Allow for o1 use through personal API token/pay per token

IsaacThoman

I regularly use the o1 family of models through OpenAI's API Playground. It's a bit of a hassle though, to always be switching between Kagi Assistant and the playground though.

I understand that offering unlimited access to the o1 family of models is terrifying for Kagi financially. I'd be happy with either of these solutions:

A) supply my own OpenAI API Token in settings

or, even better from a privacy perspective,

B) Pay Kagi on a per-token basis for use (would want to be able to set monthly price limits but would be simplest for end users)

Either would be fine, I just want a workflow where I don't need to constantly switch between apps.

Vlad

IsaacThoman Lets expand this with concrete suggetsionf for UI:

How do you imagine this option would be surfaced to the user?
How do you imagine all available models from the provider would be shown for the user?

Pay per token is also interesting, basically passing on the API cost to the user, the problem is the requires additional billing.

IsaacThoman

Would it be possible to give users a $5 monthly allocation for 'experimental' models such as o1 as part of the Ultimate plan? These models could be indicated with a little beaker icon in The Assistant UI. The $5 allocation could be communicated similarly to how search usage is currently shown on the Billing page for the Starter plan.

Maybe on the Billing page, a user could also specify an additional spending limit for additional use. I'd also be happy to pre-pay credit towards these 'experimental' models. (That's how I do it through the OpenAI API currently)