langchain-core 1.0.0a3__tar.gz → 1.0.0a5__tar.gz

{langchain_core-1.0.0a3 → langchain_core-1.0.0a5}/PKG-INFO

@@ -1,19 +1,19 @@
 Metadata-Version: 2.1
 Name: langchain-core
-Version: 1.0.0a3
+Version: 1.0.0a5
 Summary: Building applications with LLMs through composability
 License: MIT
 Project-URL: Source Code, https://github.com/langchain-ai/langchain/tree/master/libs/core
 Project-URL: Release Notes, https://github.com/langchain-ai/langchain/releases?q=tag%3A%22langchain-core%3D%3D0%22&expanded=true
 Project-URL: repository, https://github.com/langchain-ai/langchain
-Requires-Python: >=3.10
-Requires-Dist: langsmith>=0.3.45
+Requires-Python: <4.0.0,>=3.10.0
+Requires-Dist: langsmith<1.0.0,>=0.3.45
 Requires-Dist: tenacity!=8.4.0,<10.0.0,>=8.1.0
-Requires-Dist: jsonpatch<2.0,>=1.33
-Requires-Dist: PyYAML>=5.3
-Requires-Dist: typing-extensions>=4.7
-Requires-Dist: packaging>=23.2
-Requires-Dist: pydantic>=2.7.4
+Requires-Dist: jsonpatch<2.0.0,>=1.33.0
+Requires-Dist: PyYAML<7.0.0,>=5.3.0
+Requires-Dist: typing-extensions<5.0.0,>=4.7.0
+Requires-Dist: packaging<26.0.0,>=23.2.0
+Requires-Dist: pydantic<3.0.0,>=2.7.4
 Description-Content-Type: text/markdown
 
 # 🦜🍎️ LangChain Core

{langchain_core-1.0.0a3 → langchain_core-1.0.0a5}/langchain_core/_api/beta_decorator.py

@@ -174,6 +174,7 @@ def beta(
             def finalize(_wrapper: Callable[..., Any], new_doc: str) -> Any:
                 """Finalize the property."""
                 return property(fget=_fget, fset=_fset, fdel=_fdel, doc=new_doc)
+
         else:
             _name = _name or obj.__qualname__
             if not _obj_type:
@@ -226,17 +227,17 @@ def warn_beta(
 ) -> None:
     """Display a standardized beta annotation.
 
-    Arguments:
-        message : str, optional
+    Args:
+        message:
             Override the default beta message. The
             %(name)s, %(obj_type)s, %(addendum)s
             format specifiers will be replaced by the
             values of the respective arguments passed to this function.
-        name : str, optional
+        name:
             The name of the annotated object.
-        obj_type : str, optional
+        obj_type:
             The object type being annotated.
-        addendum : str, optional
+        addendum:
             Additional text appended directly to the final message.
     """
     if not message:

{langchain_core-1.0.0a3 → langchain_core-1.0.0a5}/langchain_core/_api/deprecation.py

@@ -431,35 +431,35 @@ def warn_deprecated(
 ) -> None:
     """Display a standardized deprecation.
 
-    Arguments:
-        since : str
+    Args:
+        since:
             The release at which this API became deprecated.
-        message : str, optional
+        message:
             Override the default deprecation message. The %(since)s,
             %(name)s, %(alternative)s, %(obj_type)s, %(addendum)s,
             and %(removal)s format specifiers will be replaced by the
             values of the respective arguments passed to this function.
-        name : str, optional
+        name:
             The name of the deprecated object.
-        alternative : str, optional
+        alternative:
             An alternative API that the user may use in place of the
             deprecated API. The deprecation warning will tell the user
             about this alternative if provided.
-        alternative_import: str, optional
+        alternative_import:
             An alternative import that the user may use instead.
-        pending : bool, optional
+        pending:
             If True, uses a PendingDeprecationWarning instead of a
             DeprecationWarning. Cannot be used together with removal.
-        obj_type : str, optional
+        obj_type:
             The object type being deprecated.
-        addendum : str, optional
+        addendum:
             Additional text appended directly to the final message.
-        removal : str, optional
+        removal:
             The expected removal version. With the default (an empty
             string), a removal version is automatically computed from
             since. Set to other Falsy values to not schedule a removal
             date. Cannot be used together with pending.
-        package: str, optional
+        package:
             The package of the deprecated object.
     """
     if not pending:

{langchain_core-1.0.0a3 → langchain_core-1.0.0a5}/langchain_core/callbacks/manager.py

@@ -92,7 +92,7 @@ def trace_as_chain_group(
         metadata (dict[str, Any], optional): The metadata to apply to all runs.
             Defaults to None.
 
