bump to v1.1.2
Patch release (no algorithmic changes)
Fixes
- Enforced mandatory scoring model (
--model) with no silent fallback - Restored correct progressive profile construction
- Ensured scoring model validation occurs before embedding inference
Notes
- Alignment algorithm, scoring model, and numerical behavior are unchanged
- This release fixes a regression introduced during CLI enforcement