An agentic Knowledge Graph RAG application that ingests documents into a hybrid retrieval system (Neo4j graph + pgvector), with an autonomous AI agent that reasons over the knowledge to answer queries through a mobile-first web interface.
Chat
Agent Reasoning
Documents
Knowledge Graph
Base Knowledge Ingestion
Frontend (Next.js 16) Backend (FastAPI) Infrastructure
┌──────────────────┐ ┌──────────────────────┐ ┌──────────────────┐
│ App Router │────▶│ REST / WebSocket │────▶│ PostgreSQL + │
│ shadcn/ui │ │ Domain modules: │ │ pgvector │
│ Tailwind CSS 4 │ │ auth, documents, │ │ Neo4j 5 │
│ TypeScript 5 │ │ knowledge, agents, │ │ Redis 7 │
└──────────────────┘ │ ingestion │ │ Supabase Storage│
│ Background worker │ └──────────────────┘
└──────────────────────┘
Two-tier knowledge:
- Base KG -- Permanent knowledge ingested from Google Drive via agent swarm
- User uploads -- PDF/DOCX files with 7-day TTL, auto-extracted into graph + vectors
| Layer | Tech |
|---|---|
| Frontend | Next.js 16, TypeScript 5 (strict), Tailwind CSS 4, shadcn/ui |
| Backend | FastAPI, Python 3.12+, SQLAlchemy 2.0 (async), Pydantic |
| Database | PostgreSQL 17 + pgvector, Neo4j 5, Redis 7 |
| AI | LangChain + LiteLLM, Claude Sonnet 4 (primary), GPT-4o (fallback) |
| Storage | Supabase Storage (uploads), Google Drive (base KG source) |
| Infra | Docker Compose, Vercel (frontend), VPS (backend) |
- Docker + Docker Compose
- Node.js 18+ with
pnpm - Python 3.12+ with
uv
cp .env.example .env # Edit with your API keys
# Full local stack (Postgres + Neo4j + Redis + backend + worker + frontend)
make dev
# Seed dev accounts (admin@dingdong.dev / admin123, user@dingdong.dev / user123)
make seedThe app will be available at:
- Frontend: http://localhost:3000
- Backend API: http://localhost:8000/api/v1
- API Docs: http://localhost:8000/api/v1/docs
- Neo4j Browser: http://localhost:17474
| Guide | Description |
|---|---|
| Development | Project structure, commands, API endpoints, environment variables |
| Self-Hosting | Full setup guide for your own infrastructure |
| Knowledge Seeding | Google Drive ingestion, graph export/import, user uploads |
| Deployment | Production deployment (backend VPS, frontend Vercel/Docker) |
Private