I made MemorySync since I was getting bored of having to setup dedicated vector databases, embedding pipelines, and session storages every single time I wanted to make a LLM agent persistent.
How it works: you make a call to an endpoint to store a bit of context about a particular user, and then make calls to an endpoint to fetch his memories in a conversation that takes place later. Everything is taken care of behind the scenes, so you don't have to worry about the infrastructure.
Limitations currently: it's deployed in one US region and isn't self-hostable, although I wouldn't mind if someone asked for it.
What do you think of the API interface? What must-have functionality should be included in an agent's state management?
Hey HN, creator here!
I made MemorySync since I was getting bored of having to setup dedicated vector databases, embedding pipelines, and session storages every single time I wanted to make a LLM agent persistent.
How it works: you make a call to an endpoint to store a bit of context about a particular user, and then make calls to an endpoint to fetch his memories in a conversation that takes place later. Everything is taken care of behind the scenes, so you don't have to worry about the infrastructure.
Limitations currently: it's deployed in one US region and isn't self-hostable, although I wouldn't mind if someone asked for it. What do you think of the API interface? What must-have functionality should be included in an agent's state management?
[dead]
[dead]