Apple
4601c6fca8
feat: add Vision Encoder service + Vision RAG implementation
...
- Vision Encoder Service (OpenCLIP ViT-L/14, GPU-accelerated)
- FastAPI app with text/image embedding endpoints (768-dim)
- Docker support with NVIDIA GPU runtime
- Port 8001, health checks, model info API
- Qdrant Vector Database integration
- Port 6333/6334 (HTTP/gRPC)
- Image embeddings storage (768-dim, Cosine distance)
- Auto collection creation
- Vision RAG implementation
- VisionEncoderClient (Python client for API)
- Image Search module (text-to-image, image-to-image)
- Vision RAG routing in DAGI Router (mode: image_search)
- VisionEncoderProvider integration
- Documentation (5000+ lines)
- SYSTEM-INVENTORY.md - Complete system inventory
- VISION-ENCODER-STATUS.md - Service status
- VISION-RAG-IMPLEMENTATION.md - Implementation details
- vision_encoder_deployment_task.md - Deployment checklist
- services/vision-encoder/README.md - Deployment guide
- Updated WARP.md, INFRASTRUCTURE.md, Jupyter Notebook
- Testing
- test-vision-encoder.sh - Smoke tests (6 tests)
- Unit tests for client, image search, routing
- Services: 17 total (added Vision Encoder + Qdrant)
- AI Models: 3 (qwen3:8b, OpenCLIP ViT-L/14, BAAI/bge-m3)
- GPU Services: 2 (Vision Encoder, Ollama)
- VRAM Usage: ~10 GB (concurrent)
Status: Production Ready ✅
2025-11-17 05:24:36 -08:00
Apple
1ed1181105
feat: add RAG quality metrics, optimized prompts, and evaluation tools
...
Optimized Prompts:
- Create utils/rag_prompt_builder.py with citation-optimized prompts
- Specialized for DAO tokenomics and technical documentation
- Proper citation format [1], [2] with doc_id, page, section
- Memory context integration (facts, events, summaries)
- Token count estimation
RAG Service Metrics:
- Add comprehensive logging in query_pipeline.py
- Log: question, doc_ids, scores, retrieval method, timing
- Track: retrieval_time, total_query_time, documents_found, citations_count
- Add metrics in ingest_pipeline.py: pages_processed, blocks_processed, pipeline_time
Router Improvements:
- Use optimized prompt builder in _handle_rag_query()
- Add graceful fallback: if RAG unavailable, use Memory only
- Log prompt token count, RAG usage, Memory usage
- Return detailed metadata (rag_used, memory_used, citations_count, metrics)
Evaluation Tools:
- Create tests/rag_eval.py for systematic quality testing
- Test fixed questions with expected doc_ids
- Save results to JSON and CSV
- Compare RAG Service vs Router results
- Track: citations, expected docs found, query times
Documentation:
- Create docs/RAG_METRICS_PLAN.md
- Plan for Prometheus metrics collection
- Grafana dashboard panels and alerts
- Implementation guide for metrics
2025-11-16 05:12:19 -08:00
Apple
9b86f9a694
feat: implement RAG Service MVP with PARSER + Memory integration
...
RAG Service Implementation:
- Create rag-service/ with full structure (config, document_store, embedding, pipelines)
- Document Store: PostgreSQL + pgvector via Haystack
- Embedding: BAAI/bge-m3 (multilingual, 1024 dim)
- Ingest Pipeline: Convert ParsedDocument to Haystack Documents, embed, index
- Query Pipeline: Retrieve documents, generate answers via DAGI Router
- FastAPI endpoints: /ingest, /query, /health
Tests:
- Unit tests for ingest and query pipelines
- E2E test with example parsed JSON
- Test fixtures with real PARSER output example
Router Integration:
- Add mode='rag_query' routing rule in router-config.yml
- Priority 7, uses local_qwen3_8b for RAG queries
Docker:
- Add rag-service to docker-compose.yml
- Configure dependencies (router, city-db)
- Add model cache volume
Documentation:
- Complete README with API examples
- Integration guides for PARSER and Router
2025-11-16 04:41:53 -08:00
