serena

Run the pipeline-level integration test suite for the STT recognition loop. Tests drive the full run_stt_worker → _recognition_loop → on_stt_event path using a ScriptedBackend (no real microphone or STT model). Use when you want to verify the recognition pipeline is correctly wired — wake detection, command matching, direct triggers, one-breath, and reply window.

Run an end-to-end STT test that simulates a real voice conversation with alexa-custom. Synthesises Italian speech using piper TTS, feeds it through the configured STT backend (Vosk or sherpa-onnx), and reports recognition results. Use when you want to verify STT is working without a real microphone.