Architecture (voir docs/ARCHITECTURE-LITELLM.md):
open-webui -> LiteLLM:4000
|- qwen3.6-27b -> jean-proxy:3012 -> Jean (.11:8080)
|- llama3.1-8b -> wol-proxy:3011 -> Ollama (.11:11434)
Nouveaux fichiers:
- wol-proxy/ (versionne l'app.py de prod deja en place)
- jean-proxy/ (clone du wol-proxy avec HEALTH_PATH=/health + retry post-WoL
pour services qui mettent du temps a demarrer apres WoL, ex: Jean qui
charge 14 Go en VRAM)
- litellm/config.yaml (routing par nom de modele, drop_params=true)
- docker-compose.litellm.yml (override ADDITIF: ajoute uniquement jean-proxy
et litellm, ne redemarre PAS open-webui/wol-proxy/hermes)
Teste en prod sur .27: Jean + Ollama repondent via LiteLLM:4000 avec
path-through de chat_template_kwargs.enable_thinking (raisonnement Qwen3.6
desactivable par requete). Zero impact hermes-webui/wol-proxy.
Basculer open-webui reste une etape manuelle (step 3 de la doc).
- Server API (/api/items/) does not expose note CRUD, only sync protocol
- Clipper API is the only way to read/write notes remotely
- Requires Joplin desktop open + Web Clipper enabled
- Added ping() function to check connectivity
- Valves: joplin_url, joplin_api_token
- Tool with Valves (leantime_url, leantime_api_key)
- 14 functions: projects, tickets, comments, users, timesheets, status labels
- JSON-RPC 2.0 client matching Leantime API
- Activate/deactivate per conversation in Open WebUI
- Added Valves (Pydantic) pattern for ocr_url configuration via wheel icon
- Inspired by Kreuzberg tool: multi-level file search (exact, substring, word matches)
- Support for multi-file OCR with extract_text_from_images
- Added search_files function for pattern matching
- effi-ocr runs on port 8301 (Tesseract with preprocessing)
- Original prompt kept for reference (200+ lines, too verbose)
- New optimized prompt: ~40 lines, -75% tokens
- Added table for tools usage (Kreuzberg, web_search, sub-agent, etc.)
- Optimized version includes explicit Kreuzberg usage for OCR
- README: add Tool OpenWebUI section with current features
- README: update Historique with Tool section
- Tool doc: add Limitations and Future Improvements sections
Future improvements noted:
- Support historique (données N dernières minutes)
- Filtres par période
- Graphiques et alertes automatisées
Permet à Effi de:
- get_all_stats() - Stats de tous les systèmes
- get_system_stats('folzocorp') - Stats d'un système
- check_alerts() - Vérifier les alertes
- get_gpu_status('folzocorp') - Status GPU AMD
- can_run_comfyui() - Vérifier si ComfyUI peut tourner
- Fix bug where old history entries were counted as recent activity
- Added recent_threshold parameter (default 300s)
- is_recent_history() now checks timestamp of last history item
- Add full architecture diagram with all machines (FolzoCorp, EffiServer, NAS)
- Document all services on each machine with ports and roles
- Add memory management section (auto-unload scripts)
- Add project structure tree
- Update technologies list
- Add documentation links
- Monitors VRAM, queue status, and history
- Restarts ComfyUI container after timeout if RAM > 5GB and VRAM < 1%
- More effective at freeing RAM than model unloading alone
- Corriger extraction VRAM via API correcte
- Ajouter gestion signaux propre
- Ajouter TimeoutStopSec pour arrêt propre
- Ajouter service restart Beszel au réveil
- effi-stt: Déchargement Whisper après 300s inactivité (INACTIVITY_TIMEOUT)
* Nouveaux endpoints: /unload, /reload, /stats
* Modelo rechargé à la première requête
- ComfyUI: Script monitor qui décharge VRAM après inactivité
* Vérifie usage VRAM toutes les 60s
* Exécute workflow UnloadAllModelsNode après 300s inactivité
- Kreuzberg: Manager qui redémarre le container après seuil mémoire + inactivité
* Restart si > 2GB pendant > 10min inactif
- Documentation: docs/MEMOIRE-GESTION.md
- Add download-ace-step-full.sh script for easier model downloads
- Remove effi-ocr (migrated to effi-kreuzberg for OCR functionality)
- Generate Song v0.5.4 now fully functional with JK-AceStep-Nodes
- Removed image preprocessing (Kreuzberg handles it internally)
- Improved file search (partial name matching)
- Shows available files when file not found
- Uses mimetypes for proper content-type
- Better file finding (startswith + contains)
- Normalize language codes (French->fra, English->eng, etc.)
- List all files in directory when searching
- Add extract_files() for processing multiple files one by one
- Support comma-separated and JSON array input
- Switch OCR backend from Tesseract to PaddleOCR (better accuracy)
- Update tool description
- Convert images to grayscale before OCR
- Increase contrast by 2x for better text detection
- This matches the preprocessing that effi-ocr (Tesseract) was using
- Add docs/guide-ace-step.md with complete installation guide
- Add comfyui/download-ace-step.sh script for downloading model
- Add ComfyUI-Unload-Model custom node for model memory management
ACE-Step 1.5 Turbo: MIT license, <4GB VRAM, 50+ languages
- Update TOOL-OPENWEBUI.md with correct OCR_URL for local deployment
- Add docker-compose-ocr.yml for deploying effi-ocr on effi-server
- Clarify architecture: effi-ocr must be on same server as OpenWebUI
- Update environment variable documentation for OCR_URL
- Add docs/audio/tts-research.md with Voxtral TTS, Kokoro-FastAPI, openedai-speech comparison
- Add docs/audio/stt-research.md with Voxtral STT, faster-whisper options
- Add docs/gpu/optimization.md with OLLAMA_KEEP_ALIVE and VRAM management
- Add docs/README.md with architecture overview
- Add index files for docs navigation
Research conducted April 2026 for future audio improvements
- Création du répertoire docs/prompts/
- VERSIONS.md: Historique des versions du prompt
- prompt-v1.md: Version actuelle du prompt système
Ce système permet de garder les anciennes versions
du prompt lors des modifications futures.
