Memory Braid

[distant hull]

Hi there. I wanted to use Mem0 but didn't want to give up on memory-core usage of markdown files so I combined them both into this plugin.

So basically what it does is that it uses memory-core as it is and on top of that uses auto capture and auto recall from mem0, injecting auto recalls (if any) into the prompt.

https://www.npmjs.com/package/memory-braid
https://github.com/rodrigouroz/memory-braid

I'm the author and solely maintainer but I'd be happy to receive feedback in any form: issues, PRs, you name it.

[burnt ridge]

plugins.slots.memory = "memory-lancedb" 这个配置是干什么用的？

[woeful forge]

YOu should read this 🙂 https://grok.com/share/c2hhcmQtMg_08b71fd8-c248-4e80-8e2b-203f057d7ef9

(its good, really good - include on your blurb

[pliant valve]

@distant hullDili ba ang Claw adunay duha ka set sa memory logic? Core ug LanceDB, unsa may imoha?

[rigid sapphire]

how is it different from what I see it already doing with ~/.openclaw/memory/main.sqlite ? that seems to have all the memory files indexed.

[distant hull]

It doesn’t replace that, it augments it.
For user turns it does auto capture (stores embeddings and extract entities) and auto recall (injects relevant memories in the prompt.
For agents it exposes a tool for them to remember learnings for similar auto recall in the future.
Memories are distinguished between user memories and agent learnings.

[rigid sapphire]

does it support multiple agents with separate memories?

~I can't tell if the built-in one does... it's sitting at the 'top' not inside an agent's workspace.~ 🤔 OH... yes it does -- main is the id of my first agent. 🤦‍♀️ So when the second starts working I should get developer.sqlite in there.

[distant hull]

Right, memories are per workspace/agent and user