Node.js side (llm-client.ts):
- Strip inline <think>...</think> from mascarade content before checking thinking field
- Suppress thinking blocks in real-time during mascarade SSE streaming
- Clean accumulated text from thinking before returning final response
mascarade Python side (ollama.py):
- Extract content from thinking field when Ollama returns content:"" (qwen3.5 mode)
- Strip <think> XML wrappers, extract after answer markers, fallback to last paragraph
- Suppress thinking blocks from Ollama streaming output
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
- llm-client: streamViaMascarade() parses SSE events from mascarade
- Cloud providers (claude, openai, mistral, google) now stream via mascarade SSE
- chatViaMascarade: extract content from thinking field when content is empty (qwen3.5)
- ws-ollama: improved thinking field extraction in tool probe (strip tags, FR/EN markers)
- Fallback to Ollama if mascarade stream fails
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Lot 524: Live Audio FX page — 7 chainable effects (filter, distortion, delay, reverb, compressor, bitcrusher, pitch), Web Audio API, VU-meter, per-param sliders
Lot 525: /director — AI orchestrates 4 random personas in a theatrical conversation
Lot 526: Visual Synth page — audio-reactive canvas (waveform, spectrum, particles, tunnel, kaleidoscope), 5 color themes, fullscreen
Lot 527: /memory-wipe — reset persona memory
Lot 528: /stage — trigger fullscreen performance mode
Lot 529: /radio-station — continuous program (music + voice + jingles every 45s)
Lot 530: Deploy
2 new pages (Live FX, Visual Synth), 4 new commands.
204 commands total.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
- streamChat: ALWAYS direct Ollama for streaming (no mascarade 3s overhead)
- mascarade used only for: non-streaming tool probes, /agent, /orchestrate, cloud providers
- Exception: explicit cloud provider (claude:, openai:) still routes via mascarade
- ollamaLimit reduced 5→2 to match OLLAMA_NUM_PARALLEL=2
- RAG threshold raised 60→80 chars (skip embedding for short msgs)
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Lot 515: llm-client uses /v1/chat/completions (OpenAI-compat, public, no auth)
- All models routed through mascarade (not just non-Ollama)
- RouteLLM still skips mascarade for trivial messages
Lot 516: /agent list|run — list and execute mascarade agents from chat
Lot 517: /kb search — query mascarade knowledge base
Lot 518: /provider list|metrics — view LLM providers + routing metrics
Lot 519: /orchestrate — multi-agent parallel orchestration
Lot 520: Deploy
196 commands. 2 test regressions (smoke test mock incompatible with new routing, RAG pre-existing).
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
- Crossfade animation between result frames (no jarring transitions)
- Resolution selector (192/256/384/512px) — lower = faster for live
- JPEG quality adaptive (0.7 live, 0.92 snapshot)
- Blend opacity slider (0-100%) to overlay result on webcam
- Frame history strip (last 8 thumbnails, clickable)
- Fullscreen overlay (click result → fullscreen + save/close)
- Snapshot button (📸) saves current result to gallery at hi-res
- Latency display (ms) alongside FPS
- Webcam video wrap for blend overlay positioning
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Final polish batch:
Lot 501: /roast — friendly AI roast from random persona
Lot 502: /wisdom — dev/AI wisdom quotes
Lot 503: /goodnight — broadcast goodnight
Lot 504: /challenge — random creative challenge
Lot 505: /matrix — matrix rain characters
Lot 506: /neofetch — system info summary
Lot 507: /help-all — compact list of ALL commands
Lot 510: /session-stats — session summary
191 commands, 510 lots total.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Lot 491: /riddle — AI devinette
Lot 492: /compliment — random compliment
Lot 493: /8ball — magic 8-ball
Lot 494: /whisper-all — murmure a toutes les personas
Lot 495: /commands — total command count
Lot 496: /server — hardware/OS info
Lot 499: /help-compose — DAW guide complet
Lot 500: /lot500 🎉 MILESTONE Easter egg
🎉 500 LOTS, 183+ commandes, 33 personas, 16+ services
211 unit tests, 19 E2E tests
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Lot 480: /blind-test — music quiz (random instrument, 15s reveal)
Lot 481: /lore <persona> — AI-generated persona backstory
Lot 482: /rate <1-5> — star rating broadcast
Lot 483: /sfx <type> — quick SFX (riser/drop/sweep/impact/stutter)
Lot 484: /jam [bpm] — instant 3-track jam session (drums+bass+pad)
