How saksham6/document-rag-prototype works
Document RAG Prototype is a database-backed service for organizing knowledge bases, uploading and processing PDF/TXT files, and answering questions with retrieved document sources. The principal workflow runs through FastAPI routes: upload and persist metadata, explicitly ingest into chunks, embed chunks, then search via query embedding and pgvector before generating a grounded answer. README documents a browser UI and health endpoints as additional capabilities. Edges follow the sampled route/service calls and README; unsampled route wiring to schemas and persistence is not expanded beyond documented behavior.
Main entry points: main.py FastAPI app · ingestion.py Ingestion endpoints · session.py Database sessions · rag.py Search endpoint
Components 18
API and UI
- Browser UI
static/index.html - FastAPI app
app/api/main.py - Knowledge bases
src/document_rag_prototype/api/routes/knowledge_bases.py - Document and chunk routes
src/document_rag_prototype/api/routes/documents.py - Health checks
src/document_rag_prototype/api/routes/health.py - API schemas
src/document_rag_prototype/api/schemas/models.py
Document processing
- Ingestion endpoints
src/document_rag_prototype/api/routes/ingestion.py - Runtime config
src/document_rag_prototype/core/config.py - Extract and chunk
src/document_rag_prototype/services/ingestion_service.py
Retrieval and answers
- Search endpoint
src/document_rag_prototype/api/routes/rag.py - Embedding service
src/document_rag_prototype/services/embedding_service.py - Vector retrieval
src/document_rag_prototype/services/retrieval_service.py - Answer generation
src/document_rag_prototype/services/generation_service.py
Persistence
- ORM models
src/document_rag_prototype/db/models.py - Database sessions
src/document_rag_prototype/db/session.py - PostgreSQL / pgvector
Other components
- User
- OpenAI API
Connections 25
- User → Browser UI: uses
README.md - Browser UI → FastAPI app: accesses
README.md - FastAPI app → Knowledge bases: registers
app/api/main.py - FastAPI app → Document and chunk routes: registers
app/api/main.py - FastAPI app → Ingestion endpoints: registers
app/api/main.py - FastAPI app → Search endpoint: registers
app/api/main.py - FastAPI app → Health checks: registers
app/api/main.py - Knowledge bases → Database sessions: gets sessions
src/document_rag_prototype/api/routes/knowledge_bases.py - Document and chunk routes → Database sessions: gets sessions
src/document_rag_prototype/api/routes/documents.py - Document and chunk routes → ORM models: reads and writes
src/document_rag_prototype/api/routes/documents.py - Ingestion endpoints → Database sessions: gets sessions
src/document_rag_prototype/api/routes/ingestion.py - Ingestion endpoints → ORM models: reads and writes
src/document_rag_prototype/api/routes/ingestion.py - Ingestion endpoints → Runtime config: gets upload path
src/document_rag_prototype/api/routes/ingestion.py - Ingestion endpoints → Extract and chunk: extracts and chunks
src/document_rag_prototype/api/routes/ingestion.py - Ingestion endpoints → Embedding service: embeds chunks
src/document_rag_prototype/api/routes/ingestion.py - Search endpoint → Database sessions: gets session
src/document_rag_prototype/api/routes/rag.py - Search endpoint → Vector retrieval: requests matches
src/document_rag_prototype/api/routes/rag.py - Search endpoint → Answer generation: requests answer
src/document_rag_prototype/api/routes/rag.py - Vector retrieval → Embedding service: embeds query
src/document_rag_prototype/services/retrieval_service.py - Vector retrieval → PostgreSQL / pgvector: queries vectors
src/document_rag_prototype/services/retrieval_service.py - Answer generation → OpenAI API: generates answer
src/document_rag_prototype/services/generation_service.py - Embedding service → OpenAI API: creates embeddings
src/document_rag_prototype/services/embedding_service.py - Database sessions → PostgreSQL / pgvector: connects
README.md - Health checks → Database sessions: checks database
src/document_rag_prototype/api/routes/health.py - API schemas → Search endpoint: validates requests
src/document_rag_prototype/api/routes/rag.py