assistant-ui-langgraph-fastapi Demonstration Project

This project is a demonstration of an AI agent architecture that integrates several key technologies:

• LangGraph: Orchestrates the agent logic.
• FastAPI: Serves as the backend server.
• assistant-stream: Enables real-time response streaming from the backend to the frontend.
• assistant-ui & Next.js: Provides a modern, responsive chat interface.

The architecture is split into a backend/ directory (FastAPI + LangGraph) and a frontend/ directory (Next.js client).

The frontend is a Next.js application generated via the assistant-ui CLI. Use yarn to manage dependencies and start the development server:

cd frontend
yarn install
yarn dev

The backend requires environment variables to function. Navigate to the ./backend directory and create a .env file based on the provided template:

1. cd backend
2. Create .env using the contents of .env.example as a guide.

The backend uses poetry for dependency management and the LangChain CLI. It utilizes LangGraph's create_react_agent for agent creation. Run the following commands to start the server:

cd backend
poetry install
poetry run python -m app.server

Before setting up the project, ensure your environment meets the following version requirements:

• Python: 3.11
• Node.js: v20.18.0
• npm: v10.9.2
• Yarn: v1.22.22

The backend server is a FastAPI application that exposes an AI agent via a LangGraph route. You can start the server by running the server.py module directly with Python. By default, it runs on 0.0.0.0:8000 using uvicorn.

python -m backend.app.server

The server exposes the LangGraph agent at a specific route. In the default configuration, the route is /api/chat. This endpoint is integrated into the FastAPI app instance using the add_langgraph_route utility, which connects the assistant_ui_graph to the web server.

add_langgraph_route(app, assistant_ui_graph, "/api/chat")