shard_assignments.node_to_runner.values())
runners = {
runner_ids[0]: RunnerFailed(error_message="OOM"),
runner_ids[1]: RunnerShutdown(),
}
state = State(
meta_instances={meta.meta_instance_id: meta},
instances={iid: inst},
runners=runners,
)
event = InstanceRetrying(
instance_id=iid,
meta_instance_id=meta.meta_instance_id,
failure_error="OOM",
)
new_state = apply(state, IndexedEvent(idx=0, event=event))
# Instance still exists
assert iid in new_state.instances
# Runners removed
assert runner_ids[0] not in new_state.runners
assert runner_ids[1] not in new_state.runners
def test_apply_instance_retrying_increments_failure():
"""InstanceRetrying increments consecutive_failures on the MetaInstance."""
meta = _meta_instance()
iid, inst = _instance(node_ids=["node-a"], meta_instance_id=meta.meta_instance_id)
state = State(
meta_instances={meta.meta_instance_id: meta},
instances={iid: inst},
)
event = InstanceRetrying(
instance_id=iid,
meta_instance_id=meta.meta_instance_id,
failure_error="crash",
)
new_state = apply(state, IndexedEvent(idx=0, event=event))
mi = new_state.meta_instances[meta.meta_instance_id]
assert mi.consecutive_failures == 1
assert mi.last_failure_error == "crash"
def test_apply_instance_retrying_skips_missing_runners():
"""InstanceRetrying doesn't assert if runners haven't reported yet."""
meta = _meta_instance()
iid, inst = _instance(node_ids=["node-a"], meta_instance_id=meta.meta_instance_id)
# No runners in state at all
state = State(
meta_instances={meta.meta_instance_id: meta},
instances={iid: inst},
)
event = InstanceRetrying(
instance_id=iid,
meta_instance_id=meta.meta_instance_id,
failure_error="crash",
)
# Should not raise
new_state = apply(state, IndexedEvent(idx=0, event=event))
assert iid in new_state.instances
def test_apply_instance_created_resets_failure_counter():
"""InstanceCreated resets consecutive_failures but preserves last_failure_error."""
meta = _meta_instance().model_copy(
update={"consecutive_failures": 3, "last_failure_error": "old error"}
)
_, inst = _instance(node_ids=["node-a"], meta_instance_id=meta.meta_instance_id)
state = State(meta_instances={meta.meta_instance_id: meta})
event = InstanceCreated(instance=inst)
new_state = apply(state, IndexedEvent(idx=0, event=event))
mi = new_state.meta_instances[meta.meta_instance_id]
assert mi.consecutive_failures == 0
assert mi.last_failure_error == "old error"
assert mi.placement_error is None
# --- InstanceHealthReconciler retry-vs-delete ---
async def test_health_reconciler_retries_when_under_limit():
"""InstanceHealthReconciler emits InstanceRetrying when consecutive_failures < 3."""
meta = _meta_instance()
iid, inst = _instance(node_ids=["node-a"], meta_instance_id=meta.meta_instance_id)
runner_ids = list(inst.shard_assignments.node_to_runner.values())
state = State(
meta_instances={meta.meta_instance_id: meta},
instances={iid: inst},
runners={runner_ids[0]: RunnerFailed(error_message="OOM")},
topology=_topology("node-a"),
)
reconciler = InstanceHealthReconciler()
events = await reconciler.reconcile(state)
assert len(events) == 1
assert isinstance(events[0], InstanceRetrying)
assert events[0].instance_id == iid
assert events[0].meta_instance_id == meta.meta_instance_id
async def test_health_reconciler_deletes_when_limit_reached():
"""InstanceHealthReconciler emits InstanceDeleted when consecutive_failures >= 3."""
meta = _meta_instance().model_copy(update={"consecutive_failures": 3})
iid, inst = _instance(node_ids=["node-a"], meta_instance_id=meta.meta_instance_id)
runner_ids = list(inst.shard_assignments.node_to_runner.values())
state = State(
meta_instances={meta.meta_instance_id: meta},
instances={iid: inst},
runners={runner_ids[0]: RunnerFailed(error_message="OOM")},
topology=_topology("node-a"),
)
reconciler = InstanceHealthReconciler()
events = await reconciler.reconcile(state)
assert len(events) == 1
assert isinstance(events[0], InstanceDeleted)
async def test_health_reconciler_deletes_without_meta_instance():
"""Instances without a MetaInstance are deleted immediately on runner failure."""
