How rnktyagi/pdf_rag_chatbot works
This PDF question-answering product lets users upload PDFs and ask document-grounded questions through a chat UI. Uploads are converted to contextualized chunks and indexed; chat requests enter a stateful agent, which can retrieve user-scoped passages and answer using an LLM. The README provides little architectural detail, so the map relies on the sampled source; unsampled capabilities and integrations are not inferred.
Main entry points: main.py FastAPI endpoints · graph.py RAG agent · embedding.py Embedding and indexing · retrieval.py Hybrid retriever
Components 12
User interface
- Chat UI
app.py
API and agent
- FastAPI endpoints
src/main.py - RAG agent
src/graph.py - Retrieval tool
src/graph.py
Document processing
- PDF chunking
src/ingestion.py - Embedding and indexing
src/embedding.py
Retrieval
- Hybrid retriever
src/retrieval.py
External services
- Qdrant vector store
- Cross-encoder reranker
- Postgres checkpoints
- Groq-hosted LLM
Other components
- User
Connections 13
- User → Chat UI: asks or uploads
app.py - Chat UI → FastAPI endpoints: posts requests
app.py - FastAPI endpoints → Embedding and indexing: processes upload
src/main.py - Embedding and indexing → PDF chunking: chunks PDF
src/embedding.py - Embedding and indexing → Qdrant vector store: indexes chunks
src/embedding.py - FastAPI endpoints → RAG agent: invokes chat
src/main.py - FastAPI endpoints → Postgres checkpoints: sets up checkpoints
src/main.py - RAG agent → Retrieval tool: dispatches retrieval
src/graph.py - RAG agent → Groq-hosted LLM: generates answers
src/graph.py - Retrieval tool → Hybrid retriever: searches for user
src/graph.py - Hybrid retriever → Qdrant vector store: searches chunks
src/retrieval.py - Hybrid retriever → Cross-encoder reranker: reranks matches
src/retrieval.py - Retrieval tool → RAG agent: returns passages
src/graph.py