-    .. note:
+    .. note::
         Must have ``LANGCHAIN_TRACING_V2`` env var set to true to see the trace in
         LangSmith.
 
@@ -179,7 +179,7 @@ async def atrace_as_chain_group(
     Yields:
         The async callback manager for the chain group.
 
-    .. note:
+    .. note::
         Must have ``LANGCHAIN_TRACING_V2`` env var set to true to see the trace in
         LangSmith.
 

{langchain_core-1.0.0a3 → langchain_core-1.0.0a5}/langchain_core/callbacks/usage.py

@@ -32,7 +32,7 @@ class UsageMetadataCallbackHandler(BaseCallbackHandler):
             result_2 = llm_2.invoke("Hello", config={"callbacks": [callback]})
             callback.usage_metadata
 
-        .. code-block:: none
+        .. code-block::
 
             {'gpt-4o-mini-2024-07-18': {'input_tokens': 8,
               'output_tokens': 10,
@@ -119,7 +119,7 @@ def get_usage_metadata_callback(
                 llm_2.invoke("Hello")
                 print(cb.usage_metadata)
 
-        .. code-block:: none
+        .. code-block::
 
             {'gpt-4o-mini-2024-07-18': {'input_tokens': 8,
               'output_tokens': 10,

{langchain_core-1.0.0a3 → langchain_core-1.0.0a5}/langchain_core/document_loaders/langsmith.py

@@ -31,7 +31,7 @@ class LangSmithLoader(BaseLoader):
             for doc in loader.lazy_load():
                 docs.append(doc)
 
-        .. code-block:: pycon
+        .. code-block:: python
 
             # -> [Document("...", metadata={"inputs": {...}, "outputs": {...}, ...}), ...]
 

{langchain_core-1.0.0a3 → langchain_core-1.0.0a5}/langchain_core/indexing/api.py

@@ -296,7 +296,11 @@ def index(
     For the time being, documents are indexed using their hashes, and users
     are not able to specify the uid of the document.
 
-    Important:
+    .. versionchanged:: 0.3.25
+        Added ``scoped_full`` cleanup mode.
+
+    .. important::
+
         * In full mode, the loader should be returning
           the entire dataset, and not just a subset of the dataset.
           Otherwise, the auto_cleanup will remove documents that it is not
@@ -309,7 +313,7 @@ def index(
           chunks, and we index them using a batch size of 5, we'll have 3 batches
           all with the same source id. In general, to avoid doing too much
           redundant work select as big a batch size as possible.
-        * The `scoped_full` mode is suitable if determining an appropriate batch size
+        * The ``scoped_full`` mode is suitable if determining an appropriate batch size
           is challenging or if your data loader cannot return the entire dataset at
           once. This mode keeps track of source IDs in memory, which should be fine
           for most use cases. If your dataset is large (10M+ docs), you will likely
@@ -378,10 +382,6 @@ def index(
         TypeError: If ``vectorstore`` is not a VectorStore or a DocumentIndex.
         AssertionError: If ``source_id`` is None when cleanup mode is incremental.
             (should be unreachable code).
-
-    .. version_modified:: 0.3.25
-
-        * Added `scoped_full` cleanup mode.
     """
     # Behavior is deprecated, but we keep it for backwards compatibility.
     # # Warn only once per process.
@@ -636,26 +636,30 @@ async def aindex(
     documents were deleted, which documents should be skipped.
 
