langchain 0.3.27__py3-none-any.whl → 0.4.0.dev0__py3-none-any.whl

langchain/chains/transform.py

@@ -10,6 +10,7 @@ from langchain_core.callbacks import (
     CallbackManagerForChainRun,
 )
 from pydantic import Field
+from typing_extensions import override
 
 from langchain.chains.base import Chain
 
@@ -25,6 +26,7 @@ class TransformChain(Chain):
             from langchain.chains import TransformChain
             transform_chain = TransformChain(input_variables=["text"],
              output_variables["entities"], transform=func())
+
     """
 
     input_variables: list[str]
@@ -63,6 +65,7 @@ class TransformChain(Chain):
         """
         return self.output_variables
 
+    @override
     def _call(
         self,
         inputs: dict[str, str],
@@ -70,6 +73,7 @@ class TransformChain(Chain):
     ) -> dict[str, str]:
         return self.transform_cb(inputs)
 
+    @override
     async def _acall(
         self,
         inputs: dict[str, Any],

langchain/chat_models/base.py

@@ -3,15 +3,7 @@ from __future__ import annotations
 import warnings
 from collections.abc import AsyncIterator, Iterator, Sequence
 from importlib import util
-from typing import (
-    Any,
-    Callable,
-    Literal,
-    Optional,
-    Union,
-    cast,
-    overload,
-)
+from typing import Any, Callable, Literal, Optional, Union, cast, overload
 
 from langchain_core.language_models import (
     BaseChatModel,
@@ -27,6 +19,7 @@ from langchain_core.runnables import Runnable, RunnableConfig, ensure_config
 from langchain_core.runnables.schema import StreamEvent
 from langchain_core.tools import BaseTool
 from langchain_core.tracers import RunLog, RunLogPatch
+from langchain_core.v1.chat_models import BaseChatModel as BaseChatModelV1
 from pydantic import BaseModel
 from typing_extensions import TypeAlias, override
 
@@ -47,10 +40,23 @@ def init_chat_model(
     model_provider: Optional[str] = None,
     configurable_fields: Literal[None] = None,
     config_prefix: Optional[str] = None,
+    message_version: Literal["v0"] = "v0",
     **kwargs: Any,
 ) -> BaseChatModel: ...
 
 
+@overload
+def init_chat_model(
+    model: str,
+    *,
+    model_provider: Optional[str] = None,
+    configurable_fields: Literal[None] = None,
+    config_prefix: Optional[str] = None,
+    message_version: Literal["v1"] = "v1",
+    **kwargs: Any,
+) -> BaseChatModelV1: ...
+
+
 @overload
 def init_chat_model(
     model: Literal[None] = None,
@@ -58,6 +64,7 @@ def init_chat_model(
     model_provider: Optional[str] = None,
     configurable_fields: Literal[None] = None,
     config_prefix: Optional[str] = None,
+    message_version: Literal["v0", "v1"] = "v0",
     **kwargs: Any,
 ) -> _ConfigurableModel: ...
 
@@ -69,6 +76,7 @@ def init_chat_model(
     model_provider: Optional[str] = None,
     configurable_fields: Union[Literal["any"], list[str], tuple[str, ...]] = ...,
     config_prefix: Optional[str] = None,
+    message_version: Literal["v0", "v1"] = "v0",
     **kwargs: Any,
 ) -> _ConfigurableModel: ...
 
@@ -84,8 +92,9 @@ def init_chat_model(
         Union[Literal["any"], list[str], tuple[str, ...]]
     ] = None,
     config_prefix: Optional[str] = None,
+    message_version: Literal["v0", "v1"] = "v0",
     **kwargs: Any,
-) -> Union[BaseChatModel, _ConfigurableModel]:
+) -> Union[BaseChatModel, BaseChatModelV1, _ConfigurableModel]:
     """Initialize a ChatModel in a single line using the model's name and provider.
 
