Reasoning and History Agent V2 Bug

General
1

If you create an Agent with Reasoning only, you get a warning that ‘add_history_to_context’ is True when the default is supposed to be False. If you then try an add ‘add_history_to_context=True’ you get an unexpected keyword argument.

This error implies that certain parameters should only be specified in specific situations… which currently is not defined in the documentation.

2

Hi @bills, Thank you for raising this. I’ve shared this with the team and we are looking into this.

3

Hey @bills ! Could you share more about how reasoning is used? via a reasoning model or using default reasoning=True.

Also could you share more on the Agent config to help us debug this. Thanks!

4

A very simple example:

    *self*.agent = get_agentic_agent(name=*self*.va_name,
                                   user_id=*self*.user_id,

                                   model_id=*self*.llm_model,

                                   model_location=*self*.llm_host,

                                   description=*self*.description,

                                   instructions=*self*.instructions,

                                   tools=*self*.tools,

                                   has_memory=False,

                                   has_storage=False,

                                   has_knowledge=False,

                                   has_reasoning=True,

                                   debug_mode=*self*.debug_mode,

                                   )

During initialization the following warning message is generated:

WARNING add_history_to_context is True, but no database has been assigned to the agent. History will not be added to the context.

The key thing to note is that has_memory, has_storage, has_knowledge are all set to False (i.e., not activated)… only reasoning. The agent routine that is being called is simply setting up the various values, variables, … storage areas depending on what things are being requested. All of the other states are working…

Also, how do we filter out warning messages? Agent has a ‘debug_level’ variable, but we have not been able to find any documentation that describes how to use it and what impact it might have in different situations.

Thanks for addressing this issue.