Phase 1: Reproducible Tas baseline from SAM_UI with no behaviour change (gate G0) #228

Parent: #227

Goal

Record a reproducible Tas baseline that later comparisons can rely on. No change to Tas behaviour or code paths.

Context

• Current SAM_UI Tas path: WPF/SAM.Analytical.UI.WPF/Modify/Simulate.cs:769 (TBD preparation :923-955) → Modify/RunWorkflow.cs:49 (WorkflowCalculator.Calculate()).
• The SAM_Tas benchmark-tas CLI (STA thread, built with .NET Framework MSBuild) writes B1a documents (SAM_Validation docs/benchmark/SCHEMA.md).
• Known limits: SingleBox is illegal in Tas ("illegal construction"). The only recorded live run (HungaryHouse, 2026-07-24) used mismatched Boston weather.

Scope

• Choose and commit two models:
  • Case A: a Tas-legal single zone.
  • Case B: a representative multizone building.
  • For each, record provenance and SHA256.
• Choose one EPW per case that suits the location, and record its hash.
• Run each case through (1) the current SAM_UI Simulate path and (2) benchmark-tas, with identical model, weather and design-day source.
• Record the SAM_UI, SAM_Tas and SAM commits, the Tas version, canonicalModelHash, the weather hash and the run duration.

Out of scope

Any code change to the Tas workflow; OpenStudio.

Depends on

Nothing.

Acceptance criteria

• Two consecutive runs of each case give identical B1a metrics.
• The B1a metrics from the SAM_UI Tas path match benchmark-tas for the same inputs. Any difference is explained, for example by different WorkflowSettings such as sizing, IZAMs or unmet hours.
• The WorkflowSettings used are documented, because they define the "Tas baseline" assumptions.

Validation gate G0

Evidence

documentation/validation/tas-baseline/ (models or links to them, hashes, B1a JSON, run notes). Anything only on the licensed machine is labelled machine-local.