Documentation example for tools uses incorrect tool_choice format

[liu673]

The problem I discovered while learning LangGraph.
documentation: https://langchain-ai.github.io/langgraph/agents/tools/#force-tool-use

The current documentation example for tools uses an invalid tool_choice format ({"type": "tool", "name": "greet"}) that doesn't work. After debugging, I found the correct format should follow the OpenAI-style function calling specification ({"type": "function", "function": {"name": "greet"}}). This discrepancy causes runtime errors and confusion for users.

My LangGraph Version: 0.4.8
Langchain Version: 0.3.25

from langchain_core.tools import tool

@tool(return_direct=True, name_or_callable="greet")
def greet(user_name: str) -> int:
    """Greet user."""
    return f"Hello {user_name}!"

tools = [greet]

# Documented format (fails)
agent = create_react_agent(
    model=llm.bind_tools(tools, tool_choice={"type": "tool", "name": "greet"}),
    tools=tools
)

agent.invoke(
    {"messages": [{"role": "user", "content": "Hi, I am Bob"}]}
)

run code errors:

BadRequestError: Error code: 400 - {'error': {'code': 'InvalidParameter', 'message': "The parameter `tool_choice.type` specified in the request are not valid: expected 'function', but got `tool` instead. Request id: 0217503980949656878bc4002d5738785e1bec3dfd2141146d576", 'param': 'tool_choice.type', 'type': 'BadRequest'}}

Working Solution:

# Correct format
agent = create_react_agent(
    model=llm.bind_tools(
        tools=tools,
        tool_choice={"type": "function", "function": {"name": "greet"}}
    ),
    tools=tools
)

Expected Behavior:
Documentation examples should show the correct tool_choice format that matches the underlying LLM API expectations.