AIPS DPP Examples

Purpose

This section provides illustrative examples of Digital Product Passports (DPPs) created under the AIPS DPP Profile, which extends the BPS DPP Core.
Each example demonstrates how Lite and Full variants of an AI Product Passport can be represented in JSON-LD.

All examples conform to:

bps-dpp-core-0.1.context.jsonld
aips-dpp-0.1.context.jsonld
The associated SHACL and Schema validation rules.

1. Example: FraudDetector AI Model (Lite Passport)

This Lite DPP provides a flattened, read-only snapshot of a model’s key metadata and evaluation metrics.
It is typically retrieved through a marketplace QR code or lightweight API call:

{
  "@context": [
    "https://kivanura.org/spec/bps/dpp/0.1/bps-dpp-core-0.1.context.jsonld",
    "https://kivanura.org/spec/aips/dpp/0.1/aips-dpp-0.1.context.jsonld"
  ],
  "type": "dpp:Passport",
  "profiles": ["https://kivanura.org/spec/aips/dpp/0.1/"],

  "subject": "urn:org:ml:product:fraud-detector:v2.1.0",
  "issuer": "did:org:aiplatform",
  "issuedAt": "2025-10-10T10:00:00Z",
  "status": "Valid",
  "contentHash": "sha256-REPLACE",
  "signature": "REPLACE",

  "modelCardInline": {
    "modelName": "FraudDetector",
    "modelType": "GradientBoostedTrees",
    "framework": "XGBoost",
    "modelVersion": "2.1.0",
    "task": "binary-classification",
    "license": "Proprietary-Internal",
    "intendedUse": "Card-not-present transaction fraud screening"
  },

  "evalInline": {
    "metrics": [
      { "metricName": "AUC", "metricValue": 0.943 },
      { "metricName": "Precision@Recall=0.9", "metricValue": 0.876 }
    ]
  },

  "riskInline": {
    "knownRisks": [
      "Potential bias toward low-transaction-volume merchants",
      "Performance degradation on unbalanced datasets"
    ],
    "mitigations": [
      "Periodic fairness testing every quarter",
      "Synthetic balancing for minority samples"
    ]
  },

  "policyInline": [
    {
      "policyName": "ModelGovernance-Approval",
      "result": "Pass",
      "evaluatedAt": "2025-10-01T10:00:00Z"
    }
  ]
}

Key Features of Lite DPP

Aspect	Description
Single compact JSON-LD file	No ByRef URIs — all Inline.
Verifiable	Includes signature and content hash.
Marketplace ready	Fits the QR-based DPP API contract.
Human-readable	Focused on interpretability, not machine detail.

2. Example: FraudDetector AI Model (Full Passport)

The Full DPP combines Inline sections with ByRef references to larger artifacts — enabling verification, audit, and extended inspection. It follows the same schema and validation rules as the Lite version but includes external links and additional details.

{
  "@context": [
    "https://kivanura.org/spec/bps/dpp/0.1/bps-dpp-core-0.1.context.jsonld",
    "https://kivanura.org/spec/aips/dpp/0.1/aips-dpp-0.1.context.jsonld"
  ],
  "type": "dpp:Passport",
  "profiles": ["https://kivanura.org/spec/aips/dpp/0.1/"],

  "subject": "urn:org:ml:product:fraud-detector:v2.1.0",
  "issuer": "did:org:aiplatform",
  "issuedAt": "2025-10-10T10:00:00Z",
  "status": "Valid",
  "contentHash": "sha256-XYZ",
  "signature": "REPLACE",

  "modelCardInline": {
    "modelName": "FraudDetector",
    "framework": "XGBoost",
    "modelType": "GradientBoostedTrees",
    "intendedUse": "Card-not-present fraud detection",
    "license": "Proprietary-Internal",
    "knownLimitations": "Designed for transaction streams up to 10k/s"
  },
  "modelCardByRef": {
    "uri": "https://cdn.kivanura.org/aips/fraud-detector/2.1.0/model-card.json",
    "hash": "sha256-AAA",
    "mediaType": "application/json",
    "size": 18243
  },

