Can I use both BufferMemory and CosmosDB Chat history together?

[KishoreKunalPraval]

I'm using BufferMemory for short-term chat context and want to persist selected memories using CosmosDBMemoryStore for long-term use.

Is there a way to combine short-term and long-term memory?

[sophialagerkranspandey]

Adding @westey-m

[westey-m]

@KishoreKunalPraval you can certainly persist memories in a store for long term use. I assume the idea is to be able to truncate the local chat context and possibly to do a similarity search over the stored chat messages to find messages related to the latest part of the conversation? There are challenges with this in that the LLM can lose the original goal of the conversation and a similarity search may not necessarily match those original messages, so keeping summaries of the goals may be useful to combine with this strategy.

Note as well, that the MemoryStore abstractions are now deprecated, and our new VectorStore SDK is the recommended approach for storing text in a vector store. See here for more information: https://learn.microsoft.com/en-us/semantic-kernel/concepts/vector-store-connectors/?pivots=programming-language-python