中文 | English

Local, privacy-first RAG using PostgreSQL + pgvector and Ollama. Streams responses via SSE. Simple, fast, hackable.

• Process management 【LangGraph】
• Documents transform core tools 【Docling】、supprt PDF、DOCX、PPTX、Image、HTML、AsciiDoc、Markdown
• Semantic search with pgvector (768-dim nomic-embed-text)
• REST + Streaming responses via SSE
• Optional frontend: React + Vite + TypeScript
• Local models via Ollama or openai api

• Local Setup
• Run
• Frontend (optional)
• License

• env configure

pip install -r requirements.txt
docling-tools models download 
cp env.example .env   # edit DB_* for your environment
#python scripts/init_db.py   # if not using docker init

python main.py

App: http://localhost:8000

• Swagger UI: http://localhost:8000/docs
• ReDoc: http://localhost:8000/redoc

React + Vite + TypeScript app lives in frontend-app/.

Dev

cd frontend-app
npm install          # or: pnpm i / yarn
npm run dev          # http://localhost:5173

MIT. See LICENSE.