Real-time streaming
When someone messages a bot, Gondola sends the request to your Cortex agent and streams the reply into the thread as it is generated — word by word, not as a stored-and-posted document.
Security
Gondola routes chat to your Cortex agents and streams replies back. Message content is not stored in Gondola.
How Gondola handles data
When someone messages a bot, Gondola sends the request to your Cortex agent and streams the reply into the thread as it is generated — word by word, not as a stored-and-posted document.
Questions and answers stay in your Snowflake Cortex thread. Gondola keeps only routing metadata: platform channel and thread IDs, Snowflake thread references, and usage counts.
Snowflake PATs and keys, plus Slack and Teams tokens and signing secrets, are encrypted at rest with AES-256-GCM (unique IV per value, authenticated against tampering). A 256-bit master key lives only in server environment configuration — never in the database. Secrets are decrypted in memory when needed, never written in plaintext, and scrubbed from logs and errors.
Administrative actions are recorded append-only: connection, agent, bot, and credential changes; team membership; and org-level events — with actor, timestamp, and non-secret context. Audit events never include credentials or message content.
The app runs as a single US-based cloud instance. Database and authentication are on Supabase — managed PostgreSQL with daily backups and point-in-time recovery, plus Supabase Auth for sign-in. All application traffic is served over TLS.