execution

Base class for reusable anomaly workflows.

Adapt an existing worker transport or application-specific event sink.

A validated graph of job nodes.

Raised when a job definition is structurally invalid.

One processing step in a job graph.

Raised when a job or node cannot be executed.

Outcome of one synchronous job execution.

Execute job definitions synchronously with a propagated context.

Append portable JSON events to a local file (offline by default).

Run and inspect local models with a deliberately small execution contract.

One experiment's observations, predictions, metrics, and provenance.

Outcome of executing one node.

Execute job nodes against a shared context.

Train, score, and classify a dataset with built-in library components.

Train, score, and classify a dataset with built-in library components.

One execution in a project, with a step-indexed history and summary.

Persist events synchronously; raise on failure so callers can retry.

Host-injected logging, findings, metrics, datasets, and compute services.

A validated, deterministic DAG with explicit predecessor outputs.

One Python or inference step with an explicit runtime selection.

Return the run bound to this context, or ``None``.

Finish the current standalone execution; a worker owns hosted lifecycles.

Serialize a DataFrame into JSON-friendly, index-free records.

Return the context activated by the current execution host.

Start and bind a standalone run, or reuse the active hosted execution.

Write values to the active run (see :meth:`Run.log`).

Build a valid context dict, optionally seeding the input data.

Read a boolean flag from `flags` with a top-level fallback.

Read one conventional frame key from the context.

Rebuild a DataFrame from context records.

Execute user-authored function-body code against the context.

Temporarily bind an existing run without changing its lifecycle.

Write a frame into the context in its JSON-encodable form.