     .. note::
@@ -136,6 +145,20 @@ def init_chat_model(
             - ``deepseek...``                     -> ``deepseek``
             - ``grok...``                         -> ``xai``
             - ``sonar...``                        -> ``perplexity``
+
+        message_version: The version of the BaseChatModel to return. Either ``"v0"`` for
+            a v0 :class:`~langchain_core.language_models.chat_models.BaseChatModel` or
+            ``"v1"`` for a v1 :class:`~langchain_core.v1.chat_models.BaseChatModel`. The
+            output version determines what type of message objects the model will
+            generate.
+
+            .. note::
+                Currently supported for these providers:
+
+                - ``openai``
+
+            .. versionadded:: 0.4.0
+
         configurable_fields: Which model parameters are configurable:
 
             - None: No configurable fields.
@@ -188,7 +211,7 @@ def init_chat_model(
 
             o3_mini = init_chat_model("openai:o3-mini", temperature=0)
             claude_sonnet = init_chat_model("anthropic:claude-3-5-sonnet-latest", temperature=0)
-            gemini_2_flash = init_chat_model("google_vertexai:gemini-2.0-flash", temperature=0)
+            gemini_2_flash = init_chat_model("google_vertexai:gemini-2.5-flash", temperature=0)
 
             o3_mini.invoke("what's your name")
             claude_sonnet.invoke("what's your name")
@@ -322,8 +345,9 @@ def init_chat_model(
 
     if not configurable_fields:
         return _init_chat_model_helper(
-            cast(str, model),
+            cast("str", model),
             model_provider=model_provider,
+            message_version=message_version,
             **kwargs,
         )
     if model:
@@ -341,14 +365,27 @@ def _init_chat_model_helper(
     model: str,
     *,
     model_provider: Optional[str] = None,
+    message_version: Literal["v0", "v1"] = "v0",
     **kwargs: Any,
-) -> BaseChatModel:
+) -> Union[BaseChatModel, BaseChatModelV1]:
     model, model_provider = _parse_model(model, model_provider)
+    if message_version != "v0" and model_provider not in ("openai",):
+        warnings.warn(
+            f"Model provider {model_provider} does not support "
+            f"message_version={message_version}. Defaulting to v0.",
+            stacklevel=2,
+        )
     if model_provider == "openai":
         _check_pkg("langchain_openai")
-        from langchain_openai import ChatOpenAI
+        if message_version == "v0":
+            from langchain_openai import ChatOpenAI
+
+            return ChatOpenAI(model=model, **kwargs)
+        # v1
+        from langchain_openai.v1 import ChatOpenAI as ChatOpenAIV1
+
+        return ChatOpenAIV1(model=model, **kwargs)
 
-        return ChatOpenAI(model=model, **kwargs)
     if model_provider == "anthropic":
         _check_pkg("langchain_anthropic")
         from langchain_anthropic import ChatAnthropic
@@ -632,7 +669,7 @@ class _ConfigurableModel(Runnable[LanguageModelInput, Any]):
         **kwargs: Any,
     ) -> _ConfigurableModel:
         """Bind config to a Runnable, returning a new Runnable."""
-        config = RunnableConfig(**(config or {}), **cast(RunnableConfig, kwargs))
+        config = RunnableConfig(**(config or {}), **cast("RunnableConfig", kwargs))
         model_params = self._model_params(config)
         remaining_config = {k: v for k, v in config.items() if k != "configurable"}
         remaining_config["configurable"] = {
@@ -781,7 +818,7 @@ class _ConfigurableModel(Runnable[LanguageModelInput, Any]):
         if config is None or isinstance(config, dict) or len(config) <= 1:
             if isinstance(config, list):
                 config = config[0]
-            yield from self._model(cast(RunnableConfig, config)).batch_as_completed(  # type: ignore[call-overload]
+            yield from self._model(cast("RunnableConfig", config)).batch_as_completed(  # type: ignore[call-overload]
                 inputs,
                 config=config,
                 return_exceptions=return_exceptions,
@@ -811,7 +848,7 @@ class _ConfigurableModel(Runnable[LanguageModelInput, Any]):
             if isinstance(config, list):
                 config = config[0]
             async for x in self._model(
-                cast(RunnableConfig, config),
+                cast("RunnableConfig", config),
             ).abatch_as_completed(  # type: ignore[call-overload]
                 inputs,
                 config=config,

langchain/document_loaders/blob_loaders/schema.py

@@ -1,12 +1,9 @@
-from typing import TYPE_CHECKING, Any
+from typing import Any
 
 from langchain_core.document_loaders import Blob, BlobLoader
 
 from langchain._api import create_importer
 
-if TYPE_CHECKING:
-    pass
-
 # Create a way to dynamically look up deprecated imports.
 # Used to consolidate logic for raising deprecation warnings and
 # handling optional imports.

