How to update code with functionality from another file

[piovis2023]

Does anyone have any good tips on prompting with aider (claude 3.5 sonnet)?

I have literally 2 files:

1. App.py (700 lines of code)
2. JSON drop down (200lines of code)

Both are python apps running on streamlit.

I've been trying to get aider to implement the exact functionality from file 2 into file 1.

It gets 60-70% of the code right. I spent 4 hours ans multiple prompts to do it. No luck.

I then tried to ask aider to split file 1 into multiple files to ensure that it's more efficient and robust, but it never makes it through to the end accurately.

When I try its attempt, the app is broken or not entirely the same as the single file.

I really need some help.

Thanks

[paul-gauthier]

Thanks for trying aider and filing this issue. This doc may be helpful:

https://aider.chat/docs/usage/tips.html

Also, when reporting problems, it is very helpful if you can provide the info below. It will help me give a more complete answer.

• Aider version
• LLM model you are using

Including the “announcement” lines that aider prints at startup is an easy way to share some of this helpful info.

Aider v0.37.1-dev
Models: gpt-4o with diff edit format, weak model gpt-3.5-turbo
Git repo: .git with 243 files
Repo-map: using 1024 tokens

[piovis2023]

@paul-gauthier thanks so much for a quick response.

I'll check this doc and provide the additional details soon

[piovis2023]

Still no luck with the tips.

Here is the additional info @paul-gauthier : Aider v0.55.0 Main model: claude-3-5-sonnet-20240620 with diff edit format, infinite output Weak model: claude-3-haiku-20240307 Git repo: .git with 4 files Repo-map: using 1024 tokens, auto refresh VSCode terminal detected, pretty output has been disabled.

Is there anything else that I can get you?

[piovis2023]

Could someone kindly provide some prompts or examples where they were able to achieve this?

[fry69]

I then tried to ask aider to split file 1 into multiple files to ensure that it's more efficient and robust, but it never makes it through to the end accurately.

My workflow in a situation like this:

• ask aider to create tests for your big file implementation
• check if the tests actually test the implementation (LLMs love to create tests which only check mocked functions inside the test script)
• ask aider to split up the big file, piece by piece (e.g. by function or bigger logic parts)
• run the tests to check if everything still works

[paul-gauthier]

I'm going to close this issue for now, but feel free to add a comment here and I will re-open. Or feel free to file a new issue any time.