DiscoveryActionProvider

from causaliq_discovery import DiscoveryActionProvider

DiscoveryActionProvider exposes learn_graph as a CausalIQ workflow action. When included in a CausalIQ workflow, it enables automated causal graph discovery inside pipeline orchestration.

Availability

DiscoveryActionProvider is only available when the causaliq-workflow package is installed. The causaliq_discovery package exports a WORKFLOW_AVAILABLE flag that can be checked at runtime:

from causaliq_discovery import workflow_action

if workflow_action.WORKFLOW_AVAILABLE:
    provider = workflow_action.DiscoveryActionProvider()

When causaliq-workflow is not installed, importing DiscoveryActionProvider from the main package raises ImportError.

DiscoveryActionProvider

Subclass of CausalIQActionProvider (from causaliq-workflow). Registers one workflow action: learn_graph.

learn_graph action

Action name: learn_graph

Accepts the same inputs as the learn_graph Python function, supplied as workflow action input fields:

Input field	Type	Required	Description
data	str	Yes	Path to the input CSV data file.
algorithm	str	Yes	Structure learning algorithm name.
output	str	Yes	Directory path to write result files.
variant	str	No	Algorithm variant (defaults to first registered).
hyperparameters	dict	No	Hyperparameter name/value pairs.
trace	bool	No	Whether to include an execution trace.
variable_types	str \| dict	No	Network context file path or variable type mapping.
sample_size	int	No	Number of rows to use from the input data.
knowledge	str	No	Path to a knowledge JSON file.
randomise	list	No	Randomisation options to apply.
seed	int	No	Deterministic randomisation seed.

Exceptions

Exception	Condition
ActionValidationError	An input field fails validation.
ActionExecutionError	The algorithm fails during execution.

Example workflow definition

actions:
  - name: learn_graph
    provider: causaliq_discovery
    inputs:
      data: data/sachs.csv
      algorithm: tabu-stable
      output: output/sachs
      hyperparameters:
        score: bic-g