Versions 1.5.0 and 1.6.0 of the open-source, multi-provider AI code-review tool for GitHub let teams point it at any OpenAI-compatible endpoint, including DeepSeek, Kimi, Grok, Mistral, OpenRouter, and self-hosted gateways, and add a scheduled summary of the pull requests merged each week.
A patch release makes the sampling temperature opt-in, so the AI code-review tool's Claude commands work again on Anthropic's newest models, which no longer accept that parameter.
A run of releases makes the open-source, multi-provider tool run automatically on pull requests from agents like Copilot, Claude Code, and Cursor, adds a composite GitHub Action for running it as a step in any workflow, and publishes it to the GitHub Marketplace.
A patch release, surfaced by the tool's first external adopter, keeps code reviews running when an AI provider returns a temporary error instead of failing the whole command.
The open-source, multi-provider AI reviewer for GitHub can now be installed with one thin workflow instead of copying about 30 files, and upgrades become a one-line tag bump.
The open-source multi-provider AI code-review tool for GitHub ships 66 unit tests, an internal library refactor that removes roughly 945 lines, and two dependency security fixes.