langchain-core 0.3.71__py3-none-any.whl → 0.4.0.dev0__py3-none-any.whl

langchain_core/messages/human.py

@@ -28,6 +28,7 @@ class HumanMessage(BaseMessage):
             # Instantiate a chat model and invoke it with the messages
             model = ...
             print(model.invoke(messages))
+
     """
 
     example: bool = False

langchain_core/messages/modifier.py

@@ -13,7 +13,7 @@ class RemoveMessage(BaseMessage):
 
     def __init__(
         self,
-        id: str,  # noqa: A002
+        id: str,
         **kwargs: Any,
     ) -> None:
         """Create a RemoveMessage.

langchain_core/messages/tool.py

@@ -5,9 +5,12 @@ from typing import Any, Literal, Optional, Union
 from uuid import UUID
 
 from pydantic import Field, model_validator
-from typing_extensions import NotRequired, TypedDict, override
+from typing_extensions import override
 
 from langchain_core.messages.base import BaseMessage, BaseMessageChunk, merge_content
+from langchain_core.messages.content_blocks import InvalidToolCall as InvalidToolCall
+from langchain_core.messages.content_blocks import ToolCall as ToolCall
+from langchain_core.messages.content_blocks import ToolCallChunk as ToolCallChunk
 from langchain_core.utils._merge import merge_dicts, merge_obj
 
 
@@ -59,6 +62,7 @@ class ToolMessage(BaseMessage, ToolOutputMixin):
     The tool_call_id field is used to associate the tool call request with the
     tool call response. This is useful in situations where a chat model is able
     to request multiple tool calls in parallel.
+
     """  # noqa: E501
 
     tool_call_id: str
@@ -176,41 +180,11 @@ class ToolMessageChunk(ToolMessage, BaseMessageChunk):
         return super().__add__(other)
 
 
-class ToolCall(TypedDict):
-    """Represents a request to call a tool.
-
-    Example:
-
-        .. code-block:: python
-
-            {
-                "name": "foo",
-                "args": {"a": 1},
-                "id": "123"
-            }
-
-        This represents a request to call the tool named "foo" with arguments {"a": 1}
-        and an identifier of "123".
-    """
-
-    name: str
-    """The name of the tool to be called."""
-    args: dict[str, Any]
-    """The arguments to the tool call."""
-    id: Optional[str]
-    """An identifier associated with the tool call.
-
-    An identifier is needed to associate a tool call request with a tool
-    call result in events when multiple concurrent tool calls are made.
-    """
-    type: NotRequired[Literal["tool_call"]]
-
-
 def tool_call(
     *,
     name: str,
     args: dict[str, Any],
-    id: Optional[str],  # noqa: A002
+    id: Optional[str],
 ) -> ToolCall:
     """Create a tool call.
 
@@ -222,42 +196,11 @@ def tool_call(
     return ToolCall(name=name, args=args, id=id, type="tool_call")
 
 
-class ToolCallChunk(TypedDict):
-    """A chunk of a tool call (e.g., as part of a stream).
-
-    When merging ToolCallChunks (e.g., via AIMessageChunk.__add__),
-    all string attributes are concatenated. Chunks are only merged if their
-    values of `index` are equal and not None.
-
-    Example:
-
-    .. code-block:: python
-
-        left_chunks = [ToolCallChunk(name="foo", args='{"a":', index=0)]
-        right_chunks = [ToolCallChunk(name=None, args='1}', index=0)]
-
-        (
-            AIMessageChunk(content="", tool_call_chunks=left_chunks)
-            + AIMessageChunk(content="", tool_call_chunks=right_chunks)
-        ).tool_call_chunks == [ToolCallChunk(name='foo', args='{"a":1}', index=0)]
-    """
-
-    name: Optional[str]
-    """The name of the tool to be called."""
-    args: Optional[str]
-    """The arguments to the tool call."""
-    id: Optional[str]
-    """An identifier associated with the tool call."""
-    index: Optional[int]
-    """The index of the tool call in a sequence."""
-    type: NotRequired[Literal["tool_call_chunk"]]
-
-
 def tool_call_chunk(
     *,
     name: Optional[str] = None,
     args: Optional[str] = None,
-    id: Optional[str] = None,  # noqa: A002
+    id: Optional[str] = None,
     index: Optional[int] = None,
 ) -> ToolCallChunk:
     """Create a tool call chunk.
@@ -273,29 +216,11 @@ def tool_call_chunk(
     )
 
