Add model integration tests and advisory regression reports

[jpn--]

Changes

Add automated model validation against the released dependencies in uv.lock. Every pull request runs input/contract tests and the full model sequence for a fixed sample of 2,000 complete households, retaining the full zone system and skims. Structural failures fail CI; changes in modeled distributions remain advisory.

• Check input schemas, household IDs, skim dimensions and zone mappings, population preservation, output relationships, and trip structure.
• Add weekly/manual 10,000-household and single-process runs.
• Publish diagnostic logs, advisory baseline comparisons, runtime, memory, versions, and input/configuration hashes as Actions artifacts and summaries.
• Document local execution, fixture provenance, baseline review, and how to extend the tests.

No behavioral model configurations, input population files, or dependency versions are changed. Constraint/telework component-specific tests can be added when those extensions land on main.

Validation

All jobs passed on the exact branch head in GitHub Actions on the driftlesslabs fork:

• 17 input and contract tests.
• 2,000-household multiprocess run: 139 seconds, 1.72 GiB peak summed process RSS.
• 10,000-household multiprocess run: 166 seconds, 1.99 GiB peak summed process RSS.
• 2,000-household single-process run.

Local model runs, Ruff checks, YAML lint, and whitespace checks also passed. Summed RSS may count shared pages more than once. Scheduling fallback counts remain advisory (7 trips in the smaller run and 52 in the larger run).

Weekly scheduling activates on the default branch after merge. Branch protection can then require the contracts and model checks.

[Copilot]

Copilot review overview

🟡 Changes recommended

Tour-sequence validation is incorrect for normal round trips, and extension loading and summary metadata requirements are incomplete.

Get a fresh assessment by requesting another Copilot review.

Review effort: Lite
Findings: 1 · 1

Open (2)

• Validate tour-wide sequence and counts instead of outbound legs · New
• Pass extension package name instead of filesystem path · New

What changed in this PR

Adds automated model input/contract tests, regression baselines, diagnostics, and CI workflows.

Changes:

• Adds fixed-sample and scheduled model validation.
• Adds advisory baselines and runtime/memory diagnostics.
• Documents local testing and CI workflows.

File	Description
tests/​test_model_inputs.py	Validates inputs, IDs, skims, and zones.
tests/​test_model_ci.py	Tests output-contract validation.
tests/​model/​settings.yaml	Configures test-scale model runs.
tests/​baselines/​2000.json	Stores small-run baseline metrics.
tests/​baselines/​10000.json	Stores large-run baseline metrics.
scripts/​model_ci.py	Runs and validates model integrations; tour sequencing, extension loading, and summary metadata require changes.
README.md	Documents automated model tests.
docs/​testing.md	Documents testing, fixtures, diagnostics, and baselines.
.github/​workflows/​model-tests.yml	Defines CI and scheduled validation jobs.

---

💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.

[Copilot]

Copilot review overview

🔵 Needs a closer look

Address the two moderate validation gaps before approval.

Review effort: Lite
Findings: None

Resolved since last review (2)

• Validate tour-wide sequence and counts instead of outbound legs
• Pass extension package name instead of filesystem path