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

langchain_core/language_models/_utils.py

@@ -1,8 +1,10 @@
+import copy
 import re
 from collections.abc import Sequence
 from typing import Optional
 
 from langchain_core.messages import BaseMessage
+from langchain_core.v1.messages import MessageV1
 
 
 def _is_openai_data_block(block: dict) -> bool:
@@ -51,6 +53,7 @@ def _parse_data_uri(uri: str) -> Optional[dict]:
                 "mime_type": "image/jpeg",
                 "data": "/9j/4AAQSkZJRg...",
             }
+
     """
     regex = r"^data:(?P<mime_type>[^;]+);base64,(?P<data>.+)$"
     match = re.match(regex, uri)
@@ -137,3 +140,37 @@ def _normalize_messages(messages: Sequence[BaseMessage]) -> list[BaseMessage]:
         formatted_messages.append(formatted_message)
 
     return formatted_messages
+
+
+def _normalize_messages_v1(messages: Sequence[MessageV1]) -> list[MessageV1]:
+    """Extend support for message formats.
+
+    Chat models implement support for images in OpenAI Chat Completions format, as well
+    as other multimodal data as standard data blocks. This function extends support to
+    audio and file data in OpenAI Chat Completions format by converting them to standard
+    data blocks.
+    """
+    formatted_messages = []
+    for message in messages:
+        formatted_message = message
+        if isinstance(message.content, list):
+            for idx, block in enumerate(message.content):
+                if (
+                    isinstance(block, dict)
+                    # Subset to (PDF) files and audio, as most relevant chat models
+                    # support images in OAI format (and some may not yet support the
+                    # standard data block format)
+                    and block.get("type") in {"file", "input_audio"}
+                    and _is_openai_data_block(block)  # type: ignore[arg-type]
+                ):
+                    if formatted_message is message:
+                        formatted_message = copy.copy(message)
+                        # Also shallow-copy content
+                        formatted_message.content = list(formatted_message.content)
+
+                    formatted_message.content[idx] = (  # type: ignore[call-overload]
+                        _convert_openai_format_to_data_block(block)  # type: ignore[arg-type]
+                    )
+        formatted_messages.append(formatted_message)
+
+    return formatted_messages

langchain_core/language_models/base.py

@@ -31,6 +31,7 @@ from langchain_core.messages import (
 from langchain_core.prompt_values import PromptValue
 from langchain_core.runnables import Runnable, RunnableSerializable
 from langchain_core.utils import get_pydantic_field_names
+from langchain_core.v1.messages import AIMessage as AIMessageV1
 
 if TYPE_CHECKING:
     from langchain_core.outputs import LLMResult
@@ -85,7 +86,9 @@ def _get_token_ids_default_method(text: str) -> list[int]:
 LanguageModelInput = Union[PromptValue, str, Sequence[MessageLikeRepresentation]]
 LanguageModelOutput = Union[BaseMessage, str]
 LanguageModelLike = Runnable[LanguageModelInput, LanguageModelOutput]
-LanguageModelOutputVar = TypeVar("LanguageModelOutputVar", BaseMessage, str)
+LanguageModelOutputVar = TypeVar(
+    "LanguageModelOutputVar", BaseMessage, str, AIMessageV1
+)
 
 
 def _get_verbosity() -> bool:

langchain_core/language_models/chat_models.py

@@ -11,22 +11,9 @@ from abc import ABC, abstractmethod
 from collections.abc import AsyncIterator, Iterator, Sequence
 from functools import cached_property
 from operator import itemgetter
-from typing import (
-    TYPE_CHECKING,
-    Any,
-    Callable,
-    Literal,
-    Optional,
-    Union,
-    cast,
-)
+from typing import TYPE_CHECKING, Any, Callable, Literal, Optional, Union, cast
 
