mirror of
https://github.com/Hestia-Homes/Model.git
synced 2026-07-12 13:29:04 +00:00
b37ee071b7
Review findings on PR #1526, behaviour-preserving: - No fabric groups, or phase 1 committed nothing → delegate to the plain optimise_package: the phase-2 prefix would be empty and its re-scoring would reproduce the role-1 signals the groups already carry, one full SAP-calculator run per Option for zero effect (the common case on already-insulated stock). - Phase 1 consumed every group → return the fabric package instead of a phase-2 pass that can only select the empty package. - _rescored_groups takes start_sap from the caller: the post-fabric baseline score is the phase-1 package score already in hand, not a fresh calculator run. - fabric_types → phase_one_types: the set holds everything phase 1 committed, injected dependencies included — that inclusion is what the outstanding-dependencies filter relies on, so the name must not invite narrowing it to FABRIC_MEASURE_TYPES. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> |
||
|---|---|---|
| .. | ||
| __init__.py | ||
| measure_dependency.py | ||
| optimiser.py | ||