Memory issue

[gwc4github]

I am trying to add memory to a chain. I want to use conversation summary buffer because the generated information is bound to get large.

Here is my configuration:

If I don't add the prompt I get this error: `UnboundLocalError cannot access local variable 'response' where it is not associated with a value'

When I add the prompt as shown I get:

ValidationError 1 validation error for ConversationSummaryBufferMemory root Got unexpected prompt input variables. The prompt expects ['history', 'input'], but it should have {'new_lines', 'summary'}. (type=value_error)

You can see that the config has different field names than 'new_lines' and 'summary':

So then I tried the following prompt: And was back to the error: UnboundLocalError cannot access local variable 'response' where it is not associated with a value

The documentation seems to expect dict input somewhere so I am not sure how to use that info: https://github.com/kreneskyp/ix/blob/master/docs/chains/memory.rst

[gwc4github]

I noticed that the Pirate Chain uses the same memory buffer so I changed my settings to the same. I also took out the prompt since pirate didn't use one. This removed the error but the memory doesn't work. It is more forgetful than my cat.

[kreneskyp]

Here are related docs for adding memory to the OpenAIFunctionsAgent. https://python.langchain.com/docs/modules/agents/how_to/add_memory_openai_functions

The example in the docs use a MessagesPlaceholder to add the memory variable to the agent's template. Need to implement something similar here.

One possible option is updating the agent loader to add a MessagesPlaceholder to agent_kwargs when a memory class is connected and there isn't a custom prompt.