-from pydantic import (
-    BaseModel,
-    ConfigDict,
-    Field,
-    model_validator,
-)
+from pydantic import BaseModel, ConfigDict, Field, model_validator
 from typing_extensions import override
 
 from langchain_core._api import deprecated
@@ -63,6 +50,7 @@ from langchain_core.outputs import (
     ChatGeneration,
     ChatGenerationChunk,
     ChatResult,
+    Generation,
     LLMResult,
     RunInfo,
 )
@@ -653,6 +641,34 @@ class BaseChatModel(BaseLanguageModel[BaseMessage], ABC):
     def _combine_llm_outputs(self, llm_outputs: list[Optional[dict]]) -> dict:  # noqa: ARG002
         return {}
 
+    def _convert_cached_generations(self, cache_val: list) -> list[ChatGeneration]:
+        """Convert cached Generation objects to ChatGeneration objects.
+
+        Handle case where cache contains Generation objects instead of
+        ChatGeneration objects. This can happen due to serialization/deserialization
+        issues or legacy cache data (see #22389).
+
+        Args:
+            cache_val: List of cached generation objects.
+
+        Returns:
+            List of ChatGeneration objects.
+        """
+        converted_generations = []
+        for gen in cache_val:
+            if isinstance(gen, Generation) and not isinstance(gen, ChatGeneration):
+                # Convert Generation to ChatGeneration by creating AIMessage
+                # from the text content
+                chat_gen = ChatGeneration(
+                    message=AIMessage(content=gen.text),
+                    generation_info=gen.generation_info,
+                )
+                converted_generations.append(chat_gen)
+            else:
+                # Already a ChatGeneration or other expected type
+                converted_generations.append(gen)
+        return converted_generations
+
     def _get_invocation_params(
         self,
         stop: Optional[list[str]] = None,
@@ -1010,7 +1026,8 @@ class BaseChatModel(BaseLanguageModel[BaseMessage], ABC):
                 prompt = dumps(messages)
                 cache_val = llm_cache.lookup(prompt, llm_string)
                 if isinstance(cache_val, list):
-                    return ChatResult(generations=cache_val)
+                    converted_generations = self._convert_cached_generations(cache_val)
+                    return ChatResult(generations=converted_generations)
             elif self.cache is None:
                 pass
             else:
@@ -1082,7 +1099,8 @@ class BaseChatModel(BaseLanguageModel[BaseMessage], ABC):
                 prompt = dumps(messages)
                 cache_val = await llm_cache.alookup(prompt, llm_string)
                 if isinstance(cache_val, list):
-                    return ChatResult(generations=cache_val)
+                    converted_generations = self._convert_cached_generations(cache_val)
+                    return ChatResult(generations=converted_generations)
             elif self.cache is None:
                 pass
             else:
@@ -1367,12 +1385,13 @@ class BaseChatModel(BaseLanguageModel[BaseMessage], ABC):
         """Model wrapper that returns outputs formatted to match the given schema.
 
         Args:
-            schema:
-                The output schema. Can be passed in as:
-                    - an OpenAI function/tool schema,
-                    - a JSON Schema,
-                    - a TypedDict class,
-                    - or a Pydantic class.
+            schema: The output schema. Can be passed in as:
+
+                - an OpenAI function/tool schema,
+                - a JSON Schema,
+                - a TypedDict class,
+                - or a Pydantic class.
+
                 If ``schema`` is a Pydantic class then the model output will be a
                 Pydantic instance of that class, and the model-generated fields will be
                 validated by the Pydantic class. Otherwise the model output will be a
@@ -1386,7 +1405,7 @@ class BaseChatModel(BaseLanguageModel[BaseMessage], ABC):
                 then both the raw model response (a BaseMessage) and the parsed model
                 response will be returned. If an error occurs during output parsing it
                 will be caught and returned as well. The final output is always a dict
-                with keys "raw", "parsed", and "parsing_error".
+                with keys ``'raw'``, ``'parsed'``, and ``'parsing_error'``.
 
         Returns:
             A Runnable that takes same inputs as a :class:`langchain_core.language_models.chat.BaseChatModel`.
@@ -1397,9 +1416,10 @@ class BaseChatModel(BaseLanguageModel[BaseMessage], ABC):
             Otherwise, if ``include_raw`` is False then Runnable outputs a dict.
 