langchain/embeddings/base.py

@@ -47,6 +47,7 @@ def _parse_model_string(model_name: str) -> tuple[str, str]:
     Raises:
         ValueError: If the model string is not in the correct format or
             the provider is unsupported
+
     """
     if ":" not in model_name:
         providers = _SUPPORTED_PROVIDERS
@@ -177,6 +178,7 @@ def init_embeddings(
             )
 
     .. versionadded:: 0.3.9
+
     """
     if not model:
         providers = _SUPPORTED_PROVIDERS.keys()

langchain/embeddings/cache.py

@@ -80,7 +80,7 @@ def _value_serializer(value: Sequence[float]) -> bytes:
 
 def _value_deserializer(serialized_value: bytes) -> list[float]:
     """Deserialize a value."""
-    return cast(list[float], json.loads(serialized_value.decode()))
+    return cast("list[float]", json.loads(serialized_value.decode()))
 
 
 # The warning is global; track emission, so it appears only once.
@@ -192,7 +192,7 @@ class CacheBackedEmbeddings(Embeddings):
                 vectors[index] = updated_vector
 
         return cast(
-            list[list[float]],
+            "list[list[float]]",
             vectors,
         )  # Nones should have been resolved by now
 
@@ -230,7 +230,7 @@ class CacheBackedEmbeddings(Embeddings):
                 vectors[index] = updated_vector
 
         return cast(
-            list[list[float]],
+            "list[list[float]]",
             vectors,
         )  # Nones should have been resolved by now
 

langchain/evaluation/agents/trajectory_eval_chain.py

@@ -140,6 +140,7 @@ class TrajectoryEvalChain(AgentTrajectoryEvaluator, LLMEvalChain):
         )
         print(result["score"])  # noqa: T201
         # 0
+
     """
 
     agent_tools: Optional[list[BaseTool]] = None
@@ -301,7 +302,7 @@ The following is the expected answer. Use this to measure correctness:
             chain_input,
             callbacks=_run_manager.get_child(),
         )
-        return cast(dict, self.output_parser.parse(raw_output))
+        return cast("dict", self.output_parser.parse(raw_output))
 
     async def _acall(
         self,
@@ -326,7 +327,7 @@ The following is the expected answer. Use this to measure correctness:
             chain_input,
             callbacks=_run_manager.get_child(),
         )
-        return cast(dict, self.output_parser.parse(raw_output))
+        return cast("dict", self.output_parser.parse(raw_output))
 
     @override
     def _evaluate_agent_trajectory(

langchain/evaluation/comparison/eval_chain.py

@@ -191,6 +191,7 @@ class PairwiseStringEvalChain(PairwiseStringEvaluator, LLMEvalChain, LLMChain):
     )
 
     @classmethod
+    @override
     def is_lc_serializable(cls) -> bool:
         return False
 

langchain/evaluation/criteria/eval_chain.py

@@ -236,6 +236,7 @@ class CriteriaEvalChain(StringEvaluator, LLMEvalChain, LLMChain):
     output_key: str = "results"  #: :meta private:
 
     @classmethod
+    @override
     def is_lc_serializable(cls) -> bool:
         return False
 
@@ -249,6 +250,7 @@ class CriteriaEvalChain(StringEvaluator, LLMEvalChain, LLMChain):
         return False
 
     @property
+    @override
     def requires_input(self) -> bool:
         return True
 
@@ -520,6 +522,7 @@ class LabeledCriteriaEvalChain(CriteriaEvalChain):
     """Criteria evaluation chain that requires references."""
 