 
-class InvalidToolCall(TypedDict):
-    """Allowance for errors made by LLM.
-
-    Here we add an `error` key to surface errors made during generation
-    (e.g., invalid JSON arguments.)
-    """
-
-    name: Optional[str]
-    """The name of the tool to be called."""
-    args: Optional[str]
-    """The arguments to the tool call."""
-    id: Optional[str]
-    """An identifier associated with the tool call."""
-    error: Optional[str]
-    """An error message associated with the tool call."""
-    type: NotRequired[Literal["invalid_tool_call"]]
-
-
 def invalid_tool_call(
     *,
     name: Optional[str] = None,
     args: Optional[str] = None,
-    id: Optional[str] = None,  # noqa: A002
+    id: Optional[str] = None,
     error: Optional[str] = None,
 ) -> InvalidToolCall:
     """Create an invalid tool call.

langchain_core/messages/utils.py

@@ -40,6 +40,12 @@ from langchain_core.messages.human import HumanMessage, HumanMessageChunk
 from langchain_core.messages.modifier import RemoveMessage
 from langchain_core.messages.system import SystemMessage, SystemMessageChunk
 from langchain_core.messages.tool import ToolCall, ToolMessage, ToolMessageChunk
+from langchain_core.v1.messages import AIMessage as AIMessageV1
+from langchain_core.v1.messages import AIMessageChunk as AIMessageChunkV1
+from langchain_core.v1.messages import HumanMessage as HumanMessageV1
+from langchain_core.v1.messages import MessageV1, MessageV1Types
+from langchain_core.v1.messages import SystemMessage as SystemMessageV1
+from langchain_core.v1.messages import ToolMessage as ToolMessageV1
 
 if TYPE_CHECKING:
     from langchain_text_splitters import TextSplitter
@@ -111,6 +117,7 @@ def get_buffer_string(
             ]
             get_buffer_string(messages)
             # -> "Human: Hi, how are you?\nAI: Good, how are you?"
+
     """
     string_messages = []
     for m in messages:
@@ -202,7 +209,7 @@ def message_chunk_to_message(chunk: BaseMessageChunk) -> BaseMessage:
 
 
 MessageLikeRepresentation = Union[
-    BaseMessage, list[str], tuple[str, str], str, dict[str, Any]
+    BaseMessage, list[str], tuple[str, str], str, dict[str, Any], MessageV1
 ]
 
 
@@ -212,7 +219,7 @@ def _create_message_from_message_type(
     name: Optional[str] = None,
     tool_call_id: Optional[str] = None,
     tool_calls: Optional[list[dict[str, Any]]] = None,
-    id: Optional[str] = None,  # noqa: A002
+    id: Optional[str] = None,
     **additional_kwargs: Any,
 ) -> BaseMessage:
     """Create a message from a message type and content string.
@@ -293,6 +300,130 @@ def _create_message_from_message_type(
     return message
 
