RFC-038: Domain of Concern Registry — AIGP Dialects as First-Class Artifacts — 4. Dialect Structure

AIGP Specification › RFC-038: Domain of Concern Registry — AIGP Dialects as First-Class Artifacts › 4. Dialect Structure

← 3. Core Definitions · Section index · 5. Versioning Model →

4. Dialect Structure

A dialect is a structured bundle with the following layers:

dialect/
├── MANIFEST.json          # Identity, version, dependencies, compatibility
├── domain.json            # Domain of Concern declaration (RFC-034 §14–15)
├── model/                 # Mediation Observation Model (RFC-035)
│   ├── construct.json     # Layer 1: Construct declarations
│   ├── evidence.json      # Layer 2: Evidence admissibility rules
│   ├── measurement.json   # Layer 3: Variable declarations (full §10 schema)
│   └── calculation.json   # Layer 4: Calculation interface contract
├── observers.json         # Observer accreditation requirements (RFC-037)
├── profile/               # Default Local Profile
│   ├── thresholds.json    # Per-variable, per-gap threshold defaults
│   ├── hard_stops.json    # Circuit-break conditions
│   ├── severity.json      # Severity mapping declarations
│   └── escalation.json    # Escalation path defaults
├── CHANGELOG.md           # Version history with semantic change descriptions
└── VALIDATION.json        # Self-test: structural completeness assertions

4.1 MANIFEST.json

The manifest is the dialect’s identity and dependency declaration:

{
  "dialect_id": "autonomous_systems",
  "version": "2.1.0",
  "name": "Autonomous Systems",
  "description": "Observation apparatus for AI systems that recommend, select, or execute actions within bounded operational envelopes.",
  "series": "Mediated Intelligence Quality and Concern Evaluation",
  "publisher": "kanjani-ai-research",
  "published_at": "2026-07-01T00:00:00Z",
  "license": "AIGP-Dialect-1.0",

  "dependencies": {
    "aigp_core": ">=1.0.0",
    "rfc_034": ">=1.0.0",
    "rfc_035": ">=1.0.0",
    "rfc_036": ">=1.0.0",
    "rfc_037": ">=1.0.0"
  },

  "concern_classes": [
    "unauthorized_autonomous_action",
    "out_of_envelope_behavior",
    "high_consequence_action_under_uncertainty",
    "insufficient_human_control"
  ],

  "variable_count": 10,
  "observer_modes": ["machine", "human", "hybrid"],
  "minimum_admissibility_floor": 2,

  "compatibility": {
    "backwards_compatible_with": ["2.0.0", "2.0.1"],
    "breaking_changes_from": "1.x"
  },

  "signatures": {
    "manifest_hash": "sha256:...",
    "bundle_hash": "sha256:...",
    "publisher_signature": "..."
  }
}

---

---

← 3. Core Definitions · Section index · 5. Versioning Model →