Hi everyone.
This is Self-Hosted LLM Code Review Agent.
Several agent would review your PRs from your GitLab or Forgejo (also GitHub too)
Just bring your own model. You can use your own LLM APIs from Anthropic, OpenAI(compatible) or your local model
It's for self hosted environment so you'll need Docker and a server host
Most important goal is not to make 'false positive' slop review.
It's free and open source now. there's no subscriptions. You can pick a model whatever you want and it works with GitLab, Forgejo, GitHub
github: https://github.com/seoes/proval
6 ratings
Maintainer of Proval
AI video with native sound, paid per clip. Nothing stored.
A calmer Mac menu bar, one click away.
Enterprise Software. Intelligent Automation. Global Impact.
Comments
2Maintainer of Proval
Hi everyone! feel free to ask anything
The hard part is the false-positive problem, so I'd like to know how you handle it: is there a confidence threshold, or a per-repo ignore list that the agent learns from when a reviewer dismisses a comment? The inline review and PR reply features suggest it keeps thread context, and a short note in the README on how it decides what to flag would help teams compare it with hosted reviewers.