Vision Models support

[gmag11]

Copilot do not identify vision support to vision enabled models. I've been trying to implement it by adding "capabilities":["completion","vision"] to /api/show endpoint for vision models. Doing so makes Copilot to not show any ollama model usable in model selection, even if I am able to configure ollama models. I mean, no one of selected ollama models is listed in Copilot models.

I suppose that Copilot sees something unexpected int /show response.

I checked https://github.com/microsoft/vscode-copilot-chat to find actual requirements but did not find any other than capabilities in /show.

Did you find the same issue?

[modpotato]

i dont personally have a glm coding plan, nor do i know much about the vsc chat api, but it may be something you need to configure as a modality.. if you have any news about this feel free lmk or submit a pr