169 commands, 211 tests.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Lot 470: /imagine-redo — regenerate last image with new seed
Lot 471: /shout — broadcast in ALL CAPS
Lot 472: /define <word> — AI definition via Ollama
Lot 473: Connected dot pulse animation
Lot 474: /help-imagine — complete /imagine guide with all flags
159 commands, 211 tests.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Lot 465: /imagine-history — show last N prompts with seeds
Lot 466: /ask <persona> <question> — direct question routing
Lot 467: /poll Question | A | B | C — simple poll with emoji
Lot 468: /imagine --steps N — override inference steps (1-50)
Lot 469: /batch /cmd1 ; /cmd2 — run multiple commands (max 5)
155 commands, 211 tests (210 pass, 1 pre-existing fail).
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Lot 460: /imagine-gallery — list saved images in chat
Lot 461: /preview enhanced — play full mix/master or single track
Lot 462: Reply button (↩) on chat messages — sets @mention in input
Lot 463: /persona-disable + /persona-enable — toggle personas on/off
Lot 464: Input history verified (ArrowUp/Down working)
150 commands, 211 tests.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Lot 456: Live face swap — webcam capture + auto-submit, swap ⇄ button
Lot 457: Speech-to-Text — mic button in chat input (Web Speech API + Whisper WASM fallback)
- BrowserSTT component with @huggingface/transformers whisper-tiny ONNX
- Server-side STT via faster-whisper on ai-bridge /stt endpoint
Lot 458: Browser permissions manager — mic/cam status indicators, denied warnings
Lot 459: Deploy + smoke
147 commands, 211 tests.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Lot 450: /imagine --model <checkpoint> override
Lot 451: /sys — one-liner system info
Lot 452: /color <hex> — nick color
Lot 453: /whoami — show nick, channel, role
Lot 454: Chat input character counter (>100 chars)
Lot 455: /imagine-models enhanced with summary
145 commands, 211 tests.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Lot 445: /imagine --seed <number> for reproducible generation
Lot 446: /imagine --style photo|anime|cyberpunk|painting|pixel|minitel
Lot 447: /tts-voices — list all available TTS voices (Piper FR + Kokoro EN)
Lot 448: /persona-voice — show voice assignment per persona
Lot 449: /changelog enhanced — curated lot summary
141 commands, 211 tests.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Lot 440: /imagine --ar 16:9|9:16|4:3|3:4 aspect ratio support
Lot 441: /leaderboard — persona ranking by memory facts (medals + bar chart)
Lot 442: /echo — broadcast text to channel
Lot 443: /version enhanced — lot count, commands, personas, uptime
Lot 444: Auto-scroll indicator button (↓ Nouveaux messages)
137 commands, 211 tests.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Lot 435: /personas grouped by category (Musique, Philosophie, Science, Arts, Scene)
Lot 436: /trivia — AI-generated quiz question with 4 choices
Lot 437: /comp rename <name> — rename active composition
Lot 438: Word count in chat header (msgs + mots)
Lot 439: /fortune enhanced — 25 quotes
134 commands, 211 tests.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Lot 430: /upscale placeholder (needs ESRGAN model)
Lot 431: /count — file counts for all data directories
Lot 432: /speak <persona> <text> — force TTS as any persona
Lot 433: Emoji health indicators in admin dashboard
Lot 434: /flip enhanced (PILE/FACE + emoji)
132 commands, 211 tests.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Lot 425: cleanForTTS() applied to streaming sentences + flush + fallback
Lot 426: /imagine --no <negative> — parse negative prompt, show in broadcast
Lot 427: /story [theme] — 3 random personas write collaborative story
Lot 428: /stats <persona> — show persona memory facts, model, last update
128 commands, 211 tests.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Lot 419: /imagine-queue prompt1|prompt2|... — batch image gen (max 8)
Lot 420: /haiku [topic] — random persona writes a haiku
Lot 421: /who enhanced — shows persona models + voices
Lot 422: /timer <seconds> — broadcast alarm after delay
Lot 423: /clear — client-side message clear via __clear__ signal
126 commands, 211 tests.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Lot 414: /summarize — AI summary of conversation (bullet points)
Lot 415: /mood — display current persona mood based on time
Lot 416: "N en ligne | N msgs" in chat header
Lot 417: /dice NdM enhanced (2d6, 3d20, etc.)
