Upgrades and backups
Upgrade safely, back up Postgres and .eve/, and restore a host.
Upgrades
- Back up Postgres and
.eve/ - Deploy the new image / git revision
- Boot with the same
BRAIN_DATABASE_URL— Brain applies schema on startup - Confirm
/eve/v1/healthand a test chat
Preserve .eve/ so MCP tokens and BYOA credentials survive the upgrade.
What to back up
| Asset | Why |
|---|---|
| Postgres | Users, workspaces, chats, playbooks, schedules, licenses |
.eve/ | OAuth tokens, BYOA credential files (treat as secrets) |
| Secrets store | BETTER_AUTH_SECRET, BRAIN_LICENSE_SECRET (if set), BRAIN_INTERNAL_TOKEN, BRAIN_BOOTSTRAP_TOKEN, model keys |
There is no SQLite fallback — Postgres is the source of truth.
Restore / migrate runbook
- Provision Postgres and restore from snapshot or
pg_restore/psqldump - Point
BRAIN_DATABASE_URLat the restored database - Restore the same
BETTER_AUTH_SECRET(sessions / signing) andBRAIN_LICENSE_SECRETif you used one - Remount or copy
.eve/onto the new host with restricted permissions - Set
BRAIN_PUBLIC_URL/BETTER_AUTH_URLto the new public origin (update IdP callback URLs if the host changed) - Start Brain; smoke
/eve/v1/health,/sign-in, a chat turn, and one MCP Connect
Host rename checklist
- Update OAuth app callback URLs for Slack / GitHub / …
- Re-verify SSO DNS if the public domain changed
- Rotate
BRAIN_INTERNAL_TOKENonly if you intentionally invalidate schedule callers
What does not migrate automatically
- Ephemeral sandbox workspaces (attached-repo clones)
- Local browser theme preference (
brain-theme)