Run repomix and include CLI output of another tool in one step

[coezbek]

Description

When vibe coding with non-agent tools (so directly with AI-Studio or chatgpt.com) it would be nice if repomix could automatically run a command line tool and add the output of this tool to the summary.

I would image something like

repomix -c npm install

This would run repomix as usual but also run the given command and include its output (preferably at the top). For example, this could generate:

I just ran npm install in my codebase in ${PWD}, but it FAILED with the following output:

...CLI_OUTPUT_HERE...

The rest of this file is a merged representation of the entire codebase, combined into a single document by Repomix.
The content has been processed where security check has been disabled.

File Summary

...

This would really help to package error messages together with the latest status and current working directory.

[yamadashy]

Hi, @coezbek !
Thank you for your suggestion!

So essentially, you'd like to include command execution results in the output?
I can see how including results from commands like git log, npm test, or build errors would be useful for providing "current code + what's happening" context.

However, I'd like to consider implementing this as a more abstracted feature!

[coezbek]

I think the idea would be that the static context of the repo is often not that interesting, but we are providing it to an LLM because something is broken or there is an error.

So for instance I would run npm run dev manually test an feature and there would be an issue in the logs. At the moment, I need to copy the issue, paste it into a LLM chat, then use repomix to package the codebase and paste this as well.

I would prefer if repomix --command npm run dev would do this for me in just one call (npm run dev is interactive, so a bit difficult).

[yamadashy]

How about implementing an --include-content option that allows including arbitrary content?

This would let users include command results as part of the regular repomix output:

repomix --include-content "$(npm run build 2>&1)"
repomix --include-content "$(git log --oneline -20)"
repomix --include-content "$(cat error.log)"

This approach would not only satisfy the command execution requirement, but also address the broader need to include any important context or content. Also, from a security perspective, this approach is safer than having repomix execute arbitrary commands internally, which could introduce security risks.

I'm considering where to place this in the output structure. Since this would likely be high-priority context, it might work well at the beginning or end of the output - perhaps right before the custom instruction section?

[coezbek]

I think it should be pretty early in the content. Maybe the content could be scanned for file paths which are given priority.

If repomix -ic doesn't have a parameter following it could be used to read input from stdin:

npm run build 2>&1 | repomix -ic

[yamadashy]

If placing it early makes sense, I'll put it after the header section (above the directory structure).

Since we already have --stdin for file paths, I'll use --stdin-content for consistency:

npm run build 2>&1 | repomix --stdin-content

By the way, I'm curious about your reasoning for placing it early. From my perspective, considering the "Lost in the Middle" phenomenon with LLMs, either the beginning or the end would work well. What makes you prefer the beginning specifically?