Running Pipeline.from_dict on the same payload crashes

[tstadel]

Describe the bug
When a pipeline_dict with nested instance (e.g. a document store) is passed twice to Pipeline.from_dict, a TypeError is raised.

Error message

{
	"name": "TypeError",
	"message": "argument of type 'InMemoryDocumentStore' is not iterable",
	"stack": "---------------------------------------------------------------------------
TypeError                                 Traceback (most recent call last)
Cell In[52], line 1
----> 1 p = Pipeline.from_dict(pipeline_config)

File ~/.local/share/hatch/env/virtual/deepset-cloud-custom-nodes/1TRZ-Wjw/deepset-cloud-custom-nodes/lib/python3.11/site-packages/haystack/core/pipeline/base.py:181, in PipelineBase.from_dict(cls, data, callbacks, **kwargs)
    179         # Create a new one
    180         component_class = component.registry[component_data[\"type\"]]
--> 181         instance = component_from_dict(component_class, component_data, name, callbacks)
    182     pipe.add_component(name=name, instance=instance)
    184 for connection in data.get(\"connections\", []):

File ~/.local/share/hatch/env/virtual/deepset-cloud-custom-nodes/1TRZ-Wjw/deepset-cloud-custom-nodes/lib/python3.11/site-packages/haystack/core/serialization.py:117, in component_from_dict(cls, data, name, callbacks)
    114     return default_from_dict(cls, data)
    116 if callbacks is None or callbacks.component_pre_init is None:
--> 117     return do_from_dict()
    119 with _hook_component_init(component_pre_init_callback):
    120     return do_from_dict()

File ~/.local/share/hatch/env/virtual/deepset-cloud-custom-nodes/1TRZ-Wjw/deepset-cloud-custom-nodes/lib/python3.11/site-packages/haystack/core/serialization.py:112, in component_from_dict.<locals>.do_from_dict()
    110 def do_from_dict():
    111     if hasattr(cls, \"from_dict\"):
--> 112         return cls.from_dict(data)
    114     return default_from_dict(cls, data)

File ~/.local/share/hatch/env/virtual/deepset-cloud-custom-nodes/1TRZ-Wjw/deepset-cloud-custom-nodes/lib/python3.11/site-packages/haystack/components/writers/document_writer.py:79, in DocumentWriter.from_dict(cls, data)
     77 if \"document_store\" not in init_params:
     78     raise DeserializationError(\"Missing 'document_store' in serialization data\")
---> 79 if \"type\" not in init_params[\"document_store\"]:
     80     raise DeserializationError(\"Missing 'type' in document store's serialization data\")
     82 try:

TypeError: argument of type 'InMemoryDocumentStore' is not iterable"
}

Expected behavior

• Pipeline.from_dict does not change payloads.
• Pipeline.from_dict is idempotent.

Additional context
Add any other context about the problem here, like document types / preprocessing steps / settings of reader etc.

To Reproduce

pipeline_yaml = """
components:
    text_converter:
        type: haystack.components.converters.txt.TextFileToDocument
        init_parameters:
            encoding: utf-8

    writer:
        type: haystack.components.writers.document_writer.DocumentWriter
        init_parameters:
            document_store:
                type: haystack.document_stores.in_memory.document_store.InMemoryDocumentStore
                init_parameters: {}
            policy: OVERWRITE

connections: # Defines how the components are connected
    - sender: text_converter.documents
      receiver: writer.documents

max_loops_allowed: 100
"""
import yaml

pipeline_config = yaml.safe_load(pipeline_yaml)
p = Pipeline.from_dict(pipeline_config) # works

p = Pipeline.from_dict(pipeline_config) # crashes

FAQ Check

• Have you had a look at our new FAQ page?

System:

• OS:
• GPU/CPU:
• Haystack version (commit or version number):
• DocumentStore:
• Reader:
• Retriever: