Enhancing Cody's Context Awareness in Neovim for Large Codebases

Cody
1

Hello

I’ve been exploring Cody in Neovim for managing a large monorepo & I’ve encountered a challenge with context awareness. While Cody efficiently handles smaller projects, it struggles with identifying relevant code segments and dependencies when the codebase scales up. :upside_down_face:

This becomes particularly noticeable when working with deeply nested modules and external libraries. :innocent:

I suspect the issue might stem from the way Cody indexes and references symbols across multiple files. I’ve tried adjusting the Neovim LSP configuration and modifying the workspace settings, but the problem persists. :thinking: Checked Using sg.nvim - What can Cody do and how can I find out?Splunk course guide for reference .

Has anyone faced a similar issue or found a workaround to improve Cody’s ability to understand the project’s structure and provide more accurate code suggestions?

If there are any best practices or plugins that could enhance Cody’s performance in this scenario, I’d love to hear about them.

Thank you!! :slightly_smiling_face:

4

Hey @dereekthelr

I’m not very familiar with NeoVim, but in general I can say that by working with large unfamiliar codebases, you need to give Cody some guidance about the code base. Currently,
Cody has internally a symbol finder engine included, which search code files based on keywords derived from the prompt. Basically, it uses keyword tagging and searches for similar symbols in the codebase.
For example, if your prompt includes “How does the authentication system works? Explain the purpose of it.”, it searches for “login”, “auth”, “authentication” and such to find relevant files.

Like you said, large code bases with nested modules are sometimes more challenging for Cody. The best practise is to know the modules, structure of the project and implemented API. This process can be discovered with Cody too through exploration of the code base and write down the information in a distilled code documentation. These documents can then be referenced in a new chat for context and avoiding repetitive and redundant information gathering.

I hope that helps.