Support default model overrides for prompts

[hemant1729]

It would be great if we could support overriding the default model for a given prompt. My use case is rewriting Slack messages, and I feel that a larger model like gpt-4.5 performs much better than a smaller one like gemini-2.0-flash, although I’d like to continue using smaller models for most cases.

[timlenardo]

Hey @hemant1729 - thinking about this one. When you write Slack messages, do you always use the same prompt?

I’m trying to determine the best way to configure model rules. Here are a few approaches I’m considering:

• Match on a specific prompt and switch models when there’s an exact match.
• Use keywords, similar to our current System Prompt suggestions.
• Route requests based on the app in use (e.g., all Slack requests go to gpt-4.5, others to gemini-2.0-flash). We currently support this for System Prompt suggestions as well.

Which of these options would best solve your problem?