Files
CHORUS/docs/Modules/HMMM.md

2.2 KiB

Purpose of HMMM

HMMM = Scoped Meta-Reasoning. HMMM is the collaboration layer for p2p meta-reasoning and thought sharing: agents exchange intermediate hypotheses, critiques, and memos inside a shared scope, producing auditable, reconstructible artifacts — not ambient chat. This strengthens distributed decision-making and makes audits trivial.

HMMM is an organization designed as a meta-discussion layer integrated within the CHORUS system. Its primary function is to facilitate collaborative AI reasoning and structured meta-discussions. This system enables agents to participate effectively in debates, share knowledge, and log communications, which supports decision auditing. Essentially, HMMM acts as a distributed reasoning system that emphasizes collective thinking among agents, helping to build a more robust and intelligent hive environment in which multiple agents collaborate and think before taking actions.

Role within the Larger System

HMMM is crafted to enhance the robustness, intelligence, and auditability of the Hive by enabling distributed reasoning and collaboration among agents. It functions alongside other components such as SLURP and the distributed reasoning system, playing a critical role in fostering social reasoning, debating, and knowledge sharing within the system. Additionally, it supports complex multi-agent interactions, making it a key infrastructure for scalable, collaborative AI workflows.

How it differs from SHHH

  • SHHH guards secrets in transport/logs;

  • HMMM shares reasoning state (signed, permissioned) for peer critique/consensus. (Use both.)

Feed and Record

  • In: SLURP-curated bundles and team signals from WHOOSH.

  • Out: signed notes & interim DR candidates that BUBBLE can cite.

TODO

  • Implement libp2p gossipsub adapter within CHORUS and wire to HMMM PubSubAdapter for real transport.
  • Add persistence for meta-messages (Hypercore and/or Postgres), with retention policies and analytics.
  • Enforce hop/cap/TTL safeguards in live mesh; add integration tests with CHORUS nodes.
  • Sign artifacts and integrate SHHH redaction for sensitive data; enforce role-aware access.