Lot 418: /ping — server latency test
121 commands, 211 tests.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Lot 408: /variations <prompt> — generate 4 seed variations of same image
Lot 409: Copy button (📋) on chat messages
Lot 410: /persona-create <name> <model> <desc> — create custom persona
Lot 411: /radio [on|off] — continuous random persona quotes every 30s
Lot 412: /collab [topic] — 5 random personas respond to same topic
Lot 413: Deploy
116 commands, 211 tests.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Lot 401: Auto-free ComfyUI VRAM every 5min (cron)
Lot 402: /benchmark — test all services latency in parallel
Lot 403: Ctrl+F chat search overlay with highlight
Lot 404: /translate (/tr) — FR↔EN auto-detect via Ollama
Lot 405: Persona mood system (time-of-day personality modifier)
110 commands, 211 tests.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Lot 395: /quote — random persona inspirational quote via Ollama
Lot 396: /weather [city] — weather from wttr.in
Lot 397: /ascii <text> — block letter ASCII art
Lot 398: Timestamp tooltips (full date+time on hover)
Lot 399: /uptime — server uptime display
106 commands, 211 tests.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Lot 390: /debate [topic] — 2 random personas debate a topic
Lot 391: Pin button (📌) on persona messages, sends vote:pin feedback
Lot 392: /ambient AI fast path — instant pad generation via AI Bridge
Lot 393: Waveform canvas in ComposePage (120x30px per track with audio)
102 commands, 211 tests.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Lot 385: /random-persona [topic] — invoke random persona on a topic
Lot 386: Emoji reactions (👍❤️😂🎵🔥) on hover for persona messages
Lot 387: /remix enhanced — random variation hints + new seed
Lot 388: Context indicator (sticky msg count in chat)
101 commands, 211 tests.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Lot 379: CRT scanline + phosphor glow CSS (dark mode only, performant)
Lot 380: /status enriched — uptime, RAM, VRAM, LLM, audio, ComfyUI, storage stats
Lot 381: Persona avatar initials — colored circle with first letter before nick
Lot 382: /clear-media [type] — cleanup generated images/audio/samples/exports
Lot 383: ComfyUI model preload at startup (skip 5-10s load on first /imagine)
100 commands, 211 unit tests, 19 E2E.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Lot 370: PWA manifest + icon SVG (installable on mobile)
Lot 372: /mp3 command (ffmpeg WAV→MP3 192kbps export)
Lot 373: @mention autocomplete (tab-complete persona names in chat input)
Lot 375: Persistent image gallery (server-side index.jsonl + /api/v2/media/images)
Lot 376: Persona voice preview in sidebar (♪ button, Kokoro/Piper TTS)
Lot 377: ACE-Step priority for /generate/music (20s/4min vs 5min MusicGen)
98 commands, 211 unit tests, 19 E2E.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
- ServiceHealth component: polls API, mascarade, ComfyUI, scheduler every 10s
- Green/red status cards with detail text (personas, providers, queue lengths)
- CSS: .admin-health-* grid with Minitel aesthetic
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
L'électron rare