Functionalities:
Playlist analyser: Get the songs in a playlist, get the information about them, extract information such as clusters with numerical features, and create the cluster description using an llm, then pass the clusters and filtered track info to llm to create the analysis.
The first version will contain only one llm call, but the evolution will contain a multi-agent conversation to create the review.
TODO:
- Improve prompt with clear structure of the analysis
- Test with multi-agent frameworks
- Let the user select the model (gemini or open ai)
- Improve clustering method
Playlist Creator: Use the output of the playlist analyser to generate some interesting songs that could be added to that playlist.
Song curiosities: Get the currently listening song, use an llm agent to go into the web, search wikipedia for the song, pass this to an LLM and return the results.