Open Data Governance Standard (ODGS) — LLM Bridge

AI-generated governance artefacts, deterministically enforced.

---

For engineers: See Quick Start below, or the guided walkthrough at metricprovenance.com/start.
For compliance and risk officers: The bridge produces governance artefacts — compiled rules, drift reports, S-Cert narratives — that enter the Sovereign Validation Engine as structured, schema-validated JSON.
For architectural assessment and certified packs: metricprovenance.com/#suite

---

---

What's New in v0.2.0

Change	Detail
Ed25519 offline verification	Licence JWTs signed by Metric Provenance; verified against embedded public key. Forged tokens are cryptographically rejected.
Tier gate	Community / Professional / Enterprise access from odgs-workspace.yaml. Air-gapped environments fully supported.
odgs-maturity integration	Governance maturity scoring delegates to the 8-pillar ISO/IEC 38505-1 & ISO/IEC 25012 engine when installed.

The European Data Governance Maturity Benchmark 2026 recorded an average maturity of 37.6% across 99 enterprises — a 62.4% gap against current regulatory expectation. The ODGS engine applies the same assessment methodology at the pipeline level.

---

What This Is

The ODGS LLM Bridge is a headless Python package that converts probabilistic LLM output into deterministic governance artefacts. Every LLM-generated artifact passes a JSON Schema Validation Gate before it enters the ODGS Sovereign Validation Engine — the engine never sees malformed data.

┌──────────────────────────────────────────────────────────────┐
│                      ODGS LLM Bridge                         │
│                                                              │
│  Regulation Text ──→ [Regulatory Compiler] ──→ ODGS Rules   │
│  Definition Files ──→ [Drift Watchdog]     ──→ Alerts       │
│  Rule Sets        ──→ [Conflict Detector]  ──→ Conflicts     │
│  S-Cert JSON      ──→ [Audit Narrator]     ──→ Narrative     │
│  Catalog Metadata ──→ [Binding Discoverer] ──→ Data Map      │
│                                                              │
│              ↓  All outputs pass through  ↓                  │
│           [JSON Schema Validation Gate]                      │
│              ↓  before entering          ↓                   │
│       [ODGS Runtime Enforcement Engine v6.1.0]                │
└──────────────────────────────────────────────────────────────┘

EU AI Act & High-Risk AI Systems

Organisations operating High-Risk AI Systems under EU AI Act Articles 10 and 12 require demonstrable, auditable data governance at the pipeline level. The bridge compiles regulatory text into machine-verifiable ODGS rules; each evaluation produces an S-Cert suitable for regulatory submission.

Certified Sovereign Packs and the S-Cert Registry are available through Metric Provenance certified implementation partners. For architectural assessment: metricprovenance.com/#suite.

---

Quick Start

# Core bridge — no LLM provider bundled (bring your own)
pip install odgs-llm-bridge

# With Ollama (recommended for air-gapped / sovereign deployments)
pip install odgs-llm-bridge[ollama]

# With all providers
pip install odgs-llm-bridge[all]

Python API

from odgs_llm import OdgsLlmBridge

# Auto-detects best available provider (Ollama → Google GenAI → OpenAI)
bridge = OdgsLlmBridge()

# 1. Compile regulation text → validated ODGS rules
rules = bridge.compile_regulation("""
    Article 10(3): The provider shall ensure that training,
    validation and testing data sets are relevant, sufficiently
    representative, and to the best extent possible, free of errors
    and complete in view of the intended purpose.
""")

# 2. Narrate an S-Cert for non-technical stakeholders
narrative = bridge.narrate_audit(scert_json, audience="executive")

# 3. Detect contradictions between regulatory rule sets
conflicts = bridge.detect_conflicts(all_rules)

# 4. Identify definitional drift across governance artefact versions
warnings = bridge.check_drift("./definitions/", threshold_days=90)

