feat: make it work with large codebases
Resolved 💬 4 comments Opened Apr 17, 2025 by buger Closed Aug 7, 2025
Would you be interested in PR with integrating it with Probe? https://probeai.dev/
Instead of doing this loop of grep, read portion of file, repeat, you can just treat your codebase as big database and query it, without pre-indexing required, and it will return proper AST.
Where it would take 20 steps of current classical agentinc code search, it can be just 1 query with probe to fetch all needed context.
4 Comments
This is highly needed. I've been going for a similar approach but with a rust backend for search, AST creation for maximal performance on large codebases. Still very early and not as functional as claude code, https://github.com/amrit110/oli
Yeah highest prio feature for Codex team should met code indexing in vector space, that will really make this more useful.
i think openai would want to limit third party usage. i _would_ love to see a fork of codex that does exactly this ;)
Thank you for the feedback! We're primarily focused on the native/Rust implementation of Codex CLI. Please try again with the latest version, and if the issue persists, we'd be happy to consider a PR to the
codex-rs/implementation.