[enhancement]: "Expand Current Prompt" for locally hosted installs

[BieHDC]

Is there an existing issue for this?

• I have searched the existing issues

Contact Details

No response

What should this feature add?

I really would like to have the "Expand Current Prompt" and "Upload Image for Prompt Generation" in the self hosted version. It could ether run in the context of Invoke itself (preferred) or connect/integrate into my existing Ollama instance.

Alternatives

Right now i have to manually go to the Ollama terminal to expand the text and turn images into prompts, which is annoying.

Additional Content

[fishzle]

Similar to #7909 #6719

[csehatt741]

@BieHDC, I wonder how you use Ollama with InvokeAI efficiently. Is it a multi-GPU environment or do you use Ollama hosted online?

[Fortyseven]

FWIW, if you set it up to use an OpenAI compatible endpoint, you can serve from a bunch of local providers -- including Ollama.

As for memory concerns, yeah, most of the models + whatever Invoke is running will exhaust VRAM. Personally I have a 3060 running on a headless machine on the LAN running inference requests. But if it's an OAI endpoint, you can point it at OpenRouter or any other cloud-based server, too.