[bojemoi] fix(ml-threat): auto-train on synthetic data if models missing or corrupt

Commit 450d681

Description

Fallback at startup: if pkl absent or incompatible (sklearn version mismatch), import train.train_models() and retrain with 2000 synthetic samples before loading, so the service is always operational without manual intervention.

Co-Authored-By: Claude Sonnet 4.6 noreply@anthropic.com

Files Changed

Diff Summary