A few things Stanza can do:
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.
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.
3. Developer API: Build custom tools to inject context from AI which understands your code.
A few things that are unique about Stanza:
1. Multi-Resolution Code Understanding
Stanza builds a hierarchical representation of repositories. It can “zoom in” on specific symbols or “zoom out” to understand architecture.
2. Agentic Retrieval
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.
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!
3. Commit-Level Views
Stanza maintains an index for every new commit in your repo, so the AI's context is always consistent with the version of the code you’re working with.
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.
Thanks for reading! I'd love to hear your thoughts, and happy to answer questions.
