Kitaru

client

Kitaru client for execution and artifact management.

KitaruClient provides a programmatic API for inspecting and managing executions outside flow bodies.

Example::

from kitaru import KitaruClient

client = KitaruClient() execution = client.executions.get("exec-123") print(execution.status)

func_temporary_active_stack(stack_name_or_id) -> Iterator[None]

Temporarily activate a stack while running an operation.

paramstack_name_or_idstr | None

Returns

collections.abc.Iterator[None]
func_snapshot_source_parts(run) -> tuple[str, str | None]

Return (module, attribute) from a run snapshot source.

paramrunPipelineRunResponse

Returns

tuple[str, str | None]
func_import_module_for_replay(module_name, run_id) -> Any

Import a module by name, falling back to sys.modules search.

ZenML records the pipeline source module relative to the archived source root (e.g. replay_with_overrides), but in the running process the module may be loaded under a different path. Three fallback strategies:

  1. Direct importlib.import_module (exact match).
  2. Search sys.modules for a suffix match (e.g. the module is loaded as examples.replay.replay_with_overrides).
  3. Return __main__ — when invoked via python -m pkg.mod, the module is loaded as __main__ and won't appear under its dotted name in sys.modules.
parammodule_namestr
paramrun_idstr | Any

Returns

typing.Any
func_resolve_flow_for_replay(run) -> _ReplayFlowLike

Resolve the original flow wrapper object for a replay source run.

paramrunPipelineRunResponse

Returns

kitaru.client._ReplayFlowLike
func_resolve_pipeline_for_replay(run) -> Any

Resolve the underlying pipeline object for replay fallback.

paramrunPipelineRunResponse

Returns

typing.Any
func_restart_run_from_snapshot(*, run, client, operation_name) -> None

Restart an execution from its stored snapshot metadata.

paramrunPipelineRunResponse
paramclientKitaruClient
paramoperation_namestr

Returns

None