[LOOK WHAT I BUILT] llmcc: lightweight engine for better coding context understanding

Hi Codex community,

There's been some discussion in the community around the need for some kind of indexing strategies for better context understanding. I want to share llmcc, a lightweight indexing engine I've been working on to enhance how language models understand code structure and dependencies.

Core Features

llmcc trying to solve three key problems:

1. Project Architecture Understanding - Generates a high-level architecture graph of any codebase using the PageRank algorithm for llm to quickly understand high level codebase
2. Forward Dependency Analysis - Finds all code that depends on a specific symbol (class, function, etc.) across a folder
3. Reverse Dependency Analysis - Finds all dependencies required by a specific symbol across a folder
4. No rag, no embeddings, no summary etc stuff, very simple and lightweight tool

Integration with Codex

I've created an experimental integration branch at: https://github.com/allenanswerzq/codex/pull/2/files

Writing and tweaking prompts is kind of time consuming, so may not work too well at the moment.

Note: requires OPENAI_API_KEY to use customized instructions enforced by openai.

Currently has rust and python language supported, rust tested more, Hope someone can try it out share your ideas and feedback, to see if is something worth going forward more..

Thanks.

<img width="1113" height="528" alt="Image" src="https://github.com/user-attachments/assets/ad991b9d-17e4-427a-be8c-7035ff268a3d" />