LM Studio support is flaky and needs work

[Morriz]

Describe the need of your request

Hi, thanks for the project first of all. I use LM Studio extensively and have no need for other solutions that unnecessarily copy gigabytes of models that I already have.

Now, I notice the following:

• LM Studio is a connection option for chat, but does not query the running models for us to choose?
• LM Studio is not a connection option for autocomplete?

Proposed solution

LM Studio fully supported, also for autocomplete.

Additional context

No response

[xardbaiz]

Hello, @Morriz
If it's still relevant I described how to get more from CodeGPT + local LM Studio (including autocomplete as you asked) in the nightbour thread:
#384 (comment)

[Morriz]

@xardbaiz Your solution seems like another way to connect an endpoint, but it does NOT offer autocompletion as it needs to be chosen from a fixed list that does NOT include either "Custom" or "LM Studio"

[xardbaiz]

@Morriz actually - it does. Autocompletion FIM template you choosing depending on the model in use

Just as example

If you load Qwen to LM Studio then choose "CodeQwen" from fixed dropdown list

CodeGPT

LM Studio

[Morriz]

Are we talking about vscode here? I am seeing screenshots of some web application. Can you provide vscode walkthrough video/screenshots describing how to get autocomplete working?

[carlrobertoh]

@Morriz Are you sure you're using the correct extension?

https://plugins.jetbrains.com/plugin/21056-codegpt

[Morriz]

Yes, when I open it, it shows a black sidebar, I just can't get the autocompletion to work.

You show a full page something, not the sidebar...and you keep showing screens that I can't get to.

[Morriz]

there is only one CodeGPT in vscode...now I see your plugin is for jetbrains? What a waste of time

[carlrobertoh]

Perhaps next time it would make our lives easier if you just read the README or the project's description.

The weird thing is that somehow you must have gotten to the GitHub issue page.

[Morriz]

just too many clones nowadays ;)