Theory-Aware Music Assistant: Turning Your Playlist Into an Ear Training Coach

I built an AI agent that personalizes ear training by connecting your weaknesses to music you already love. It takes your ear training test results, identifies the intervals or chord progressions you struggle with, then pulls your Spotify playlist and locates the exact moments in real songs where those concepts appear — with reasoned explanations of why that musical passage will help the concept stick.
For the demo, I’ll walk through each node of the agent live: test input → weakness detection → Spotify lookup → timestamped music moments → chain-of-thought reasoning. Then I’ll take the audience through my development journey — where LLMs completely fell apart trying to work directly with audio files, why plugging in a real LALM like MusicFlamingo turned out to be too heavy to be practical, and the API rate-limit management pattern I landed on to keep everything running smoothly inside the agent on free tiers.