• AI processing takes too long.
  • This is because I have to make 3 separate calls. 1 to the lyrics endpoint and 2 others to Gemini.
  • It feels long because I didn’t implement streaming. The kind of experience you see on ChatGPT where you don’t wait to see the full response, but rather see the responses as AI is generating them.
  • I didn’t implement caching, where I’d save the search results in a database for faster retrieval next time a user makes a similar request
• I’m using lyrics.ovh API for fetching lyrics of songs which is highly unreliable. See Tech stack, LLM and Lyrics API for more. ⚠️ For that reason, the app will keep crashing

<aside> ✨

This exercise was meant to be a small one, and I didn’t invest much time to handle complex edge cases and fix bugs. But I’ll keep the source codes public in case you might be interested to fix one of the issues.

</aside>