     @classmethod
+    @override
     def is_lc_serializable(cls) -> bool:
         return False
 

langchain/evaluation/embedding_distance/base.py

@@ -14,6 +14,7 @@ from langchain_core.callbacks.manager import (
 from langchain_core.embeddings import Embeddings
 from langchain_core.utils import pre_init
 from pydantic import ConfigDict, Field
+from typing_extensions import override
 
 from langchain.chains.base import Chain
 from langchain.evaluation.schema import PairwiseStringEvaluator, StringEvaluator
@@ -317,6 +318,7 @@ class EmbeddingDistanceEvalChain(_EmbeddingDistanceChainMixin, StringEvaluator):
         return True
 
     @property
+    @override
     def evaluation_name(self) -> str:
         return f"embedding_{self.distance_metric.value}_distance"
 
@@ -329,6 +331,7 @@ class EmbeddingDistanceEvalChain(_EmbeddingDistanceChainMixin, StringEvaluator):
         """
         return ["prediction", "reference"]
 
+    @override
     def _call(
         self,
         inputs: dict[str, Any],
@@ -353,6 +356,7 @@ class EmbeddingDistanceEvalChain(_EmbeddingDistanceChainMixin, StringEvaluator):
         score = self._compute_score(vectors)
         return {"score": score}
 
+    @override
     async def _acall(
         self,
         inputs: dict[str, Any],
@@ -380,6 +384,7 @@ class EmbeddingDistanceEvalChain(_EmbeddingDistanceChainMixin, StringEvaluator):
         score = self._compute_score(vectors)
         return {"score": score}
 
+    @override
     def _evaluate_strings(
         self,
         *,
@@ -414,6 +419,7 @@ class EmbeddingDistanceEvalChain(_EmbeddingDistanceChainMixin, StringEvaluator):
         )
         return self._prepare_output(result)
 
+    @override
     async def _aevaluate_strings(
         self,
         *,
@@ -473,8 +479,10 @@ class PairwiseEmbeddingDistanceEvalChain(
 
     @property
     def evaluation_name(self) -> str:
+        """Return the evaluation name."""
         return f"pairwise_embedding_{self.distance_metric.value}_distance"
 
+    @override
     def _call(
         self,
         inputs: dict[str, Any],
@@ -502,6 +510,7 @@ class PairwiseEmbeddingDistanceEvalChain(
         score = self._compute_score(vectors)
         return {"score": score}
 
+    @override
     async def _acall(
         self,
         inputs: dict[str, Any],
@@ -529,6 +538,7 @@ class PairwiseEmbeddingDistanceEvalChain(
         score = self._compute_score(vectors)
         return {"score": score}
 
+    @override
     def _evaluate_string_pairs(
         self,
         *,
@@ -564,6 +574,7 @@ class PairwiseEmbeddingDistanceEvalChain(
         )
         return self._prepare_output(result)
 
+    @override
     async def _aevaluate_string_pairs(
         self,
         *,

langchain/evaluation/exact_match/base.py

@@ -1,6 +1,8 @@
 import string
 from typing import Any
 
+from typing_extensions import override
+
 from langchain.evaluation.schema import StringEvaluator
 
 
@@ -27,8 +29,18 @@ class ExactMatchStringEvaluator(StringEvaluator):
         ignore_case: bool = False,
         ignore_punctuation: bool = False,
         ignore_numbers: bool = False,
-        **kwargs: Any,
+        **_: Any,
     ):
+        """Initialize the ExactMatchStringEvaluator.
+
+        Args:
+            ignore_case: Whether to ignore case when comparing strings.
+                Defaults to False.
+            ignore_punctuation: Whether to ignore punctuation when comparing strings.
+                Defaults to False.
+            ignore_numbers: Whether to ignore numbers when comparing strings.
+                Defaults to False.
+        """
         super().__init__()
         self.ignore_case = ignore_case
         self.ignore_punctuation = ignore_punctuation
@@ -68,6 +80,7 @@ class ExactMatchStringEvaluator(StringEvaluator):
         """
         return "exact_match"
 