iid, inst = _instance(node_ids=["node-a"])
runner_ids = list(inst.shard_assignments.node_to_runner.values())
state = State(
instances={iid: inst},
runners={runner_ids[0]: RunnerFailed(error_message="crash")},
topology=_topology("node-a"),
)
reconciler = InstanceHealthReconciler()
events = await reconciler.reconcile(state)
assert len(events) == 1
assert isinstance(events[0], InstanceDeleted)
async def test_health_reconciler_network_failure_always_deletes():
"""Network failure always triggers InstanceDeleted regardless of retry count."""
meta = _meta_instance()
iid, inst = _instance(
node_ids=["node-a", "node-b"], meta_instance_id=meta.meta_instance_id
)
state = State(
meta_instances={meta.meta_instance_id: meta},
instances={iid: inst},
topology=_topology("node-a"), # node-b missing
)
reconciler = InstanceHealthReconciler()
events = await reconciler.reconcile(state)
assert len(events) == 1
assert isinstance(events[0], InstanceDeleted)
assert events[0].failure_error == "Network connection lost"
+26 -4
View File
@@ -1,4 +1,6 @@
from exo_pyo3_bindings import PyFromSwarm
from enum import Enum
from exo_pyo3_bindings import ConnectionUpdate, ConnectionUpdateType
from exo.shared.types.common import NodeId
from exo.utils.pydantic_ext import CamelCaseModel
@@ -6,10 +8,30 @@ from exo.utils.pydantic_ext import CamelCaseModel
"""Serialisable types for Connection Updates/Messages"""
class ConnectionMessageType(Enum):
Connected = 0
Disconnected = 1
@staticmethod
def from_update_type(update_type: ConnectionUpdateType):
match update_type:
case ConnectionUpdateType.Connected:
return ConnectionMessageType.Connected
case ConnectionUpdateType.Disconnected:
return ConnectionMessageType.Disconnected
class ConnectionMessage(CamelCaseModel):
node_id: NodeId
connected: bool
connection_type: ConnectionMessageType
remote_ipv4: str
remote_tcp_port: int
@classmethod
def from_update(cls, update: PyFromSwarm.Connection) -> "ConnectionMessage":
return cls(node_id=NodeId(update.peer_id), connected=update.connected)
def from_update(cls, update: ConnectionUpdate) -> "ConnectionMessage":
return cls(
node_id=NodeId(update.peer_id),
connection_type=ConnectionMessageType.from_update_type(update.update_type),
remote_ipv4=update.remote_ipv4,
remote_tcp_port=update.remote_tcp_port,
)
-161
View File
@@ -1,161 +0,0 @@
from dataclasses import dataclass, field
from random import random
import anyio
from anyio import BrokenResourceError, ClosedResourceError
from anyio.abc import CancelScope
from loguru import logger
from exo.shared.types.commands import ForwarderCommand, RequestEventLog
from exo.shared.types.common import SessionId, SystemId
from exo.shared.types.events import (
Event,
EventId,
GlobalForwarderEvent,
IndexedEvent,
LocalForwarderEvent,
)
from exo.utils.channels import Receiver, Sender, channel
from exo.utils.event_buffer import OrderedBuffer
from exo.utils.task_group import TaskGroup
@dataclass
class EventRouter:
session_id: SessionId
command_sender: Sender[ForwarderCommand]
external_inbound: Receiver[GlobalForwarderEvent]
external_outbound: Sender[LocalForwarderEvent]
_system_id: SystemId = field(init=False, default_factory=SystemId)
internal_outbound: list[Sender[IndexedEvent]] = field(
init=False, default_factory=list
)
event_buffer: OrderedBuffer[Event] = field(
init=False, default_factory=OrderedBuffer
)
out_for_delivery: dict[EventId, tuple[float, LocalForwarderEvent]] = field(
init=False, default_factory=dict
)
_tg: TaskGroup = field(init=False, default_factory=TaskGroup)
_nack_cancel_scope: CancelScope | None = field(init=False, default=None)
_nack_attempts: int = field(init=False, default=0)
_nack_base_seconds: float = field(init=False, default=0.5)
_nack_cap_seconds: float = field(init=False, default=10.0)
async def run(self):
try:
async with self._tg as tg:
tg.start_soon(self._run_ext_in)
tg.start_soon(self._simple_retry)
finally:
self.external_outbound.close()
for send in self.internal_outbound:
send.close()
# can make this better in future
async def _simple_retry(self):
while True:
await anyio.sleep(1 + random())
# list here is a shallow clone for shared mutation
for e_id, (time, event) in list(self.out_for_delivery.items()):
if anyio.current_time() > time + 5:
self.out_for_delivery[e_id] = (anyio.current_time(), event)
await self.external_outbound.send(event)
def sender(self) -> Sender[Event]:
send, recv = channel[Event]()
if self._tg.is_running():
self._tg.start_soon(self._ingest, SystemId(), recv)
else:
self._tg.queue(self._ingest, SystemId(), recv)
return send
def receiver(self) -> Receiver[IndexedEvent]:
send, recv = channel[IndexedEvent]()
self.internal_outbound.append(send)
return recv
def shutdown(self) -> None:
self._tg.cancel_tasks()
async def _ingest(self, system_id: SystemId, recv: Receiver[Event]):
idx = 0
with recv as events:
async for event in events:
f_ev = LocalForwarderEvent(
origin_idx=idx,
origin=system_id,
session=self.session_id,
event=event,
)
idx += 1
await self.external_outbound.send(f_ev)
self.out_for_delivery[event.event_id] = (anyio.current_time(), f_ev)
async def _run_ext_in(self):
buf = OrderedBuffer[Event]()
with self.external_inbound as events:
async for event in events:
if event.session != self.session_id:
continue
if event.origin != self.session_id.master_node_id:
continue
buf.ingest(event.origin_idx, event.event)
event_id = event.event.event_id
if event_id in self.out_for_delivery:
self.out_for_delivery.pop(event_id)
drained = buf.drain_indexed()
if drained:
self._nack_attempts = 0
if self._nack_cancel_scope:
self._nack_cancel_scope.cancel()
if not drained and (
self._nack_cancel_scope is None
or self._nack_cancel_scope.cancel_called
):
# Request the next index.
self._tg.start_soon(self._nack_request, buf.next_idx_to_release)
continue
for idx, event in drained:
to_clear = set[int]()
for i, sender in enumerate(self.internal_outbound):
try:
await sender.send(IndexedEvent(idx=idx, event=event))
except (ClosedResourceError, BrokenResourceError):
to_clear.add(i)
for i in sorted(to_clear, reverse=True):
self.internal_outbound.pop(i)
async def _nack_request(self, since_idx: int) -> None:
# We request all events after (and including) the missing index.
# This function is started whenever we receive an event that is out of sequence.
# It is cancelled as soon as we receiver an event that is in sequence.
if since_idx < 0:
logger.warning(f"Negative value encountered for nack request {since_idx=}")
since_idx = 0
with CancelScope() as scope:
self._nack_cancel_scope = scope
delay: float = self._nack_base_seconds * (2.0**self._nack_attempts)
delay = min(self._nack_cap_seconds, delay)
self._nack_attempts += 1
try:
await anyio.sleep(delay)
logger.info(
f"Nack attempt {self._nack_attempts}: Requesting Event Log from {since_idx}"
)
await self.command_sender.send(
ForwarderCommand(
origin=self._system_id,
command=RequestEventLog(since_idx=since_idx),
)
)
finally:
if self._nack_cancel_scope is scope:
self._nack_cancel_scope = None
+33 -50
View File
@@ -8,16 +8,16 @@ from typing import cast
from anyio import (
BrokenResourceError,
ClosedResourceError,
create_task_group,
move_on_after,
sleep_forever,
)
from anyio.abc import TaskGroup
from exo_pyo3_bindings import (
AllQueuesFullError,
Keypair,
MessageTooLargeError,
NetworkingHandle,
NoPeersSubscribedToTopicError,
PyFromSwarm,
)
from filelock import FileLock
from loguru import logger
@@ -25,7 +25,6 @@ from loguru import logger
from exo.shared.constants import EXO_NODE_ID_KEYPAIR
from exo.utils.channels import Receiver, Sender, channel
from exo.utils.pydantic_ext import CamelCaseModel
from exo.utils.task_group import TaskGroup
from .connection_message import ConnectionMessage
from .topics import CONNECTION_MESSAGES, PublishPolicy, TypedTopic
@@ -112,9 +111,10 @@ class Router:
