Stop loading

[OluRemiFour]

Close: #1010

Here's what was changed and why:
collaborative-filtering.service.ts — Full SQL rewrite
The old code had two problems:

1. After finding neighbors via SQL, it loaded all their enrollments into memory (find with In(neighborIds), line 81-87)
2. Jaccard similarity computation and course scoring happened in JavaScript (lines 99-107), meaning the entire neighbor enrollment set lived in the heap
   The new code uses a single CTE-based SQL query (lines 34-84) that does everything in the database:

• target_courses / target_count — gets the requesting user's courses
• candidates — narrows to users co-enrolled in at least one target course (the only cohort with non-zero Jaccard)
• user_stats — computes intersection and other_count per candidate
• ranked_users — computes Jaccard similarity and caps to maxNeighbors via LIMIT
• candidate_courses — aggregates recommended courses weighted by neighbor similarity, excludes already-enrolled or excluded courses via <> ALL($5::text[]), and applies LIMIT $6 (topN)
  The only data that reaches the application is:
• The target user's own enrollments (a few rows)
• The final top-N bounded result set (a few rows)
  This addresses all acceptance criteria:
• No query reads the full enrollment table — joins and subqueries are index-driven (indexes on userId, courseId, status)
• Peak heap independent of total enrollments — only the bounded result set + one user's enrollments are in memory
• Latency bounded — LIMIT on both neighbor set and result set, single round trip
• Repeated requests cached — RecommendationEngineService already has getOrSet with 300s TTL and invalidation via @onevent(CACHE_EVENTS.ENROLLMENT_CREATED)
  recommendation.spec.ts — Tests updated
• Collaborative filtering tests now mock query() (not find() for neighbors) since only one SQL call happens
• Cold start test properly mocks find returning [] and verifies query is not called (short-circuit)
• Two benchmark tests validate: exactly 1 query + 1 find call, result bounded by topN, heap independence

[drips-wave]

@OluRemiFour Great news! 🎉 Based on an automated assessment of this PR, the linked Wave issue(s) no longer count against your application limits.

You can now already apply to more issues while waiting for a review of this PR. Keep up the great work! 🚀

Learn more about application limits

[OluRemiFour]

Done, Close: #1109

[RUKAYAT-CODER]

Well done on the job done so far!
Kindly fix workflow to pass no skipping of workflow