Greetings,
I maintain a public MIT GitHub project and want fully automated PR review + merge with no human in the loop.
I already use a custom Kimi K3 reviewer (to which I have free access), and also have free access to DeepSeek V4 Flash, GLM-5.3, and Qwen3.8. CodeRabbit seems free for public OSS repos.
Has anyone compared CodeRabbit/Greptile/Qodo with strong open-weight models for real-world PR review? Which gives better bug/security detection and fewer false positives?
Would you trust one as an autonomous merge gate, or use CodeRabbit plus a second independent model?
Thanks,
Thomas