Skip to content

Public Python API

Automo 0.3.0a2 is an alpha. The package distinguishes stable extension surfaces from implementation details so external projects do not need to depend on filesystem services or orchestration internals.

Root-level public contracts

The following are intentionally importable from automo:

  • data: DataSource, DataSnapshot, DataInput;
  • features: FeatureSpec, FeatureSetSpec, FeatureComputer, FeatureSetInput;
  • objectives/evaluation: ObjectiveSpec, MetricSpec, MetricDirection, MetricScope, EvaluationSpec, EvaluationContext, Evaluator;
  • models: ModelSpec, ModelTrainer, TrainingRequest, TrainingResult, ModelOutputBatch;
  • graphs: ModelGraphSpec, ModelNodeSpec, ModelOutputInput, CrossFitSpec;
  • plugin/runtime: ResearchPlugin, ResearchRuntime;
  • research governance: ResearchGovernance, ResearchHypothesis, ResearchProvenance, ModelComponent, ModelRelation, ModelCandidate, CandidateInput, CompositionExperiment, HypothesisObjective, EvaluationDepth.

These contracts are the preferred extension boundary for project packages.

Subpackage public contracts

Some lifecycle abstractions are intentionally namespaced:

python from automo.registry import ModelRegistry, ModelStatus from automo.refresh import ModelPoolSpec, SplitStrategy from automo.research import ResearchSearchSpace, ResearchBudget

Experimental/internal modules

Filesystem stores, orchestration services, local execution helpers, and modules named service, local, or similar are implementation details unless explicitly documented otherwise. External packages should not rely on their concrete file layout during the alpha series.

Compatibility policy

Persistent public artifacts have explicit artifact_type and schema_version fields. Python APIs may still evolve during 0.3.0aN, but changes to the documented extension contracts should be deliberate and recorded in the changelog.