PDF Q&A Bot

RAG-based PDF question-answering app with:

• Frontend: React app (frontend/)
• Backend API: Node + Express (server.js)
• RAG Service: FastAPI + Hugging Face + FAISS (rag-service/)

Upload a PDF, ask questions from its content, and generate a short summary.

Architecture

1. Frontend uploads file to Node backend (/upload)
2. Node forwards file path to FastAPI (/process-pdf)
3. FastAPI loads/splits PDF, builds vector index with embeddings
4. For /ask and /summarize, FastAPI retrieves relevant chunks and generates output with a Hugging Face model

Project Structure

.
├── frontend/           # React UI
├── rag-service/        # FastAPI RAG service
├── server.js           # Node API gateway
├── uploads/            # Uploaded files (runtime)
└── CONTRIBUTING.md

Prerequisites

• Node.js 18+ (LTS recommended)
• Python 3.10+
• pip

1) Clone and Install Dependencies

From repository root:

npm install
cd frontend && npm install
cd ../rag-service && python -m pip install -r requirements.txt

2) Environment Variables

Create .env in repo root (or edit existing):

# Optional model override
HF_GENERATION_MODEL=google/flan-t5-base

Notes:

• OPENAI_API_KEY is not required for current Hugging Face RAG flow.
• Keep real secrets out of git.

3) Run the App (3 terminals)

Terminal A — RAG service (port 5000)

cd rag-service
uvicorn main:app --host 0.0.0.0 --port 5000 --reload

Terminal B — Node backend (port 4000)

cd /workspaces/pdf-qa-bot
node server.js

Terminal C — Frontend (port 3000)

cd /workspaces/pdf-qa-bot/frontend
npm start

Open: http://localhost:3000

API Endpoints

Node backend (http://localhost:4000):

• POST /upload (multipart form-data, field: file)
• POST /ask ({ "question": "..." })
• POST /summarize ({})

FastAPI RAG service (http://localhost:5000):

• POST /process-pdf
• POST /ask
• POST /summarize

Interactive docs: http://localhost:5000/docs

Troubleshooting

• Cannot POST /upload from frontend

  • Restart frontend dev server after config changes: npm start
  • Ensure Node backend is running on port 4000

• Upload fails / connection refused

  • Ensure FastAPI is running on port 5000

• Slow first request

  • Hugging Face model downloads on first run (can take time)

• Port already in use

  • Stop old processes or change ports consistently in frontend/backend/service

Development Notes

• RAG index is in-memory (rebuilds after restart)
• Summarization and QA use retrieved context from the last processed PDF

Contributing

See CONTRIBUTING.md.