 
+def _create_message_from_message_type_v1(
+    message_type: str,
+    content: str,
+    name: Optional[str] = None,
+    tool_call_id: Optional[str] = None,
+    tool_calls: Optional[list[dict[str, Any]]] = None,
+    id: Optional[str] = None,
+    **kwargs: Any,
+) -> MessageV1:
+    """Create a message from a message type and content string.
+
+    Args:
+        message_type: (str) the type of the message (e.g., "human", "ai", etc.).
+        content: (str) the content string.
+        name: (str) the name of the message. Default is None.
+        tool_call_id: (str) the tool call id. Default is None.
+        tool_calls: (list[dict[str, Any]]) the tool calls. Default is None.
+        id: (str) the id of the message. Default is None.
+        kwargs: (dict[str, Any]) additional keyword arguments.
+
+    Returns:
+        a message of the appropriate type.
+
+    Raises:
+        ValueError: if the message type is not one of "human", "user", "ai",
+            "assistant", "tool", "system", or "developer".
+    """
+    if name is not None:
+        kwargs["name"] = name
+    if tool_call_id is not None:
+        kwargs["tool_call_id"] = tool_call_id
+    if kwargs and (response_metadata := kwargs.pop("response_metadata", None)):
+        kwargs["response_metadata"] = response_metadata
+    if id is not None:
+        kwargs["id"] = id
+    if tool_calls is not None:
+        kwargs["tool_calls"] = []
+        for tool_call in tool_calls:
+            # Convert OpenAI-format tool call to LangChain format.
+            if "function" in tool_call:
+                args = tool_call["function"]["arguments"]
+                if isinstance(args, str):
+                    args = json.loads(args, strict=False)
+                kwargs["tool_calls"].append(
+                    {
+                        "name": tool_call["function"]["name"],
+                        "args": args,
+                        "id": tool_call["id"],
+                        "type": "tool_call",
+                    }
+                )
+            else:
+                kwargs["tool_calls"].append(tool_call)
+    if message_type in {"human", "user"}:
+        message: MessageV1 = HumanMessageV1(content=content, **kwargs)
+    elif message_type in {"ai", "assistant"}:
+        message = AIMessageV1(content=content, **kwargs)
+    elif message_type in {"system", "developer"}:
+        if message_type == "developer":
+            kwargs["custom_role"] = "developer"
+        message = SystemMessageV1(content=content, **kwargs)
+    elif message_type == "tool":
+        artifact = kwargs.pop("artifact", None)
+        message = ToolMessageV1(content=content, artifact=artifact, **kwargs)
+    else:
+        msg = (
+            f"Unexpected message type: '{message_type}'. Use one of 'human',"
+            f" 'user', 'ai', 'assistant', 'function', 'tool', 'system', or 'developer'."
+        )
+        msg = create_message(message=msg, error_code=ErrorCode.MESSAGE_COERCION_FAILURE)
+        raise ValueError(msg)
+    return message
+
+
+def convert_from_v1_message(message: MessageV1) -> BaseMessage:
+    """Compatibility layer to convert v1 messages to current messages.
+
+    Args:
+        message: MessageV1 instance to convert.
+
+    Returns:
+        BaseMessage: Converted message instance.
+    """
+    content = cast("Union[str, list[str | dict]]", message.content)
+    if isinstance(message, AIMessageV1):
+        return AIMessage(
+            content=content,
+            id=message.id,
+            name=message.name,
+            tool_calls=message.tool_calls,
+            response_metadata=cast("dict", message.response_metadata),
+        )
+    if isinstance(message, AIMessageChunkV1):
+        return AIMessageChunk(
+            content=content,
+            id=message.id,
+            name=message.name,
+            tool_call_chunks=message.tool_call_chunks,
+            response_metadata=cast("dict", message.response_metadata),
+        )
+    if isinstance(message, HumanMessageV1):
+        return HumanMessage(
+            content=content,
+            id=message.id,
+            name=message.name,
+        )
+    if isinstance(message, SystemMessageV1):
+        return SystemMessage(
+            content=content,
+            id=message.id,
+        )
+    if isinstance(message, ToolMessageV1):
+        return ToolMessage(
+            content=content,
+            id=message.id,
+            tool_call_id=message.tool_call_id,
+            artifact=message.artifact,
+            name=message.name,
+            status=message.status,
+        )
+    message = f"Unsupported message type: {type(message)}"
+    raise NotImplementedError(message)
+
+
 def _convert_to_message(message: MessageLikeRepresentation) -> BaseMessage:
     """Instantiate a message from a variety of message formats.
 
