Skip to content

Releases: Cyriellewu/openpatch-lab

v0.3.1 - fix cross-repo false positives

Choose a tag to compare

@Cyriellewu Cyriellewu released this 24 Jul 06:03

Bug fix found by dogfooding.

  • Fix false-positive competing PRs: a cross-referenced PR from a different repository (e.g. a fork) was reported as competing work, producing a wrong 'review it first' recommendation. Cross-references are now filtered to the same repo.
  • Discovered while using OpenPatch to vet Textualize/rich#4196.
  • 23 tests; CI green.

v0.3.0 - batch issue scan

Choose a tag to compare

@Cyriellewu Cyriellewu released this 24 Jul 05:22

Adds batch triage.

  • openpatch scan owner/repo --top N ranks a repo's open issues by startability (open + unassigned + no competing PR + not stale rank highest). text / json / md output.
  • Pull requests are filtered out of the issue list.
  • 22 offline tests; CI green on Python 3.10-3.12.

Facts from the GitHub API; ranking is rule-driven; model-derived fields stay honestly unavailable without an LLM key.

v0.2.0 - Markdown export + GitHub Action

Choose a tag to compare

@Cyriellewu Cyriellewu released this 24 Jul 04:49

Turns OpenPatch into something other repos can install as an Action.

  • Markdown export: openpatch inspect owner/repo#123 --format md.
  • Reusable GitHub Action: add Cyriellewu/openpatch-lab@v0.2.0 to a workflow to auto-comment an OpenPatch report on every newly-opened issue (see examples/workflow-openpatch.yml).
  • build_report() / render() extracted as a public API.
  • 19 offline tests; CI green on Python 3.10-3.12.

Model-derived fields remain honestly 'unavailable' without an LLM key; the recommendation stays rule-driven.

v0.1.0 - first release

Choose a tag to compare

@Cyriellewu Cyriellewu released this 24 Jul 04:03

First release of OpenPatch Lab.

openpatch inspect owner/repo#123 produces a reviewable contribution package before you start work:

  • Issue status, assignees, competing PRs (from the issue timeline), third-party claim signals, likely files, age/skip-labels - all FACTUAL, read from the GitHub API, never inferred.
  • Optional model layer degrades to 'unavailable' without an LLM key (never a fabricated number).
  • Deterministic recommended action: ask maintainer first / coordinate / review competing PR first / skip.

Zero runtime deps; auth via existing gh login or GITHUB_TOKEN. --json for machine output. 15 offline tests, CI on Python 3.10-3.12.