Changelog

All notable changes to this project are documented here. Format loosely follows Keep a Changelog, but this is an early alpha so versions track the hackathon delivery cycle rather than strict semver.

[Unreleased]

Added

• Promoted demo scripts to top-level entry points and updated pyproject metadata so installs expose phoenix-extract and related commands.
• Cursor support bot fabrication case added to the real-harvested dataset (one more failure pattern documented end-to-end).

Changed

• Documented the Content-Length parse fallback inside the body-size middleware.
• Revised SECURITY.md with supported versions and the reporting flow.

Maintenance

• Dependabot rolled actions/checkout from 4 to 6, codecov/codecov-action from 4 to 6, and github/codeql-action from 3 to 4.

[0.1.0] - 2026-05

Initial alpha release for the Google Cloud Rapid Agent Hackathon (Arize track).

Added

• Phoenix LLM trace ingest: read an Arize Phoenix trace export and walk the spans to surface failure patterns.
• Regression test extraction: emit a runnable pytest module that replays the failing input and asserts the corrected behaviour.
• Optional Gemini-assisted assertion generation behind a --use-gemini flag (off by default; works without any LLM call).
• FastAPI demo surface with a body-size middleware for the hosted walkthrough.
• pytest + ruff + pre-commit setup with a small anti-marker scanner for prose hygiene.
• CI matrix for Python 3.11, 3.12, and 3.13. CodeQL workflow and codecov upload included from the start.

Known limitations

• Alpha. CLI flags are subject to change.
• Trace ingest currently assumes the Phoenix v3 export shape.
• The Gemini assertion path is best-effort and not yet covered by integration tests.

See README.md for the hackathon framing and stack notes.