# 5. Auto-generate physical data map from catalog metadata
bindings = bridge.discover_bindings(snowflake_catalog)

CLI

odgs-llm compile regulation.txt -o rules.json
odgs-llm drift ./definitions/ --threshold 90
odgs-llm conflicts rules.json
odgs-llm narrate scert.json --audience legal -o report.md
odgs-llm discover catalog.json --metrics metrics.json -o bindings.json
odgs-llm health --provider ollama

---

Capabilities

Community (Free — no account required)

Capability	Description
Schema validation	Validate LLM-generated artefacts against ODGS JSON schemas
Governance scoring	8-pillar ISO/IEC 38505-1 & ISO/IEC 25012 maturity score (0–100) with gap analysis
Conformance check	Run ODGS conformance self-check against workspace definitions

Professional (Certified Pack Licence)

Capability	Description
compile_regulation	Convert legislative text (EU AI Act, DORA, GDPR) → ODGS rule JSON
check_drift	Detect semantic drift in governance definitions across versions
detect_conflicts	Identify contradictions and shadows between regulatory rule sets
narrate_audit	Convert S-Cert JSON → narrative for executive, legal, or technical audiences
discover_bindings	Auto-generate physical_data_map.json from Snowflake, Databricks, dbt catalogs

Enterprise (Certified Pack Licence + Partner Agreement)

Capability	Description
sync_catalog	Pull and ingest metadata from Databricks / Snowflake / Collibra
harvest_sovereign_rules	Extract and mint sovereign rules directly from data stores (Flint Bridge)

Enterprise Rollout & Certified Packs — see metricprovenance.com/#suite or read the Executive Brief.

---

Workspace Licence Configuration

For air-gapped and offline deployments, licences are issued as Ed25519-signed JWTs and placed in odgs-workspace.yaml:

license:
  key: "eyJ..."   # Ed25519-signed JWT issued by Metric Provenance
  tier: enterprise

The bridge verifies the JWT signature offline against the embedded public key — no network call required. Licence issuance is handled through metricprovenance.com/#suite.

---

Model Providers

Provider	Extra	Recommended for
Ollama (local)	pip install odgs-llm-bridge[ollama]	Full sovereignty — zero egress, air-gapped environments
Google GenAI	pip install odgs-llm-bridge[google]	Gemini cloud API
OpenAI-compatible	pip install odgs-llm-bridge[openai]	GPT-NL (TNO), Mistral, GPT-4o
LiteLLM	pip install odgs-llm-bridge[litellm]	Universal router (100+ providers)

Auto-detection priority: Ollama → Google GenAI → OpenAI → error.

Custom Provider

from odgs_llm.providers import ModelProvider, ModelResponse

class MyProvider(ModelProvider):
    name = "my-provider"

    def generate(self, system_prompt, user_message, **kwargs) -> ModelResponse:
        return ModelResponse(text="...", model="my-model", provider=self.name)

bridge = OdgsLlmBridge(provider=MyProvider())

---

Normative References

• ODGS v6.1.0 — Sovereign Validation Engine specification
• EU AI Act (2024/1689) — Articles 10, 12 (data governance obligations for high-risk AI)
• ISO/IEC 42001:2023 — AI Management System
• ISO/IEC 38505-1:2026 & ISO/IEC 25012:2008 — 8-pillar data governance & quality maturity framework
• GPT4NL (TNO) — Dutch sovereign LLM initiative

---

About ODGS

The Open Data Governance Standard (ODGS) is an open protocol for deterministic data governance enforcement. ODGS is on a path toward formal standardization.

• Protocol specification — pip install odgs
• MCP Server — pip install odgs-mcp-server
• Maturity engine — pip install odgs-maturity
• Research paper (SSRN 6205478)
• Metric Provenance

Licence

Apache 2.0 — see LICENSE.

The protocol engine and LLM bridge are open source. Certified Regulation Packs are issued under a commercial licence through Metric Provenance certified partners.