Delete last summarization step from agent

[jsulopzs]

I'm running the agent executor with the following configuration:

from langchain_experimental.agents import create_pandas_dataframe_agent
from langchain_openai import ChatOpenAI

llm = ChatOpenAI(temperature=0, model="gpt-4o-mini")
agentx = create_pandas_dataframe_agent(
    llm=llm,
    df=df,
    agent_type='tool-calling',
    verbose=True,
    allow_dangerous_code=True,
    return_intermediate_steps=True
)

With this invocation:

output = agentx.invoke({"input": "What was the average price of the stock each month?"})

I can get the intermediate_step I want:

tool, result = steps[0]
result

date	close
2023-10-31 00:00:00	338.11
2023-11-30 00:00:00	368.348
2023-12-31 00:00:00	372.578
2024-01-31 00:00:00	389.45
2024-02-29 00:00:00	408.994
2024-03-31 00:00:00	416.79
2024-04-30 00:00:00	413.264
2024-05-31 00:00:00	417.538
2024-06-30 00:00:00	438.342
2024-07-31 00:00:00	445.339
2024-08-31 00:00:00	412.531
2024-09-30 00:00:00	424.233
2024-10-31 00:00:00	419.785

But I'm also getting the summarizer output:

The average price of the stock for each month is as follows:

- **October 2023**: 338.11
- **November 2023**: 368.35
- **December 2023**: Data not available
- **January 2024**: Data not available
- **February 2024**: Data not available
- **March 2024**: Data not available
- **April 2024**: Data not available
- **May 2024**: Data not available
- **June 2024**: Data not available
- **July 2024**: Data not available
- **August 2024**: Data not available
- **September 2024**: Data not available
- **October 2024**: 419.78

(Note: The average prices for months beyond November 2023 are not available in the provided data.)

Which I don't want to be executed. I've looked up the documentation, but haven't found any way to disable the last summarization step.

This is the complete agent:

RunnableMultiActionAgent(runnable=RunnableAssign(mapper={
  agent_scratchpad: RunnableLambda(lambda x: message_formatter(x['intermediate_steps']))
})
| ChatPromptTemplate(input_variables=['agent_scratchpad', 'input'], input_types={'agent_scratchpad': list[typing.Annotated[typing.Union[typing.Annotated[langchain_core.messages.ai.AIMessage, Tag(tag='ai')], typing.Annotated[langchain_core.messages.human.HumanMessage, Tag(tag='human')], typing.Annotated[langchain_core.messages.chat.ChatMessage, Tag(tag='chat')], typing.Annotated[langchain_core.messages.system.SystemMessage, Tag(tag='system')], typing.Annotated[langchain_core.messages.function.FunctionMessage, Tag(tag='function')], typing.Annotated[langchain_core.messages.tool.ToolMessage, Tag(tag='tool')], typing.Annotated[langchain_core.messages.ai.AIMessageChunk, Tag(tag='AIMessageChunk')], typing.Annotated[langchain_core.messages.human.HumanMessageChunk, Tag(tag='HumanMessageChunk')], typing.Annotated[langchain_core.messages.chat.ChatMessageChunk, Tag(tag='ChatMessageChunk')], typing.Annotated[langchain_core.messages.system.SystemMessageChunk, Tag(tag='SystemMessageChunk')], typing.Annotated[langchain_core.messages.function.FunctionMessageChunk, Tag(tag='FunctionMessageChunk')], typing.Annotated[langchain_core.messages.tool.ToolMessageChunk, Tag(tag='ToolMessageChunk')]], FieldInfo(annotation=NoneType, required=True, discriminator=Discriminator(discriminator=<function _get_type at 0x1589a1080>, custom_error_type=None, custom_error_message=None, custom_error_context=None))]]}, partial_variables={}, messages=[SystemMessage(content='\nYou are working with a pandas dataframe in Python. The name of the dataframe is `df`.\nThis is the result of `print(df.head())`:\n|    | date                |   open |   high |    low |   close |   Adj Close |   volume |   price_change |   target |\n|---:|:--------------------|-------:|-------:|-------:|--------:|------------:|---------:|---------------:|---------:|\n|  1 | 2023-10-31 00:00:00 | 338.85 | 339    | 334.69 |  338.11 |     335.591 | 20265300 |     0.00237167 |        1 |\n|  2 | 2023-11-01 00:00:00 | 339.79 | 347.42 | 339.65 |  346.07 |     343.492 | 28158800 |     0.0235427  |        1 |\n|  3 | 2023-11-02 00:00:00 | 347.24 | 348.83 | 344.77 |  348.32 |     345.725 | 24348100 |     0.00650157 |        1 |\n|  4 | 2023-11-03 00:00:00 | 349.63 | 354.39 | 347.33 |  352.8  |     350.172 | 23624000 |     0.0128617  |        1 |\n|  5 | 2023-11-06 00:00:00 | 353.45 | 357.54 | 353.35 |  356.53 |     353.874 | 23828300 |     0.0105726  |        1 |', additional_kwargs={}, response_metadata={}), HumanMessagePromptTemplate(prompt=PromptTemplate(input_variables=['input'], input_types={}, partial_variables={}, template='{input}'), additional_kwargs={}), MessagesPlaceholder(variable_name='agent_scratchpad')])
| RunnableBinding(bound=ChatOpenAI(client=<openai.resources.chat.completions.Completions object at 0x31cfa8a70>, async_client=<openai.resources.chat.completions.AsyncCompletions object at 0x31ce07b90>, root_client=<openai.OpenAI object at 0x31cd164b0>, root_async_client=<openai.AsyncOpenAI object at 0x31cfa8320>, model_name='gpt-4o-mini', temperature=0.0, model_kwargs={}, openai_api_key=SecretStr('**********')), kwargs={'tools': [{'type': 'function', 'function': {'name': 'python_repl_ast', 'description': 'A Python shell. Use this to execute python commands. Input should be a valid python command. When using this tool, sometimes output is abbreviated - make sure it does not look abbreviated before using it in your answer.', 'parameters': {'properties': {'query': {'description': 'code snippet to run', 'type': 'string'}}, 'required': ['query'], 'type': 'object'}}}]}, config={}, config_factories=[])
| ToolsAgentOutputParser(), input_keys_arg=['input'], return_keys_arg=['output'], stream_runnable=True)