One YAML per feed, replacing layer-keyed config and three Python tupl… #17
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| name: ci | |
| # Lint, the full test suite, an end-to-end smoke build on a reduced dataset, | |
| # and an ops-agent dry run -- all on the pure-Python path (PyIceberg + DuckDB). | |
| # PySpark is not installed here, so the cross-engine parity tests skip cleanly; | |
| # that is the expected offline outcome, not a gap (see ADR-0002). | |
| # | |
| # HARD RULE: no step in this workflow may gate on a forecast-quality metric. | |
| # The smoke build runs at N_TRANSACTIONS=5000, where gold.daily_merchant_spend | |
| # covers only 1,532 of 3,756 ticker-days -- roughly 59% of training rows have | |
| # every spend feature NULL and the spend-target correlation collapses to | |
| # ~0.003. Any model-quality number computed at that scale is noise, and on | |
| # this dataset the model is *expected* to lose to its baseline at full scale | |
| # too (docs/forecast-report.md). A job that goes red when the model loses | |
| # teaches people to ignore CI. `make forecast` fails only when the pipeline | |
| # produced no predictions at all; that is the contract this workflow relies on. | |
| on: [push, pull_request] | |
| jobs: | |
| test: | |
| runs-on: ubuntu-latest | |
| steps: | |
| - uses: actions/checkout@v4 | |
| - uses: astral-sh/setup-uv@v3 | |
| - run: uv sync --extra dev | |
| - name: Lint | |
| run: uv run ruff check src tests | |
| - name: Tests | |
| run: uv run pytest -v | |
| - name: Smoke the full pipeline on a reduced dataset | |
| run: N_TRANSACTIONS=5000 uv run make all | |
| - name: Ops agent (dry-run, no API key -- rules-only path) | |
| run: uv run make agent | |
| - name: The smoke build must not modify tracked files | |
| # `make all` regenerates docs/forecast-report.md, and at reduced scale | |
| # it necessarily differs from the committed full-scale report -- so | |
| # restore that one file, then require the rest of the tree to be | |
| # untouched. Anything else the pipeline rewrote is a real finding. | |
| run: | | |
| git checkout -- docs/forecast-report.md | |
| git diff --exit-code |