Python 3.14 support

[iddobergerbioz]

Checked other resources

• This is a feature request, not a bug report or usage question.
• I added a clear and descriptive title that summarizes the feature request.
• I used the GitHub search to find a similar feature request and didn't find it.
• I checked the LangChain documentation and API reference to see if this feature already exists.
• This is not related to the langchain-community package.

Package (Required)

• langchain
• langchain-openai
• langchain-anthropic
• langchain-classic
• langchain-core
• langchain-cli
• langchain-model-profiles
• langchain-tests
• langchain-text-splitters
• langchain-chroma
• langchain-deepseek
• langchain-exa
• langchain-fireworks
• langchain-groq
• langchain-huggingface
• langchain-mistralai
• langchain-nomic
• langchain-ollama
• langchain-perplexity
• langchain-prompty
• langchain-qdrant
• langchain-xai
• Other / not sure / general

Feature Description

Any estimate on when will you have support for Python 3.14?
(I saw there is a line: "Python 3.14 support is coming soon!" in "https://blog.langchain.com/langchain-langgraph-1dot0/", but it was published in October, without any updates)

Use Case

We would be able to move our code to Python 3.14

Proposed Solution

No response

Alternatives Considered

No response

Additional Context

No response

[gawry-axur]

Are there any plans to add the suport?

langchain_core/_api/deprecation.py:26: UserWarning: Core Pydantic V1 functionality isn't compatible with Python 3.14 or greater.
    from pydantic.v1.fields import FieldInfo as FieldInfoV1

[ulgens]

I tried to understand what is missing for 3.14 support and how I can help, but I wasn't able to do so.

I found several merged PRs, related to 3.14 support:

• fix(core): support for Python 3.14 #33461
• fix(langchain_v1): support Python 3.14 #33559
• feat(langchain_v1): Python 3.14 support #33560
• fix: support python 3.14 in various projects #33575
• fix(cli): support Python 3.14 #33598
• fix(nomic): support Python 3.14 #33655
• release(nomic): 1.0.1 #33948
• fix(core): use get_type_hints for Python 3.14 TypedDict compatibility #34390
• chore(text-splitters): bump dependency locks for python 3.14 #34522
• chore(langchain-classic): bump markupsafe lock for python 3.14 #34523
• chore(deps): bump the uv group across 3 directories with 1 update #34741
• chore(deps): bump langgraph-checkpoint from 2.1.2 to 3.0.0 in /libs/cli in the uv group across 1 directory #34787
• chore(deps): bump the uv group across 20 directories with 3 updates #34941

but I'm not sure what is left for the "official" support. One thing I could find was that 3.14 is missing in the integration tests: #34993

[cbornet]

langchain-chroma is the only one not working atm. All the other packages are unit tested in CI on Python 3.14:

langchain/.github/scripts/check_diff.py

Line 138 in 91bb474

elif dir_ in {"libs/partners/chroma"}:

[cbornet]

Gustavo Gawry (@gawry-axur), pydantic v1 models don't work on Python 3.14. This is independant of Langchain (if there is still a pydantic v1 model used by langchain, that's a bug). Can you check if you use pydantic v1 models in your code ?

[gawry-axur]

@gawry-axur, pydantic v1 models don't work on Python 3.14. This is independant of Langchain (if there is still a pydantic v1 model used by langchain, that's a bug). Can you check if you use pydantic v1 models in your code ?

I just ran bump-pydantic in my codebase, And it really seems like the warning is coming from langchain-core.

Then I ran bump-pydantic on langchain-core and got this:

130 % uv run bump-pydantic libs/core
[11:09:27] Start bump-pydantic.                                                                                                                                                main.py:61
           Found 331 files to process.                                                                                                                                         main.py:78
[11:15:13] Refactored 6 files.                                                                                                                                                main.py:144
           Run successfully!

git diff

diff --git a/libs/core/langchain_core/structured_query.py b/libs/core/langchain_core/structured_query.py
index 84dcd9a5c2..036111a473 100644
--- a/libs/core/langchain_core/structured_query.py
+++ b/libs/core/langchain_core/structured_query.py
@@ -132,7 +132,7 @@ class Comparison(FilterDirective):
     attribute: str
     """The attribute to compare."""
 
-    value: Any
+    value: Any = None
     """The value to compare to."""
 
     def __init__(
@@ -179,10 +179,10 @@ class StructuredQuery(Expr):
     query: str
     """Query string."""
 
-    filter: FilterDirective | None
+    filter: FilterDirective | None = None
     """Filtering expression."""
 
-    limit: int | None
+    limit: int | None = None
     """Limit on the number of results."""
 
     def __init__(
diff --git a/libs/core/langchain_core/utils/pydantic.py b/libs/core/langchain_core/utils/pydantic.py
index aa47ae37f5..90447ab87a 100644
--- a/libs/core/langchain_core/utils/pydantic.py
+++ b/libs/core/langchain_core/utils/pydantic.py
@@ -21,13 +21,11 @@ from packaging import version
 
