TE
TechEcho
Home24h TopNewestBestAskShowJobs
GitHubTwitter
Home

TechEcho

A tech news platform built with Next.js, providing global tech news and discussions.

GitHubTwitter

Home

HomeNewestBestAskShowJobs

Resources

HackerNews APIOriginal HackerNewsNext.js

© 2025 TechEcho. All rights reserved.

Show HN: Stanza – AI Code Reviews, Codebase Chat, and Developer API

2 pointsby bregr5 months ago
Hey HN! My name is Brendan, and I&#x27;ve spent the last few months building Stanza. Stanza automates parts of code review and helps engineers understand large, complex codebases more easily.<p>A few things Stanza can do:<p>1. Automated Code Reviews: near-instant feedback on pull requests - usually within about a minute. It checks for bugs, security issues, test coverage, and adherence to custom rules or coding patterns.<p>2. Codebase Chat: Chat with multiple codebases at once, and zoom in on particular commits, file patterns, or subdirectories. Stanza can even generate technical diagrams on-demand.<p>3. Developer API: Build custom tools to inject context from AI which understands your code.<p>A few things that are unique about Stanza:<p>1. Multi-Resolution Code Understanding<p>Stanza builds a hierarchical representation of repositories. It can “zoom in” on specific symbols or “zoom out” to understand architecture.<p>2. Agentic Retrieval<p>Stanza goes beyond simple RAG. It can navigate repositories to find relevant files. This iterative traversal allows Stanza to gather “multi-hop” context. Agentic retrieval is particularly useful for large monorepos, where relationships between disparate parts of the codebase can be non-obvious.<p>As an example: if a function `foo` calls `bar`, and `bar` calls `baz`, sometimes it can be necessary to understand what `baz` does in order to understand `bar` . Simple one-hop retrieval might get lucky and include context on `baz`, but often, it won’t be sufficient and will only find `bar`. Agentic retrieval is a way to solve this!<p>3. Commit-Level Views<p>Stanza maintains an index for every new commit in your repo, so the AI&#x27;s context is always consistent with the version of the code you’re working with.<p>For example: if you’re reviewing a PR, you mainly care about the commits involved in the PR. If you’re trying to gather information to debug a production issue, you care about the version of the code which is deployed in production. Consistent context is critical to ensuring the AI gives correct information and makes accurate suggestions.<p>Thanks for reading! I&#x27;d love to hear your thoughts, and happy to answer questions.<p><a href="https:&#x2F;&#x2F;stanza.sh&#x2F;" rel="nofollow">https:&#x2F;&#x2F;stanza.sh&#x2F;</a>

1 comment

mjcatoen5 months ago
Really nicely done. Eager to give this a try
评论 #42613493 未加载