🔬 Open Deep Research UI

Built with ❤️ by @artreimus from Ylang Labs

A powerful, full-stack deep research application - Combining a LangGraph backend and a Next.js assistant-UI frontend to deliver comprehensive research capabilities with multiple model providers, search APIs, and optional MCP servers. Local dev friendly; can target LangGraph Cloud/Platform or self-hosted servers.

📸 App

🏗️ Monorepo Layout

• backend/ — LangGraph server for Open Deep Research (Python, uv, langgraph-cli)
• frontend/ — Next.js app using assistant‑ui and LangGraph SDK

🚀 Quick Start (Local)

• Backend
  • cd backend
  • cp .env.example .env then set provider keys (e.g., OPENAI_API_KEY, TAVILY_API_KEY, etc.)
  • uv sync
  • Start server: uvx --refresh --from "langgraph-cli[inmem]" --with-editable . --python 3.11 langgraph dev --allow-blocking
  • Server runs at http://127.0.0.1:2024 (Studio, API, docs)
• Frontend
  • cd frontend
  • Create .env.local with:
    • LANGGRAPH_API_URL=http://127.0.0.1:2024
    • NEXT_PUBLIC_LANGGRAPH_ASSISTANT_ID="Deep Researcher" (matches backend/langgraph.json)
    • LANGCHAIN_API_KEY=... if your server requires an x-api-key (LangGraph Cloud/Platform)
    • Optional: NEXT_PUBLIC_LANGGRAPH_API_URL=... to bypass the proxy (requires CORS + client key handling)
  • pnpm install
  • pnpm dev → open http://localhost:3000

Ask a research question in the chat. Use the right‑hand Settings tab to tweak models, search provider, concurrency, and MCP tooling; these are forwarded to the backend as LangGraph configurable fields per run.

🌍 Environment

• Backend (backend/.env)
  • Model providers: OPENAI_API_KEY, ANTHROPIC_API_KEY, GOOGLE_API_KEY, OPENROUTER_API_KEY, GROQ_API_KEY, etc.
  • Search: TAVILY_API_KEY (default), others optional per your config
  • LangSmith/Platform: LANGSMITH_API_KEY, LANGSMITH_PROJECT, LANGSMITH_TRACING
  • Options: AUTO_MODEL_SELECTION=true, STRICT_PROVIDER_MATCH=false
  • Optional OAP/Supabase integration: SUPABASE_URL, SUPABASE_KEY, GET_API_KEYS_FROM_CONFIG
• Frontend (frontend/.env.local)
  • LANGGRAPH_API_URL — Base URL of your LangGraph server
  • LANGCHAIN_API_KEY — Added as x-api-key by the Next.js proxy when calling the server
  • NEXT_PUBLIC_LANGGRAPH_ASSISTANT_ID — Assistant/graph ID (default: "Deep Researcher")
  • NEXT_PUBLIC_LANGGRAPH_API_URL — Optional direct browser→server base URL (CORS + API key handling required)

🛠️ Dev Commands

• Backend
  • Install: uv sync
  • Run: uvx --refresh --from "langgraph-cli[inmem]" --with-editable . --python 3.11 langgraph dev --allow-blocking
  • Tests: uv run -m pytest -q
  • Lint/Type: uv run ruff check . · uv run mypy src
• Frontend
  • Dev/Build/Start: pnpm dev · pnpm build · pnpm start
  • Lint: pnpm lint

🚀 Deployment

• Backend
  • LangGraph Cloud/Platform recommended for durable hosting
  • Self‑hosted: run LangGraph Server behind a load balancer; ensure x-api-key or auth is enforced as needed
• Frontend
  • Any Node host (e.g., Vercel). Set env vars to point at your hosted LangGraph server
  • Prefer keeping calls through the built‑in /api proxy to attach x-api-key

💖 Acknowledgments

• LangChain Team - For the amazing Open Deep Research framework and LangGraph
• Assistant UI - For the excellent assistant-ui components and framework
• Community - For feedback, contributions, and support

🔧 Troubleshooting

• 401/403 from server: check LANGCHAIN_API_KEY and that you are using the /api proxy
• Assistant not found: confirm NEXT_PUBLIC_LANGGRAPH_ASSISTANT_ID matches backend
• CORS errors: remove NEXT_PUBLIC_LANGGRAPH_API_URL (use proxy) or enable CORS on the server

📄 License

This project is licensed under the Apache License 2.0. See the LICENSE file for details.