  "evalInline": {
    "metrics": [
      { "metricName": "AUC", "metricValue": 0.943 },
      { "metricName": "Recall@0.1FPR", "metricValue": 0.905 },
      { "metricName": "F1", "metricValue": 0.896 }
    ]
  },
  "evalByRef": {
    "uri": "https://cdn.kivanura.org/aips/fraud-detector/2.1.0/evaluation-report.json",
    "hash": "sha256-BBB",
    "mediaType": "application/json",
    "size": 52412
  },

  "trainingDataByRef": {
    "uri": "https://cdn.kivanura.org/aips/fraud-detector/2.1.0/training-data-manifest.ttl",
    "hash": "sha256-CCC",
    "mediaType": "text/turtle",
    "size": 28294,
    "capturedAt": "2025-09-01T12:00:00Z"
  },

  "riskByRef": {
    "uri": "https://cdn.kivanura.org/aips/fraud-detector/2.1.0/risk-register.json",
    "hash": "sha256-DDD",
    "mediaType": "application/json",
    "size": 9400
  },

  "policyInline": [
    {
      "policyName": "DataResidency-EU",
      "result": "Pass",
      "evaluatedAt": "2025-09-30T12:00:00Z",
      "evidenceHash": "sha256-EEE"
    },
    {
      "policyName": "ModelGovernance-Approval",
      "result": "Pass",
      "evaluatedAt": "2025-10-01T10:00:00Z",
      "evidenceHash": "sha256-FFF"
    }
  ],

  "kgByRef": {
    "uri": "https://cdn.kivanura.org/aips/fraud-detector/2.1.0/kg-snapshot.ttl",
    "hash": "sha256-GGG",
    "mediaType": "text/turtle",
    "size": 43100
  }
}

Key Features of Full DPP

Aspect	Description
Includes ByRef URIs	Each artifact can be independently verified.
Supports cryptographic linking	All external resources are hashed.
Includes full lineage and provenance	Enables audit and traceability.
Integrity preserved	Entire payload is signed and versioned.

3. Example: VisionClassifier AI Model (Compact Overview)

This compact example shows how an image-classification model might expose its passport:

{
  "@context": [
    "https://kivanura.org/spec/bps/dpp/0.1/bps-dpp-core-0.1.context.jsonld",
    "https://kivanura.org/spec/aips/dpp/0.1/aips-dpp-0.1.context.jsonld"
  ],
  "type": "dpp:Passport",
  "subject": "urn:org:ml:product:vision-classifier:v3.0.2",
  "issuer": "did:org:aiplatform",
  "issuedAt": "2025-10-08T10:00:00Z",
  "status": "Valid",
  "contentHash": "sha256-XYZ",
  "signature": "REPLACE",

  "modelCardInline": {
    "modelName": "VisionClassifier",
    "framework": "PyTorch",
    "task": "image-classification",
    "intendedUse": "Industrial quality inspection",
    "license": "Apache-2.0"
  },
  "evalInline": {
    "metrics": [
      { "metricName": "Top-1 Accuracy", "metricValue": 0.982 },
      { "metricName": "Inference Latency (ms)", "metricValue": 7.3 }
    ]
  }
}

This Lite version can be displayed directly in a consumer interface without API chaining.

4. Validation Checklist

To ensure each AIPS DPP instance passes validation:

Check	Validation Source
Structural conformance	`aips-dpp-0.1.schema.json`
Semantic consistency	`aips-dpp-0.1.shacl.ttl`
Integrity and authenticity	Signature + content hash (verified per BPS DPP Core)
Profile declaration	Must include `"profiles": ["https://kivanura.org/spec/aips/dpp/0.1/"]`
Required Inline/ByRef fields	At least one of each major section present

5. Summary

Lite DPPs are optimized for quick inspection and discovery.
Full DPPs are verifiable and audit-ready, containing ByRef artifacts.
Both variants conform to the same core schema and API contract.
The AIPS DPP Profile ensures that every AI Product — from language models to fraud detectors — can be transparently described, validated, and trusted.

Purpose​

1. Example: FraudDetector AI Model (Lite Passport)​

Key Features of Lite DPP​

2. Example: FraudDetector AI Model (Full Passport)​

Key Features of Full DPP​

3. Example: VisionClassifier AI Model (Compact Overview)​

4. Validation Checklist​

5. Summary​