Public chat
Messages sent in public chat are sent to NVIDIA-backed model endpoints to generate a response. Public chat is not connected to private profile memory.
Data use
Cory OS uses a small set of third-party services to provide chat, authentication, contact forwarding, rate limiting, and private memory.
Messages sent in public chat are sent to NVIDIA-backed model endpoints to generate a response. Public chat is not connected to private profile memory.
Private chat requires sign-in through Clerk. Private messages are sent to NVIDIA-backed model endpoints and may include user-approved private memories as additional context.
Messages that look like direct contact requests may be screened by the model. Approved contact-intent messages can be forwarded to Discord so Cory can review and respond.
Signed-in users can create profile memories from private chat. New memory candidates stay pending until the user approves them, and pending candidates can include extracted memory content plus a short source snippet from the private chat message that created them. Only approved memories can be reused in that user's private chat context.