Version updated for https://github.com/sun-praise/pi-review-agent to version v1.5.0. This action is used across all versions by ? repositories. Action Type This is a Composite action. Go to the GitHub Marketplace to find the latest changes. Action Summary Multi-persona PR review agent using Pi and Litellm AI to automate inline reviews, cache management, and team collaboration on GitHub or Gitea platforms. What’s Changed Highlights Two-layer verifier suppresses hallucinated findings — rule-based line/file checks plus LLM re-confirmation. Demoted items appear in a collapsible section. Regex grep tool — the walkGrep matcher now accepts regex patterns by default. Cross-model fallback — configure a comma-separated fallback model list (PI_REVIEW_FALLBACK_MODELS) so the agent retries on another model when the primary fails. What’s Changed Added Two-layer verifier to suppress hallucinated findings (#21) Regex support in grep tool Cross-model fallback support (#29) Fixed parseDiffPath handles file paths containing spaces (#25) filterDiff truncates at section boundaries and excludes build artifacts by default (#28) walkGrep glob matching normalizes path separators for Windows compatibility Full Changelog: https://github.com/sun-praise/pi-review-agent/compare/v1.4.0...v1.5.0

Version updated for https://github.com/sun-praise/pi-review-agent to version v1.4.0. This action is used across all versions by ? repositories. Action Type This is a Composite action. Go to the GitHub Marketplace to find the latest changes. What’s Changed What’s new Repository style-guide support: pi-review-agent can now load a repo-level style-guide and inject it into reviewer prompts. Auto-detected paths: STYLE_GUIDE.md, .github/STYLE_GUIDE.md, docs/style-guide.md, .github/style-guide.md. New built-in style persona dedicated to style-guide enforcement. quality persona now receives the style-guide by default. Custom personas can opt in via use-style-guide: true in .github/reviewers/*.yaml. Explicit override via --style-guide CLI flag or style-guide action input. Usage - uses: sun-praise/pi-review-agent@v1 with: team: "quality:1,style:1,security:1" style-guide: "./docs/STYLE_GUIDE.md" litellm-url: ${{ secrets.LITELLM_URL }} litellm-api-key: ${{ secrets.LITELLM_API_KEY }} Refer to README.md for full documentation.

Version updated for https://github.com/sun-praise/pi-review-agent to version v1.3.1. This action is used across all versions by ? repositories. Action Type This is a Composite action. Go to the GitHub Marketplace to find the latest changes. What’s Changed What’s Changed Fixed Team mode now respects the model input (#16). Previously the model was registered into the LiteLLM provider but never forwarded to runReview, so every persona and the coordinator fell back to the hardcoded default deepseek-v4-flash and failed for any non-default model such as mimo-v2.5. Single mode model input also fixed: it had the same omission and would fail for non-default models because the provider only registers the user-selected model. Verification Tested end-to-end in PR #18 with team: quality:1,security:1,performance:1 and model: mimo-v2.5: all three personas plus the coordinator completed successfully and posted a review comment. New Contributors None.