DMG — Decision Moment Graph

The Reference Implementation of the Decision Moment Standard (DMS). Store decision moments, not raw chats, objects or data.

---

What is DMG?

DMG is the standard graph encoding for the DMS Standard.

It provides the schemas, SDKs, and validators to ensure your decisions are MERIT-compliant:

Principle	Verified By	Grade
📐 Measured	OUTCOME checks vs predictions	Platinum
🔍 Evidenced	MEMO options (≥3) & rationale	Bronze
🔙 Reversible	DOORS owner & rollback plan	Gold
👁️ Inspectable	MOMENT event hash chain	Silver
🔗 Traceable	TRACE source citations	Platinum

"DMS defines the rules. DMG stores the proof."

---

The DMG Lifecycle

The DMG Lifecycle

DMG implements the 7-Step Heptagon lifecycle defined by DMS:

FRAME → SPAR → GATE → COMMIT → ENACT → YIELD → GAUGE

---

Quick Start

1. Install

pip install decision-moment-graph
# For embeddings: pip install sentence-transformers

2. Run DMG Lifecycle

python dmg/cli/lifecycle.py "Should we migrate to microservices?" --dry-run

Output:

📍 Phase 1: FRAME, Structure question & options
📍 Phase 2: SPAR, Deliberate & Challenge
📍 Phase 3: GATE, Check Governance & Reversibility
📍 Phase 4: COMMIT, Sign & Hash (The "Capture")
📍 Phase 5: ENACT, Execute instructions
📍 Phase 6: YIELD, Measure results
📍 Cycle Complete: GAUGE, Calibrate confidence & learn

3. Use as Library

from decision_moment_graph import AgenticSPARAdapter
from agentic_kit.dispatchers import LoggingDispatcher

adapter = AgenticSPARAdapter()
dispatcher = LoggingDispatcher()

result = adapter.run_loop(
    spar_output,
    action_executor=dispatcher.execute,
    auto_execute=True
)

---

The DMG Lifecycle

The decision lifecycle maps directly to DMG kits and objects:

FRAME → SPAR → GATE → COMMIT → ENACT → YIELD → GAUGE

Step	Phase	Primitive	Kit/Module	Action
1	FRAME	MEMO	dmg/frame	Define context & options
2	SPAR	DISSENT	spar-kit	Deliberate & Challenge
3	GATE	RAMP	dmg/gate	Check Governance & Reversibility
4	COMMIT	MOMENT	dmg/core	Sign & Hash (The "Capture")
5	ENACT	—	—	Execute instructions
6	YIELD	OUTCOME	dmg/yield	Measure results
7	GAUGE	RETRO	dmg/gauge	Calibrate confidence & learn

SDK Modules

Module	Purpose
agentic_adapter.py	Core DMG Lifecycle adapter
spar_adapter.py	Converts SPAR to DMG
sparkit_runner.py	Embedded SPAR runner
swarm_coordinator.py	Parallel agent swarm orchestration
audit_export.py	Compliance reports

Note: Core primitives (Memory, Dispatchers, Hub, Reflection) are now imported from agentic-kit.

Repository Structure

decision-moment-graph/
├── dmg/                     # Python Package
│   ├── schema/              # JSON Schema definitions
│   ├── sdk/                 # Reference implementations
│   └── cli/                 # Validator CLI
├── conformance/             # Test fixtures + runner
└── docs/                    # Implementation Docs

Note: Normative specs have moved to decision-moment-standard/spec.

Agentic Integration (MCP)

DMG provides a Context Engine MCP Server for AI Agents:

• Tools:
  • validate_moment(path): Run MERIT validation on any file
  • create_moment_scaffold(title): Generate compliant JSON
• Resources: dmg://graph/active, dmg://moment/{id}
• Setup: Run pip install mcp and point your client to dmg/mcp/server.py.

New Modules (Jan 2026)

Module	Path	Purpose
GAUGE Emitter	src/gauge-emitter.js	Ecosystem event emission at phase completion
Decision Replay	src/replay.js	Counterfactual analysis and training data

// Decision replay example
const { captureSnapshot, replayDecision, REPLAY_MODES } = require('dmg/src/replay');

const snapshotId = captureSnapshot(decision);
const replay = replayDecision(snapshotId, {
    mode: REPLAY_MODES.COUNTERFACTUAL,
    modifications: { context: { budget: newBudget } }
});

---

Governance Gates

For consequential decisions (RAMP ≥ 3) marked Final/Approved:

✅ Scheduled outcome check required
✅ At least 1 dissent captured (or ZK-Dissent ref)
✅ Rollback owner named
✅ ≥2 reversal signals defined
✅ At least 1 prediction with confidence

---

The Open Protocol

DMG is the open standard toolkit. ARANGAM provides premium infrastructure.

What You Get	DMG (OSS Kit)	ARANGAM Platform
DMG Core Schemas	✅	✅
Validator CLI	✅	✅
Reference Libraries	✅	✅
Hosted MOMENT Ledger	❌	✅ Hash-chained audit trail
Team Collaboration	❌	✅ Approvals + workflows
Enterprise Governance	❌	✅ SSO + retention policies

The kit is free. The infrastructure is premium.

See ATTRIBUTION.md for attribution guidelines.

---

The SYNTHAI Ecosystem

Component	Role
Decision Moment Graph	The Standard for reversible, auditable decisions.
VAULT-KIT	The Protocol for privacy-first coordination.
agentic-kit	The Infrastructure for reliable agent systems.

Built by SYNTHAI, Decision Intelligence for the AI Era.