Deployment¶

Astra supports four deployment targets. All use the same codebase; backend selection (Metal, CUDA, CPU) is explicit via environment config.

Environments¶

Environment	Typical target	Secrets
Local dev	Docker Compose + native services	`.env` / local only — not for prod secrets
Staging	Smaller k8s or single node	Vault or sealed secrets
Production	GKE / EKS / bare metal + managed data	Vault at runtime (PRD §18)

Page	Target
Local (docker-compose)	Development — infra in Docker, services native
Kubernetes / Helm	Cloud production — Helm, namespaces, mTLS
GCP Managed Services	GCP production — Cloud SQL, Memorystore, GCS
macOS Production	Mac Mini / Mac Studio production deployments

Namespace	Services
`control-plane`	`api-gateway`, `identity`, `access-control`
`kernel`	`scheduler-service`, `task-service`, `agent-service`, `goal-service`, `planner-service`, `memory-service`
`workers`	`execution-worker`, `browser-worker`, `tool-runtime`, `worker-manager`, `llm-router`, `prompt-manager`, `evaluation-service`
`infrastructure`	Postgres, Redis, Memcached, MinIO (local)
`observability`	Prometheus, Grafana, OpenTelemetry collector, Loki

Component	Strategy
Stateless services	HPA based on CPU / request rate
Workers	HPA based on Redis queue depth + scheduler hints
Redis	Cluster mode, shard count based on throughput
Postgres	Primary for writes, read replicas for heavy reads
Memcached	Memory-optimised nodes, horizontal scaling