fix: don't modify the input dictionary in from_dict

haystack/components/agents/agent.py

@@ -61,7 +61,7 @@
 )
 from haystack.utils.async_utils import _execute_component_async
 from haystack.utils.callable_serialization import deserialize_callable, serialize_callable
-from haystack.utils.deserialization import deserialize_component_inplace
+from haystack.utils.deserialization import _copy_serialized_data, deserialize_component_inplace
 
 logger = logging.getLogger(__name__)
 
@@ -638,6 +638,7 @@ def from_dict(cls, data: dict[str, Any]) -> "Agent":
         :param data: Dictionary to deserialize from.
         :returns: Deserialized agent.
         """
+        data = _copy_serialized_data(data)  # `from_dict` must not modify the caller's data
         init_params = data.get("init_parameters", {})
 
         deserialize_component_inplace(init_params, key="chat_generator")

haystack/components/builders/chat_prompt_builder.py

@@ -10,6 +10,7 @@
 from haystack.dataclasses.chat_message import ChatMessage, ChatRole, TextContent
 from haystack.lazy_imports import LazyImport
 from haystack.utils import Jinja2TimeExtension
+from haystack.utils.deserialization import _copy_serialized_data
 from haystack.utils.jinja2_chat_extension import ChatMessageExtension
 from haystack.utils.jinja2_extensions import _extract_template_variables_and_assignments
 from haystack.utils.jinja2_sandbox import HaystackSandboxedEnvironment
@@ -347,6 +348,7 @@ def from_dict(cls, data: dict[str, Any]) -> "ChatPromptBuilder":
         :returns:
             The deserialized component.
         """
+        data = _copy_serialized_data(data)  # `from_dict` must not modify the caller's data
         init_parameters = data["init_parameters"]
         template = init_parameters.get("template")
         if template:

haystack/components/converters/docx.py

@@ -15,6 +15,7 @@
 from haystack.components.converters.utils import LinkFormat, get_bytestream_from_source, normalize_metadata
 from haystack.dataclasses import ByteStream
 from haystack.lazy_imports import LazyImport
+from haystack.utils.deserialization import _copy_serialized_data
 
 logger = logging.getLogger(__name__)
 
@@ -167,6 +168,7 @@ def from_dict(cls, data: dict[str, Any]) -> "DOCXToDocument":
         :returns:
             The deserialized component.
         """
+        data = _copy_serialized_data(data)  # `from_dict` must not modify the caller's data
         if "table_format" in data["init_parameters"]:
             data["init_parameters"]["table_format"] = DOCXTableFormat.from_str(data["init_parameters"]["table_format"])
         if "link_format" in data["init_parameters"]:

haystack/components/converters/output_adapter.py

@@ -15,6 +15,7 @@
 from haystack.core.errors import DeserializationError
 from haystack.core.serialization_security import _is_unsafe_deserialization
 from haystack.utils import deserialize_callable, deserialize_type, serialize_callable, serialize_type
+from haystack.utils.deserialization import _copy_serialized_data
 from haystack.utils.jinja2_extensions import _extract_template_variables_and_assignments
 from haystack.utils.jinja2_sandbox import HaystackSandboxedEnvironment
 
@@ -171,6 +172,7 @@ def from_dict(cls, data: dict[str, Any]) -> "OutputAdapter":
         :returns:
             The deserialized component.
         """
+        data = _copy_serialized_data(data)  # `from_dict` must not modify the caller's data
         init_params = data.get("init_parameters", {})
 
         # `unsafe=True` swaps the Jinja sandbox for a NativeEnvironment that executes arbitrary code.

haystack/components/converters/pdfminer.py

@@ -13,6 +13,7 @@
 from haystack.components.converters.utils import LinkFormat, get_bytestream_from_source, normalize_metadata
 from haystack.dataclasses import ByteStream
 from haystack.lazy_imports import LazyImport
+from haystack.utils.deserialization import _copy_serialized_data
 
