Phase 3: Add pytest test suite and GitHub Actions CI

[WolfpackOfOne]

Summary

Adds automated testing and continuous integration so the project validates every push and pull request. Closes #3.

• Adds requirements-dev.txt (runtime deps + pytest + ruff); no migration to Poetry/Hatch.
• Configures Ruff (E, F, I, UP, B) and pytest in pyproject.toml, with src on the pytest pythonpath so tests run without an editable install.
• Adds a tests/ suite using only tiny synthetic data (no competition data):
  • test_features.py — engineered columns, row-count preservation, no-NaN-after-fill, timestamp sorting, and clear errors when timestamp/value are missing.
  • test_models.py — pipeline construction, reproducible seed, fit/predict, prediction length.
  • test_prediction_output.py — submission schema (timestamp, has_structural_break), row count, binary predictions.
  • test_cli.py — subprocess end-to-end run on tmp_path CSVs plus a missing-file error path.
• Adds .github/workflows/ci.yml (Python 3.11, single Linux job) running ruff check . and pytest.
• Adds CI + license badges and a Development section to the README.

Motivation

Structural-break pipelines break silently — feature leakage across time, output-schema drift, path handling. A small, fast test suite plus CI proves the baseline works, makes future refactors (Phase 4 detectors) safer, and signals to reviewers that this is a maintained project.

Validation

In a fresh virtual environment:

pip install -r requirements-dev.txt && pip install -e .
ruff check .   # All checks passed!
pytest         # 17 passed

CI on this PR runs the same two commands. The CI badge will resolve once the workflow runs on main.

Follow-up Work

• Phase 4 (research/add-change-point-models): CUSUM, rolling z-score, and PELT detectors on a shared interface, each with synthetic-data tests.

https://claude.ai/code/session_01Ddjy7PGk1UdXXcn95NyrGt

---

Generated by Claude Code