open-code-review

Why we included this project

Code review is usually the slowest part of shipping, and this tool goes straight at that bottleneck. It is a command-line assistant that reads a Git diff, hands the changed files to an LLM you configure, and posts comments tied to specific line numbers. The interesting part is how it splits work: fixed logic decides which files get bundled and which rules apply, so nothing gets skipped, while the agent handles the fuzzy parts like reading full file context and searching the codebase. That split keeps token usage and latency low enough for CI, and it has been battle-tested at Alibaba's scale. It also ships with a multi-language ruleset for common failure classes like null-pointer handling, thread-safety, XSS, and SQL injection, and it works with any OpenAI- or Anthropic-compatible endpoint, so teams can point it at their own model and keep data in-house.