How mayooear/ai-pdf-chatbot-langchain works
This is a PDF-ingestion and question-answering chatbot: users upload PDFs for indexing, then ask questions and receive streamed answers with sources. The README documents the frontend/backend interaction and major capabilities; sampled graph code shows ingestion, routing, retrieval, and response generation. Some frontend-to-backend wiring and integrations are documented rather than visible in the excerpts, so those links cite the README; no unsampled implementation details are assumed.
Main entry points: retrieval_graph/graph.ts Query routing and answers · ingestion_graph/graph.ts Document ingestion · retrieval.ts Vector retrieval · page.tsx Chat and upload UI
Components 19
Chat interface
- Chat and upload UI
frontend/app/page.tsx - LangGraph client
frontend/lib/langgraph-base.ts
Application endpoints
- Chat endpoint
frontend/app/api/chat/route.ts - Ingest endpoint
frontend/app/api/ingest/route.ts
LangGraph workflows
- Document ingestion
backend/src/ingestion_graph/graph.ts - Ingestion state
backend/src/ingestion_graph/state.ts - Ingestion configuration
backend/src/ingestion_graph/configuration.ts - Query routing and answers
backend/src/retrieval_graph/graph.ts - Query state
backend/src/retrieval_graph/state.ts - Query configuration
backend/src/retrieval_graph/configuration.ts - Routing and answer prompts
backend/src/retrieval_graph/prompts.ts - Document formatting
backend/src/retrieval_graph/utils.ts
Models and retrieval
- Retriever configuration
backend/src/shared/configuration.ts - Vector retrieval
backend/src/shared/retrieval.ts - Document reducer
backend/src/shared/state.ts - Chat model loader
backend/src/shared/utils.ts
Other components
- User
- Supabase vector store
- LLM provider
Connections 25
- User → Chat and upload UI: uses
- Chat and upload UI → Ingest endpoint: uploads PDFs
- Chat and upload UI → Chat endpoint: asks questions
- Chat endpoint → LangGraph client: streams answers
- Ingest endpoint → LangGraph client: submits documents
- Ingest endpoint → Document ingestion: invokes graph
- Chat endpoint → Query routing and answers: invokes graph
- Document ingestion → Ingestion state: uses state
backend/src/ingestion_graph/graph.ts - Document ingestion → Ingestion configuration: loads configuration
backend/src/ingestion_graph/graph.ts - Document ingestion → Document reducer: reduces documents
backend/src/ingestion_graph/graph.ts - Document ingestion → Vector retrieval: adds documents
backend/src/ingestion_graph/graph.ts - Ingestion state → Document reducer: reduces state
backend/src/ingestion_graph/state.ts - Query routing and answers → Query state: uses state
backend/src/retrieval_graph/graph.ts - Query routing and answers → Query configuration: loads configuration
backend/src/retrieval_graph/graph.ts - Query routing and answers → Vector retrieval: retrieves documents
backend/src/retrieval_graph/graph.ts - Query routing and answers → Document formatting: formats context
backend/src/retrieval_graph/graph.ts - Query routing and answers → Routing and answer prompts: uses prompts
backend/src/retrieval_graph/graph.ts - Query routing and answers → Chat model loader: loads chat model
backend/src/retrieval_graph/graph.ts - Query state → Document reducer: reduces documents
backend/src/retrieval_graph/state.ts - Ingestion configuration → Retriever configuration: extends configuration
backend/src/ingestion_graph/configuration.ts - Query configuration → Retriever configuration: extends configuration
backend/src/retrieval_graph/configuration.ts - Vector retrieval → Retriever configuration: validates configuration
backend/src/shared/retrieval.ts - Vector retrieval → Supabase vector store: reads and writes embeddings
backend/src/shared/retrieval.ts - Vector retrieval → LLM provider: creates embeddings
backend/src/shared/retrieval.ts - Chat model loader → LLM provider: loads model