Overview#
Asya 🎭 rests on a small set of ideas. Each page below explains one capability — what it enables, why it matters, and where to go for details.
Core Architecture#
| Concept | One-liner |
|---|---|
| Actor Mesh | Choreography over orchestration — actors scale and fail independently |
| Message Knows the Way | Routing lives in the envelope, not in a central coordinator |
| Virtual Actors | Stateless Deployments with optional persistent memory via state proxy |
Developer Model#
| Concept | One-liner |
|---|---|
| Separation of Concerns | Data scientists write Python; platform teams configure infrastructure |
| Dynamic Routing | Actors rewrite the route at runtime — branches, loops, human-in-the-loop |
| Flow Compiler | Familiar Python control flow compiled to flat actor graphs |
Infrastructure#
| Concept | One-liner |
|---|---|
| Kubernetes Native | AsyncActor is a CRD — kubectl, Helm, GitOps, RBAC all work |
| Scale Zero to Infinity | KEDA watches queues — GPU pods cost nothing between batches |
| Pluggable Transport | Swap SQS, RabbitMQ, or Pub/Sub via Crossplane without changing user code |
| Pluggable Storage | S3, GCS, Redis, NATS KV — state proxy connectors for virtual actor memory |
| Built-in Resiliency | Durable queues, retries, DLQ, SLA deadlines, timeouts |
Integration#
| Concept | One-liner |
|---|---|
| HTTP Gateway | Bridges async actors with synchronous HTTP, A2A, and MCP |
| Agentic Native | Agent swarms as distributed actors with streaming and pause/resume |
Operations#
| Concept | One-liner |
|---|---|
| Built-in Observability | OpenTelemetry tracing, Prometheus metrics, structured logging |
| Developer Experience | Fits any dev flow — GitOps, any OCI image, CI/CD, local testing |