fix: fix parallel agent invocation by using create_react_agent with v1 version

[ohayolink]

Summary

This PR fixes parallel agent invocation in the supervisor architecture by changing the create_react_agent version from v2 to v1 when parallel_tool_calls=True. ref Issue #230

Problem

When parallel_tool_calls=True is enabled in the supervisor, parallel agent handoffs were failing due to how different versions of create_react_agent handle tool calls:
v1: Handles multiple tool calls in a way that allows ToolNode to combine Send commands properly
v2: Creates separate Send commands for each tool call individually, preventing proper command combination

Root Cause Analysis

1. ToolNode Command Combination: The ToolNode in LangGraph has special logic to combine multiple Send commands from parallel tool calls into a single parent command (see _combine_tool_outputs method in tool_node.py https://github.com/langchain-ai/langgraph/blob/main/libs/prebuilt/langgraph/prebuilt/tool_node.py#L403-L432)
2. v2 Behavior Prevents Combination: When using v2, create_react_agent generates individual Send commands for each tool call separately, which prevents the ToolNode from properly combining them for parallel execution (see https://github.com/langchain-ai/langgraph/blob/main/libs/prebuilt/langgraph/prebuilt/chat_agent_executor.py#L805-L809)
3. Handoff Tool Design: The handoff tools are specifically designed to work with this command combination mechanism. (see https://github.com/langchain-ai/langgraph-supervisor-py/blob/main/langgraph_supervisor/handoff.py#L106-L110)

[cluttered-cabinet]

@sydney-runkle do you know the right person to route this to?

[dy93]

ping

[naufal-fachri]

Dude, I have been debugging the source code for THISS, I thought there was an error in my code since I was having the same issue. UP for this !!