@@ -340,6 +471,66 @@ def _convert_to_message(message: MessageLikeRepresentation) -> BaseMessage:
         message_ = _create_message_from_message_type(
             msg_type, msg_content, **msg_kwargs
         )
+    elif isinstance(message, MessageV1Types):
+        message_ = convert_from_v1_message(message)
+    else:
+        msg = f"Unsupported message type: {type(message)}"
+        msg = create_message(message=msg, error_code=ErrorCode.MESSAGE_COERCION_FAILURE)
+        raise NotImplementedError(msg)
+
+    return message_
+
+
+def _convert_to_message_v1(message: MessageLikeRepresentation) -> MessageV1:
+    """Instantiate a message from a variety of message formats.
+
+    The message format can be one of the following:
+
+    - BaseMessagePromptTemplate
+    - BaseMessage
+    - 2-tuple of (role string, template); e.g., ("human", "{user_input}")
+    - dict: a message dict with role and content keys
+    - string: shorthand for ("human", template); e.g., "{user_input}"
+
+    Args:
+        message: a representation of a message in one of the supported formats.
+
+    Returns:
+        an instance of a message or a message template.
+
+    Raises:
+        NotImplementedError: if the message type is not supported.
+        ValueError: if the message dict does not contain the required keys.
+    """
+    if isinstance(message, MessageV1Types):
+        if isinstance(message, AIMessageChunkV1):
+            message_: MessageV1 = message.to_message()
+        else:
+            message_ = message
+    elif isinstance(message, str):
+        message_ = _create_message_from_message_type_v1("human", message)
+    elif isinstance(message, Sequence) and len(message) == 2:
+        # mypy doesn't realise this can't be a string given the previous branch
+        message_type_str, template = message  # type: ignore[misc]
+        message_ = _create_message_from_message_type_v1(message_type_str, template)
+    elif isinstance(message, dict):
+        msg_kwargs = message.copy()
+        try:
+            try:
+                msg_type = msg_kwargs.pop("role")
+            except KeyError:
+                msg_type = msg_kwargs.pop("type")
+            # None msg content is not allowed
+            msg_content = msg_kwargs.pop("content") or ""
+        except KeyError as e:
+            msg = f"Message dict must contain 'role' and 'content' keys, got {message}"
+            msg = create_message(
+                message=msg, error_code=ErrorCode.MESSAGE_COERCION_FAILURE
+            )
+            raise ValueError(msg) from e
+        message_ = _create_message_from_message_type_v1(
+            msg_type, msg_content, **msg_kwargs
+        )
     else:
         msg = f"Unsupported message type: {type(message)}"
         msg = create_message(message=msg, error_code=ErrorCode.MESSAGE_COERCION_FAILURE)
@@ -367,6 +558,25 @@ def convert_to_messages(
     return [_convert_to_message(m) for m in messages]
 
 
+def convert_to_messages_v1(
+    messages: Union[Iterable[MessageLikeRepresentation], PromptValue],
+) -> list[MessageV1]:
+    """Convert a sequence of messages to a list of messages.
+
+    Args:
+        messages: Sequence of messages to convert.
+
+    Returns:
+        list of messages (BaseMessages).
+    """
+    # Import here to avoid circular imports
+    from langchain_core.prompt_values import PromptValue
+
+    if isinstance(messages, PromptValue):
+        return messages.to_messages(message_version="v1")
+    return [_convert_to_message_v1(m) for m in messages]
+
+
 def _runnable_support(func: Callable) -> Callable:
     @overload
     def wrapped(
@@ -463,6 +673,7 @@ def filter_messages(
                 SystemMessage("you're a good assistant."),
                 HumanMessage("what's your name", id="foo", name="example_user"),
             ]
+
     """  # noqa: E501
     messages = convert_to_messages(messages)
     filtered: list[BaseMessage] = []
@@ -869,6 +1080,7 @@ def trim_messages(
                     HumanMessage("This is a 4 token text. The full message is 10 tokens.", id="first"),
                     AIMessage( [{"type": "text", "text": "This is the FIRST 4 token block."}], id="second"),
                 ]
+
     """  # noqa: E501
     # Validate arguments
     if start_on and strategy == "first":
@@ -1004,10 +1216,11 @@ def convert_to_openai_messages(
 
     oai_messages: list = []
 
-    if is_single := isinstance(messages, (BaseMessage, dict, str)):
+    if is_single := isinstance(messages, (BaseMessage, dict, str, MessageV1Types)):
         messages = [messages]
 
