deepset-ai/haystack · error · PipelineInvalidPipelineSnapshotError

Invalid pipeline snapshot: components {invalid_input_data} i

Error message

Invalid pipeline snapshot: components {invalid_input_data} in 'input_data' are not part of the current pipeline.

What it means

_validate_pipeline_snapshot_against_pipeline raises PipelineInvalidPipelineSnapshotError when keys of the snapshot's original_input_data (serialized_data) reference components that do not exist in the current pipeline. The input data must be addressed to components of the pipeline being resumed.

Source

Thrown at haystack/core/pipeline/breakpoint.py:80

    :param pipeline_snapshot: The saved state to validate.
    """

    pipeline_state = pipeline_snapshot.pipeline_state
    valid_components = set(graph.nodes.keys())

    # Check if the ordered_component_names are valid components in the pipeline
    invalid_ordered_components = set(pipeline_snapshot.ordered_component_names) - valid_components
    if invalid_ordered_components:
        raise PipelineInvalidPipelineSnapshotError(
            f"Invalid pipeline snapshot: components {invalid_ordered_components} in 'ordered_component_names' "
            f"are not part of the current pipeline."
        )

    # Check if the original_input_data is valid components in the pipeline
    serialized_input_data = pipeline_snapshot.original_input_data["serialized_data"]
    invalid_input_data = set(serialized_input_data.keys()) - valid_components
    if invalid_input_data:
        raise PipelineInvalidPipelineSnapshotError(
            f"Invalid pipeline snapshot: components {invalid_input_data} in 'input_data' "
            f"are not part of the current pipeline."
        )

    # Validate 'component_visits'
    invalid_component_visits = set(pipeline_state.component_visits.keys()) - valid_components
    if invalid_component_visits:
        raise PipelineInvalidPipelineSnapshotError(
            f"Invalid pipeline snapshot: components {invalid_component_visits} in 'component_visits' "
            f"are not part of the current pipeline."
        )

    component_name = pipeline_snapshot.break_point.component_name
    visit_count = pipeline_snapshot.pipeline_state.component_visits[component_name]

    logger.info(
        "Resuming pipeline from {component} with visit count {visits}", component=component_name, visits=visit_count
    )

View on GitHub (pinned to e318778c9b)

Solutions

  1. Use a snapshot whose original_input_data keys match the current pipeline's components; regenerate it from the current pipeline.
  2. Remove or rename the stale component keys in original_input_data['serialized_data'].
  3. Pre-check in code: set(snapshot.original_input_data['serialized_data']) <= set(pipeline.graph.nodes).

Example fix

// before
snapshot.original_input_data["serialized_data"] = {"old_retriever": {...}}
// after
snapshot.original_input_data["serialized_data"] = {"retriever": {...}}
Defensive patterns

Strategy: validation

Validate before calling

invalid = set(snapshot.original_input_data["serialized_data"]) - set(pipeline.graph.nodes)
assert not invalid, f"input_data references unknown components: {invalid}"

Type guard

def input_data_valid(snapshot, pipeline) -> bool:
    return set(snapshot.original_input_data["serialized_data"]) <= set(pipeline.graph.nodes)

Try / catch

try:
    result = pipeline.run(data, pipeline_snapshot=snapshot)
except PipelineInvalidPipelineSnapshotError as e:
    logger.error("Snapshot input_data invalid: %s", e)

Prevention

When it happens

Trigger: Resuming with a snapshot whose original_input_data contains component keys from another/older pipeline; hand-editing a snapshot JSON and adding or renaming component keys.

Common situations: Sharing snapshot files between different pipeline definitions; refactoring component names after capturing a state; manually constructing snapshots for tests with wrong keys.

Related errors

AI-assisted analysis of deepset-ai/haystack@e318778c9b (2026-08-30). Data as JSON: /api/errors/b884b95c5978323c. Report an issue: GitHub.