concentrate-ai wants to be Mike's fourth AI provider
A vendor is proposing its own hosted service as an opt-in model provider, with a privacy routing trick as the pitch - and it hasn't been merged yet.
Read this one with the author's angle in mind: it's a vendor proposing its own paid service into someone else's project, and the request even dangles an offer of free tokens. What the change actually adds:
- A live model menu: instead of a fixed, hard-coded list, the picker pulls each connected provider's current catalogue, keeps only the chat-capable models, and merges them into one grouped list you can refresh on demand.
- Privacy-aware routing: for models flagged Zero Data Retention (the provider keeps nothing after answering), prompts can be routed through Concentrate so the underlying AI company never sees the plaintext. A small badge shows when that's in effect.
- Strictly optional: leave the Concentrate key blank and nothing about today's behaviour changes.
The author is candid that the privacy routing only kicks in after the model list has loaded once, so it isn't foolproof out of the box.
So what Anyone worried about where client-confidential prompts actually travel should track how this proposal lands - while remembering who's asking.
Spotted something wrong? Or know the PR text has fresher detail than the writeup above?