7. Streaming and Multi-turn

The HelloWorld sample proves basics. Real systems need richer task state, incremental updates, and multi-turn continuation.

Why this step matters

• Beginners see how A2A task state works in practice.
• Developers learn how to emit status and artifact events correctly.
• Advanced teams validate continuation and recovery behavior.

Run the LangGraph sample

1. Configure GOOGLE_API_KEY for the sample.
2. Start the server in the LangGraph sample directory.
3. Run test_client.py from another terminal.

What to observe

• TaskStatusUpdateEvent during intermediate processing.
• TaskArtifactUpdateEvent when final outputs are produced.
• Task continuation after input_required responses.

Multi-turn continuation pattern

1. Client sends an initial ambiguous request.
2. Agent returns input_required with clarification prompt.
3. Client sends another message with the same taskId and contextId.
4. Server continues the same task until completion.

Important v1.0 alignment

• Do not rely on a final flag in status events.
• Use task state transitions plus stream closure to detect completion.

Next step

Continue to 8. Next Steps .