 with LazyImport("Run 'pip install pdfminer.six'") as pdfminer_import:
     from pdfminer.converter import PDFPageAggregator
@@ -150,6 +151,7 @@ def from_dict(cls, data: dict[str, Any]) -> "PDFMinerToDocument":
         :returns:
             Deserialized component.
         """
+        data = _copy_serialized_data(data)  # `from_dict` must not modify the caller's data
         if "link_format" in data.get("init_parameters", {}):
             data["init_parameters"]["link_format"] = LinkFormat.from_str(data["init_parameters"]["link_format"])
         return default_from_dict(cls, data)

haystack/components/converters/pypdf.py

@@ -12,6 +12,7 @@
 from haystack.components.converters.utils import LinkFormat, get_bytestream_from_source, normalize_metadata
 from haystack.dataclasses import ByteStream
 from haystack.lazy_imports import LazyImport
+from haystack.utils.deserialization import _copy_serialized_data
 
 with LazyImport("Run 'pip install pypdf'") as pypdf_import:
     from pypdf import PdfReader
@@ -167,6 +168,7 @@ def from_dict(cls, data: dict[str, Any]) -> "PyPDFToDocument":
         :returns:
             Deserialized component.
         """
+        data = _copy_serialized_data(data)  # `from_dict` must not modify the caller's data
         if "link_format" in data.get("init_parameters", {}):
             data["init_parameters"]["link_format"] = LinkFormat.from_str(data["init_parameters"]["link_format"])
         return default_from_dict(cls, data)

haystack/components/embedders/azure_document_embedder.py

@@ -10,6 +10,7 @@
 from haystack import component, default_from_dict, default_to_dict, logging
 from haystack.components.embedders import OpenAIDocumentEmbedder
 from haystack.utils import Secret, deserialize_callable, serialize_callable
+from haystack.utils.deserialization import _copy_serialized_data
 from haystack.utils.http_client import init_http_client
 
 logger = logging.getLogger(__name__)
@@ -250,6 +251,7 @@ def from_dict(cls, data: dict[str, Any]) -> "AzureOpenAIDocumentEmbedder":
         :returns:
             Deserialized component.
         """
+        data = _copy_serialized_data(data)  # `from_dict` must not modify the caller's data
         serialized_azure_ad_token_provider = data["init_parameters"].get("azure_ad_token_provider")
         if serialized_azure_ad_token_provider:
             data["init_parameters"]["azure_ad_token_provider"] = deserialize_callable(

haystack/components/embedders/azure_text_embedder.py

@@ -10,6 +10,7 @@
 from haystack import component, default_from_dict, default_to_dict
 from haystack.components.embedders import OpenAITextEmbedder
 from haystack.utils import Secret, deserialize_callable, serialize_callable
+from haystack.utils.deserialization import _copy_serialized_data
 from haystack.utils.http_client import init_http_client
 
 
@@ -226,6 +227,7 @@ def from_dict(cls, data: dict[str, Any]) -> "AzureOpenAITextEmbedder":
         :returns:
             Deserialized component.
         """
+        data = _copy_serialized_data(data)  # `from_dict` must not modify the caller's data
         serialized_azure_ad_token_provider = data["init_parameters"].get("azure_ad_token_provider")
         if serialized_azure_ad_token_provider:
             data["init_parameters"]["azure_ad_token_provider"] = deserialize_callable(

haystack/components/embedders/mock_document_embedder.py

@@ -15,6 +15,7 @@
     _estimate_usage,
 )
 from haystack.utils import deserialize_callable, serialize_callable
+from haystack.utils.deserialization import _copy_serialized_data
 
 
 @component
@@ -122,6 +123,7 @@ def to_dict(self) -> dict[str, Any]:
     @classmethod
     def from_dict(cls, data: dict[str, Any]) -> MockDocumentEmbedder:
         """Deserialize the component from a dictionary."""
+        data = _copy_serialized_data(data)  # `from_dict` must not modify the caller's data
         init_params = data.get("init_parameters", {})
         embedding_fn = init_params.get("embedding_fn")
         if embedding_fn:

haystack/components/embedders/mock_text_embedder.py

@@ -14,6 +14,7 @@
     _estimate_usage,
 )
 from haystack.utils import deserialize_callable, serialize_callable
+from haystack.utils.deserialization import _copy_serialized_data
 
 
 @component
@@ -105,6 +106,7 @@ def to_dict(self) -> dict[str, Any]:
     @classmethod
     def from_dict(cls, data: dict[str, Any]) -> MockTextEmbedder:
         """Deserialize the component from a dictionary."""
+        data = _copy_serialized_data(data)  # `from_dict` must not modify the caller's data
         init_params = data.get("init_parameters", {})
         embedding_fn = init_params.get("embedding_fn")
         if embedding_fn:

haystack/components/evaluators/context_relevance.py

@@ -11,6 +11,7 @@
 from haystack.components.generators.chat.types import ChatGenerator
 from haystack.core.serialization import component_to_dict
 from haystack.utils import deserialize_chatgenerator_inplace
+from haystack.utils.deserialization import _copy_serialized_data
 
 logger = logging.getLogger(__name__)
 
@@ -264,6 +265,7 @@ def from_dict(cls, data: dict[str, Any]) -> "ContextRelevanceEvaluator":
         :returns:
             The deserialized component instance.
         """
+        data = _copy_serialized_data(data)  # `from_dict` must not modify the caller's data
         if data["init_parameters"].get("chat_generator"):
             deserialize_chatgenerator_inplace(data["init_parameters"], key="chat_generator")
         return default_from_dict(cls, data)

haystack/components/evaluators/faithfulness.py

@@ -12,6 +12,7 @@
 from haystack.components.generators.chat.types import ChatGenerator
 from haystack.core.serialization import component_to_dict
 from haystack.utils import deserialize_chatgenerator_inplace
+from haystack.utils.deserialization import _copy_serialized_data
 
 logger = logging.getLogger(__name__)
 
@@ -263,6 +264,7 @@ def from_dict(cls, data: dict[str, Any]) -> "FaithfulnessEvaluator":
         :returns:
             The deserialized component instance.
         """
+        data = _copy_serialized_data(data)  # `from_dict` must not modify the caller's data
         if data["init_parameters"].get("chat_generator"):
             deserialize_chatgenerator_inplace(data["init_parameters"], key="chat_generator")
         return default_from_dict(cls, data)

haystack/components/evaluators/llm_evaluator.py

@@ -17,6 +17,7 @@
 from haystack.core.serialization import component_to_dict
 from haystack.dataclasses.chat_message import ChatMessage
 from haystack.utils import deserialize_chatgenerator_inplace, deserialize_type, serialize_type
+from haystack.utils.deserialization import _copy_serialized_data
 from haystack.utils.misc import _parse_dict_from_json
 
 logger = logging.getLogger(__name__)
@@ -419,6 +420,7 @@ def from_dict(cls, data: dict[str, Any]) -> "LLMEvaluator":
         :returns:
             The deserialized component instance.
         """
+        data = _copy_serialized_data(data)  # `from_dict` must not modify the caller's data
         data["init_parameters"]["inputs"] = [
             (name, deserialize_type(type_)) for name, type_ in data["init_parameters"]["inputs"]
         ]

haystack/components/extractors/image/llm_document_content_extractor.py

@@ -21,6 +21,7 @@
 from haystack.dataclasses.chat_message import ChatMessage
 from haystack.utils import deserialize_chatgenerator_inplace
 from haystack.utils.async_utils import _execute_component_async
+from haystack.utils.deserialization import _copy_serialized_data
 from haystack.utils.misc import _parse_dict_from_json
 
 logger = logging.getLogger(__name__)
@@ -237,6 +238,7 @@ def from_dict(cls, data: dict[str, Any]) -> "LLMDocumentContentExtractor":
         :returns:
             An instance of the component.
         """
+        data = _copy_serialized_data(data)  # `from_dict` must not modify the caller's data
         init_params = data.get("init_parameters", {})
         deserialize_chatgenerator_inplace(init_params, key="chat_generator")
 

