anthropic[patch]: support parallel_tool_calls

libs/partners/anthropic/langchain_anthropic/chat_models.py

@@ -819,6 +819,7 @@ def bind_tools(
         tool_choice: Optional[
             Union[Dict[str, str], Literal["any", "auto"], str]
         ] = None,
+        parallel_tool_calls: Optional[bool] = None,
         **kwargs: Any,
     ) -> Runnable[LanguageModelInput, BaseMessage]:
         r"""Bind tool-like objects to this chat model.
@@ -832,6 +833,10 @@ def bind_tools(
                 - name of the tool as a string or as dict ``{"type": "tool", "name": "<<tool_name>>"}``: calls corresponding tool;
                 - ``"auto"``, ``{"type: "auto"}``, or None: automatically selects a tool (including no tool);
                 - ``"any"`` or ``{"type: "any"}``: force at least one tool to be called;
+            parallel_tool_calls: Set to ``False`` to disable parallel tool use.
+                Defaults to ``None`` (no specification, which allows parallel tool use).
+
+                .. versionadded:: 0.3.2
             kwargs: Any additional parameters are passed directly to
                 :meth:`~langchain_anthropic.chat_models.ChatAnthropic.bind`.
 
@@ -968,6 +973,19 @@ class GetPrice(BaseModel):
                 f"Unrecognized 'tool_choice' type {tool_choice=}. Expected dict, "
                 f"str, or None."
             )
+
+        if parallel_tool_calls is not None:
+            disable_parallel_tool_use = not parallel_tool_calls
+            if "tool_choice" in kwargs:
+                kwargs["tool_choice"]["disable_parallel_tool_use"] = (
+                    disable_parallel_tool_use
+                )
+            else:
+                kwargs["tool_choice"] = {
+                    "type": "any",
+                    "disable_parallel_tool_use": disable_parallel_tool_use,
+                }
+
         return self.bind(tools=formatted_tools, **kwargs)
 
     def with_structured_output(

libs/partners/anthropic/tests/integration_tests/test_chat_models.py

@@ -444,6 +444,25 @@ def test_tool_use() -> None:
     assert len(chunks) > 1
 
 
+class GenerateUsername(BaseModel):
+    "Get a username based on someone's name and hair color."
+
+    name: str
+    hair_color: str
+
+
+def test_disable_parallel_tool_calling() -> None:
+    llm = ChatAnthropic(model="claude-3-5-sonnet-20241022")
+    llm_with_tools = llm.bind_tools([GenerateUsername], parallel_tool_calls=False)
+    result = llm_with_tools.invoke(
+        "Use the GenerateUsername tool to generate user names for:\n\n"
+        "Sally with green hair\n"
+        "Bob with blue hair"
+    )
+    assert isinstance(result, AIMessage)
+    assert len(result.tool_calls) == 1
+
+
 def test_anthropic_with_empty_text_block() -> None:
     """Anthropic SDK can return an empty text block."""
 

libs/partners/openai/tests/integration_tests/chat_models/test_base.py

@@ -630,6 +630,18 @@ def test_bind_tools_tool_choice() -> None:
     assert not msg.tool_calls
 
 
+def test_disable_parallel_tool_calling() -> None:
+    llm = ChatOpenAI(model="gpt-4o-mini")
+    llm_with_tools = llm.bind_tools([GenerateUsername], parallel_tool_calls=False)
+    result = llm_with_tools.invoke(
+        "Use the GenerateUsername tool to generate user names for:\n\n"
+        "Sally with green hair\n"
+        "Bob with blue hair"
+    )
+    assert isinstance(result, AIMessage)
+    assert len(result.tool_calls) == 1
+
+
 @pytest.mark.parametrize("model", ["gpt-4o-mini", "o1"])
 def test_openai_structured_output(model: str) -> None:
     class MyModel(BaseModel):