self._net: NetworkingHandle = handle
self._tmp_networking_sender: Sender[tuple[str, bytes]] | None = send
self._id_count = count()
self._tg: TaskGroup = TaskGroup()
self._tg: TaskGroup | None = None
async def register_topic[T: CamelCaseModel](self, topic: TypedTopic[T]):
assert self._tg is None, "Attempted to register topic after setup time"
send = self._tmp_networking_sender
if send:
self._tmp_networking_sender = None
@@ -122,8 +122,7 @@ class Router:
send = self.networking_receiver.clone_sender()
router = TopicRouter[T](topic, send)
self.topic_routers[topic.topic] = cast(TopicRouter[CamelCaseModel], router)
if self._tg.is_running():
await self._networking_subscribe(topic.topic)
await self._networking_subscribe(str(topic.topic))
def sender[T: CamelCaseModel](self, topic: TypedTopic[T]) -> Sender[T]:
router = self.topic_routers.get(topic.topic, None)
@@ -149,15 +148,14 @@ class Router:
async def run(self):
logger.debug("Starting Router")
try:
async with self._tg as tg:
async with create_task_group() as tg:
self._tg = tg
for topic in self.topic_routers:
router = self.topic_routers[topic]
tg.start_soon(router.run)
tg.start_soon(self._networking_recv)
tg.start_soon(self._networking_recv_connection_messages)
tg.start_soon(self._networking_publish)
# subscribe to pending topics
for topic in self.topic_routers:
await self._networking_subscribe(topic)
# Router only shuts down if you cancel it.
await sleep_forever()
finally:
@@ -167,7 +165,9 @@ class Router:
async def shutdown(self):
logger.debug("Shutting down Router")
self._tg.cancel_tasks()
if not self._tg:
return
self._tg.cancel_scope.cancel()
async def _networking_subscribe(self, topic: str):
await self._net.gossipsub_subscribe(topic)
@@ -178,41 +178,28 @@ class Router:
logger.info(f"Unsubscribed from {topic}")
async def _networking_recv(self):
try:
while True:
from_swarm = await self._net.recv()
logger.debug(from_swarm)
match from_swarm:
case PyFromSwarm.Message(origin, topic, data):
logger.trace(
f"Received message on {topic} from {origin} with payload {data}"
)
if topic not in self.topic_routers:
logger.warning(
f"Received message on unknown or inactive topic {topic}"
)
continue
router = self.topic_routers[topic]
await router.publish_bytes(data)
case PyFromSwarm.Connection():
message = ConnectionMessage.from_update(from_swarm)
logger.trace(
f"Received message on connection_messages with payload {message}"
)
if CONNECTION_MESSAGES.topic in self.topic_routers:
router = self.topic_routers[CONNECTION_MESSAGES.topic]
assert router.topic.model_type == ConnectionMessage
router = cast(TopicRouter[ConnectionMessage], router)
await router.publish(message)
case _:
logger.critical(
"failed to exhaustively check FromSwarm messages - logic error"
)
except Exception as exception:
logger.opt(exception=exception).error(
"Gossipsub receive loop terminated unexpectedly"
while True:
topic, data = await self._net.gossipsub_recv()
logger.trace(f"Received message on {topic} with payload {data}")
if topic not in self.topic_routers:
logger.warning(f"Received message on unknown or inactive topic {topic}")
continue
router = self.topic_routers[topic]
await router.publish_bytes(data)
async def _networking_recv_connection_messages(self):
while True:
update = await self._net.connection_update_recv()
message = ConnectionMessage.from_update(update)
logger.trace(
f"Received message on connection_messages with payload {message}"
)
raise
if CONNECTION_MESSAGES.topic in self.topic_routers:
router = self.topic_routers[CONNECTION_MESSAGES.topic]
assert router.topic.model_type == ConnectionMessage
router = cast(TopicRouter[ConnectionMessage], router)
await router.publish(message)
async def _networking_publish(self):
with self.networking_receiver as networked_items:
@@ -224,10 +211,6 @@ class Router:
pass
except AllQueuesFullError:
logger.warning(f"All peer queues full, dropping message on {topic}")
