tasq/node_modules/@claude-flow/neural/examples/sona-usage.ts

319 lines
10 KiB
TypeScript

/**
* Example: Using SONA Integration with V3 Neural Module
*
* Demonstrates:
* - Creating a SONA learning engine
* - Learning from trajectories
* - Adapting behavior based on context
* - Performance monitoring (<0.05ms learning target)
*/
import {
createSONALearningEngine,
type Context,
type Trajectory,
type TrajectoryStep,
} from '../src/index.js';
import { getModeConfig } from '../src/sona-manager.js';
// =============================================================================
// Example 1: Basic SONA Learning
// =============================================================================
async function basicLearningExample() {
console.log('\n=== Example 1: Basic SONA Learning ===\n');
// Create SONA engine with balanced mode
const modeConfig = getModeConfig('balanced');
const sona = createSONALearningEngine('balanced', modeConfig);
console.log(`Created SONA engine (balanced mode)`);
console.log(`- LoRA Rank: ${modeConfig.loraRank}`);
console.log(`- Learning Rate: ${modeConfig.learningRate}`);
console.log(`- Quality Threshold: ${modeConfig.qualityThreshold}`);
// Create a sample trajectory
const trajectory: Trajectory = {
trajectoryId: 'traj-001',
context: 'Implement authentication middleware',
domain: 'code',
steps: [
{
stepId: 'step-1',
timestamp: Date.now(),
action: 'analyze requirements',
stateBefore: new Float32Array(768).fill(0.1),
stateAfter: new Float32Array(768).fill(0.2),
reward: 0.8,
},
{
stepId: 'step-2',
timestamp: Date.now() + 1000,
action: 'implement JWT validation',
stateBefore: new Float32Array(768).fill(0.2),
stateAfter: new Float32Array(768).fill(0.3),
reward: 0.9,
},
{
stepId: 'step-3',
timestamp: Date.now() + 2000,
action: 'write tests',
stateBefore: new Float32Array(768).fill(0.3),
stateAfter: new Float32Array(768).fill(0.4),
reward: 0.95,
},
],
qualityScore: 0.88,
isComplete: true,
startTime: Date.now(),
endTime: Date.now() + 3000,
};
// Learn from trajectory
await sona.learn(trajectory);
console.log(`\nLearned from trajectory:`);
console.log(`- Steps: ${trajectory.steps.length}`);
console.log(`- Quality: ${trajectory.qualityScore}`);
console.log(`- Learning time: ${sona.getLearningTime().toFixed(4)}ms`);
console.log(`- Target: <0.05ms ✓`);
// Get statistics
const stats = sona.getStats();
console.log(`\nEngine statistics:`);
console.log(`- Total trajectories: ${stats.totalTrajectories}`);
console.log(`- Patterns learned: ${stats.patternsLearned}`);
console.log(`- Avg quality: ${stats.avgQuality.toFixed(3)}`);
console.log(`- Enabled: ${stats.enabled}`);
}
// =============================================================================
// Example 2: Context Adaptation
// =============================================================================
async function adaptationExample() {
console.log('\n=== Example 2: Context Adaptation ===\n');
// Create SONA engine with real-time mode (fastest)
const modeConfig = getModeConfig('real-time');
const sona = createSONALearningEngine('real-time', modeConfig);
console.log(`Created SONA engine (real-time mode)`);
console.log(`- Max Latency: ${modeConfig.maxLatencyMs}ms`);
// First, learn from some trajectories to build patterns
const trajectories: Trajectory[] = [
{
trajectoryId: 'traj-auth-1',
context: 'Implement JWT authentication',
domain: 'code',
steps: createMockSteps(3, 0.9),
qualityScore: 0.9,
isComplete: true,
startTime: Date.now(),
},
{
trajectoryId: 'traj-auth-2',
context: 'Add OAuth2 flow',
domain: 'code',
steps: createMockSteps(4, 0.85),
qualityScore: 0.85,
isComplete: true,
startTime: Date.now(),
},
{
trajectoryId: 'traj-api-1',
context: 'Build REST API endpoint',
domain: 'code',
steps: createMockSteps(5, 0.8),
qualityScore: 0.8,
isComplete: true,
startTime: Date.now(),
},
];
// Learn from all trajectories
for (const traj of trajectories) {
await sona.learn(traj);
}
console.log(`\nLearned from ${trajectories.length} trajectories`);
