Skip to content

README.md

Latest commit

 

History

History
91 lines (70 loc) · 3.09 KB

README.md

File metadata and controls

91 lines (70 loc) · 3.09 KB

LangGraph A2A Conversational Agent (TypeScript)

This project demonstrates an Agent-to-Agent (A2A) conversation using LangGraph with the A2A protocol in TypeScript.

Setup

  1. Install dependencies:

    cd typescript
npm install

  2. Configure environment variables:

    cp .env.example .env

    Edit .env and add your OPENAI_API_KEY.

Running the Agents

  1. Start the first agent server:

    npx @langchain/langgraph-cli dev --port 2024

    Copy the assistant_id from the output.

  2. In another terminal, start the second agent server:

    npx @langchain/langgraph-cli dev --port 2025

    Copy the assistant_id from this output as well.

    Access API Documentation:

    • LangGraph Studio: The server output will show a Studio UI link (e.g., https://smith.langchain.com/studio?baseUrl=http://localhost:2025). This is the primary way to view API documentation, test endpoints, and visualize your graph in the TypeScript version.
    • Note: The TypeScript LangGraph dev server does not expose a /docs endpoint like the Python version. Use LangGraph Studio for API exploration.

  3. Configure the assistant IDs: Add the following to your .env file:

    AGENT_A_ID=<assistant_id_from_port_2024>
AGENT_B_ID=<assistant_id_from_port_2025>

  4. Run the conversation simulation:

    npm run conversation

This will simulate a conversation between the two agents, with each agent responding to the other's messages.

Architecture

The implementation consists of two main components:

1. LangGraph Agent (src/langgraph_agent.ts)

  • Defines a conversational agent using LangGraph's StateGraph
  • Uses OpenAI's GPT-4o-mini for responses
  • Maintains message history in state
  • Configured for brief, engaging responses (max 100 tokens)

2. A2A Conversation Orchestrator (src/a2a_conversation.ts)

  • Sends messages between two agent instances using the A2A protocol
  • Uses JSON-RPC 2.0 format for agent communication
  • Orchestrates 3 rounds of back-and-forth conversation
  • Displays conversation with colored output for each agent

Project Structure

typescript/
├── src/
│   ├── langgraph_agent.ts       # Agent definition and logic
│   └── a2a_conversation.ts      # Conversation orchestrator
├── package.json                 # Dependencies and scripts
├── tsconfig.json                # TypeScript configuration
├── .env.example                 # Environment template
└── README.md                    # This file

Scripts

  • npm run dev - Watch mode for development
  • npm run build - Build TypeScript to JavaScript
  • npm run conversation - Run the A2A conversation simulation

Notes

  • Each agent runs as a separate LangGraph server on different ports (2024 and 2025)
  • Both agents use the same logic (GPT-4o-mini) but maintain independent state
  • The A2A protocol enables standardized communication between agents
  • Responses are kept brief (100 token limit) for concise conversations