How ankit0514/rag_project works
This is an end-to-end document RAG demo: a user uploads a PDF, DOCX, or TXT, then asks a question and receives an answer with source snippets. Uploads are loaded, chunked, embedded, and indexed; chat combines vector and BM25 retrieval, reranks results, builds a prompt, and generates an answer using Gemini or a local fallback. The graph also includes conversation memory and the repository’s database models. Some subsystems are represented by paths rather than inspected implementation details; edges are limited to relationships shown in excerpts, the source index, or README.
Main entry points: chat_service.py Chat service · vector_search.py Vector search · hybrid_search.py Hybrid search · session.py Database session
Components 23
API and interface
- Demo frontend
frontend/index.html - FastAPI app
backend/app/main.py - RAG routes
backend/app/api/routes/rag.py
Document ingestion
- Format loaders
backend/app/ingestion/pipeline.py - PDF DOCX TXT loaders
backend/app/ingestion/loaders - Text chunking
backend/app/ingestion/chunking/recursive_chunker.py - Embedding service
backend/app/embeddings/embedding_service.py
Search and ranking
- Vector search
backend/app/retrieval/vector_search.py - FAISS index
backend/app/vectorstore/faiss_store.py - Hybrid search
backend/app/retrieval/hybrid_search.py - BM25 retrieval
backend/app/retrieval/bm25.py - Result reranker
backend/app/retrieval/reranker.py
Chat and generation
- Chat service
backend/app/services/chat_service.py - Prompt builder
backend/app/llm/prompt_builder.py - Prompt templates
backend/app/llm/prompts.py - Gemini and fallback
backend/app/llm/gemini_client.py - Conversation memory
backend/app/memory/conversation_memory.py
State and configuration
- Settings
backend/app/core/config.py - Database session
backend/app/database/session.py - Chat history model
backend/app/database/models/chat.py - Document model
backend/app/database/models/document.py - User model
backend/app/database/models/user.py
Other components
- User
Connections 22
- User → Demo frontend: uses
README.md - Demo frontend → FastAPI app: opens demo
README.md - FastAPI app → RAG routes: registers routes
backend/app/main.py - User → RAG routes: uploads and asks
README.md - RAG routes → Chat service: delegates
backend/app/api/routes/rag.py - Chat service → Hybrid search: ingests and searches
backend/app/services/chat_service.py - Format loaders → PDF DOCX TXT loaders: dispatches by format
backend/app/ingestion/pipeline.py - Vector search → Format loaders: loads document
backend/app/retrieval/vector_search.py - Vector search → Text chunking: chunks text
backend/app/retrieval/vector_search.py - Vector search → Embedding service: embeds text
backend/app/retrieval/vector_search.py - Vector search → FAISS index: indexes and searches
backend/app/retrieval/vector_search.py - Hybrid search → Vector search: gets vector results
backend/app/retrieval/hybrid_search.py - Hybrid search → BM25 retrieval: gets keyword results
backend/app/retrieval/hybrid_search.py - Hybrid search → Result reranker: reranks combined results
backend/app/retrieval/hybrid_search.py - Chat service → Prompt builder: builds prompt
backend/app/services/chat_service.py - Prompt builder → Prompt templates: uses templates
backend/app/llm/prompt_builder.py - Chat service → Gemini and fallback: generates answer
backend/app/services/chat_service.py - Chat service → Conversation memory: reads and updates
backend/app/services/chat_service.py - Database session → Settings: reads settings
backend/app/database/session.py - Chat history model → Database session: imports base
backend/app/database/models/chat.py - Document model → Database session: imports base
backend/app/database/models/document.py - User model → Database session: imports base
backend/app/database/models/user.py