🐛 Bug Report: TypeError: unsupported operand type(s) for +=: 'async_generator_asend' and 'async_generator_asend'

Which component is this bug for?

Langchain Instrumentation

📜 Description

When using LangChain instrumentation on an AgentExecutor, the error specified in the title occurs.

👟 Reproduction steps

1. Add LangChain Instrumentation to application, either manually or through automated instrumentation.
2. Make a call to any LangChain AgentExecutor, e.g. in the following LangServe context for a simple chat app:

agent = create_tool_calling_agent(model, tools, agent_prompt)

agent_executor = AgentExecutor(
    agent=agent,
    tools=tools,
)

add_routes(
    app,
    agent_executor.with_types(input_type=Input, output_type=Output),
    path="/chat",
)

3. By default, stream_runnable is set to True, which then throws the error in the title. Setting stream_runnable to False allows the agent executor to run without issue.

👍 Expected behavior

No error - AgentExecutor chain should complete without issue, traces should be available at destination specified in instrumentation.

👎 Actual Behavior with Screenshots

chat-app-langchain  |   File "/usr/local/lib/python3.11/site-packages/langchain/agents/agent.py", line 560, in aplan
chat-app-langchain  |     final_output += chunk
chat-app-langchain  | TypeError: unsupported operand type(s) for +=: 'async_generator_asend' and 'async_generator_asend'

Which links to the following LangChain agent code: https://github.com/langchain-ai/langchain/blob/master/libs/langchain/langchain/agents/agent.py#L560C18-L561C14

🤖 Python Version

Tried with both 3.11 and 3.12

📃 Provide any additional context for the Bug.

Tried with opentelemetry-instrumentation==0.46b0 and both langchain==0.2.1 and langchain==0.2.2.
Exception is thrown across multiple agent types.

👀 Have you spent some time to check if this bug has been raised before?

• I checked and didn't find similar issue

Are you willing to submit PR?

Yes I am willing to submit a PR!

Reproduce the failure with LangChain AgentExecutor and stream_runnable=True, then inspect the linked langchain/agents/agent.py entry point around the aplan failure and the LangChain instrumentation path. Done means the AgentExecutor completes with streaming enabled, no TypeError occurs on Python 3.11 or 3.12, and traces remain available at the configured destination.