haystack/components/extractors/llm_metadata_extractor.py

@@ -23,6 +23,7 @@
 from haystack.dataclasses import ChatMessage
 from haystack.utils import deserialize_chatgenerator_inplace, expand_page_range
 from haystack.utils.async_utils import _execute_component_async
+from haystack.utils.deserialization import _copy_serialized_data
 from haystack.utils.misc import _parse_dict_from_json
 
 logger = logging.getLogger(__name__)
@@ -264,6 +265,7 @@ def from_dict(cls, data: dict[str, Any]) -> "LLMMetadataExtractor":
         :returns:
             An instance of the component.
         """
+        data = _copy_serialized_data(data)  # `from_dict` must not modify the caller's data
 
         deserialize_chatgenerator_inplace(data["init_parameters"], key="chat_generator")
         return default_from_dict(cls, data)

haystack/components/generators/chat/azure.py

@@ -21,6 +21,7 @@
     warm_up_tools,
 )
 from haystack.utils import Secret, deserialize_callable, serialize_callable
+from haystack.utils.deserialization import _copy_serialized_data
 from haystack.utils.http_client import init_http_client
 
 
@@ -364,6 +365,7 @@ def from_dict(cls, data: dict[str, Any]) -> "AzureOpenAIChatGenerator":
         :returns:
             The deserialized component instance.
         """
+        data = _copy_serialized_data(data)  # `from_dict` must not modify the caller's data
         deserialize_tools_or_toolset_inplace(data["init_parameters"], key="tools")
         init_params = data.get("init_parameters", {})
         serialized_callback_handler = init_params.get("streaming_callback")

haystack/components/generators/chat/azure_responses.py

@@ -14,6 +14,7 @@
 from haystack.dataclasses.streaming_chunk import StreamingCallbackT
 from haystack.tools import ToolsType, deserialize_tools_or_toolset_inplace, serialize_tools_or_toolset
 from haystack.utils import Secret, deserialize_callable, serialize_callable
+from haystack.utils.deserialization import _copy_serialized_data
 
 
 @component
@@ -249,6 +250,7 @@ def from_dict(cls, data: dict[str, Any]) -> "AzureOpenAIResponsesChatGenerator":
         :returns:
             The deserialized component instance.
         """
+        data = _copy_serialized_data(data)  # `from_dict` must not modify the caller's data
         # If api_key is a str, it's a callable (Secrets are handled automatically by default_from_dict)
         serialized_api_key = data["init_parameters"].get("api_key")
         if isinstance(serialized_api_key, str):

haystack/components/generators/chat/fallback.py

@@ -13,7 +13,7 @@
 from haystack.dataclasses import ChatMessage, StreamingCallbackT
 from haystack.tools import ToolsType
 from haystack.utils.async_utils import _execute_component_async
-from haystack.utils.deserialization import deserialize_component_inplace
+from haystack.utils.deserialization import _copy_serialized_data, deserialize_component_inplace
 
 logger = logging.getLogger(__name__)
 
@@ -73,6 +73,7 @@ def to_dict(self) -> dict[str, Any]:
     @classmethod
     def from_dict(cls, data: dict[str, Any]) -> FallbackChatGenerator:
         """Rebuild the component from a serialized representation, restoring nested chat generators."""
+        data = _copy_serialized_data(data)  # `from_dict` must not modify the caller's data
         # Reconstruct nested chat generators from their serialized dicts
         init_params = data.get("init_parameters", {})
         serialized = init_params.get("chat_generators") or []

haystack/components/generators/chat/llm.py

@@ -10,7 +10,7 @@
 from haystack.core.serialization import component_to_dict, default_from_dict, default_to_dict
 from haystack.dataclasses import ChatMessage, StreamingCallbackT
 from haystack.utils.callable_serialization import deserialize_callable, serialize_callable
-from haystack.utils.deserialization import deserialize_component_inplace
+from haystack.utils.deserialization import _copy_serialized_data, deserialize_component_inplace
 
 logger = logging.getLogger(__name__)
 
