For "Project AI", we'll likely be utilizing Tanstack AI. This will be the backbone of how we tackle things. Ideally we'll be using Mistral, but a provider does not exist as of yet so we'll need to consider building a custom provider, or see what Mistral offer straight out.

The reasoning behind using the "Tanstack AI" provider is that we plan on allowing Pro/CE to bring their own API so they can connect up any providers they'd like.

We'll need to provide a method of using other AIs outside of the default provided. Depending on the costs incurred, we'll potentally need to build in a costs/meters with our billing. This should be easy enough to handle, but also hopefully we can design something within Mistral that's uber low cost that it's not a major concern.

References: