I'm one of the cofounders of Sourcebot, an open source alternative to Sourcegraph. Sourcebot lets you index thousands of repos across multiple platforms (GitHub, GitLab, Bitbucket), and gives you a powerful interface to search across them. You can learn more in our original HN launch post: https://news.ycombinator.com/item?id=41711032
We just added an AI code review agent that reviews your PRs and automatically detects issues that a human reviewer may have missed. We've been using an AI code review agent for a few weeks now, and it regularly catches issues that we would've merged to prod.
The review agent automatically fetches relevant context from code you've indexed in Sourcebot to provide accurate reviews. We’ve found that fetching this context is critical for the LLM to provide meaningful suggestions.
Would love any feedback if y'all get the chance to try it out! We're planning on expanding the context fetching capabilities to support: - Fetching definitions from function calls in a code snippet - Fetching all usages of a function across all your repos to ensure proper usage patterns - Any other code context fetching y'all think would be useful!
Michael Sukkarieh