@@ -116,6 +116,7 @@ def from_dict(cls, data: dict[str, Any]) -> "LLM":
         :param data: Dictionary to deserialize from.
         :return: Deserialized LLM instance.
         """
+        data = _copy_serialized_data(data)  # `from_dict` must not modify the caller's data
         init_params = data.get("init_parameters", {})
 
         deserialize_component_inplace(init_params, key="chat_generator")

haystack/components/generators/chat/mock.py

@@ -25,6 +25,7 @@
 from haystack.dataclasses.streaming_chunk import ToolCallDelta, _invoke_streaming_callback
 from haystack.tools import ToolsType
 from haystack.utils import deserialize_callable, serialize_callable
+from haystack.utils.deserialization import _copy_serialized_data
 
 logger = logging.getLogger(__name__)
 
@@ -186,6 +187,7 @@ def to_dict(self) -> dict[str, Any]:
     @classmethod
     def from_dict(cls, data: dict[str, Any]) -> MockChatGenerator:
         """Deserialize the component from a dictionary."""
+        data = _copy_serialized_data(data)  # `from_dict` must not modify the caller's data
         init_params = data.get("init_parameters", {})
         responses = init_params.get("responses")
         if responses is not None:

haystack/components/generators/chat/openai.py

@@ -49,6 +49,7 @@
     warm_up_tools,
 )
 from haystack.utils import Secret, deserialize_callable, serialize_callable
+from haystack.utils.deserialization import _copy_serialized_data
 from haystack.utils.http_client import init_http_client
 
 logger = logging.getLogger(__name__)
@@ -331,6 +332,7 @@ def from_dict(cls, data: dict[str, Any]) -> "OpenAIChatGenerator":
         :returns:
             The deserialized component instance.
         """
+        data = _copy_serialized_data(data)  # `from_dict` must not modify the caller's data
         deserialize_tools_or_toolset_inplace(data["init_parameters"], key="tools")
         init_params = data.get("init_parameters", {})
         serialized_callback_handler = init_params.get("streaming_callback")

haystack/components/generators/chat/openai_responses.py

@@ -38,6 +38,7 @@
     warm_up_tools,
 )
 from haystack.utils import Secret, deserialize_callable, serialize_callable
+from haystack.utils.deserialization import _copy_serialized_data
 from haystack.utils.http_client import init_http_client
 
 logger = logging.getLogger(__name__)
@@ -344,6 +345,7 @@ def from_dict(cls, data: dict[str, Any]) -> "OpenAIResponsesChatGenerator":
         :returns:
             The deserialized component instance.
         """
+        data = _copy_serialized_data(data)  # `from_dict` must not modify the caller's data
         # we only deserialize the tools if they are haystack tools
         # because openai tools are not serialized in the same way
         tools = data["init_parameters"].get("tools")

haystack/components/joiners/branch.py

@@ -7,6 +7,7 @@
 from haystack import component, default_from_dict, default_to_dict
 from haystack.core.component.types import GreedyVariadic
 from haystack.utils import deserialize_type, serialize_type
+from haystack.utils.deserialization import _copy_serialized_data
 
 
 @component
@@ -113,6 +114,7 @@ def from_dict(cls, data: dict[str, Any]) -> "BranchJoiner":
         :returns:
             A deserialized `BranchJoiner` instance.
         """
+        data = _copy_serialized_data(data)  # `from_dict` must not modify the caller's data
         data["init_parameters"]["type_"] = deserialize_type(data["init_parameters"]["type_"])
         return default_from_dict(cls, data)
 

haystack/components/joiners/list_joiner.py

@@ -8,6 +8,7 @@
 from haystack import component, default_from_dict, default_to_dict
 from haystack.core.component.types import Variadic
 from haystack.utils import deserialize_type, serialize_type
+from haystack.utils.deserialization import _copy_serialized_data
 
 
 @component
@@ -96,6 +97,7 @@ def from_dict(cls, data: dict[str, Any]) -> "ListJoiner":
         :param data: Dictionary to deserialize from.
         :returns: Deserialized component.
         """
+        data = _copy_serialized_data(data)  # `from_dict` must not modify the caller's data
         init_parameters = data.get("init_parameters")
         if init_parameters is not None and init_parameters.get("list_type_") is not None:
             data["init_parameters"]["list_type_"] = deserialize_type(data["init_parameters"]["list_type_"])