except MessageTooLargeError:
logger.warning(
f"Message too large for gossipsub on {topic} ({len(data)} bytes), dropping"
)
def get_node_id_keypair(
@@ -258,6 +241,6 @@ def get_node_id_keypair(
# if no valid credentials, create new ones and persist
with open(path, "w+b") as f:
keypair = Keypair.generate()
keypair = Keypair.generate_ed25519()
f.write(keypair.to_bytes())
return keypair
+119 -3
View File
@@ -4,7 +4,7 @@ from datetime import datetime
from loguru import logger
from exo.shared.types.common import NodeId
from exo.shared.types.common import MetaInstanceId, NodeId
from exo.shared.types.events import (
ChunkGenerated,
Event,
@@ -12,6 +12,10 @@ from exo.shared.types.events import (
InputChunkReceived,
InstanceCreated,
InstanceDeleted,
InstanceRetrying,
MetaInstanceCreated,
MetaInstanceDeleted,
MetaInstancePlacementFailed,
NodeDownloadProgress,
NodeGatheredInfo,
NodeTimedOut,
@@ -28,6 +32,7 @@ from exo.shared.types.events import (
TracesCollected,
TracesMerged,
)
from exo.shared.types.meta_instance import MetaInstance
from exo.shared.types.profiling import (
NodeIdentity,
NodeNetworkInfo,
@@ -72,6 +77,14 @@ def event_apply(event: Event, state: State) -> State:
return apply_instance_created(event, state)
case InstanceDeleted():
return apply_instance_deleted(event, state)
case InstanceRetrying():
return apply_instance_retrying(event, state)
case MetaInstanceCreated():
return apply_meta_instance_created(event, state)
case MetaInstanceDeleted():
return apply_meta_instance_deleted(event, state)
case MetaInstancePlacementFailed():
return apply_meta_instance_placement_failed(event, state)
case NodeTimedOut():
return apply_node_timed_out(event, state)
case NodeDownloadProgress():
@@ -174,20 +187,123 @@ def apply_task_failed(event: TaskFailed, state: State) -> State:
return state.model_copy(update={"tasks": new_tasks})
def _update_meta_instance(
state: State, mid: MetaInstanceId, **fields: object
) -> Mapping[MetaInstanceId, MetaInstance]:
mi = state.meta_instances[mid]
return {**state.meta_instances, mid: mi.model_copy(update=fields)}
def apply_instance_created(event: InstanceCreated, state: State) -> State:
instance = event.instance
new_instances: Mapping[InstanceId, Instance] = {
**state.instances,
instance.instance_id: instance,
}
return state.model_copy(update={"instances": new_instances})
update: dict[str, object] = {"instances": new_instances}
# Reset failure tracking when a new instance is created for a meta-instance
if instance.meta_instance_id and instance.meta_instance_id in state.meta_instances:
mi = state.meta_instances[instance.meta_instance_id]
if mi.placement_error is not None or mi.consecutive_failures > 0:
update["meta_instances"] = _update_meta_instance(
state,
instance.meta_instance_id,
placement_error=None,
consecutive_failures=0,
)
return state.model_copy(update=update)
def apply_instance_deleted(event: InstanceDeleted, state: State) -> State:
deleted_instance = state.instances.get(event.instance_id)
new_instances: Mapping[InstanceId, Instance] = {
iid: inst for iid, inst in state.instances.items() if iid != event.instance_id
}
return state.model_copy(update={"instances": new_instances})
update: dict[str, object] = {"instances": new_instances}
# Track failure on the MetaInstance itself
if (
event.failure_error
and deleted_instance
and deleted_instance.meta_instance_id
and deleted_instance.meta_instance_id in state.meta_instances
):
mid = deleted_instance.meta_instance_id
mi = state.meta_instances[mid]
update["meta_instances"] = {
**state.meta_instances,
mid: mi.model_copy(
update={
"consecutive_failures": mi.consecutive_failures + 1,
"last_failure_error": event.failure_error,
}
),
}
return state.model_copy(update=update)
def apply_instance_retrying(event: InstanceRetrying, state: State) -> State:
"""Runners failed but retry limit not reached — remove runners, keep instance."""
