Skip to content

README.md

Latest commit

 

History

History
40 lines (27 loc) · 1.84 KB

README.md

File metadata and controls

40 lines (27 loc) · 1.84 KB

FastAPI With Langgraph

This is an example application demonstrating how to integrate a LangGraph chat agent with the Thesys C1 SDK (@thesysai/genui-sdk) using the LangGraph CLI for a persistent, multi-turn chat experience with tool usage.

Built with Thesys

Getting Started

  1. Run the Backend Server: See backend/README.md for instructions on using langgraph dev.
  2. Run the Frontend UI: See ui/README.md for instructions.

Overview

This project integrates LangGraph with the Thesys C1 SDK as follows:

  1. LangGraph Server (via CLI):

    • Uses langgraph-cli to serve the graph defined in graph.py.
    • Exposes a standard LangGraph API (defaulting to port 2024) for thread management, state persistence, and streaming.
    • Configured via langgraph.json.

  2. LangGraph Agent (graph.py):

    • Defines agent state (AgentState) including messages and response_id.
    • Uses the built-in add_messages reducer for handling message updates/appends.
    • Configures a graph with an agent node (calling the C1 LLM with tools) and a tools node.
    • Assigns the response_id from the state to the final AIMessage ID within the agent node.
    • State is automatically persisted by the LangGraph server.

  3. Frontend Integration (ui/src/App.tsx):

    • Uses @thesysai/genui-sdk hooks (useThreadListManager, useThreadManager) and the C1Chat component.
    • Communicates with the LangGraph API for thread persistence and streaming.
    • Uses a custom processMessage handler to transform the LangGraph SSE stream into raw text chunks for the SDK.
    • Synchronizes the selected thread ID with the browser URL using window.history.

Example Prompt

What is the weather in California?