(HKA-66) Cache song search results to reduce repeated fetch

[hkalbertkim]

Summary

Adds an in-memory TTL cache to the searchYouTube function in server/src/services/youtube.ts. Repeated searches for the same query now return cached results instead of hitting the YouTube Data API again.

Why

Every search triggers 2 YouTube API calls (search + video details). The YouTube API has a daily quota (10,000 units); each search+details pair costs ~101 units. Users often re-search the same terms. Caching eliminates redundant API calls, reduces latency, and preserves quota.

What changed

• server/src/services/youtube.ts — Added:
  • searchCache: a Map<string, CacheEntry> keyed by lowercased query
  • getCachedSearch(query): returns cached results if within TTL (5 min)
  • setCachedSearch(query, results): stores results with timestamp; evicts oldest entry when cache exceeds 100 entries
  • Cache check runs before any API call in searchYouTube()
  • Structured log lines: [search-cache] HIT / [search-cache] MISS

How to test

1. Start the server: cd server && npm run dev
2. Search for a song: curl 'http://localhost:4000/api/search?q=bohemian+rhapsody'
3. Check server logs — should see [search-cache] MISS
4. Repeat the same search within 5 minutes
5. Check server logs — should see [search-cache] HIT and response is instant
6. Wait >5 minutes and search again — should see MISS (TTL expired)

Risks / rollback

• Low risk: Cache is in-memory only, no persistence. Server restart clears it.
• Rollback: Revert this single commit; no schema/data changes.
• Edge case: Stale results for up to 5 minutes after a video is published/removed — acceptable for song search.

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

Project	Deployment	Actions	Updated (UTC)
sing-sync	Ready	Preview, Comment	Feb 11, 2026 4:57am