`# Dify Documentation Expert (RAG Chatbot)

A Retrieval-Augmented Generation (RAG) chatbot for Dify documentation, using MiniLM for semantic search and DistilBERT for answer generation. Built with Python, Flask, LangChain, and FAISS. Deployable on Vercel.

---

## Features
- Scrapes and processes Dify documentation
- Builds a FAISS vector store with MiniLM embeddings
- Uses DistilBERT for question answering
- Fast API with `/ask` and `/health` endpoints
- Ready for serverless deployment (Vercel)

---

## Project Structure
```
Dify-Expert/
├── api/
│   └── dify_api.py         # Flask API (entry point for Vercel)
├── dify_docs_scraper/
│   ├── dify_rag.py         # RAG logic (vector store + QA)
│   ├── dify_faiss_index/   # FAISS index files
│   └── ...                 # Scraper, processing scripts, etc.
├── requirements.txt        # Python dependencies
├── vercel.json             # Vercel config
├── README.md               # This file
└── futurescope.md          # Project roadmap/ideas
```

---

## Quickstart

### 1. Install dependencies
```bash
pip install -r requirements.txt
```

### 2. Build the vector store (if not already built)
```bash
python3 dify_docs_scraper/build_vector_store.py
```

### 3. Run locally (API)
```bash
python3 api/dify_api.py
```

### 4. Query the API
```bash
curl -X POST http://localhost:8000/ask \
  -H "Content-Type: application/json" \
  -d '{"question": "What is Dify?"}'
```

---

## Deployment (Vercel)
1. Push to GitHub.
2. Connect repo to Vercel.
3. Deploy (Vercel auto-detects Python from requirements.txt).
4. Endpoints:
   - `/ask` (POST)
   - `/health` (GET)

---

## API Endpoints
- `POST /ask` — `{ "question": "..." }` → `{ "answer": ..., "score": ..., "sources": [...] }`
- `GET /health` — `{ "status": "ok" }`

---

## Roadmap
See [`futurescope.md`](futurescope.md) for planned features and ideas.

---

## License
MIT