             If ``include_raw`` is True, then Runnable outputs a dict with keys:
-                - ``"raw"``: BaseMessage
-                - ``"parsed"``: None if there was a parsing error, otherwise the type depends on the ``schema`` as described above.
-                - ``"parsing_error"``: Optional[BaseException]
+
+            - ``'raw'``: BaseMessage
+            - ``'parsed'``: None if there was a parsing error, otherwise the type depends on the ``schema`` as described above.
+            - ``'parsing_error'``: Optional[BaseException]
 
         Example: Pydantic schema (include_raw=False):
             .. code-block:: python
@@ -1465,6 +1485,7 @@ class BaseChatModel(BaseLanguageModel[BaseMessage], ABC):
         .. versionchanged:: 0.2.26
 
                 Added support for TypedDict class.
+
         """  # noqa: E501
         _ = kwargs.pop("method", None)
         _ = kwargs.pop("strict", None)

langchain_core/language_models/fake_chat_models.py

@@ -3,7 +3,7 @@
 import asyncio
 import re
 import time
-from collections.abc import AsyncIterator, Iterator
+from collections.abc import AsyncIterator, Iterable, Iterator
 from typing import Any, Optional, Union, cast
 
 from typing_extensions import override
@@ -16,6 +16,10 @@ from langchain_core.language_models.chat_models import BaseChatModel, SimpleChat
 from langchain_core.messages import AIMessage, AIMessageChunk, BaseMessage
 from langchain_core.outputs import ChatGeneration, ChatGenerationChunk, ChatResult
 from langchain_core.runnables import RunnableConfig
+from langchain_core.v1.chat_models import BaseChatModel as BaseChatModelV1
+from langchain_core.v1.messages import AIMessage as AIMessageV1
+from langchain_core.v1.messages import AIMessageChunk as AIMessageChunkV1
+from langchain_core.v1.messages import MessageV1
 
 
 class FakeMessagesListChatModel(BaseChatModel):
@@ -367,3 +371,69 @@ class ParrotFakeChatModel(BaseChatModel):
     @property
     def _llm_type(self) -> str:
         return "parrot-fake-chat-model"
+
+
+class GenericFakeChatModelV1(BaseChatModelV1):
+    """Generic fake chat model that can be used to test the chat model interface."""
+
+    messages: Optional[Iterator[Union[AIMessageV1, str]]] = None
+    message_chunks: Optional[Iterable[Union[AIMessageChunkV1, str]]] = None
+
+    @override
+    def _invoke(
+        self,
+        messages: list[MessageV1],
+        run_manager: Optional[CallbackManagerForLLMRun] = None,
+        **kwargs: Any,
+    ) -> AIMessageV1:
+        """Top Level call."""
+        if self.messages is None:
+            error_msg = "Messages iterator is not set."
+            raise ValueError(error_msg)
+        message = next(self.messages)
+        return AIMessageV1(content=message) if isinstance(message, str) else message
+
+    @override
+    def _stream(
+        self,
+        messages: list[MessageV1],
+        run_manager: Optional[CallbackManagerForLLMRun] = None,
+        **kwargs: Any,
+    ) -> Iterator[AIMessageChunkV1]:
+        """Top Level call."""
+        if self.message_chunks is None:
+            error_msg = "Message chunks iterator is not set."
+            raise ValueError(error_msg)
+        for chunk in self.message_chunks:
+            if isinstance(chunk, str):
+                yield AIMessageChunkV1(chunk)
+            else:
+                yield chunk
+
+    @property
+    def _llm_type(self) -> str:
+        return "generic-fake-chat-model"
+
+
+class ParrotFakeChatModelV1(BaseChatModelV1):
+    """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
+    def _invoke(
+        self,
+        messages: list[MessageV1],
+        stop: Optional[list[str]] = None,
+        run_manager: Optional[CallbackManagerForLLMRun] = None,
+        **kwargs: Any,
+    ) -> AIMessageV1:
+        """Top Level call."""
+        if isinstance(messages[-1], AIMessageV1):
+            return messages[-1]
+        return AIMessageV1(content=messages[-1].content)
+
+    @property
+    def _llm_type(self) -> str:
+        return "parrot-fake-chat-model"

