run-locally

To test streaming responses during local agent development, launch the server with uv run start-app and send curl requests with JSON payloads to the localhost /invocations endpoint to verify streaming and non-streaming behavior.

Port conflicts or missing modules during local server startup are common configuration errors. Troubleshoot these local errors by verifying port availability, ensuring proper module installation, and checking your .env authentication setup.

Yes, you can use MLflow to run evaluation scorers on a locally hosted agent. Start the server locally, then invoke the evaluation tests to validate response quality and track experiment configurations using MLflow.

The best way to verify agent configuration and API correctness on localhost is to run the server locally, test invocations with curl, and execute unit tests with pytest to confirm the endpoint behaves correctly.

Yes, you need uv run commands to start the development server for API testing. Launching the agent on localhost requires using uv-run commands to start the app and invoke the Responses-compatible endpoint.

To troubleshoot .env authentication issues when running an agent locally, verify your .env file configuration and ensure all required environment variables are loaded correctly before starting the local development server.