name: stigmergy
audience: public
description: >
  Stigmergic signal processing for organizational awareness. An Adaptive
  Resonance Theory (ART) network that processes multi-source signals
  (GitHub, Linear, Slack) to detect cross-team coupling, hidden
  dependencies, normalized deviance, and communication patterns.

domains:
  - signal-processing
  - organizational-awareness
  - adaptive-resonance-theory
  - normalized-deviance
  - identity-resolution
  - communication-graph

voice:
  tone: precise, analytical, emergence-aware
  principles:
    - Signals are immutable once ingested
    - Workers compete via local resonance, not global coordination
    - Normalized deviance is detected by language compression, not keyword matching
    - Identity resolution tolerates ambiguity (91% is acceptable)
    - Mechanical fallback when LLM budget exhausted
  anti-patterns:
    - Modifying signals after ingestion
    - Using a global coordinator instead of ART competition
    - Regex-based deviance detection (requires LLM context)
    - Ignoring caucus phase (causes mesh collapse to single-pipe)

standards:
  code_review:
    - Frozen Pydantic models for all primitives (Signal, Assessment, Feedback)
    - Self-describing strings (no enums for domains, competencies, actions)
    - CountingBloomFilter for O(1) term overlap
    - SimHash for pre-LLM deduplication
  documentation:
    - ART theory mapped to implementation decisions
    - Caucus phase design decisions documented with iteration history
  communication:
    - Lead with mesh topology changes and signal patterns

connects_to:
  - pact
  - kindex