langchain_core/language_models/llms.py

@@ -1418,6 +1418,7 @@ class BaseLLM(BaseLanguageModel[str], ABC):
             .. code-block:: python
 
                 llm.save(file_path="path/llm.yaml")
+
         """
         # Convert file to Path object.
         save_path = Path(file_path)

langchain_core/memory.py

@@ -53,6 +53,7 @@ class BaseMemory(Serializable, ABC):
 
                 def clear(self) -> None:
                     pass
+
     """  # noqa: E501
 
     model_config = ConfigDict(

langchain_core/messages/__init__.py

@@ -33,6 +33,24 @@ if TYPE_CHECKING:
     )
     from langchain_core.messages.chat import ChatMessage, ChatMessageChunk
     from langchain_core.messages.content_blocks import (
+        Annotation,
+        AudioContentBlock,
+        Citation,
+        CodeInterpreterCall,
+        CodeInterpreterOutput,
+        CodeInterpreterResult,
+        ContentBlock,
+        DataContentBlock,
+        FileContentBlock,
+        ImageContentBlock,
+        NonStandardAnnotation,
+        NonStandardContentBlock,
+        PlainTextContentBlock,
+        ReasoningContentBlock,
+        TextContentBlock,
+        VideoContentBlock,
+        WebSearchCall,
+        WebSearchResult,
         convert_to_openai_data_block,
         convert_to_openai_image_block,
         is_data_content_block,
@@ -65,24 +83,42 @@ if TYPE_CHECKING:
 __all__ = (
     "AIMessage",
     "AIMessageChunk",
+    "Annotation",
     "AnyMessage",
+    "AudioContentBlock",
     "BaseMessage",
     "BaseMessageChunk",
     "ChatMessage",
     "ChatMessageChunk",
+    "Citation",
+    "CodeInterpreterCall",
+    "CodeInterpreterOutput",
+    "CodeInterpreterResult",
+    "ContentBlock",
+    "DataContentBlock",
+    "FileContentBlock",
     "FunctionMessage",
     "FunctionMessageChunk",
     "HumanMessage",
     "HumanMessageChunk",
+    "ImageContentBlock",
     "InvalidToolCall",
     "MessageLikeRepresentation",
+    "NonStandardAnnotation",
+    "NonStandardContentBlock",
+    "PlainTextContentBlock",
+    "ReasoningContentBlock",
     "RemoveMessage",
     "SystemMessage",
     "SystemMessageChunk",
+    "TextContentBlock",
     "ToolCall",
     "ToolCallChunk",
     "ToolMessage",
     "ToolMessageChunk",
+    "VideoContentBlock",
+    "WebSearchCall",
+    "WebSearchResult",
     "_message_from_dict",
     "convert_to_messages",
     "convert_to_openai_data_block",
@@ -103,25 +139,43 @@ __all__ = (
 _dynamic_imports = {
     "AIMessage": "ai",
     "AIMessageChunk": "ai",
+    "Annotation": "content_blocks",
+    "AudioContentBlock": "content_blocks",
     "BaseMessage": "base",
     "BaseMessageChunk": "base",
     "merge_content": "base",
     "message_to_dict": "base",
     "messages_to_dict": "base",
+    "Citation": "content_blocks",
+    "ContentBlock": "content_blocks",
     "ChatMessage": "chat",
     "ChatMessageChunk": "chat",
+    "CodeInterpreterCall": "content_blocks",
+    "CodeInterpreterOutput": "content_blocks",
+    "CodeInterpreterResult": "content_blocks",
+    "DataContentBlock": "content_blocks",
+    "FileContentBlock": "content_blocks",
     "FunctionMessage": "function",
     "FunctionMessageChunk": "function",
     "HumanMessage": "human",
     "HumanMessageChunk": "human",
+    "NonStandardAnnotation": "content_blocks",
+    "NonStandardContentBlock": "content_blocks",
+    "PlainTextContentBlock": "content_blocks",
+    "ReasoningContentBlock": "content_blocks",
     "RemoveMessage": "modifier",
     "SystemMessage": "system",
     "SystemMessageChunk": "system",
+    "WebSearchCall": "content_blocks",
+    "WebSearchResult": "content_blocks",
+    "ImageContentBlock": "content_blocks",
     "InvalidToolCall": "tool",
+    "TextContentBlock": "content_blocks",
     "ToolCall": "tool",
     "ToolCallChunk": "tool",
     "ToolMessage": "tool",
     "ToolMessageChunk": "tool",
+    "VideoContentBlock": "content_blocks",
     "AnyMessage": "utils",
     "MessageLikeRepresentation": "utils",
     "_message_from_dict": "utils",

