I built a small Workflow in make.com that handles GPT-3.5 requests from my trusted iMessage application. This workflow unfortunately doesn't utilise the iMessage protocol, but standard SMS. This means that each request is quite expensive to send and that there is a significant delay. Nonetheless, I enjoy having an advanced AI assistant right in one of my most used applications on the Mac and iPhone.
Things to improve on this project:
Upgrade to GPT-4 and make it context-aware
Upgrade to iMessage to get costs and delay down significantly.