video1.0<iframe src="https://www.loom.com/embed/111a6eb5e5b047bcabf5c3b307aba7cf" frameborder="0" width="1480" height="1110" webkitallowfullscreen mozallowfullscreen allowfullscreen></iframe>11101480Loomhttps://www.loom.com11101480https://cdn.loom.com/sessions/thumbnails/111a6eb5e5b047bcabf5c3b307aba7cf-810e8b30ad5a6bf3.gif311.7333In this Loom, I walk through my beat scouter, an AI music recommender that turns plain English requests into structured song inputs like genre, mood, and energy, then returns Spotify tracks. I show the parsing step with the Gemma model, a quality check that scores results and retries until the top options pass, and the agentic logic that surfaces the top 5 chord matches with a short reason why. I also run a reliability harness with 7 predefined inputs and a unit test suite with 11 units, and all checks pass. No viewer action was requested.