@openclaw/governance-runtime

IGF-compliant AI agent governance runtime with execution engine, federated learning, and policy enforcement.

Features

• ✅ Governance Runtime - IGF v9.3 compliant governance
• ✅ Execution Engine - 14-state execution lifecycle
• ✅ Federated Learning - Privacy-preserving distributed learning
• ✅ Conflict Resolution - Auto-resolve model conflicts
• ✅ Human Approval - Multi-level review workflow
• ✅ Policy Enforcement - Runtime policy checks
• ✅ Risk Scoring - Real-time risk assessment

Installation

npm install @openclaw/governance-runtime

Quick Start

import { createServer } from '@openclaw/governance-runtime';

// Start the governance API server
const { runtime, executionEngine } = createServer({
  port: 4010,
  mode: 'normal'
});

// Check health
console.log(runtime.getHealthStatus());

Usage

Governance Runtime

import { GovernanceRuntime } from '@openclaw/governance-runtime';

const runtime = new GovernanceRuntime({
  mode: 'normal',
  riskThreshold: 0.7
});

await runtime.initialize();

// Check status
const status = runtime.getHealthStatus();
console.log(status);

Execution Engine

import { ExecutionEngine } from '@openclaw/governance-runtime';

const engine = new ExecutionEngine();

// Create execution
const execution = await engine.createExecution({
  type: 'skill',
  skill: 'ai-agent',
  payload: { action: 'process' },
  priority: 5
});

// Start execution
await engine.startExecution(execution.id);

// Complete execution
await engine.completeExecution(execution.id, { result: 'success' });

Federated Learning

import { FederatedLearningEngine } from '@openclaw/governance-runtime';

const fl = new FederatedLearningEngine({
  minClients: 3,
  rounds: 10
});

// Register clients
fl.registerClient('client-1', [1, 2, 3]);
fl.registerClient('client-2', [4, 5, 6]);
fl.registerClient('client-3', [7, 8, 9]);

// Train
await fl.train();

// Get global model
const model = fl.getGlobalModel();

Human Approval

import { HumanApprovalSystem } from '@openclaw/governance-runtime';

const approval = new HumanApprovalSystem({
  requiredApprovals: 2
});

// Create request
const request = approval.createRequest({
  type: 'model_deployment',
  title: 'Deploy new model',
  description: 'Deploy federated model to production',
  requester: 'system'
});

// Approve
approval.approve(request.id, 'admin-1', 'LGTM');
approval.approve(request.id, 'admin-2', 'Approved');

API Reference

GovernanceRuntime

| Method | Description | |--------|-------------| | initialize() | Initialize runtime | | getHealthStatus() | Get health status | | validateExecution() | Validate execution | | registerSkill() | Register a skill |

ExecutionEngine

| Method | Description | |--------|-------------| | createExecution() | Create execution | | startExecution() | Start execution | | completeExecution() | Complete execution | | failExecution() | Fail execution | | getExecution() | Get execution | | getAllExecutions() | Get all executions |

FederatedLearningEngine

| Method | Description | |--------|-------------| | registerClient() | Register client | | trainLocalModel() | Train local model | | aggregateModels() | Aggregate models | | runRound() | Run one round | | train() | Full training |

License

MIT