Integrate MemGPT

[torrinworx]

Feature request

Implement MemGPT for better memory management with local and hosted models: https://github.com/cpacker/MemGPT

Motivation

This can help automate the manually creation and retrieval of vector stores.

Your contribution

I'm not really familiar with the langchain repo at the moment, I could provide feedback on the implementation from a users perspective.

[HassanA01]

@torrinworx can you give an estimate on how long this would take to complete? Assuming I have a few people to work with and this will be our first time contributing to langchain

[josher19]

@HassanA01 I'd guess it would take about a week if we had 3 developers who are familiar with LLM's and strong python programming skills. I have time this week if you are still interested in working on this.

[akashAD98]

@hwchase17 any opinion on this? is memgpt & langchain both are differnt ? is there any plans to support the memgpt with langchian?

[stevenbaert]

Highly interested also. Would appreciate any update if implentatation is in pipeline(?)

[Coriolan-Bataille]

Closely watching as well, memgpt seems to be a very interesting architecture.

[chrishart0]

https://memgpt.readme.io/docs/python_client

MemGPT has a python client. Someone should be able to leverage this with LangChain's memory system. Don't think it would be too hard.

[brettforbes]

hey, this integration with MemGPT seems essential.

we need a rationale (explanation on how both libraries will work together) and an implementation. any ideas?

[kaijietti]

It seems that langchain is building its own LTM module. Check this https://blog.langchain.dev/langfriend/