langchain-ai/langgraph
1
0
Fork 0
mirror of https://github.com/langchain-ai/langgraph.git synced 2026-07-15 12:35:43 -04:00
Code Issues 314 Packages Projects Releases 17 Wiki Activity

[PR #6232] feat(langgraph): supports calling (multiple) handoffs and non-handoff tools at the same time #4919

New Issue
Closed
opened 2026-02-20 17:50:57 -05:00 by yindo · 0 comments
yindo commented 2026-02-20 17:50:57 -05:00
Owner
Copy Link
Copy Source

Original Pull Request: https://github.com/langchain-ai/langgraph/pull/6232

State: closed
Merged: No

Description:

This PR extends ToolNode to support executing non-handoff tools in parallel while treating handoff tool calls (handoff_prefix, default "transfer_to") with special handling. Specifically:

Non-handoff tool calls are executed concurrently (threadpool or asyncio).

Multiple handoff calls are detected: only the first one is processed; subsequent ones are ignored with an error message.

Handoff calls can incorporate previous tool outputs into their state.messages.

Added error handling via a custom tool_error_handler, with explicit handling of AtaConnectionKickedOffError.

In chat_agent_executor, introduced a v3 mode to forward all tool_calls together, distinguishing handoffs more cleanly.

This improves agent flexibility in scenarios where multiple tool invocations and handoff operations need to be coordinated in the same step.

Issue:

N/A (new feature; not tied to a specific issue, but can help agent developers who need multi-tool and handoff orchestration).

Dependencies:

No new external dependencies introduced.

Reuses existing LangGraph execution utilities (get_executor_for_config, _run_one, _arun_one).

Tests & Docs:

from dotenv import load_dotenv
load_dotenv()

import asyncio
from typing import Annotated
from langchain.tools import tool
from langchain_openai import AzureChatOpenAI
from langgraph.prebuilt import create_react_agent, ToolNode
from langgraph.checkpoint.memory import InMemorySaver
from langgraph.store.memory import InMemoryStore
from langgraph_swarm import create_handoff_tool, create_swarm   # test with langgraph_swarm==0.0.14 to avoid implementing swarm again

model = AzureChatOpenAI(
    deployment_name="gpt-4o",
)


@tool
def get_weather(city: Annotated[str, "The city to get the weather for"]) -> dict:
    """Get the current weather information for a specified city."""
    print("[debug] get_weather called")
    return {
        "city": city,
        "temperature_range": "14-20C",
        "conditions": "Sunny with wind.",
    }


@tool
def search_web(query: Annotated[str, "The content you awant to search"]) -> dict:
    """Search the web for information retrieval."""
    print("[debug] search_web called")
    return {"content": "The nvidia stock price is $185"}


@tool
def query_device(device_id: Annotated[str, "The Device id"]) -> dict:
    """Get the current status of a device."""
    print("[debug] query_device! called")
    return {"online": True, "device_name": "Happy Device", "password": "a123456789"}


# Configure the agent with Azure OpenAI
agent = create_react_agent(
    model,
    ToolNode(
        [
            query_device,
            create_handoff_tool(agent_name="weather_agent"),
            create_handoff_tool(agent_name="web_agent"),
        ]
    ),
    prompt="You are a helpful assistant",
    name="agent",
)

weather_agent = create_react_agent(
    model,
    ToolNode(
        [
            get_weather,
            create_handoff_tool(agent_name="agent"),
            create_handoff_tool(agent_name="web_agent"),
        ]
    ),
    prompt="You are a helpful assistant that can search the weather",
    name="weather_agent",
)

web_agent = create_react_agent(
    model,
    ToolNode(
        [
            search_web,
            create_handoff_tool(agent_name="agent"),
            create_handoff_tool(agent_name="weather_agent"),
        ]
    ),
    prompt="You are a helpful assistant that can search the web",
    name="web_agent",
)

workflow = create_swarm([agent, weather_agent, web_agent], default_active_agent="agent")


