Per-Tool AI Controls
Part of this shipped. Since 1.0.22 every AI tool has its own card where you choose the model it runs on, so you can give tagging something small and fast and summaries something stronger.
Next, in 1.8.0, every AI tool gets its own switch and its own engine. Turn summaries off and keep tagging on. Run drafts on PressedMail AI and replies on your own provider. Saving one card no longer touches the others, and the site-wide AI switch only moves when you move it.
What is still open is the provider underneath. Today there is one API key and one base URL for all of it. Splitting that means a separate provider, key, and base URL per tool, which is what you want if summaries go to a hosted model and tagging runs on your own hardware.
Planned.
What already works
- A model per tool, chosen on that tool’s own card
- Editable system prompts per tool, with a reset to default
What is planned
- A switch and an engine choice (your provider or PressedMail AI) for each tool, in 1.8.0
- A separate provider, key, and base URL per tool
- Validating each provider independently, so one bad key does not read as all of AI being broken
Availability
Included with every paid plan.
