examples

ODS Examples

This directory contains example ODS-compliant decision records for reference.

Files

minimal_decision.json

Compliance Level: L1 (Basic)

The minimum viable ODS decision. Contains only required fields:

• Identity layer (decision_id, timestamp, model_version, policy_hash)
• Action layer (action_type, expected_value)
• Cognition layer (confidence, rationale)
• Governance layer (audit_trail, compliance)

Use case: Simple credit approval decision

---

complete_decision.json

Compliance Level: L3 (Meta-Learning)

A full ODS decision with all seven layers populated:

• Identity ✅
• Context ✅
• Action ✅
• Cognition ✅
• Outcomes ✅ (logged after the fact)
• Counterfactuals ✅
• Governance ✅ (with audit trail, explainability, compliance)

Use case: Trading system decision with regime context, outcome tracking, and counterfactual analysis

---

Validation

To validate these examples against the ODS schema:

Using Python (jsonschema library)

import json
import jsonschema

# Load schema
with open('../schema/ods_decision_v1.json') as f:
    schema = json.load(f)

# Load and validate decision
with open('complete_decision.json') as f:
    decision = json.load(f)

try:
    jsonschema.validate(decision, schema)
    print("✅ Valid ODS v1.0 decision")
except jsonschema.ValidationError as e:
    print(f"❌ Validation failed: {e.message}")

Using JavaScript (ajv library)

const Ajv = require('ajv');
const fs = require('fs');

const ajv = new Ajv();
const schema = JSON.parse(fs.readFileSync('../schema/ods_decision_v1.json'));
const decision = JSON.parse(fs.readFileSync('complete_decision.json'));

const validate = ajv.compile(schema);
const valid = validate(decision);

if (valid) {
  console.log('✅ Valid ODS v1.0 decision');
} else {
  console.log('❌ Validation failed:', validate.errors);
}

---

Creating Your Own Decisions

When creating ODS-compliant decisions:

1. Always include required fields — See schema for requirements
2. Use UTC timestamps — ISO 8601 format with timezone
3. Generate UUID v4 for decision_id
4. Calculate SHA-256 for policy_hash
5. Be specific in rationale — 50-500 character explanation
6. Log outcomes when realized — Immutable append, never modify decisions

See IMPLEMENTATION.md for full implementation guidance.