Long term memory

[bezlashkari]

Hi,

So I'm starting to play around with this amazing technology and have a fundamental question. Say I would like to carry out some market research on x topic on a weekly basis. I'd like it to retain and reuse the knowledge it gains week by week so it becomes an expert on x topic over time.

I would have several different agents becoming experts on different topics. How would I achieve this? Is it tied to the role? I understand that I should use a vector database of sorts in order to store the data. But the bit that I don't understand is how to differentiate between the different agents. I guess my questions are:

1. How do I define a separation in AutoGPT between areas of research?
2. How can I on subsequent runs tell Auto-GPT to build on it's previous knowledge for x-topic
3. How can I keep topics separate?

Thanks,
Bez