+    @override
     def _evaluate_strings(  # type: ignore[override]
         self,
         *,

langchain/evaluation/loading.py

@@ -58,6 +58,7 @@ def load_dataset(uri: str) -> list[dict]:
 
         from langchain.evaluation import load_dataset
         ds = load_dataset("llm-math")
+
     """
     try:
         from datasets import load_dataset

langchain/evaluation/parsing/base.py

@@ -35,18 +35,22 @@ class JsonValidityEvaluator(StringEvaluator):
         {'score': 0, 'reasoning': 'Expecting property name enclosed in double quotes'}
     """
 
-    def __init__(self, **kwargs: Any) -> None:
+    def __init__(self, **_: Any) -> None:
+        """Initialize the JsonValidityEvaluator."""
         super().__init__()
 
     @property
+    @override
     def requires_input(self) -> bool:
         return False
 
     @property
+    @override
     def requires_reference(self) -> bool:
         return False
 
     @property
+    @override
     def evaluation_name(self) -> str:
         return "json_validity"
 
@@ -110,19 +114,28 @@ class JsonEqualityEvaluator(StringEvaluator):
 
     """
 
-    def __init__(self, operator: Optional[Callable] = None, **kwargs: Any) -> None:
+    def __init__(self, operator: Optional[Callable] = None, **_: Any) -> None:
+        """Initialize the JsonEqualityEvaluator.
+
+        Args:
+            operator: A custom operator to compare the parsed JSON objects.
+                Defaults to equality (`eq`).
+        """
         super().__init__()
         self.operator = operator or eq
 
     @property
+    @override
     def requires_input(self) -> bool:
         return False
 
     @property
+    @override
     def requires_reference(self) -> bool:
         return True
 
     @property
+    @override
     def evaluation_name(self) -> str:
         return "json_equality"
 
@@ -153,7 +166,7 @@ class JsonEqualityEvaluator(StringEvaluator):
             dict: A dictionary containing the evaluation score.
         """
         parsed = self._parse_json(prediction)
-        label = self._parse_json(cast(str, reference))
+        label = self._parse_json(cast("str", reference))
         if isinstance(label, list):
             if not isinstance(parsed, list):
                 return {"score": 0}

langchain/evaluation/parsing/json_distance.py

@@ -15,13 +15,6 @@ class JsonEditDistanceEvaluator(StringEvaluator):
     after parsing them and converting them to a canonical format (i.e., whitespace and key order are normalized).
     It can be customized with alternative distance and canonicalization functions.
 
-    Args:
-        string_distance (Optional[Callable[[str, str], float]]): A callable that computes the distance between two strings.
-            If not provided, a Damerau-Levenshtein distance from the `rapidfuzz` package will be used.
-        canonicalize (Optional[Callable[[Any], Any]]): A callable that converts a parsed JSON object into its canonical string form.
-            If not provided, the default behavior is to serialize the JSON with sorted keys and no extra whitespace.
-        **kwargs (Any): Additional keyword arguments.
-
     Attributes:
         _string_distance (Callable[[str, str], float]): The internal distance computation function.
         _canonicalize (Callable[[Any], Any]): The internal canonicalization function.
@@ -40,8 +33,23 @@ class JsonEditDistanceEvaluator(StringEvaluator):
         self,
         string_distance: Optional[Callable[[str, str], float]] = None,
         canonicalize: Optional[Callable[[Any], Any]] = None,
-        **kwargs: Any,
+        **_: Any,
     ) -> None:
+        """Initialize the JsonEditDistanceEvaluator.
+
+        Args:
+            string_distance: A callable that computes the distance between two strings.
+                If not provided, a Damerau-Levenshtein distance from the `rapidfuzz`
+                package will be used.
+            canonicalize: A callable that converts a parsed JSON object into its
+                canonical string form.
+                If not provided, the default behavior is to serialize the JSON with
+                sorted keys and no extra whitespace.
+
+        Raises:
+            ImportError: If the `rapidfuzz` package is not installed and no
+            `string_distance` function is provided.
+        """
         super().__init__()
         if string_distance is not None:
             self._string_distance = string_distance
@@ -67,14 +75,17 @@ class JsonEditDistanceEvaluator(StringEvaluator):
             )
 
