Bitbucket PR Reviews with Gemini CLI

This repository demonstrates how to use the Gemini CLI to automatically review Bitbucket pull requests.

Overview

The process works as follows:

1. A pull request is created in Bitbucket.
2. A Bitbucket Pipeline is triggered by the pull request.
3. The pipeline installs the Gemini CLI.
4. The Gemini CLI is used to perform a code review, following the instructions in the REVIEW.md file.
5. The code review is posted as a comment on the pull request. Currently, curl is used to post the comment, but this will be replaced when Bitbucket MCP server is available.

Files

• bitbucket-pipelines.yml: This file defines the Bitbucket Pipeline that automates the code review process.
• REVIEW.md: This file contains the instructions for the Gemini CLI to perform the code review. It specifies that the CLI should run a git diff, review the changes, and then format the output as a JSON payload with a detailed summary of the changes, the impact of the new changes, readiness to merge, test coverage, and a security review.
• README.md: This file (the one you're reading right now).

Setup

1. Add Environment Variables to your Bitbucket repository:

   Use Vertex AI(Enterprise users):

   • GOOGLE_API_KEY: Your Google Cloud API key. Needs to be configured in Bitbucket repository settings under "Repository variables". For Enterprise users, configure GOOGLE_API_KEY. Details here.
   • GOOGLE_GENAI_USE_VERTEXAI: Set to true to use Vertex AI for code reviews. Needs to be configured in Bitbucket repository settings under "Repository variables". Details here.

   Use Gemini API Key

   • GEMINI_API_KEY: Your Gemini API key. Needs to be configured in Bitbucket repository settings under "Repository variables". For Enterprise users, configure GOOGLE_API_KEY instead.

   Bitbucket specific variables:

   • REPO_ACCESS_TOKEN: Your Bitbucket repository access token with permissions to comment on pull requests(PR Write). Needs to be configured in Bitbucket repository settings under "Repository variables".
   • BITBUCKET_WORKSPACE: The workspace where your repository is located. Runtime variable provided by Bitbucket Pipelines.
   • BITBUCKET_REPO_slug: The slug of your repository. Runtime variable provided by Bitbucket Pipelines.
   • BITBUCKET_PR_ID: The ID of the pull request. This is automatically set by Bitbucket Pipelines.

2. Enable Bitbucket Pipelines: If you haven't already, enable Bitbucket Pipelines in your repository settings.

Once you've completed the setup, any new pull requests will be automatically reviewed by the Gemini CLI.

Example

Here is an example of the pipeline in action:

And here is an example of the code review summary posted as a comment on the pull request: