Open erkkimon opened 1 year ago
Ditto, I need this feature.
I want to multiple-select multiple code files (.html and .js) and ask the AI some questions, like ECMA2015 compatibility, and optimization suggestions, as a code tutor. I love the large context size.
As it stands, I have to copy and paste individually if I am asking questions about more than one file. I paste 50 kilobytes of code into one single question in a chat.
Yes, the API costs a few tens of pennies for a question that is 50 kilobytes in one shot, but the answers have been worth the tutoring in refactoring a difficult JavaScript project.
Semi-duplicate: #34, #130 I'll settle for multi-selecting files already in my disk.
Can someone please explain how to keep a context at all? I've asked to generate docker-compose.yml file and then i've asked to produce a command for launching tests.. and it generated docker-compose.yml again with completely different naming. Why chatgpt doesn't even remember what he did a second ago? I'm using gpt4o model
Describe the feature
Genie AI's context is limited to a single code file which is slightly troublesome when the code is modular as it should be.
With GPT Repository loader the repository can be turned into a text file. From the text file it would be easy to pick relevant files and file structure of the project as context for Genie AI.
Especially now that we have GPT-4-1106 available with 128K context, we should utilize it!
https://github.com/mpoon/gpt-repository-loader