- Effi-OCR: Service OCR avec Tesseract (multilingue FR/EN)
* API FastAPI sur port 8300
* Prétraitement images (contraste, niveaux de gris)
* Tool OpenWebUI intégré
- Web Scraper: Service Python (BeautifulSoup) sur port 11235
* Alternative à Crawl4AI (incompatible ARM64/J5040)
* Scraping HTML basique mais fonctionnel
* Intégré au docker-compose d'effi-server
- Documentation:
* ROADMAP-FONCTIONNALITES.md: Plan des 9 fonctionnalités
* WEB-SCRAPER-CONFIG.md: Configuration OpenWebUI
* WEB-SCRAPER-TOOL.md: Code du Tool
- Script start-all-services.sh mis à jour (6 services)
- check-services.sh: Verify all services after WOL
- start-all-services.sh: Start all containers after boot
- WOL-ET-DEMARRAGE-AUTO.md: Complete guide for auto-start
- Ensure ComfyUI starts automatically with Docker
- Clean menu interface with clear headers
- No terminal clearing, all messages preserved
- Color-coded status messages (green/red/yellow)
- Step-by-step progress tracking
- Final summary with installed models list
- Better wget progress display without messy output
Use absolute path instead of relative /bin
Ensures models are downloaded to correct location regardless of
current working directory when running the script
- Docker Compose setup with ROCm for RX 7800 XT (gfx1101)
- Auto-download script for SDXL models
- Basic SDXL workflow for testing
- Full documentation in README.md
- Update .gitignore to exclude model files (too large)
- Update PROMPT-REPRISE.md for new ComfyUI objective
- Create detailed PROMPT-REPRISE-COMFYUI.md with installation plan
- Mark STT/TTS as completed
- Include verification steps and key points
- Ready for next agent to take over image generation
- Document all 8 original priority items
- Mark STT/TTS as completed
- Detail remaining tasks: Joplin, ComfyUI, Web Search, etc.
- Provide prioritization and time estimates
- Suggest next steps based on user preference
- Add comprehensive session report (05/04/2026)
- Document all achievements: STT, TTS, French voices
- Update TODO with completed phases and future ideas
- Include useful commands and configurations
- Ready for future optimizations
- FastAPI server with OpenAI-compatible API
- Integration with faster-whisper (large-v3)
- Full Docker support with docker-compose
- Support for multiple languages (FR/EN/Auto)
- Clean, maintainable codebase (~100 lines)
- Health endpoint and error handling
- Volume caching for models
- Add detailed journal of all attempts (successes and failures)
- Explain speaches v0.8+ incompatibility with model cards
- Research and document whisper-asr-webservice as stable alternative
- Provide complete deployment plan with docker-compose
- Include validation checklist
- install-stt.sh: Complete automated setup for speaches (STT)
- install-voices-fr.sh: Optional French voices for Piper TTS
- Both scripts handle SELinux permissions (:Z labels)
- Include health checks and user-friendly output
- Complete installation plan for openedai-speech (TTS) on folzo-Corp
- Complete installation plan for speaches (STT) on folzo-Corp
- Docker compose configurations for both services
- Testing procedures and troubleshooting
- Phase breakdown with current status (Phase 1 complete)
- Update LLM from qwen2.5:14b to qwen3.5:9b (32K context)
- Remove CouchDB, mcpo, pipelines from architecture (services removed)
- Fix memory section: manual memory instead of mem0/auto-memory
- Fix STT: Whisper runs on CPU (ROCm build failed)
- Fix TTS: XTTS with nova voice (not Piper)
- Add note about removed services and link to full history
- Update service count: 7 active services
- Add reference to HISTORIQUE-COMPLET-EFFI.md
- Synthesize 8 Perplexity conversation files into comprehensive history
- Document technology choices: AFFiNE → SiYuan → Obsidian
- Detail encountered problems and solutions (MCP, TTS, STT, etc.)
- Current validated configuration with qwen3.5:9b
- Prioritized TODO list for next phases
- Key learnings and recommendations
- Liste complete des taches prioritaires (TTS, Whisper, Joplin, Images, etc.)
- Details techniques pour chaque optimisation
- Ordre de priorite suggere
- Prompt de reprise pour prochain assistant
- Reference documentation existante
Refs: preparation-thread-2026-04-05
- Script audit-production.sh pour audit complet
- Script sync-from-production.sh pour synchronisation facile
- Guide WORKFLOW.md pour processus de modification
- README pour le dossier scripts
- Documentation complète du workflow Git
- Structure complète du projet
- Docker compose de production
- Documentation organisée par thèmes
- Guides d'installation et configuration
- Section dépannage détaillée
- Journal historique
Folzo