v0.1.0
Initial release.
Features
- Workflow registry:
@register_workflowdecorator for registering pure
workflow functions with optional Pydantic parameter validation - Pipeline abstraction:
Pipeline,Step, andForEachfor declarative
multi-step DAG definitions - WorkflowPlan: Static, serializable DAG representation compiled from
pipelines via topological sort - Content-addressed storage: Deterministic prefix computation with
IdentityKeyannotations for cache control - WorkflowContext: Unified file I/O interface (JSON, xarray, raw bytes)
with dependency access and progress reporting
Execution backends
- LocalBackend: Synchronous in-process execution for testing and CLI use
- CeleryBackend: Parallel DAG execution via Celery chord/group primitives
- StepFunctionsBackend: AWS Step Functions orchestration with Lambda
Storage backends
- LocalStorageBackend: Filesystem-based storage with write-once semantics
and path traversal protection - S3StorageBackend: AWS S3 storage backend
Utilities
run_plan_locally()helper for pipeline integration testingResourceManagerandresolve_urifor transparent local/remote resource
fetching- Extended JSON encoder with NaN, numpy, and datetime support
- xarray Dataset serialization helpers