UI Hints for Model Behaviours

[humphd]

Spin-off from #664 (comment)

Not all models support the same features. For example, some can handle images or audio, some can do function calling, etc. However, you have to know about this to use it.

We should come up with a way to hint to users what a model can do. Maybe in the list of models we add icons? Maybe once you've selected a model, we update the prompt area with some icons to indicate what can or can't be done?

Needs thought.

[Amnish04]

@humphd I think at some point we started showing only chat models in ask menu list. Is this still relevant?

chatcraft.org/src/components/PromptForm/PromptSendButton.tsx

Lines 324 to 328 in 26e39ea

	{models
	.filter((model) => isChatModel(model.id))
	.filter((model) =>
	model.name.toLowerCase().includes(debouncedSearchQuery.toLowerCase())
	)

[humphd]

Maybe not according to how we define a chat model now. But I want to make sure we filter out the models that won't work for chat completions (OpenAI has some pre-chat-completion stuff that's useless).

Maybe we should make a list that blacklists models, or maybe I'm overthinking and we should show them all and just error when stuff fails.

[tarasglek]

I made this explicit cos before we only did if openai & .includes("gpt"). I needed this to not show whisper on groq