-    messages = convert_to_messages(messages)
+    # TODO: resolve type ignore here
+    messages = convert_to_messages(messages)  # type: ignore[arg-type]
 
     for i, message in enumerate(messages):
         oai_msg: dict = {"role": _get_message_openai_role(message)}
@@ -1176,7 +1389,9 @@ def convert_to_openai_messages(
                                 "id": block["id"],
                                 "function": {
                                     "name": block["name"],
-                                    "arguments": json.dumps(block["input"]),
+                                    "arguments": json.dumps(
+                                        block["input"], ensure_ascii=False
+                                    ),
                                 },
                             }
                         )
@@ -1550,7 +1765,7 @@ def _convert_to_openai_tool_calls(tool_calls: list[ToolCall]) -> list[dict]:
             "id": tool_call["id"],
             "function": {
                 "name": tool_call["name"],
-                "arguments": json.dumps(tool_call["args"]),
+                "arguments": json.dumps(tool_call["args"], ensure_ascii=False),
             },
         }
         for tool_call in tool_calls

langchain_core/output_parsers/base.py

@@ -11,6 +11,7 @@ from typing import (
     Optional,
     TypeVar,
     Union,
+    cast,
 )
 
 from typing_extensions import override
@@ -20,19 +21,22 @@ from langchain_core.messages import AnyMessage, BaseMessage
 from langchain_core.outputs import ChatGeneration, Generation
 from langchain_core.runnables import Runnable, RunnableConfig, RunnableSerializable
 from langchain_core.runnables.config import run_in_executor
+from langchain_core.v1.messages import AIMessage, MessageV1, MessageV1Types
 
 if TYPE_CHECKING:
     from langchain_core.prompt_values import PromptValue
 
 T = TypeVar("T")
-OutputParserLike = Runnable[LanguageModelOutput, T]
+OutputParserLike = Runnable[Union[LanguageModelOutput, AIMessage], T]
 
 
 class BaseLLMOutputParser(ABC, Generic[T]):
     """Abstract base class for parsing the outputs of a model."""
 
     @abstractmethod
-    def parse_result(self, result: list[Generation], *, partial: bool = False) -> T:
+    def parse_result(
+        self, result: Union[list[Generation], AIMessage], *, partial: bool = False
+    ) -> T:
         """Parse a list of candidate model Generations into a specific format.
 
         Args:
@@ -46,7 +50,7 @@ class BaseLLMOutputParser(ABC, Generic[T]):
         """
 
     async def aparse_result(
-        self, result: list[Generation], *, partial: bool = False
+        self, result: Union[list[Generation], AIMessage], *, partial: bool = False
     ) -> T:
         """Async parse a list of candidate model Generations into a specific format.
 
@@ -71,7 +75,7 @@ class BaseGenerationOutputParser(
     @override
     def InputType(self) -> Any:
         """Return the input type for the parser."""
-        return Union[str, AnyMessage]
+        return Union[str, AnyMessage, MessageV1]
 
     @property
     @override
@@ -84,7 +88,7 @@ class BaseGenerationOutputParser(
     @override
     def invoke(
         self,
-        input: Union[str, BaseMessage],
+        input: Union[str, BaseMessage, MessageV1],
         config: Optional[RunnableConfig] = None,
         **kwargs: Any,
     ) -> T:
@@ -97,9 +101,16 @@ class BaseGenerationOutputParser(
                 config,
                 run_type="parser",
             )
+        if isinstance(input, MessageV1Types):
+            return self._call_with_config(
+                lambda inner_input: self.parse_result(inner_input),
+                input,
+                config,
+                run_type="parser",
+            )
         return self._call_with_config(
             lambda inner_input: self.parse_result([Generation(text=inner_input)]),
-            input,
+            cast("str", input),
             config,
             run_type="parser",
         )
@@ -120,6 +131,13 @@ class BaseGenerationOutputParser(
                 config,
                 run_type="parser",
             )
+        if isinstance(input, MessageV1Types):
+            return await self._acall_with_config(
+                lambda inner_input: self.aparse_result(inner_input),
+                input,
+                config,
+                run_type="parser",
+            )
         return await self._acall_with_config(
             lambda inner_input: self.aparse_result([Generation(text=inner_input)]),
             input,
@@ -129,7 +147,7 @@ class BaseGenerationOutputParser(
 
 
 class BaseOutputParser(
-    BaseLLMOutputParser, RunnableSerializable[LanguageModelOutput, T]
+    BaseLLMOutputParser, RunnableSerializable[Union[LanguageModelOutput, AIMessage], T]
 ):
     """Base class to parse the output of an LLM call.
 
