fix(langchain): correctly initialize huggingface models in init_chat_model

libs/langchain/langchain/chat_models/base.py

@@ -38,17 +38,20 @@
     *,
     model_provider: Optional[str] = None,
     configurable_fields: None = None,
+    configurable_fields: None = None,
     config_prefix: Optional[str] = None,
     **kwargs: Any,
 ) -> BaseChatModel: ...
 
 
 @overload
 def init_chat_model(
+    model: None = None,
     model: None = None,
     *,
     model_provider: Optional[str] = None,
     configurable_fields: None = None,
+    configurable_fields: None = None,
     config_prefix: Optional[str] = None,
     **kwargs: Any,
 ) -> _ConfigurableModel: ...
@@ -128,6 +131,16 @@
             - ``deepseek...``                     -> ``deepseek``
             - ``grok...``                         -> ``xai``
             - ``sonar...``                        -> ``perplexity``
+            - ``gpt-3...`` | ``gpt-4...`` | ``o1...`` -> ``openai``
+            - ``claude...``               -> ``anthropic``
+            - ``amazon...``               -> ``bedrock``
+            - ``gemini...``               -> ``google_vertexai``
+            - ``command...``              -> ``cohere``
+            - ``accounts/fireworks...``   -> ``fireworks``
+            - ``mistral...``              -> ``mistralai``
+            - ``deepseek...``             -> ``deepseek``
+            - ``grok...``                 -> ``xai``
+            - ``sonar...``                -> ``perplexity``
         configurable_fields: Which model parameters are configurable:
 
             - None: No configurable fields.
@@ -277,6 +290,10 @@
                     GetWeather,
                     GetPopulation,
                 ]
+                [
+                    GetWeather,
+                    GetPopulation,
+                ]
             )
             configurable_model_with_tools.invoke(
                 "Which city is hotter today and which is bigger: LA or NY?"
@@ -415,10 +432,32 @@
 
         return ChatMistralAI(model=model, **kwargs)  # type: ignore[call-arg,unused-ignore]
     if model_provider == "huggingface":
-        _check_pkg("langchain_huggingface")
-        from langchain_huggingface import ChatHuggingFace
+        try:
+            from langchain_huggingface.chat_models import ChatHuggingFace
+            from langchain_huggingface.llms import HuggingFacePipeline
+        except ImportError as e:
+            raise ImportError(
+                "Please install langchain-huggingface to use HuggingFace models."
+            ) from e
+
+        # The 'task' kwarg is required by from_model_id but not the base constructor.
+        # We pop it from kwargs to avoid the Pydantic 'extra_forbidden' error.
+        task = kwargs.pop("task", None)
+        if not task:
+            raise ValueError(
+                "The 'task' keyword argument is required for HuggingFace models. "
+                "For example: task='text-generation'."
+            )
+
+        # Initialize the base LLM pipeline with the model and arguments
+        llm = HuggingFacePipeline.from_model_id(
+            model_id=model,
+            task=task,
+            **kwargs,  # Pass remaining kwargs like `device`
+        )
 
-        return ChatHuggingFace(model_id=model, **kwargs)
+        # Pass the initialized LLM to the chat wrapper
+        return ChatHuggingFace(llm=llm)
     if model_provider == "groq":
         _check_pkg("langchain_groq")
         from langchain_groq import ChatGroq
@@ -497,7 +536,8 @@
 
 
 def _attempt_infer_model_provider(model_name: str) -> Optional[str]:
+    if any(model_name.startswith(pre) for pre in ("gpt-", "o1", "o3")):
     if any(model_name.startswith(pre) for pre in ("gpt-", "o1", "o3")):
         return "openai"
     if model_name.startswith("claude"):
         return "anthropic"
@@ -958,3 +998,4 @@
         **kwargs: Any,
     ) -> Runnable[LanguageModelInput, Union[dict, BaseModel]]:
         return self.__getattr__("with_structured_output")(schema, **kwargs)
+

libs/langchain/tests/unit_tests/chat_models/test_base.py

@@ -6,6 +6,7 @@
 from langchain_core.language_models import BaseChatModel
 from langchain_core.prompts import ChatPromptTemplate
 from langchain_core.runnables import RunnableConfig, RunnableSequence
+from langchain_huggingface.chat_models import ChatHuggingFace
 from pydantic import SecretStr
 
 from langchain.chat_models.base import __all__, init_chat_model
@@ -289,3 +290,16 @@ def test_configurable_with_default() -> None:
     prompt = ChatPromptTemplate.from_messages([("system", "foo")])
     chain = prompt | model_with_config
     assert isinstance(chain, RunnableSequence)
+
+
+def test_init_chat_model_huggingface() -> None:
+    """Test that init_chat_model works with huggingface."""
+    model_name = "google-bert/bert-base-uncased"
+
+    llm = init_chat_model(
+        model=model_name,
+        model_provider="huggingface",
+        task="text-generation",
+    )
+    assert isinstance(llm, ChatHuggingFace)
+    assert llm.llm.model_id == model_name