     For the time being, documents are indexed using their hashes, and users
-     are not able to specify the uid of the document.
-
-    Important:
-       * In full mode, the loader should be returning
-         the entire dataset, and not just a subset of the dataset.
-         Otherwise, the auto_cleanup will remove documents that it is not
-         supposed to.
-       * In incremental mode, if documents associated with a particular
-         source id appear across different batches, the indexing API
-         will do some redundant work. This will still result in the
-         correct end state of the index, but will unfortunately not be
-         100% efficient. For example, if a given document is split into 15
-         chunks, and we index them using a batch size of 5, we'll have 3 batches
-         all with the same source id. In general, to avoid doing too much
-         redundant work select as big a batch size as possible.
-       * The `scoped_full` mode is suitable if determining an appropriate batch size
-         is challenging or if your data loader cannot return the entire dataset at
-         once. This mode keeps track of source IDs in memory, which should be fine
-         for most use cases. If your dataset is large (10M+ docs), you will likely
-         need to parallelize the indexing process regardless.
+    are not able to specify the uid of the document.
+
+    .. versionchanged:: 0.3.25
+        Added ``scoped_full`` cleanup mode.
+
+    .. important::
+
+        * In full mode, the loader should be returning
+          the entire dataset, and not just a subset of the dataset.
+          Otherwise, the auto_cleanup will remove documents that it is not
+          supposed to.
+        * In incremental mode, if documents associated with a particular
+          source id appear across different batches, the indexing API
+          will do some redundant work. This will still result in the
+          correct end state of the index, but will unfortunately not be
+          100% efficient. For example, if a given document is split into 15
+          chunks, and we index them using a batch size of 5, we'll have 3 batches
+          all with the same source id. In general, to avoid doing too much
+          redundant work select as big a batch size as possible.
+        * The ``scoped_full`` mode is suitable if determining an appropriate batch size
+          is challenging or if your data loader cannot return the entire dataset at
+          once. This mode keeps track of source IDs in memory, which should be fine
+          for most use cases. If your dataset is large (10M+ docs), you will likely
+          need to parallelize the indexing process regardless.
 
     Args:
         docs_source: Data loader or iterable of documents to index.
@@ -720,10 +724,6 @@ async def aindex(
         TypeError: If ``vector_store`` is not a VectorStore or DocumentIndex.
         AssertionError: If ``source_id_key`` is None when cleanup mode is
             incremental or ``scoped_full`` (should be unreachable).
-
-    .. version_modified:: 0.3.25
-
-        * Added `scoped_full` cleanup mode.
     """
     # Behavior is deprecated, but we keep it for backwards compatibility.
     # # Warn only once per process.

{langchain_core-1.0.0a3 → langchain_core-1.0.0a5}/langchain_core/language_models/chat_models.py

@@ -48,7 +48,6 @@ from langchain_core.messages import (
     message_chunk_to_message,
 )
 from langchain_core.messages.block_translators.openai import (
-    convert_to_openai_data_block,
     convert_to_openai_image_block,
 )
 from langchain_core.output_parsers.openai_tools import (
@@ -149,16 +148,18 @@ def _format_for_tracing(messages: list[BaseMessage]) -> list[BaseMessage]:
                         block.get("type") == "file"
                         and is_data_content_block(block)  # v0 (image/audio/file) or v1
                         and "base64" in block
-                        # Narrows to old Base64ContentBlock or new FileContentBlock
+                        # Backward compat: convert v1 base64 blocks to v0
                     ):
                         if message_to_trace is message:
                             # Shallow copy
                             message_to_trace = message.model_copy()
                             message_to_trace.content = list(message_to_trace.content)
 
-                        message_to_trace.content[idx] = convert_to_openai_data_block(  # type: ignore[index]
-                            block
-                        )
+                        message_to_trace.content[idx] = {  # type: ignore[index]
+                            **{k: v for k, v in block.items() if k != "base64"},
+                            "data": block["base64"],
+                            "source_type": "base64",
+                        }
                     elif len(block) == 1 and "type" not in block:
                         # Tracing assumes all content blocks have a "type" key. Here
                         # we add this key if it is missing, and there's an obvious
@@ -518,7 +519,7 @@ class BaseChatModel(BaseLanguageModel[AIMessage], ABC):
         **kwargs: Any,
     ) -> Iterator[AIMessageChunk]:
         if not self._should_stream(async_api=False, **{**kwargs, "stream": True}):
-            # model doesn't implement streaming, so use default implementation
+            # Model doesn't implement streaming, so use default implementation
             yield cast(
                 "AIMessageChunk",
                 self.invoke(input, config=config, stop=stop, **kwargs),

{langchain_core-1.0.0a3 → langchain_core-1.0.0a5}/langchain_core/language_models/fake_chat_models.py

@@ -19,7 +19,7 @@ from langchain_core.runnables import RunnableConfig
 
 
 class FakeMessagesListChatModel(BaseChatModel):
-    """Fake ChatModel for testing purposes."""
+    """Fake ``ChatModel`` for testing purposes."""
 
     responses: list[BaseMessage]
     """List of responses to **cycle** through in order."""
@@ -222,10 +222,11 @@ class GenericFakeChatModel(BaseChatModel):
     """Generic fake chat model that can be used to test the chat model interface.
 
     * Chat model should be usable in both sync and async tests
-    * Invokes on_llm_new_token to allow for testing of callback related code for new
+    * Invokes ``on_llm_new_token`` to allow for testing of callback related code for new
       tokens.
     * Includes logic to break messages into message chunk to facilitate testing of
       streaming.
+
     """
 
     messages: Iterator[Union[AIMessage, str]]
@@ -240,6 +241,7 @@ class GenericFakeChatModel(BaseChatModel):
     .. warning::
         Streaming is not implemented yet. We should try to implement it in the future by
         delegating to invoke and then breaking the resulting output into message chunks.
+
     """
 
     @override
@@ -367,6 +369,7 @@ class ParrotFakeChatModel(BaseChatModel):
     """Generic fake chat model that can be used to test the chat model interface.
 
     * Chat model should be usable in both sync and async tests
+
     """
 
     @override

{langchain_core-1.0.0a3 → langchain_core-1.0.0a5}/langchain_core/load/serializable.py

@@ -111,7 +111,7 @@ class Serializable(BaseModel, ABC):
 
     # Remove default BaseModel init docstring.
     def __init__(self, *args: Any, **kwargs: Any) -> None:
-        """"""  # noqa: D419
+        """"""  # noqa: D419  # Intentional blank docstring
         super().__init__(*args, **kwargs)
 
     @classmethod

{langchain_core-1.0.0a3 → langchain_core-1.0.0a5}/langchain_core/messages/__init__.py

@@ -41,9 +41,6 @@ if TYPE_CHECKING:
         Annotation,
         AudioContentBlock,
         Citation,
-        CodeInterpreterCall,
-        CodeInterpreterOutput,
-        CodeInterpreterResult,
         ContentBlock,
         DataContentBlock,
         FileContentBlock,
@@ -53,10 +50,11 @@ if TYPE_CHECKING:
         NonStandardContentBlock,
         PlainTextContentBlock,
         ReasoningContentBlock,
+        ServerToolCall,
+        ServerToolCallChunk,
+        ServerToolResult,
         TextContentBlock,
         VideoContentBlock,
-        WebSearchCall,
-        WebSearchResult,
         is_data_content_block,
     )
     from langchain_core.messages.function import FunctionMessage, FunctionMessageChunk
@@ -96,9 +94,6 @@ __all__ = (
     "ChatMessage",
     "ChatMessageChunk",
     "Citation",
-    "CodeInterpreterCall",
-    "CodeInterpreterOutput",
-    "CodeInterpreterResult",
     "ContentBlock",
     "DataContentBlock",
     "FileContentBlock",
@@ -114,6 +109,9 @@ __all__ = (
     "PlainTextContentBlock",
     "ReasoningContentBlock",
     "RemoveMessage",
+    "ServerToolCall",
+    "ServerToolCallChunk",
+    "ServerToolResult",
     "SystemMessage",
     "SystemMessageChunk",
     "TextContentBlock",
@@ -122,8 +120,6 @@ __all__ = (
     "ToolMessage",
     "ToolMessageChunk",
     "VideoContentBlock",
-    "WebSearchCall",
-    "WebSearchResult",
     "_message_from_dict",
     "convert_to_messages",
     "convert_to_openai_data_block",
@@ -156,9 +152,6 @@ _dynamic_imports = {
     "ContentBlock": "content",
     "ChatMessage": "chat",
     "ChatMessageChunk": "chat",
-    "CodeInterpreterCall": "content",
-    "CodeInterpreterOutput": "content",
-    "CodeInterpreterResult": "content",
     "DataContentBlock": "content",
     "FileContentBlock": "content",
     "FunctionMessage": "function",
@@ -170,10 +163,11 @@ _dynamic_imports = {
     "PlainTextContentBlock": "content",
     "ReasoningContentBlock": "content",
     "RemoveMessage": "modifier",
+    "ServerToolCall": "content",
+    "ServerToolCallChunk": "content",
+    "ServerToolResult": "content",
     "SystemMessage": "system",
     "SystemMessageChunk": "system",
-    "WebSearchCall": "content",
-    "WebSearchResult": "content",
     "ImageContentBlock": "content",
     "InvalidToolCall": "tool",
     "TextContentBlock": "content",