@@ -155,13 +173,14 @@ class BaseOutputParser(
                 @property
                 def _type(self) -> str:
                     return "boolean_output_parser"
+
     """  # noqa: E501
 
     @property
     @override
     def InputType(self) -> Any:
         """Return the input type for the parser."""
-        return Union[str, AnyMessage]
+        return Union[str, AnyMessage, MessageV1]
 
     @property
     @override
@@ -188,7 +207,7 @@ class BaseOutputParser(
     @override
     def invoke(
         self,
-        input: Union[str, BaseMessage],
+        input: Union[str, BaseMessage, MessageV1],
         config: Optional[RunnableConfig] = None,
         **kwargs: Any,
     ) -> T:
@@ -201,9 +220,16 @@ class BaseOutputParser(
                 config,
                 run_type="parser",
             )
+        if isinstance(input, MessageV1Types):
+            return self._call_with_config(
+                lambda inner_input: self.parse_result(inner_input),
+                input,
+                config,
+                run_type="parser",
+            )
         return self._call_with_config(
             lambda inner_input: self.parse_result([Generation(text=inner_input)]),
-            input,
+            cast("str", input),
             config,
             run_type="parser",
         )
@@ -211,7 +237,7 @@ class BaseOutputParser(
     @override
     async def ainvoke(
         self,
-        input: Union[str, BaseMessage],
+        input: Union[str, BaseMessage, MessageV1],
         config: Optional[RunnableConfig] = None,
         **kwargs: Optional[Any],
     ) -> T:
@@ -224,15 +250,24 @@ class BaseOutputParser(
                 config,
                 run_type="parser",
             )
+        if isinstance(input, MessageV1Types):
+            return await self._acall_with_config(
+                lambda inner_input: self.aparse_result(inner_input),
+                input,
+                config,
+                run_type="parser",
+            )
         return await self._acall_with_config(
             lambda inner_input: self.aparse_result([Generation(text=inner_input)]),
-            input,
+            cast("str", input),
             config,
             run_type="parser",
         )
 
     @override
-    def parse_result(self, result: list[Generation], *, partial: bool = False) -> T:
+    def parse_result(
+        self, result: Union[list[Generation], AIMessage], *, partial: bool = False
+    ) -> T:
         """Parse a list of candidate model Generations into a specific format.
 
         The return value is parsed from only the first Generation in the result, which
@@ -247,6 +282,8 @@ class BaseOutputParser(
         Returns:
             Structured output.
         """
+        if isinstance(result, AIMessage):
+            return self.parse(result.text)
         return self.parse(result[0].text)
 
     @abstractmethod
@@ -261,7 +298,7 @@ class BaseOutputParser(
         """
 
     async def aparse_result(
-        self, result: list[Generation], *, partial: bool = False
+        self, result: Union[list[Generation], AIMessage], *, partial: bool = False
     ) -> T:
         """Async parse a list of candidate model Generations into a specific format.
 

langchain_core/output_parsers/json.py

@@ -21,6 +21,7 @@ from langchain_core.utils.json import (
     parse_json_markdown,
     parse_partial_json,
 )
+from langchain_core.v1.messages import AIMessage
 
 # Union type needs to be last assignment to PydanticBaseModel to make mypy happy.
 PydanticBaseModel = Union[BaseModel, pydantic.BaseModel]
@@ -53,7 +54,9 @@ class JsonOutputParser(BaseCumulativeTransformOutputParser[Any]):
             return pydantic_object.schema()
         return None
 
-    def parse_result(self, result: list[Generation], *, partial: bool = False) -> Any:
+    def parse_result(
+        self, result: Union[list[Generation], AIMessage], *, partial: bool = False
+    ) -> Any:
         """Parse the result of an LLM call to a JSON object.
 
         Args:
@@ -70,7 +73,7 @@ class JsonOutputParser(BaseCumulativeTransformOutputParser[Any]):
         Raises:
             OutputParserException: If the output is not valid JSON.
         """
-        text = result[0].text
+        text = result.text if isinstance(result, AIMessage) else result[0].text
         text = text.strip()
         if partial:
             try: