Infrastructure
Deployment Models
Three deployment options built around the single-binary Runtime. All models embed Temporal and a real-time media server, requiring only PostgreSQL as an external dependency. Cloud connectivity is needed for third-party AI services.
Standard Cloud
COMPUTE
Operator-Managed VMs
Runtime container, auto-provisioned
DATABASE
Azure PostgreSQL Flexible
Managed, AES-256, HA replicas
MEDIA
Cloud Media Server
Managed WebRTC infrastructure
ORCHESTRATION
Embedded Temporal
Durable execution, per-org isolation
NETWORK
Cloudflare CDN + LB
Edge protection, L7 routing
SECURITY
AI Governance + RBAC
OPA policies, budget enforcement
Advantages
Single binary deployment
Auto-scaling via Operator
Embedded Temporal & Media Server
AI spend governance built-in
Considerations
Cloud dependency for VMs
Requires Azure Flexible Server
Hybrid
COMPUTE
Operator-Managed VMs
Runtime container, cloud-hosted
DATABASE
On-Premises PostgreSQL
Local data residency
MEDIA
Self-hosted Media Server
On-prem media server
ORCHESTRATION
Embedded Temporal
Local or cloud PostgreSQL backend
NETWORK
VPN + Cloudflare
Secure tunnel to on-prem data
SECURITY
Split Key Management
Local HSM + Cloud Key Vault
Advantages
Sensitive data stays on-prem
Self-hosted media routing
Meets strict data residency
Flexible migration path
Considerations
Higher complexity
VPN/ExpressRoute costs
Split monitoring required
Self-Hosted
COMPUTE
Self-Managed VMs
Runtime on bare metal or K8s
DATABASE
Self-hosted PostgreSQL
Local hardware, full control
MEDIA
Self-hosted Media Server
Full media plane control
ORCHESTRATION
Embedded Temporal
Self-managed worker fleet
NETWORK
Hardware Firewall
On-prem load balancer
SECURITY
On-Prem HSM + RBAC
Full key & policy management
Advantages
Complete data sovereignty
No cloud dependency (except AI)
Full infrastructure control
Air-gapped option possible
Considerations
Highest operational cost
Self-managed HA/DR
AI providers still need cloud API