How vishnu-manalan/self-healing-rag works
This is a web application for interactive self-healing retrieval-augmented generation. A user submits a query in the Playground; the backend API hands it to the RAG orchestrator, which can enhance the query, retrieve candidates, validate them, rerank results, and generate an answer. Feedback supports a learning loop. The README also documents Architecture, Documentation, and Statistics views. Some subsystem links are based on explicit README descriptions or indexed imports; unsampled API and orchestration details are not assumed beyond that evidence.
Main entry points: self_healing_rag.py RAG orchestrator · App.jsx App shell · Playground.jsx Query playground · api_server.py Backend API
Components 17
Web interface
- Web entry
frontend/src/main.jsx - App shell
frontend/src/App.jsx - Query playground
frontend/src/components/Playground.jsx - Architecture view
frontend/src/components/Architecture.jsx - Documentation view
frontend/src/components/Documentation.jsx - Statistics view
frontend/src/components/Statistics.jsx
API service
- Backend API
backend/api_server.py
RAG workflow
- RAG orchestrator
backend/self_healing_rag.py - HyDE enhancement
backend/hyde.py - Query decomposition
backend/query_decomposition.py - Vector retrieval
backend/self_healing_rag.py - CRAG validation
backend/corrective_rag.py - Cross-encoder ranking
backend/reranker.py
Learning
- Example learning
backend/dynamic_prompting.py - Example index
backend/dynamic_prompting.py
Other components
- User
- Tavily search
Connections 20
- User → Query playground: submits query
README.md - Web entry → App shell: renders
frontend/src/main.jsx - App shell → Query playground: shows
frontend/src/App.jsx - App shell → Architecture view: shows
frontend/src/App.jsx - App shell → Documentation view: shows
frontend/src/App.jsx - App shell → Statistics view: shows
frontend/src/App.jsx - Query playground → Backend API: posts feedback
frontend/src/components/Playground.jsx - Query playground → Backend API: submits query
- Statistics view → Backend API: requests metrics
- Backend API → RAG orchestrator: serves RAG API
README.md - RAG orchestrator → HyDE enhancement: enhances query
backend/self_healing_rag.py - RAG orchestrator → Query decomposition: decomposes query
backend/self_healing_rag.py - RAG orchestrator → Vector retrieval: retrieves candidates
README.md - RAG orchestrator → CRAG validation: validates documents
backend/self_healing_rag.py - RAG orchestrator → Cross-encoder ranking: reranks documents
backend/self_healing_rag.py - RAG orchestrator → Example learning: uses examples
backend/self_healing_rag.py - CRAG validation → Tavily search: falls back
README.md - Example learning → Example index: stores examples
backend/dynamic_prompting.py - Query playground → Example learning: sends feedback
- Backend API → Example learning: records feedback