fix: wire ML retraining pipeline to actual training and persistence (KI-83)

fix: wire ML retraining pipeline to actual training and persistence #83

Merged

wasabi opened 2 days ago

Summary

Connects the ML retraining scaffolding to actual model training and persistence.

Changes

trigger_retrain(): now calls fit_prophet/fit_isolation_forest and save_models based on model name
Updates last_retrain_at and records accuracy_after/improvement_pct
Auto-retrain: when schedule has auto_retrain=True, automatically triggers retrain on detection
Training failures are caught and logged without crashing

Tests

9 tests covering training invocation, model persistence, auto-retrain, and failure handling.

Commits were merged into target branch

fix: wire ML retraining pipeline to actual training and persistence

- trigger_retrain() now calls fit_prophet/fit_isolation_forest and save_models
- Update last_retrain_at and record accuracy_after/improvement_pct
- Auto-retrain when schedule has auto_retrain=True
- Add 9 tests covering training invocation, persistence, auto-retrain

2 days ago

8d09f72f

wasabi merged 2 days ago
Login to comment

Submitter	wasabi
Target	main
Source	fix/ml-retraining-pipeline

Jobs

Merge Strategy

Create Merge Commit

Watchers (1)

Reference

pull request KI-83