AI-Codereview-Gitlab

Why we included this project

This project hooks into a GitLab instance through webhooks and runs LLM review on every merge request and push, posting results as notes directly on the MR or commit, so developers stay inside the review thread instead of a separate tool. Model support spans DeepSeek, OpenAI, Anthropic, ZhipuAI, Qwen, and Ollama, which is useful for teams that must standardize on one vendor or run everything behind a firewall. Beyond the review pass, it can route results to DingTalk, WeCom, or Feishu, build daily progress reports from GitLab, GitHub, or Gitea commit history, and keep a dashboard of review records with project and developer statistics. An optional agentic mode hands the model read_file and sandboxed shell tools to explore the cloned repository for a deeper pass, falling back to a plain diff review if anything fails. Teams already on GitLab who want review feedback woven into their existing workflow should find this straightforward to try.