# Compile with checkpointer/store
checkpointer = InMemorySaver()
store = InMemoryStore()

app = workflow.compile(checkpointer=checkpointer, store=store)


async def main():
    res = await app.ainvoke(
        {
            "messages": [
                (
                    "user",
                    "What is the stock price of nvidia; What is the status of device '123'",
                )
            ],
        },
        config={"configurable": {"thread_id": "123"}},
    )

    for m in res["messages"]:
        m.pretty_print()


if __name__ == "__main__":
    asyncio.run(main())
**Original Pull Request:** https://github.com/langchain-ai/langgraph/pull/6232 **State:** closed **Merged:** No --- ### Description: This PR extends ToolNode to support executing non-handoff tools in parallel while treating handoff tool calls (handoff_prefix, default "transfer_to") with special handling. Specifically: Non-handoff tool calls are executed concurrently (threadpool or asyncio). Multiple handoff calls are detected: only the first one is processed; subsequent ones are ignored with an error message. Handoff calls can incorporate previous tool outputs into their state.messages. Added error handling via a custom tool_error_handler, with explicit handling of AtaConnectionKickedOffError. In chat_agent_executor, introduced a v3 mode to forward all tool_calls together, distinguishing handoffs more cleanly. This improves agent flexibility in scenarios where multiple tool invocations and handoff operations need to be coordinated in the same step. ### Issue: N/A (new feature; not tied to a specific issue, but can help agent developers who need multi-tool and handoff orchestration). ### Dependencies: No new external dependencies introduced. Reuses existing LangGraph execution utilities (get_executor_for_config, _run_one, _arun_one). ### Tests & Docs: ``` python from dotenv import load_dotenv load_dotenv() import asyncio from typing import Annotated from langchain.tools import tool from langchain_openai import AzureChatOpenAI from langgraph.prebuilt import create_react_agent, ToolNode from langgraph.checkpoint.memory import InMemorySaver from langgraph.store.memory import InMemoryStore from langgraph_swarm import create_handoff_tool, create_swarm # test with langgraph_swarm==0.0.14 to avoid implementing swarm again model = AzureChatOpenAI( deployment_name="gpt-4o", ) @tool def get_weather(city: Annotated[str, "The city to get the weather for"]) -> dict: """Get the current weather information for a specified city.""" print("[debug] get_weather called") return { "city": city, "temperature_range": "14-20C", "conditions": "Sunny with wind.", } @tool def search_web(query: Annotated[str, "The content you awant to search"]) -> dict: """Search the web for information retrieval.""" print("[debug] search_web called") return {"content": "The nvidia stock price is $185"} @tool def query_device(device_id: Annotated[str, "The Device id"]) -> dict: """Get the current status of a device.""" print("[debug] query_device! called") return {"online": True, "device_name": "Happy Device", "password": "a123456789"} # Configure the agent with Azure OpenAI agent = create_react_agent( model, ToolNode( [ query_device, create_handoff_tool(agent_name="weather_agent"), create_handoff_tool(agent_name="web_agent"), ] ), prompt="You are a helpful assistant", name="agent", ) weather_agent = create_react_agent( model, ToolNode( [ get_weather, create_handoff_tool(agent_name="agent"), create_handoff_tool(agent_name="web_agent"), ] ), prompt="You are a helpful assistant that can search the weather", name="weather_agent", ) web_agent = create_react_agent( model, ToolNode( [ search_web, create_handoff_tool(agent_name="agent"), create_handoff_tool(agent_name="weather_agent"), ] ), prompt="You are a helpful assistant that can search the web", name="web_agent", ) workflow = create_swarm([agent, weather_agent, web_agent], default_active_agent="agent") # Compile with checkpointer/store checkpointer = InMemorySaver() store = InMemoryStore() app = workflow.compile(checkpointer=checkpointer, store=store) async def main(): res = await app.ainvoke( { "messages": [ ( "user", "What is the stock price of nvidia; What is the status of device '123'", ) ], }, config={"configurable": {"thread_id": "123"}}, ) for m in res["messages"]: m.pretty_print() if __name__ == "__main__": asyncio.run(main()) ```
yindo added the pull-request label 2026-02-20 17:50:57 -05:00
yindo closed this issue 2026-02-20 17:50:57 -05:00
Sign in to join this conversation.
No Branch/Tag Specified
Branches Tags
Labels
Clear labels
agents
awaiting author revision
bug
change
chat models
checkpointer
ci
code quality
code quality
dependencies
documentation
drawing
duplicate
enhancement
good first issue
help wanted
invalid
maintainer
meta
missing information
not maintained by langchain
open-swe-auto
open-swe-max
open-swe-max-auto
pending
performance
pull-request
Mirrored from GitHub Pull Request
 question
requires design
sdk
server
stale
studio
tests
type checking
usage question
v0.5
v1.1
No Label pull-request
Milestone
No items
No Milestone
Projects
Clear projects
No project
Assignees
Clear assignees
yindo
No Assignees
1 Participants
Notifications
Due Date
No due date set.
Dependencies

No dependencies set.

Reference: langchain-ai/langgraph#4919
Delete Branch "%!s()"

Deleting a branch is permanent. Although the deleted branch may continue to exist for a short time before it actually gets removed, it CANNOT be undone in most cases. Continue?