Blueprint Pudim Code Reviewer

Version updated for https://github.com/luismr/blueprint-pudim-code-reviewer to version v1.0.3.

  • 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

The GitHub Action blueprint-pudim-code-reviewer automates the process of reviewing pull request diffs using a swappable language model backend via LangGraph/LangChain’s model abstraction. It supports three LLM providers: Anthropic, OpenAI, and Gemini, allowing users to choose their preferred model for code reviews. The action integrates with GitHub workflows, enabling developers to integrate AI-based code review into their CI/CD processes.

What’s Changed

Fixed

  • Raw JSON posted as review comment when LLM emits trailing commas — some models occasionally append a trailing comma after the last field of an inline comment object (e.g. "body": "...",\n}), which is invalid JSON and causes json.loads to raise JSONDecodeError. parse_review_output then returns None and the full raw JSON blob is posted as an issue comment instead of a structured review. A _sanitize_json step now strips trailing commas before parsing so well-formed reviews are never silently discarded.