This project provides an AI-assisted code review tool specifically for GitLab, aiming to enhance developer productivity by leveraging Large Language Models (LLMs). It targets development teams seeking to automate and improve their code review process, offering integration with various LLMs and GitLab's merge request workflow.

How It Works

The system integrates with GitLab via webhooks, triggering LLM-based code reviews on merge requests. It utilizes a modular architecture with a UnionLLM component, compatible with LiteLLM, to support a wide array of LLMs, including OpenAI, DeepSeek, and Google's models. This multi-model approach allows for flexibility and the potential to leverage the best model for specific review tasks. Customization of review logic and notification mechanisms is facilitated through custom Response and Review Handle classes.

Quick Start & Requirements

• Install/Run: Docker or direct source installation.
  • Docker: docker run -d --network bridge --add-host=host.docker.internal:host-gateway -v ./config:/workspace/config -p 8080:80 --name codereview xuxin1/llmcodereview:latest
  • Source: pip install -r requirements.txt then python3 app.py
• Prerequisites: Python 3.x, Docker (optional), GitLab webhook configuration.
• Links: English README, 中文 README, 官方文档

Highlighted Details

• GitLab Merge Request trigger for automated comments.
• Support for numerous LLMs via UnionLLM and LiteLLM compatibility.
• Extensible architecture for custom review logic and notification methods (e.g., DingTalk, email).
• Planned support for GitHub, Gitee, and Bitbucket.

Maintenance & Community

• Developed by the technical team at Mobvista.
• Community interaction via Issues and PRs. Contact email: mixuxin@163.com.

Licensing & Compatibility

• MIT License. Permissive for commercial use and closed-source linking.

Limitations & Caveats

The project is actively under development, with features like commit and tag triggers, Feishu notifications, and static code analysis integration listed as pending. While it supports many LLMs, the effectiveness of the code review is dependent on the chosen LLM's capabilities.