MemGPT Client / SDK Javascript/Typescript client -> CLI Client?

[philiprhoades]

People,

While I have been waiting for a bug to be resolved I have been looking at other environments and the main thing that has annoyed me about most of the setups is that there is not a dead-simple CLI Client - Oogabooga and its kind are nice and necessary for some situations but quite frequently after installing something, I just want a quick chat with the LLM to see what it is like. Then I installed https://ollama.com/ and found it did exactly what I was looking for - a simple install and then the download of a model - which is immediately followed by a CLI prompt to start chatting to the just downloaded model.

This CLI interface seems like it shouldn't be too hard to do and it would be perfect for what I have been doing so far with MemGPT - has anyone thought about / proposed doing this already? - to me it seems like something that would have been done before thinking about WebUIs etc . .

[cpacker]

@philiprhoades what is the workflow you would have in mind with MemGPT?

Since MemGPT still requires some additional configuration of the LLM inference engine under the hood.

Maybe something like a faster quickstart that's also local? Like:

1. memgpt quickstart ollama --model hermes-pro (sets up MemGPT to use ollama as the backend, grabs the model hermes-pro and sets the context window + persona + user to a default)
2. memgpt run (starts the CLI)

Currently you can do a quickstart with the free hosted endpoint + OpenAI.