core[minor]: support pydantic v2 models in PydanticOutputParser

libs/core/langchain_core/output_parsers/pydantic_v2.py

@@ -0,0 +1,55 @@
+import json
+from typing import Generic, List, Type, TypeVar
+
+from langchain_core.exceptions import OutputParserException
+from langchain_core.output_parsers.json import JsonOutputParser
+from langchain_core.output_parsers.pydantic import _PYDANTIC_FORMAT_INSTRUCTIONS
+from langchain_core.outputs.generation import Generation
+from langchain_core.pydantic_v2 import BaseModel, ValidationError
+
+TBaseModel = TypeVar("TBaseModel", bound=BaseModel)
+
+
+class PydanticV2OutputParser(JsonOutputParser, Generic[TBaseModel]):
+    """Parse an output using a pydantic model."""
+
+    pydantic_v2_object: Type[TBaseModel]
+    """The pydantic model to parse."""
+
+    def parse_result(
+        self, result: List[Generation], *, partial: bool = False
+    ) -> TBaseModel:
+        json_object = super().parse_result(result)
+        try:
+            return self.pydantic_v2_object.model_validate(json_object)
+        except ValidationError as e:
+            name = self.pydantic_v2_object.__name__
+            msg = f"Failed to parse {name} from completion {json_object}. Got: {e}"
+            raise OutputParserException(msg, llm_output=json_object)
+
+    def parse(self, text: str) -> TBaseModel:
+        return super().parse(text)
+
+    def get_format_instructions(self) -> str:
+        # Copy schema to avoid altering original Pydantic schema.
+        schema = {k: v for k, v in self.pydantic_v2_object.model_json_schema().items()}
+
+        # Remove extraneous fields.
+        reduced_schema = schema
+        if "title" in reduced_schema:
+            del reduced_schema["title"]
+        if "type" in reduced_schema:
+            del reduced_schema["type"]
+        # Ensure json in context is well-formed with double quotes.
+        schema_str = json.dumps(reduced_schema)
+
+        return _PYDANTIC_FORMAT_INSTRUCTIONS.format(schema=schema_str)
+
+    @property
+    def _type(self) -> str:
+        return "pydantic"
+
+    @property
+    def OutputType(self) -> Type[TBaseModel]:
+        """Return the pydantic model."""
+        return self.pydantic_v2_object

libs/core/langchain_core/pydantic_v2/__init__.py

@@ -0,0 +1,24 @@
+import warnings
+
+from langchain_core.utils.pydantic import get_pydantic_major_version
+
+## Create namespaces for pydantic v1 and v2.
+# This code must stay at the top of the file before other modules may
+# attempt to import pydantic since it adds pydantic_v1 and pydantic_v2 to sys.modules.
+#
+# This hack is done for the following reasons:
+# * Langchain will attempt to remain compatible with both pydantic v1 and v2 since
+#   both dependencies and dependents may be stuck on either version of v1 or v2.
+# * Creating namespaces for pydantic v1 and v2 should allow us to write code that
+#   unambiguously uses either v1 or v2 API.
+# * This change is easier to roll out and roll back.
+
+
+if get_pydantic_major_version() < 2:
+    warnings.warn(
+        "The pydantic_v2 namespace only supports Pydantic v2 and later. \
+Please use pydantic_v1 namespace.",
+        ImportWarning,
+    )
+
+from pydantic import *  # noqa: F403 # type: ignore

libs/core/scripts/check_pydantic.sh

@@ -14,7 +14,10 @@ fi
 repository_path="$1"
 
 # Search for lines matching the pattern within the specified repository
-result=$(git -C "$repository_path" grep -E '^import pydantic|^from pydantic')
+result=$(
+  git -C "$repository_path" grep -E '^[[:space:]]*import pydantic|^[[:space:]]*from pydantic' \
+  -- ':!langchain_core/pydantic_*' ':!langchain_core/utils'  # exclude internal namespace and utils
+)
 
 # Check if any matching lines were found
 if [ -n "$result" ]; then