Skip to content

The agent engineering platform.

xai[patch]: support reasoning content (#30758)

ccurme9cfb95e6216c
diff --git a/libs/partners/xai/langchain_xai/chat_models.py b/libs/partners/xai/langchain_xai/chat_models.py
index 5690843f26..650ea5dd8b 100644
--- a/libs/partners/xai/langchain_xai/chat_models.py
+++ b/libs/partners/xai/langchain_xai/chat_models.py
@@ -13,6 +13,8 @@ from langchain_core.language_models.chat_models import (
LangSmithParams,
LanguageModelInput,
)
+from langchain_core.messages import AIMessageChunk
+from langchain_core.outputs import ChatGenerationChunk, ChatResult
from langchain_core.runnables import Runnable
from langchain_core.utils import secret_from_env
from langchain_openai.chat_models.base import BaseChatOpenAI
@@ -369,6 +371,44 @@ class ChatXAI(BaseChatOpenAI): # type: ignore[override]
)
return self
+ def _create_chat_result(
+ self,
+ response: Union[dict, openai.BaseModel],
+ generation_info: Optional[dict] = None,
+ ) -> ChatResult:
+ rtn = super()._create_chat_result(response, generation_info)
+
+ if not isinstance(response, openai.BaseModel):
+ return rtn
+
+ if hasattr(response.choices[0].message, "reasoning_content"): # type: ignore
+ rtn.generations[0].message.additional_kwargs["reasoning_content"] = (
+ response.choices[0].message.reasoning_content # type: ignore
+ )
+
+ return rtn
+
+ def _convert_chunk_to_generation_chunk(
+ self,
+ chunk: dict,
+ default_chunk_class: type,
+ base_generation_info: Optional[dict],
+ ) -> Optional[ChatGenerationChunk]:
+ generation_chunk = super()._convert_chunk_to_generation_chunk(
+ chunk,
+ default_chunk_class,
+ base_generation_info,
+ )
+ if (choices := chunk.get("choices")) and generation_chunk:
+ top = choices[0]
+ if isinstance(generation_chunk.message, AIMessageChunk):
+ if reasoning_content := top.get("delta", {}).get("reasoning_content"):
+ generation_chunk.message.additional_kwargs["reasoning_content"] = (
+ reasoning_content
+ )
+
+ return generation_chunk
+
def with_structured_output(
self,
schema: Optional[_DictOrPydanticClass] = None,
diff --git a/libs/partners/xai/tests/integration_tests/test_chat_models_standard.py b/libs/partners/xai/tests/integration_tests/test_chat_models_standard.py
index 632b518f71..1e14116c05 100644
--- a/libs/partners/xai/tests/integration_tests/test_chat_models_standard.py
+++ b/libs/partners/xai/tests/integration_tests/test_chat_models_standard.py
@@ -1,7 +1,9 @@
"""Standard LangChain interface tests"""
-import pytest # type: ignore[import-not-found]
+from typing import Optional
+
from langchain_core.language_models import BaseChatModel
+from langchain_core.messages import AIMessageChunk, BaseMessageChunk
from langchain_core.rate_limiters import InMemoryRateLimiter
from langchain_tests.integration_tests import ( # type: ignore[import-not-found]
ChatModelIntegrationTests, # type: ignore[import-not-found]
@@ -24,10 +26,25 @@ class TestXAIStandard(ChatModelIntegrationTests):
@property
def chat_model_params(self) -> dict:
return {
- "model": "grok-2",
+ "model": "grok-3",
"rate_limiter": rate_limiter,
+ "stream_usage": True,
}
- @pytest.mark.xfail(reason="Not yet supported.")
- def test_usage_metadata_streaming(self, model: BaseChatModel) -> None:
- super().test_usage_metadata_streaming(model)
+
+def test_reasoning_content() -> None:
+ """Test reasoning content."""
+ chat_model = ChatXAI(
+ model="grok-3-mini-beta",
+ reasoning_effort="low",
+ )
+ response = chat_model.invoke("What is 3^3?")
+ assert response.content
+ assert response.additional_kwargs["reasoning_content"]
+
+ # Test streaming
+ full: Optional[BaseMessageChunk] = None
+ for chunk in chat_model.stream("What is 3^3?"):
+ full = chunk if full is None else full + chunk
+ assert isinstance(full, AIMessageChunk)
+ assert full.additional_kwargs["reasoning_content"]