{langchain_core-1.0.0a3 → langchain_core-1.0.0a5}/langchain_core/messages/ai.py

@@ -13,6 +13,7 @@ from langchain_core.messages import content as types
 from langchain_core.messages.base import (
     BaseMessage,
     BaseMessageChunk,
+    _extract_reasoning_from_additional_kwargs,
     merge_content,
 )
 from langchain_core.messages.content import InvalidToolCall
@@ -39,7 +40,6 @@ class InputTokenDetails(TypedDict, total=False):
     Does *not* need to sum to full input token count. Does *not* need to have all keys.
 
     Example:
-
         .. code-block:: python
 
             {
@@ -66,6 +66,7 @@ class InputTokenDetails(TypedDict, total=False):
 
     Since there was a cache hit, the tokens were read from the cache. More precisely,
     the model state given these tokens was read from the cache.
+
     """
 
 
@@ -75,7 +76,6 @@ class OutputTokenDetails(TypedDict, total=False):
     Does *not* need to sum to full output token count. Does *not* need to have all keys.
 
     Example:
-
         .. code-block:: python
 
             {
@@ -94,6 +94,7 @@ class OutputTokenDetails(TypedDict, total=False):
 
     Tokens generated by the model in a chain of thought process (i.e. by OpenAI's o1
     models) that are not returned as part of model output.
+
     """
 
 
@@ -103,7 +104,6 @@ class UsageMetadata(TypedDict):
     This is a standard representation of token usage that is consistent across models.
 
     Example:
-
         .. code-block:: python
 
             {
@@ -142,6 +142,7 @@ class UsageMetadata(TypedDict):
     """Breakdown of output token counts.
 
     Does *not* need to sum to full output token count. Does *not* need to have all keys.
+
     """
 
 
@@ -153,6 +154,7 @@ class AIMessage(BaseMessage):
     This message represents the output of the model and consists of both
     the raw output as returned by the model together standardized fields
     (e.g., tool calls, usage metadata) added by the LangChain framework.
+
     """
 
     tool_calls: list[ToolCall] = []
@@ -163,6 +165,7 @@ class AIMessage(BaseMessage):
     """If provided, usage metadata for a message, such as token counts.
 
     This is a standard representation of token usage that is consistent across models.
+
     """
 
     type: Literal["ai"] = "ai"
@@ -189,7 +192,15 @@ class AIMessage(BaseMessage):
         content_blocks: Optional[list[types.ContentBlock]] = None,
         **kwargs: Any,
     ) -> None:
-        """Specify ``content`` as positional arg or ``content_blocks`` for typing."""
+        """Initialize ``AIMessage``.
+
+        Specify ``content`` as positional arg or ``content_blocks`` for typing.
+
+        Args:
+            content: The content of the message.
+            content_blocks: Typed standard content.
+            kwargs: Additional arguments to pass to the parent class.
+        """
         if content_blocks is not None:
             # If there are tool calls in content_blocks, but not in tool_calls, add them
             content_tool_calls = [
@@ -215,7 +226,12 @@ class AIMessage(BaseMessage):
 
     @property
     def content_blocks(self) -> list[types.ContentBlock]:
-        """Return content blocks of the message."""
+        """Return content blocks of the message.
+
+        If the message has a known model provider, use the provider-specific translator
+        first before falling back to best-effort parsing. For details, see the property
+        on ``BaseMessage``.
+        """
         if self.response_metadata.get("output_version") == "v1":
             return cast("list[types.ContentBlock]", self.content)
 
