asya
Active·★ 59·Apache-2.0·Updated 2026-09-17
Asya is a queue-based actor framework designed for orchestrating AI/ML workloads on Kubernetes, enabling independent scaling and dynamic pipelines. It allows data scientists to write pure Python functions while handling queues, routing, scaling, and monitoring with zero infrastructure code.
#Actor Framework#Kubernetes#AI/ML Orchestration#Queue-based Processing#Autoscaling#Python#Event-driven Architecture#Microservices
01
Features
01Independent scaling: Each actor scales 0→N based on its own queue depth.
02Zero infrastructure code: Write pure Python functions without dependencies for queues, routing, or retries.
03Dynamic pipelines: Routes are data, not code, allowing modification at runtime.
04Cost efficiency: KEDA autoscaling from zero to max, paying only for active processing.
02
Compatibility
Kubernetes
Orchestration
Verified via docs
Python
Handler Runtime
Verified via docs
KEDA
Autoscaling
Verified via docs
AWS SQS
Message Transport
Verified via docs
RabbitMQ
Message Transport
Verified via docs
Go
Sidecar/Operator
Verified via docs
03
Quick start
1
$ helm install asya deliveryhero/asya --namespace asya-system --create-namespace
04
Use cases
↳Multi-step AI/ML pipelines (e.g., document processing, image pipelines, LLM workflows).
↳Event-driven workloads (e.g., webhook processing, batch predictions, async API backends).
↳Cost-sensitive deployments (e.g., GPU inference, bursty traffic, dev/staging environments).
05
Alternatives
Comments
Log in to leave a comment
No comments yet. Be the first!