Any time soon to bring back the multimodal support for server?

[victorx98]

since it was removed in March.

[ggerganov]

Mostly depends on help from the community. Requires good understanding of server and updating the llava / clip APIs to allow batched decoding support, proper image prompt handling, improved memory management of image buffers, etc.

[domasofan]

That would be great if someone could have a look at it.
Blind people would profit a lot with llava and the web ui.

Greetings,
Simon