@@ -256,6 +272,15 @@ class AIMessage(BaseMessage):
                         tool_call_block["extras"] = tool_call["extras"]  # type: ignore[typeddict-item]
                     blocks.append(tool_call_block)
 
+        # Best-effort reasoning extraction from additional_kwargs
+        # Only add reasoning if not already present
+        # Insert before all other blocks to keep reasoning at the start
+        has_reasoning = any(block.get("type") == "reasoning" for block in blocks)
+        if not has_reasoning and (
+            reasoning_block := _extract_reasoning_from_additional_kwargs(self)
+        ):
+            blocks.insert(0, reasoning_block)
+
         return blocks
 
     # TODO: remove this logic if possible, reducing breaking nature of changes
@@ -315,6 +340,7 @@ class AIMessage(BaseMessage):
 
         Returns:
             A pretty representation of the message.
+
         """
         base = super().pretty_repr(html=html)
         lines = []
@@ -354,7 +380,10 @@ class AIMessageChunk(AIMessage, BaseMessageChunk):
     # non-chunk variant.
     type: Literal["AIMessageChunk"] = "AIMessageChunk"  # type: ignore[assignment]
     """The type of the message (used for deserialization).
-    Defaults to "AIMessageChunk"."""
+
+    Defaults to ``AIMessageChunk``.
+
+    """
 
     tool_call_chunks: list[ToolCallChunk] = []
     """If provided, tool call chunks associated with the message."""
@@ -417,6 +446,15 @@ class AIMessageChunk(AIMessage, BaseMessageChunk):
                     tc["index"] = idx
                 blocks.append(tc)
 
+        # Best-effort reasoning extraction from additional_kwargs
+        # Only add reasoning if not already present
+        # Insert before all other blocks to keep reasoning at the start
+        has_reasoning = any(block.get("type") == "reasoning" for block in blocks)
+        if not has_reasoning and (
+            reasoning_block := _extract_reasoning_from_additional_kwargs(self)
+        ):
+            blocks.insert(0, reasoning_block)
+
         return blocks
 
     @model_validator(mode="after")
@@ -424,7 +462,10 @@ class AIMessageChunk(AIMessage, BaseMessageChunk):
         """Initialize tool calls from tool call chunks.
 
         Returns:
-            This ``AIMessageChunk``.
+            The values with tool calls initialized.
+
+        Raises:
+            ValueError: If the tool call chunks are malformed.
         """
         if not self.tool_call_chunks:
             if self.tool_calls:
@@ -508,6 +549,31 @@ class AIMessageChunk(AIMessage, BaseMessageChunk):
 
         return self
 
+    @model_validator(mode="after")
+    def init_server_tool_calls(self) -> Self:
+        """Parse server_tool_call_chunks."""
+        if (
+            self.chunk_position == "last"
+            and self.response_metadata.get("output_version") == "v1"
+            and isinstance(self.content, list)
+        ):
+            for idx, block in enumerate(self.content):
+                if (
+                    isinstance(block, dict)
+                    and block.get("type")
+                    in ("server_tool_call", "server_tool_call_chunk")
+                    and (args_str := block.get("args"))
+                    and isinstance(args_str, str)
+                ):
+                    try:
+                        args = json.loads(args_str)
+                        if isinstance(args, dict):
+                            self.content[idx]["type"] = "server_tool_call"  # type: ignore[index]
+                            self.content[idx]["args"] = args  # type: ignore[index]
+                    except json.JSONDecodeError:
+                        pass
+        return self
+
     @overload  # type: ignore[override]  # summing BaseMessages gives ChatPromptTemplate
     def __add__(self, other: "AIMessageChunk") -> "AIMessageChunk": ...
 
@@ -677,9 +743,9 @@ def add_usage(
 def subtract_usage(
     left: Optional[UsageMetadata], right: Optional[UsageMetadata]
 ) -> UsageMetadata:
-    """Recursively subtract two UsageMetadata objects.
+    """Recursively subtract two ``UsageMetadata`` objects.
 
-    Token counts cannot be negative so the actual operation is max(left - right, 0).
+    Token counts cannot be negative so the actual operation is ``max(left - right, 0)``.
 
     Example:
         .. code-block:: python