TabbyAPI, tool calling and image generation

[frenzybiscuit]

To get images to work (sorta, they dont work) you have to add the following to the chat_template:

https://github.com/theroyallab/llm-prompt-templates

https://github.com/theroyallab/llm-prompt-templates/blob/main/tool_calls/chatml_with_headers.jinja

However, when I type "Generate an image of a woman walking down the street" the following is returned...

LibreChat         | {
LibreChat         |   object: "chat.completion",
LibreChat         |   usage: null,
LibreChat         |   id: "chatcmpl-e1e4d5e87bf94fbd9f201577f3b8cc14",
LibreChat         |   // 1 choice(s)
LibreChat         |   choices: [{"message":{"tool_calls_json":null,"role":"assistant","content":"So the House of Gadgetry is another... [truncated]],
LibreChat         |   created: 1742571416,
LibreChat         |   model: "legion-v1a-llama-70b-4.0bpw-h8-exl2",
LibreChat         | }

This is still indefinitely better, as its actually throwing errors now instead of generating text images through the LLM itself.

I will keep experimenting. Hopefully I can figure it out! :)