I'm using OpenClaw and ClaudeCode integrations and i'm seeing the next problem
Here's my memory graph.
Most of separated documents are different sessions of claude code for the same project and codebase. Many memories are heavly duplicated and sessions/containers are fully separated even tho it's the same project and parts can be indexed/updated to have more proper relations and shared memories between containers + deduplication.
I guess the proper way for it is just to find most similar vectors with extremly high similarity rate etc and if they match create a relation between them or create one memory instead of both and have multiple relations. Would be a really nice thing to have. Could be done once a day or on new memory creation.