langchain_core/messages/ai.py

@@ -8,11 +8,7 @@ from typing import Any, Literal, Optional, Union, cast
 from pydantic import model_validator
 from typing_extensions import NotRequired, Self, TypedDict, override
 
-from langchain_core.messages.base import (
-    BaseMessage,
-    BaseMessageChunk,
-    merge_content,
-)
+from langchain_core.messages.base import BaseMessage, BaseMessageChunk, merge_content
 from langchain_core.messages.tool import (
     InvalidToolCall,
     ToolCall,
@@ -20,15 +16,9 @@ from langchain_core.messages.tool import (
     default_tool_chunk_parser,
     default_tool_parser,
 )
-from langchain_core.messages.tool import (
-    invalid_tool_call as create_invalid_tool_call,
-)
-from langchain_core.messages.tool import (
-    tool_call as create_tool_call,
-)
-from langchain_core.messages.tool import (
-    tool_call_chunk as create_tool_call_chunk,
-)
+from langchain_core.messages.tool import invalid_tool_call as create_invalid_tool_call
+from langchain_core.messages.tool import tool_call as create_tool_call
+from langchain_core.messages.tool import tool_call_chunk as create_tool_call_chunk
 from langchain_core.utils._merge import merge_dicts, merge_lists
 from langchain_core.utils.json import parse_partial_json
 from langchain_core.utils.usage import _dict_int_op
@@ -37,6 +27,16 @@ logger = logging.getLogger(__name__)
 
 
 _LC_ID_PREFIX = "run-"
+"""Internal tracing/callback system identifier.
+
+Used for:
+- Tracing. Every LangChain operation (LLM call, chain execution, tool use, etc.)
+  gets a unique run_id (UUID)
+- Enables tracking parent-child relationships between operations
+"""
+
+_LC_AUTO_PREFIX = "lc_"
+"""LangChain auto-generated ID prefix for messages and content blocks."""
 
 
 class InputTokenDetails(TypedDict, total=False):
@@ -57,6 +57,7 @@ class InputTokenDetails(TypedDict, total=False):
     .. versionadded:: 0.3.9
 
     May also hold extra provider-specific keys.
+
     """
 
     audio: int
@@ -89,6 +90,7 @@ class OutputTokenDetails(TypedDict, total=False):
             }
 
     .. versionadded:: 0.3.9
+
     """
 
     audio: int
@@ -128,6 +130,7 @@ class UsageMetadata(TypedDict):
     .. versionchanged:: 0.3.9
 
         Added ``input_token_details`` and ``output_token_details``.
+
     """
 
     input_tokens: int
@@ -425,17 +428,27 @@ def add_ai_message_chunks(
 
     chunk_id = None
     candidates = [left.id] + [o.id for o in others]
-    # first pass: pick the first non-run-* id
+    # first pass: pick the first provider-assigned id (non-run-* and non-lc_*)
     for id_ in candidates:
-        if id_ and not id_.startswith(_LC_ID_PREFIX):
+        if (
+            id_
+            and not id_.startswith(_LC_ID_PREFIX)
+            and not id_.startswith(_LC_AUTO_PREFIX)
+        ):
             chunk_id = id_
             break
     else:
-        # second pass: no provider-assigned id found, just take the first non-null
+        # second pass: prefer lc_* ids over run-* ids
         for id_ in candidates:
-            if id_:
+            if id_ and id_.startswith(_LC_AUTO_PREFIX):
                 chunk_id = id_
                 break
+        else:
+            # third pass: take any remaining id (run-* ids)
+            for id_ in candidates:
+                if id_:
+                    chunk_id = id_
+                    break
 
     return left.__class__(
         example=left.example,