smoke-gateway

You can run AI Gateway smoke tests by providing a virtual key and optional routing flags to execute a full test suite across all catalog models. The harness outputs a Markdown report detailing the validation results.

Yes, AI Gateway smoke testing supports non-stream, SSE, and embedding workflows. The test harness validates full-surface coverage across all catalog models, ensuring routing rules and traffic metrics function correctly across these specific request types.

Automated smoke testing validates routing rules through per-rule isolation and automatic management. It cross-checks traffic_event database rows against Prometheus counters, triggering auto-fix loops when metric validation failures are detected.

When a smoke test detects a failure in the AI Gateway Control Plane, a self-healing loop initiates. This loop automatically rebuilds, restarts the gateway, and reruns the tests until a green status is achieved, ensuring reliability without manual intervention.

You can use this smoke test harness for regression validation across development, test, and production-like environments. It provides cross-model coverage across P3, P3E embeddings, and routing scenarios to verify system stability.

Executing a full smoke suite against the AI Gateway requires inputs such as a virtual key, routing flags, model selections, and reporting parameters. These inputs allow the test harness to configure and validate the necessary workflows.