domino

domino is a lightweight Hydra-based serial workflow orchestration library.

Install

pip install .

Run

domino --config-path examples/conf --config-name config
python -m domino --config-path examples/conf --config-name config

Config

work_dir: ""

ctx:
  work_dir: ${work_dir}

workflow:
  func1:
    callable: "examples/steps.py:make_base"
    kwargs: {}
    return_key: null
  func2:
    callable: "examples/steps.py:split_base"
    kwargs: {}
    return_key: ["base", "length"]

Callable specs support two forms:

• module:target: load an importable module or .py file path with load_callable; the target side can be a callable attribute name or a dotted attribute path such as client:pro.trade_cal.
• ctx:object.method: resolve object from the runtime ctx, then follow the dotted attribute path to a callable with load_context_callable.

Workflow execution should call resolve_callable, which dispatches to the normal module/file loader or the runtime-ctx loader based on the spec.

Step function parameters are resolved from step kwargs first, then ctx. All step kwargs are passed through to the function; ctx only fills missing signature parameters. Parameters missing from both step kwargs and ctx are omitted. If a step declares a parameter named ctx, domino passes the whole context unless the step explicitly sets kwargs.ctx.

Step kwargs values written as ${ctx.some_key} are resolved at the start of each step against the current runtime ctx, so later steps can reference values stored by earlier steps.