The Concordance Engine v0.1.1 (Tier B): Reference Implementation of the Logotic Technique Catalogue
Authors/Creators
Description
Tier B reference implementation of the Logotic Technique Catalogue (DOI: 10.5281/zenodo.19831619). A complete reference architecture with one runnable loop, not yet a production service.
The Concordance Engine implements reusable coordination patterns first identified in adversarial AI-agent architectures, then reclassified for cooperative production, verification, and defensive provenance work within the Crimson Hexagonal Archive's witness-teaming methodology.
In plain terms: Concordance Engine is a small reference system for running structured AI-assisted production loops where critique, revision, verification, human override, and provenance are treated as first-class artefacts. The Tier B release proves the pattern with one runnable Critique Loop and ships the surrounding prompts, schemas, and templates needed to extend the same pattern into future operatives.
Bundle contents (75 files):
- Runnable Python package (
concordance_engine/) with the Critique Loop implementation, type definitions, MANUS authority interface, Anthropic API substrate wrapper, and a mock substrate enabling zero-secret demonstration - Five agent system prompts (Critic, Reviser, Verifier, Chartering Agent, Archivist) at calibrated register
- Eight skill specifications (four cooperative, four defensive) with frontmatter-loadable structure per the catalogue's V.9 / VI.9 patterns
- Five middleware specifications (Skills, Citation, Provenance, Verification, Collision-Detection)
- Aperture Atlas Cypher schema deployable to Neo4j 5.x, with three standing queries (ghost basins, competitor responses, outbound references) and seed JSON
- Four templates (Zenodo deposit, license header, hex coordinate, DISC-NNN record)
- Project governance documents: Charter, Plan-State, Citation Plan, Construction Plan
- Sample draft and captured Critique Loop session
- Smoke tests: 12 invariants verified without API calls
What runs: The Critique Loop (catalogue §V.12) executes end-to-end against the Anthropic API. A mock-substrate mode enables full demonstration without API keys: python -m concordance_engine.demo_critique_loop --mock.
What is specified but deferred to Tier C: Charter Generator, Verifier, Basin Hardening Cycle, Term Collision Audit, Profile Integrity Guardian, Depth-Proof Validator. These have complete agent prompts and skill specifications; runnable Python wrappers await infrastructure (retrieval-layer probing, Zenodo deposit-creation integration, live Neo4j instance).
Honest scope note: The runnable Tier B loop uses Anthropic as its only live substrate. Multi-provider verification is architecturally anticipated (config/substrates.yaml exposes the seam) but not implemented in this release. The catalogue's two-source verification protocol requires substrate independence; Tier B's Critique Loop demonstrates the loop pattern, not the cross-substrate verification.
Provenance integrity: Every revision in the Critique Loop is recorded with SHA-256 hashes of pre- and post-revision drafts, forming a tamper-evident chain auditable from the deposit's workspace/deposits/critique_session_*.json.
License: CC BY 4.0 for documentation, methodology, agent prompts, and operative documents. MIT for runnable Python source code. The dual structure supports both academic citation (CC BY) and developer adoption (MIT).
Document classification: EA-CE-01 v0.1.1
Hex: 06.LOG.CONCORDANCE.ENGINE.01
Catalogue companion: The Logotic Technique Catalogue (DOI: 10.5281/zenodo.19831619)
∮ = 1
No comments:
Post a Comment