How eklakh-ai-engineer/enterprise-rag works
Enterprise RAG is an evidence-grounded document search and question-answering product, with a React workspace and FastAPI backend. Users authenticate, upload PDFs for asynchronous ingestion and indexing, then query the indexed evidence; answers carry mapped citations and undergo faithfulness and quality checks. The graph separates document preparation, retrieval and answer validation, plus persistence and external services. README documents capabilities beyond the sampled code; those links are cited there, while unsampled implementation details are not inferred.
Main entry points: pipeline.py Query orchestration · main.py FastAPI routes · worker.py Ingestion worker · App.jsx Workspace UI
Components 31
Workspace and API
- Workspace UI
frontend/src/App.jsx - Auth client
frontend/src/api/auth.js - RAG client
frontend/src/api/rag.js - FastAPI routes
app/api/main.py - Token authentication
app/auth/service.py
Document ingestion
- PDF parsing
app/parsing/pdf_parser.py - Evidence chunking
app/chunking/recursive_chunker.py - Ingestion pipeline
app/ingestion/pipeline.py - Ingestion worker
app/ingestion/worker.py - Document indexing
app/indexing/pipeline.py - Embedding provider
app/indexing/embeddings.py
Retrieval and answers
- Query orchestration
app/query/pipeline.py - Dense retrieval
app/retrieval/dense/retriever.py - BM25 retrieval
app/retrieval/bm25/retriever.py - Hybrid fusion
app/retrieval/hybrid/retriever.py - Cross-encoder reranking
app/reranking/cross_encoder.py - Context assembly
app/query/context.py - Grounded generation
app/generation/rag_generator.py - OpenRouter client
app/generation/openrouter.py
Evidence and evaluation
- Citation mapping
app/citations/mapper.py - Citation validation
app/citations/verifier.py - Faithfulness check
app/faithfulness/checker.py - Answer evaluation
app/evaluation/answer_evaluator.py - Pipeline metrics
app/observability/metrics.py
Persistence and integrations
- Domain records
app/persistence/entities.py - Repositories
app/persistence/repositories.py - Supabase persistence
app/persistence/supabase.py - Supabase file storage
app/storage/supabase.py
Other components
- User
- Supabase Auth
- OpenRouter service
Connections 32
- User → Workspace UI: uses
README.md - Workspace UI → Auth client: signs in
frontend/src/App.jsx - Workspace UI → RAG client: uploads
frontend/src/App.jsx - RAG client → FastAPI routes: requests
- FastAPI routes → Token authentication: authenticates
app/api/main.py - Token authentication → Supabase Auth: verifies token
app/auth/service.py - FastAPI routes → Query orchestration: runs query
app/api/main.py - FastAPI routes → Supabase file storage: stores upload
app/api/main.py - FastAPI routes → Supabase persistence: creates records
app/api/main.py - Ingestion worker → Ingestion pipeline: processes document
app/ingestion/worker.py - Ingestion pipeline → PDF parsing: parses PDF
app/ingestion/pipeline.py - Ingestion pipeline → Evidence chunking: chunks pages
app/ingestion/pipeline.py - Ingestion worker → Document indexing: indexes chunks
app/ingestion/worker.py - Document indexing → Embedding provider: encodes chunks
app/indexing/pipeline.py - Document indexing → Supabase persistence: persists vectors
- Ingestion worker → Repositories: updates jobs
app/ingestion/worker.py - Repositories → Domain records: uses records
app/persistence/repositories.py - Query orchestration → Dense retrieval: retrieves candidates
app/query/pipeline.py - Query orchestration → BM25 retrieval: retrieves candidates
app/query/pipeline.py - Dense retrieval → Hybrid fusion: provides candidates
app/retrieval/hybrid/retriever.py - BM25 retrieval → Hybrid fusion: provides candidates
app/retrieval/hybrid/retriever.py - Query orchestration → Hybrid fusion: fuses candidates
app/query/pipeline.py - Query orchestration → Cross-encoder reranking: reranks results
app/query/pipeline.py - Query orchestration → Context assembly: assembles context
README.md - Query orchestration → Grounded generation: generates answer
app/query/pipeline.py - Grounded generation → OpenRouter client: requests completion
app/generation/rag_generator.py - OpenRouter client → OpenRouter service: calls model
app/generation/openrouter.py - Query orchestration → Citation mapping: maps citations
app/query/pipeline.py - Query orchestration → Citation validation: validates citations
README.md - Query orchestration → Faithfulness check: checks claims
app/query/pipeline.py - Query orchestration → Answer evaluation: evaluates answer
app/query/pipeline.py - Query orchestration → Pipeline metrics: records metrics
app/query/pipeline.py