A2A-Stateless-Message

# A2A Adapter for a LangChain Bot This example demonstrates how to create an **Agent-to-Agent (A2A) adapter** for an existing *'legacy'* AI service. The architecture consists of two separate services: - `legacy.py`: A simple FastAPI server that uses LangChain and OpenAI to provide the core chat logic. It knows nothing about the A2A protocol. - `server.py`: An A2A-compliant server that acts as a proxy. It translates incoming A2A messages, calls the `legacy.py` service, and returns the response in the standard A2A format. The `client.py` communicates exclusively with the `server.py` adapter. --- ## Prerequisites - Python and `uv` (Universal Virtualenv Runner) - An `.env` file in the root directory containing your `OPENAI_API_KEY="sk-..."` - Three separate terminal windows --- ## How to Run You need to start both services before running the client. ### Terminal 1: Start the Legacy Bot Service This is the internal AI logic service. ```bash uv run python legacy.py ``` --- ### Terminal 2: Start the A2A Adapter Server This server listens for A2A requests and forwards them to the legacy service. ```bash uv run python server.py ``` --- ### Terminal 3: Start the Client The client communicates with the A2A adapter server. ```bash uv run python client.py ``` --- ## Usage You can now type messages in the client terminal (Terminal 3), for example: - `hello` - `What is on your menu?` - `Are you open on Mondays?` --- 💡 **Note**: This example highlights a powerful pattern: using A2A to expose a non-A2A-compliant service to a wider network of agents **without modifying the original service's code**.