instance = state.instances.get(event.instance_id)
if instance is None:
# Instance was already deleted (e.g. cascade from DeleteMetaInstance).
# The InstanceDeleted handler already incremented consecutive_failures
# on the MetaInstance, so skipping here avoids double-counting.
return state
# Remove all runners belonging to this instance from state
runner_ids_to_remove = set(instance.shard_assignments.node_to_runner.values())
new_runners: Mapping[RunnerId, RunnerStatus] = {
rid: rs for rid, rs in state.runners.items() if rid not in runner_ids_to_remove
}
update: dict[str, object] = {"runners": new_runners}
# Increment failure count on the MetaInstance
if event.meta_instance_id in state.meta_instances:
update["meta_instances"] = _update_meta_instance(
state,
event.meta_instance_id,
consecutive_failures=state.meta_instances[
event.meta_instance_id
].consecutive_failures
+ 1,
last_failure_error=event.failure_error,
)
return state.model_copy(update=update)
def apply_meta_instance_created(event: MetaInstanceCreated, state: State) -> State:
new_meta: Mapping[MetaInstanceId, MetaInstance] = {
**state.meta_instances,
event.meta_instance.meta_instance_id: event.meta_instance,
}
return state.model_copy(update={"meta_instances": new_meta})
def apply_meta_instance_deleted(event: MetaInstanceDeleted, state: State) -> State:
new_meta: Mapping[MetaInstanceId, MetaInstance] = {
mid: mi
for mid, mi in state.meta_instances.items()
if mid != event.meta_instance_id
}
return state.model_copy(update={"meta_instances": new_meta})
def apply_meta_instance_placement_failed(
event: MetaInstancePlacementFailed, state: State
) -> State:
if event.meta_instance_id not in state.meta_instances:
return state
return state.model_copy(
update={
"meta_instances": _update_meta_instance(
state, event.meta_instance_id, placement_error=event.reason
)
}
)
def apply_runner_status_updated(event: RunnerStatusUpdated, state: State) -> State:
-2
View File
@@ -78,6 +78,4 @@ EXO_ENABLE_IMAGE_MODELS = (
os.getenv("EXO_ENABLE_IMAGE_MODELS", "false").lower() == "true"
)
EXO_OFFLINE = os.getenv("EXO_OFFLINE", "false").lower() == "true"
EXO_TRACING_ENABLED = os.getenv("EXO_TRACING_ENABLED", "false").lower() == "true"
+13 -4
View File
@@ -4,8 +4,10 @@ import anyio
from anyio import (
CancelScope,
Event,
create_task_group,
get_cancelled_exc_class,
)
from anyio.abc import TaskGroup
from loguru import logger
from exo.routing.connection_message import ConnectionMessage
@@ -13,7 +15,6 @@ from exo.shared.types.commands import ForwarderCommand
from exo.shared.types.common import NodeId, SessionId
from exo.utils.channels import Receiver, Sender
from exo.utils.pydantic_ext import CamelCaseModel
from exo.utils.task_group import TaskGroup
DEFAULT_ELECTION_TIMEOUT = 3.0
@@ -81,12 +82,13 @@ class Election:
self._candidates: list[ElectionMessage] = []
self._campaign_cancel_scope: CancelScope | None = None
self._campaign_done: Event | None = None
self._tg = TaskGroup()
self._tg: TaskGroup | None = None
async def run(self):
logger.info("Starting Election")
try:
async with self._tg as tg:
async with create_task_group() as tg:
self._tg = tg
tg.start_soon(self._election_receiver)
tg.start_soon(self._connection_receiver)
tg.start_soon(self._command_counter)
@@ -122,7 +124,12 @@ class Election:
)
async def shutdown(self) -> None:
self._tg.cancel_tasks()
if not self._tg:
logger.warning(
"Attempted to shutdown election service that was not running"
)
return
self._tg.cancel_scope.cancel()
async def _election_receiver(self) -> None:
with self._em_receiver as election_messages:
@@ -136,6 +143,7 @@ class Election:
if message.clock > self.clock:
self.clock = message.clock
logger.debug(f"New clock: {self.clock}")
assert self._tg is not None
logger.debug("Starting new campaign")
candidates: list[ElectionMessage] = [message]
logger.debug(f"Candidates: {candidates}")
@@ -170,6 +178,7 @@ class Election:
# These messages are strictly peer to peer
self.clock += 1
logger.debug(f"New clock: {self.clock}")
assert self._tg is not None
candidates: list[ElectionMessage] = []
self._candidates = candidates
logger.debug("Starting new campaign")
