Initial brainstorming

[ggerganov]

References:

• p1 : LLM-based code completion engine at the edge p1#1
• https://github.com/ggml-org/llama.vim

This repo will be an attempt to implement a llama.vim-like extension for VS Code. In this discussion we can brainstorm on the specific implementation. I don't have any experience with writing and using VS Code extensions, so I might not be able to provide a lot of specific insights. But overall, I think the implementation should follow closely the logic of the llama.vim plugin.

One of the main differences would probably be the mechanism for collecting the extra context. In llama.vim we use Vim events for entering/leaving files, entering/leaving insert mode, etc. Hopefully there is some similar functionality available in VS Code.

So far, @VJHack has expressed interest in working on this extension, so I've added them to the ggml-org/p1 team. Everyone else is welcome to join too.

[gfvvz]

Great idea, hope can support VSCode SSH remote access as well.

[ggerganov]

Yes, the extension will be compatible with SSH remote access. It will send requests from the local client to a user-configured llama-server instance, so it does not matter if the editing is using SSH or not.

[ggerganov]

@gfvvz @VJHack First version of the extension is now available in the marketplace. See the README for instructions. Any feedback is appreciated.

[ElectriPixie]

This looks cool I'm going to try it out.