     @property
+    @override
     def requires_input(self) -> bool:
         return False
 
     @property
+    @override
     def requires_reference(self) -> bool:
         return True
 
     @property
+    @override
     def evaluation_name(self) -> str:
         return "json_edit_distance"
 

langchain/evaluation/parsing/json_schema.py

@@ -33,12 +33,9 @@ class JsonSchemaEvaluator(StringEvaluator):
 
     """  # noqa: E501
 
-    def __init__(self, **kwargs: Any) -> None:
+    def __init__(self, **_: Any) -> None:
         """Initializes the JsonSchemaEvaluator.
 
-        Args:
-            kwargs: Additional keyword arguments.
-
         Raises:
             ImportError: If the jsonschema package is not installed.
         """

langchain/evaluation/qa/eval_chain.py

@@ -80,18 +80,22 @@ class QAEvalChain(LLMChain, StringEvaluator, LLMEvalChain):
     )
 
     @classmethod
+    @override
     def is_lc_serializable(cls) -> bool:
         return False
 
     @property
+    @override
     def evaluation_name(self) -> str:
         return "correctness"
 
     @property
+    @override
     def requires_reference(self) -> bool:
         return True
 
     @property
+    @override
     def requires_input(self) -> bool:
         return True
 
@@ -214,6 +218,7 @@ class ContextQAEvalChain(LLMChain, StringEvaluator, LLMEvalChain):
     """LLM Chain for evaluating QA w/o GT based on context"""
 
     @classmethod
+    @override
     def is_lc_serializable(cls) -> bool:
         return False
 
@@ -242,6 +247,7 @@ class ContextQAEvalChain(LLMChain, StringEvaluator, LLMEvalChain):
             raise ValueError(msg)
 
     @property
+    @override
     def evaluation_name(self) -> str:
         return "Contextual Accuracy"
 
@@ -344,10 +350,12 @@ class CotQAEvalChain(ContextQAEvalChain):
     """LLM Chain for evaluating QA using chain of thought reasoning."""
 
     @classmethod
+    @override
     def is_lc_serializable(cls) -> bool:
         return False
 
     @property
+    @override
     def evaluation_name(self) -> str:
         return "COT Contextual Accuracy"
 

langchain/evaluation/qa/generate_chain.py

@@ -7,6 +7,7 @@ from typing import Any
 from langchain_core.language_models import BaseLanguageModel
 from langchain_core.output_parsers import BaseLLMOutputParser
 from pydantic import Field
+from typing_extensions import override
 
 from langchain.chains.llm import LLMChain
 from langchain.evaluation.qa.generate_prompt import PROMPT
@@ -25,6 +26,7 @@ class QAGenerateChain(LLMChain):
     output_key: str = "qa_pairs"
 
     @classmethod
+    @override
     def is_lc_serializable(cls) -> bool:
         return False
 

langchain/evaluation/regex_match/base.py

@@ -1,6 +1,8 @@
 import re
 from typing import Any
 
+from typing_extensions import override
+
 from langchain.evaluation.schema import StringEvaluator
 
 
@@ -27,7 +29,12 @@ class RegexMatchStringEvaluator(StringEvaluator):
         )  # This will return {'score': 1.0} as the prediction matches the second pattern in the union
     """  # noqa: E501
 
-    def __init__(self, *, flags: int = 0, **kwargs: Any):  # Default is no flags
+    def __init__(self, *, flags: int = 0, **_: Any):  # Default is no flags
+        """Initialize the RegexMatchStringEvaluator.
+
+        Args:
+            flags: Flags to use for the regex match. Defaults to 0 (no flags).
+        """
         super().__init__()
         self.flags = flags
 
@@ -65,6 +72,7 @@ class RegexMatchStringEvaluator(StringEvaluator):
         """
         return "regex_match"
 
+    @override
     def _evaluate_strings(  # type: ignore[override]
         self,
         *,