Imported document text, support messages, source metadata, classifications, draft replies, citations, and usage events are stored in Postgres.
Document content may be split into chunks and represented as embeddings for workspace-scoped retrieval.
AI request metadata may include provider, model, token estimate, retrieved source ids, draft status, and reviewer information.
Contact form submissions are stored so the team can respond to business, billing, security, or support inquiries.
Service providers
ZhaiBro uses infrastructure and integration providers to operate the service, including authentication, billing, document import, AI model calls, database, and email or support handling.
OAuth tokens are encrypted before storage and are never exposed in browser responses or normal logs.
AI model requests run only from the server or worker, not from browser-side keys.
Session and workspace membership metadata, Notion or Google document metadata, and AWS Bedrock model metadata may be processed when those integrations are configured.
Deletion
Workspace administrators can delete imported sources, and the system is designed to delete imported source text, chunks, and embeddings tied to that workspace source.
Deletion does not remove records that must be retained for security, billing, abuse prevention, or audit integrity.
Customers can request account or data questions through [email protected].
Backups and logs may expire on operational retention schedules.
