Install & upgrade
Production checklist
The pre-production gate: unique secrets, rotated admin password, CORS, HTTPS, PostgreSQL and backups.
Work through this before exposing CHouse UI beyond localhost.
Secrets & first-run
- Generate unique
JWT_SECRET(min 32 bytes) — Secrets generation - Generate unique
RBAC_ENCRYPTION_KEY(32-byte hex) - Generate unique
RBAC_ENCRYPTION_SALT(32-byte hex) - Change the default admin password (
admin@localhost/admin123!)
Networking
- Set
CORS_ORIGINto your actual domain (default*is unsafe) - Serve HTTPS via a reverse proxy or ingress
- Restrict access with firewall rules —
5521serves the UI, API and MCP (/mcp) - Make the proxy set
X-Forwarded-Forto the real client IP (sign-in rate limiting depends on it) - Register the SSO redirect URI at your IdP if using SSO
Storage
- Use PostgreSQL (
RBAC_DB_TYPE=postgres+RBAC_POSTGRES_URL) for multi-instance deployments — SQLite cannot span pods - With SQLite, mount a persistent volume at
/app/data - Schedule database backups — including the secrets needed to decrypt stored passwords
Runtime
- Set
NODE_ENV=productionandLOG_LEVEL=info(orwarn) - Run Check privileges on each connection (Admin › Connections › Edit) so the collector can read the system tables it needs
- To use fixes, give each connection a remediation credential with only the grants the fixes need
- Set
PUBLIC_BASE_URLto the address people use - Leave MCP off (AI Governance › MCP) unless agents need access; turn on only the tools they need, and set agent budgets
- Verify migrations ran on boot:
docker logs chouse-ui | grep RBAC
AI (optional)
- Configure AI providers via Admin › AI models before enabling AI features
- AI calls spend provider tokens — set spending limits at the provider
- Gate who can spend tokens with
ai:optimize/ai:chat/doctor:run
Verification
bash
# Health of the deployment
curl -s http://localhost:5521/api/health | head -1 # or your host
docker logs chouse-ui | grep -i error | headDone? CHouse UI is production-ready. Ongoing care: audit log review, Migrations & upgrades for safe version bumps, and Alerting to let the system page you when a cluster misbehaves.