// the take
GitLab-specific LLM code review bot that hooks into merge request webhooks and posts AI-generated review comments. Django backend handles webhook events, calls configurable LLM providers (OpenAI, DeepSeek, Claude CLI, etc.), and dispatches notifications to Feishu/DingTalk/WeChat. Aimed at teams self-hosting on GitLab who want automated PR feedback without paying for GitHub Copilot or similar.
Supports multiple LLM backends including local/private deployments via a pluggable provider abstraction, which is the main reason you'd pick this over a SaaS tool. The webhook rule system lets you configure per-project or per-event prompts rather than applying one global prompt to everything. Docker Compose setup is genuinely one command to get running. The claude_cli_service integration for Claude subscription users is a practical workaround to avoid API costs.
The codebase has a `.env` checked into the repo alongside `.env.example`, which is a credential leak waiting to happen and signals loose security hygiene throughout. The notification_dispatcher.py has a `.backup` file sitting in the same directory as the live code, suggesting manual file-copy 'version control' rather than proper git practices. Multi-agent support is listed as 'in development' in the README but there's no evidence of it in the tree — it's vaporware at this point. No test suite visible anywhere in the directory structure, so you're flying blind when modifying the review pipeline logic.
We read the README, the directory tree and a key source file in June 2026. Nobody pays to be listed. How we judge →