 # root_validator is deprecated but we need it for backward compatibility of @pre_init
 from pydantic import (  # type: ignore[deprecated]
-    BaseModel,
+    model_validator, BaseModel,
     ConfigDict,
     Field,
     PydanticDeprecationWarning,
-    RootModel,
-    root_validator,
-)
+    RootModel)
 from pydantic import (
     create_model as _create_model_base,
 )
@@ -141,8 +139,9 @@ def pre_init(func: Callable) -> Any:
         # Ideally we would use @model_validator(mode="before") but this would change the
         # order of the validators. See https://github.com/pydantic/pydantic/discussions/7434.
         # So we keep root_validator for backward compatibility.
-        @root_validator(pre=True)  # type: ignore[deprecated]
+        @model_validator(mode="before")  # type: ignore[deprecated]
         @wraps(func)
+        @classmethod
         def wrapper(cls: type[BaseModel], values: dict[str, Any]) -> Any:
             """Decorator to run a function before model initialization.
 
diff --git a/libs/core/tests/unit_tests/load/test_secret_injection.py b/libs/core/tests/unit_tests/load/test_secret_injection.py
index 7810638e62..7231b87753 100644
--- a/libs/core/tests/unit_tests/load/test_secret_injection.py
+++ b/libs/core/tests/unit_tests/load/test_secret_injection.py
@@ -176,7 +176,7 @@ class TestPydanticModelTopLevel:
         """Pydantic model containing a `Serializable` with secret-like dict."""
 
         class MyModel(BaseModel):
-            message: Any
+            message: Any = None
 
         msg = HumanMessage(
             content="Hello",
diff --git a/libs/core/tests/unit_tests/runnables/test_runnable_events_v1.py b/libs/core/tests/unit_tests/runnables/test_runnable_events_v1.py
index 06f13e035b..4f5bd03c02 100644
--- a/libs/core/tests/unit_tests/runnables/test_runnable_events_v1.py
+++ b/libs/core/tests/unit_tests/runnables/test_runnable_events_v1.py
@@ -1963,7 +1963,7 @@ async def test_runnable_with_message_history() -> None:
         # the content in the store!
 
         # Using Any type here rather than list[BaseMessage] due to pydantic issue!
-        messages: Any
+        messages: Any = None
 
         def add_message(self, message: BaseMessage) -> None:
             """Add a self-created message to the store."""
diff --git a/libs/core/tests/unit_tests/runnables/test_runnable_events_v2.py b/libs/core/tests/unit_tests/runnables/test_runnable_events_v2.py
index 49342d2fc1..ed61b946a1 100644
--- a/libs/core/tests/unit_tests/runnables/test_runnable_events_v2.py
+++ b/libs/core/tests/unit_tests/runnables/test_runnable_events_v2.py
@@ -1911,7 +1911,7 @@ async def test_runnable_with_message_history() -> None:
         # the content in the store!
 
         # Using Any type here rather than list[BaseMessage] due to pydantic issue!
-        messages: Any
+        messages: Any = None
 
         def add_message(self, message: BaseMessage) -> None:
             """Add a self-created message to the store."""
diff --git a/libs/core/tests/unit_tests/test_tools.py b/libs/core/tests/unit_tests/test_tools.py
index 76506ae2a4..69ba80bd42 100644
--- a/libs/core/tests/unit_tests/test_tools.py
+++ b/libs/core/tests/unit_tests/test_tools.py
@@ -2190,7 +2190,7 @@ def test__get_all_basemodel_annotations_v2(*, use_v1_namespace: bool) -> None:
     D = TypeVar("D", bound=str | int)
 
     class ModelD(ModelC, Generic[D]):
-        d: D | None
+        d: D | None = None
 
     expected = {
         "a": str,

[cbornet]

bump-pydantic didn't show any issue linked to using a v1 model: only minor stuff and a recommendation to use model_validator instead of root-validator but it is documented that root-validator is kept atm for backward compatibility (and it works with Python 3.14).
I just checked and didn't find any v1 model in langchain-core (except in tests that are excluded from Python 3.14 testing)
If you find a v1 model class, can you report it ?

[cbornet]

Also, there's an issue in pydantic that it emits warnings even if you don't use v1 models: pydantic/pydantic#12618 . So if you are sure to not use v1 models in your own code, you can safely ignore the warning. The issue will be fixed in Pydantic 2.13.
I'll see if it's possible to force remove the warning from langchain.

[cbornet]

I'll see if it's possible to force remove the warning from langchain.

So it's possible but it will remove the warning completely, even for imports done in user code if they are done after importing langchain.
Mason Daugherty (@mdrxy) Eugene Yurtsev (@eyurtsev) should we do it ?

[itaru2622]

Hi all, I found workaround in user program level, as below.

not sure the place to add below additional code in langchain lib(s).
maybe somewhere in __init__.py or others, before the line of from pydantic.v1...

#!/usr/bin/env python3
# test.py
 
+# starts workaround : add  code to ignore warning
+import warnings
+warnings.filterwarnings("ignore", message=".*Pydantic V1 functionality isn't compatible with.*")
+# ends workaround : add  code to ignore warning

from langchain_ollama import ChatOllama
model = ChatOllama(model="llama3.2")
result = model.invoke("Hi, who is it?")
print(result.content)

without workaround:

./test.py 
/usr/local/lib/python3.14/site-packages/langchain_core/_api/deprecation.py:25: UserWarning: Core Pydantic V1 functionality isn't compatible with Python 3.14 or greater.
  from pydantic.v1.fields import FieldInfo as FieldInfoV1
I'm an artificial intelligence model known as Llama. Llama stands for "Large Language Model Meta AI."

with workaround:

./test.py 
I'm an artificial intelligence model known as Llama. Llama stands for "Large Language Model Meta AI."