Skip to content

Improve memory weighting/chunking at scale for shared memory system #3

Description

@CryptoJym

Raised by: James Brady (in discussion with Bo Wang)

Context: During the Open Source Strategy Meeting, James raised the ongoing challenge of properly weighting and evaluating memories at scale in Borg's shared memory system (mem0 + Graphiti). Bo Wang confirmed this is one of the two biggest open problems: how to weight, chunk, and decide what's most relevant/important to save.

Discussion: James has trained several 4B models using LoRA adapters to help pull context from Codex/Claude threads into memory, but the weighting/evaluation-at-scale problem remains unsolved.

Acceptance criteria (implied):

  • Define a scoring/weighting approach for candidate memories.
  • Define chunking strategy for large threads.
  • Validate against real usage (multiple agents/threads) before wider rollout.

Created automatically from Otter.ai meeting: Open Source Strategy Meeting (2026-09-15)

Activity

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type

    Projects

    No projects

      Milestone

      No milestone

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions