Preserve and use resolved model/data provenance in policyengine-app-v2

Problem

policyengine-app-v2 already carries some version-ish metadata, but it does not preserve or render a full immutable execution bundle for reports and simulations.

Current gaps:

• the old metadata flow uses a single opaque version mostly as a load sentinel, not as reproducibility provenance
• report and simulation domain models preserve apiVersion, but not model/data release identity
• v2 request builders know about dataset IDs, but adapters drop provenance when hydrating responses into domain models
• the reproducibility tab reconstructs code from current metadata defaults and floating dataset conventions instead of the report’s actual resolved bundle
• there is still app-side reliance on “latest” model-version helpers for flows that should be historical and pinned

Relevant code paths include:

• metadata loading and guard logic: app/src/api/metadata.ts, app/src/types/metadata.ts, app/src/reducers/metadataReducer.ts, app/src/routing/guards/MetadataGuard.tsx
• report/simulation models and adapters: app/src/types/ingredients/Report.ts, app/src/types/ingredients/Simulation.ts, app/src/adapters/ReportAdapter.ts, app/src/adapters/SimulationAdapter.ts
• v2 hydration points: app/src/api/v2/reportFull.ts, app/src/api/v2/simulations.ts, app/src/api/v2/economyAnalysis.ts
• reproducibility UI and code generation: app/src/pages/report-output/SocietyWideReportOutput.tsx, app/src/pages/report-output/reproduce-in-python/PolicyReproducibility.tsx, app/src/utils/reproducibilityCode.ts
• latest-version helper: app/src/api/v2/taxBenefitModels.ts

Desired contract

Once downstream APIs expose a structured execution bundle, the app should:

• preserve that bundle in report/simulation domain objects
• render it where users expect reproducibility information
• use it when generating reproduce-in-Python code
• stop reconstructing historical provenance from mutable “latest” metadata

At minimum the app should retain and display:

• model package/version
• data package/version
• resolved dataset artifact or manifest revision
• bundle or manifest ID if available

What should change

1. Extend app domain models and adapters to preserve structured release provenance from API responses.
2. Hydrate provenance at the report/simulation fetch layer instead of dropping it.
3. Update the reproduce tab to generate code from the report’s actual resolved bundle, not current Redux defaults.
4. Add a small provenance display on relevant report/simulation output pages.
5. Remove reliance on app-side “latest” resolution for historical reports and replay flows.
6. Keep backward compatibility for older saved objects where possible, but prefer server-supplied pinned provenance whenever available.

Acceptance criteria

• Report and Simulation objects retain structured model/data provenance from API responses.
• Historical report pages render or otherwise surface the resolved execution bundle.
• Reproduce-in-Python uses the resolved bundle from the report/simulation being viewed.
• App-side latest-version helpers are not used for replay or historical provenance.
• Client-side caching and saved-object hydration do not discard release-bundle metadata.

Upstream dependencies

This depends on API/schema work in:

• PolicyEngine/policyengine-api#3393
• PolicyEngine/policyengine-api-v2#434
• PolicyEngine/policyengine-api-v2-alpha#219

And on the data/orchestration contracts in:

• PolicyEngine/policyengine-us-data#724
• PolicyEngine/policyengine-uk-data#322
• PolicyEngine/policyengine.py#270

Start by reading the report and simulation models and adapters, then trace hydration through app/src/api/v2/reportFull.ts, simulations.ts, and economyAnalysis.ts. Review the reproducibility components and app/src/utils/reproducibilityCode.ts alongside the upstream API dependencies. Done means resolved provenance survives hydration, appears in historical output, and drives reproduce-in-Python without latest-version reconstruction.