-4
View File
@@ -90,7 +90,6 @@ class ModelCard(CamelCaseModel):
base_model: str = ""
capabilities: list[str] = []
uses_cfg: bool = False
trust_remote_code: bool = True
@field_validator("tasks", mode="before")
@classmethod
@@ -138,7 +137,6 @@ class ModelCard(CamelCaseModel):
hidden_size=config_data.hidden_size or 0,
supports_tensor=config_data.supports_tensor,
tasks=[ModelTask.TextGeneration],
trust_remote_code=False,
)
await mc.save_to_custom_dir()
_card_cache[model_id] = mc
@@ -190,8 +188,6 @@ class ConfigData(BaseModel):
["DeepseekV3ForCausalLM"],
["Qwen3NextForCausalLM"],
["Qwen3MoeForCausalLM"],
["Qwen3_5MoeForConditionalGeneration"],
["Qwen3_5ForConditionalGeneration"],
["MiniMaxM2ForCausalLM"],
["LlamaForCausalLM"],
["GptOssForCausalLM"],
+9 -2
View File
@@ -1,7 +1,7 @@
import pytest
from anyio import create_task_group, fail_after, move_on_after
from exo.routing.connection_message import ConnectionMessage
from exo.routing.connection_message import ConnectionMessage, ConnectionMessageType
from exo.shared.election import Election, ElectionMessage, ElectionResult
from exo.shared.types.commands import ForwarderCommand, TestCommand
from exo.shared.types.common import NodeId, SessionId, SystemId
@@ -327,7 +327,14 @@ async def test_connection_message_triggers_new_round_broadcast() -> None:
tg.start_soon(election.run)
# Send any connection message object; we close quickly to cancel before result creation
await cm_tx.send(ConnectionMessage(node_id=NodeId(), connected=True))
await cm_tx.send(
ConnectionMessage(
node_id=NodeId(),
connection_type=ConnectionMessageType.Connected,
remote_ipv4="",
remote_tcp_port=0,
)
)
# Expect a broadcast for the new round at clock=1
while True:
+17 -11
View File
@@ -6,7 +6,7 @@ from uuid import uuid4
from pydantic import BaseModel, Field, field_validator
from exo.shared.models.model_cards import ModelCard, ModelId
from exo.shared.types.common import CommandId, NodeId
from exo.shared.types.common import CommandId, MetaInstanceId, NodeId
from exo.shared.types.memory import Memory
from exo.shared.types.worker.instances import Instance, InstanceId, InstanceMeta
from exo.shared.types.worker.shards import Sharding, ShardMetadata
@@ -262,9 +262,24 @@ class DeleteInstanceResponse(BaseModel):
instance_id: InstanceId
class CancelCommandResponse(BaseModel):
class CreateMetaInstanceParams(BaseModel):
model_id: ModelId
sharding: Sharding = Sharding.Pipeline
instance_meta: InstanceMeta = InstanceMeta.MlxRing
min_nodes: int = 1
node_ids: list[NodeId] | None = None
class CreateMetaInstanceResponse(BaseModel):
message: str
command_id: CommandId
meta_instance_id: MetaInstanceId
class DeleteMetaInstanceResponse(BaseModel):
message: str
command_id: CommandId
meta_instance_id: MetaInstanceId
ImageSize = Literal[
@@ -442,12 +457,3 @@ class TraceListItem(CamelCaseModel):
class TraceListResponse(CamelCaseModel):
traces: list[TraceListItem]
class DeleteTracesRequest(CamelCaseModel):
task_ids: list[str]
class DeleteTracesResponse(CamelCaseModel):
deleted: list[str]
not_found: list[str]
+12 -1
View File
@@ -6,7 +6,8 @@ from exo.shared.types.api import (
ImageGenerationTaskParams,
)
from exo.shared.types.chunks import InputImageChunk
from exo.shared.types.common import CommandId, NodeId, SystemId
from exo.shared.types.common import CommandId, MetaInstanceId, NodeId, SystemId
from exo.shared.types.meta_instance import MetaInstance
from exo.shared.types.text_generation import TextGenerationTaskParams
from exo.shared.types.worker.instances import Instance, InstanceId, InstanceMeta
from exo.shared.types.worker.shards import Sharding, ShardMetadata
@@ -52,6 +53,14 @@ class TaskCancelled(BaseCommand):
cancelled_command_id: CommandId
class CreateMetaInstance(BaseCommand):
meta_instance: MetaInstance
class DeleteMetaInstance(BaseCommand):
meta_instance_id: MetaInstanceId
class TaskFinished(BaseCommand):
finished_command_id: CommandId
@@ -94,6 +103,8 @@ Command = (
| CreateInstance
| DeleteInstance
| TaskCancelled
| CreateMetaInstance
| DeleteMetaInstance
| TaskFinished
| SendInputChunk
)

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