Feature/svd online updating

[Prateek2007-cmd]

Description

This PR resolves Issue #1596 by implementing Bayesian online updating for the Collaborative Filtering model. Previously, the system lacked a mechanism to efficiently integrate new interactions, often requiring expensive, full Truncated SVD matrix reconstructions for every transaction or going stale.
Closes #1639
Changes included:

• In-Memory Micro-Updates: Modified the /api/purchases endpoint in backend/main.py to trigger the online_update() SGD method within the loaded CollaborativeRecommender instance. This dynamically nudges the specific user and item latent vectors towards the new interaction error gradient without a full matrix recalculation.
• Scheduled Nightly Rebuilds: Added a celery_app.conf.beat_schedule in celery_app.py to trigger a new tasks.rebuild_models_task via cron (daily at 3:00 AM UTC). This runs the heavy, full matrix factorization offline during off-peak hours to recalibrate the baseline factors and expand the matrix dimensionality for newly joined users/items.

Related Issues

• Resolves #1596

Type of Change

• New Feature (Online Model Updating / Scheduled Retraining)
• Performance Improvement

Testing

• Confirmed /api/purchases correctly intercepts the incoming payload and triggers collab_model.online_update in the global cache.
• Verified celery_app.conf.beat_schedule correctly registers the rebuild_models_task for execution via celery beat.

[github-actions]

🎉 Welcome to Hybrid Recommender, @Prateek2007-cmd! This is your first contribution here!

Labels added: gssoc:approved | mentor:leonagoel | status:review-needed

PR Description Checklist:

NO - What changed section
NO - Why section
NO - How to test section
YES - Related issue linked

⚠️ Some required sections are missing. Please update your PR description.

What happens next:

1. @leonagoel will review your changes
2. CI checks must pass
3. Once approved, this PR will be auto-merged

⏱️ Please respond to review comments within 48 hours.