// Now adapt to a new context
const context: Context = {
domain: 'code',
queryEmbedding: new Float32Array(768).fill(0.15),
metadata: {
task: 'Implement user session management',
},
};
const adapted = await sona.adapt(context);
console.log(`\nAdapted behavior:`);
console.log(`- Transformation applied: micro-LoRA`);
console.log(`- Similar patterns found: ${adapted.patterns.length}`);
console.log(`- Suggested route: ${adapted.suggestedRoute}`);
console.log(`- Confidence: ${adapted.confidence.toFixed(3)}`);
console.log(`- Adaptation time: ${sona.getAdaptationTime().toFixed(4)}ms`);
if (adapted.patterns.length > 0) {
console.log(`\nTop pattern:`);
console.log(`- Type: ${adapted.patterns[0].patternType}`);
console.log(`- Quality: ${adapted.patterns[0].avgQuality.toFixed(3)}`);
console.log(`- Cluster size: ${adapted.patterns[0].clusterSize}`);
}
}
// =============================================================================
// Example 3: Pattern Discovery
// =============================================================================
async function patternDiscoveryExample() {
console.log('\n=== Example 3: Pattern Discovery ===\n');
const modeConfig = getModeConfig('research');
const sona = createSONALearningEngine('research', modeConfig);
console.log(`Created SONA engine (research mode)`);
console.log(`- Pattern Clusters: ${modeConfig.patternClusters}`);
// Learn from multiple high-quality trajectories
const domains: Array<'code' | 'math' | 'reasoning'> = ['code', 'math', 'reasoning'];
for (let i = 0; i < 10; i++) {
const domain = domains[i % domains.length];
const trajectory: Trajectory = {
trajectoryId: `traj-${domain}-${i}`,
context: `${domain} task ${i}`,
domain,
steps: createMockSteps(3 + (i % 3), 0.7 + Math.random() * 0.3),
qualityScore: 0.7 + Math.random() * 0.3,
isComplete: true,
startTime: Date.now(),
};
await sona.learn(trajectory);
}
console.log(`\nLearned from 10 trajectories across ${domains.length} domains`);
// Force a learning cycle to cluster patterns
const learnStatus = sona.forceLearning();
console.log(`\nForced learning cycle: ${learnStatus}`);
// Find patterns for each domain
for (const domain of domains) {
const queryEmbedding = new Float32Array(768).fill(0.1 + Math.random() * 0.1);
const patterns = sona.findPatterns(queryEmbedding, 3);
console.log(`\n${domain.toUpperCase()} patterns:`);
patterns.forEach((pattern, idx) => {
console.log(` ${idx + 1}. Quality: ${pattern.avgQuality.toFixed(3)}, Cluster: ${pattern.clusterSize} trajectories`);
});
}
const stats = sona.getStats();
console.log(`\nFinal statistics:`);
console.log(`- Total trajectories: ${stats.totalTrajectories}`);
console.log(`- Patterns learned: ${stats.patternsLearned}`);
console.log(`- Avg quality: ${stats.avgQuality.toFixed(3)}`);
}
// =============================================================================
// Example 4: Performance Monitoring
// =============================================================================
async function performanceMonitoringExample() {
console.log('\n=== Example 4: Performance Monitoring ===\n');
const modes: Array<'real-time' | 'balanced' | 'edge'> = ['real-time', 'balanced', 'edge'];
for (const mode of modes) {
const modeConfig = getModeConfig(mode);
const sona = createSONALearningEngine(mode, modeConfig);
// Benchmark learning
const trajectory: Trajectory = {
trajectoryId: `perf-${mode}`,
context: 'Performance test',
domain: 'general',
steps: createMockSteps(5, 0.8),
qualityScore: 0.8,
isComplete: true,
startTime: Date.now(),
};
const iterations = 100;
const times: number[] = [];
for (let i = 0; i < iterations; i++) {
await sona.learn(trajectory);
times.push(sona.getLearningTime());
}
const avgTime = times.reduce((sum, t) => sum + t, 0) / times.length;
const minTime = Math.min(...times);
const maxTime = Math.max(...times);
const meetsTarget = avgTime < 0.05;
console.log(`\n${mode.toUpperCase()} mode (${iterations} iterations):`);
console.log(`- Avg learning time: ${avgTime.toFixed(4)}ms ${meetsTarget ? '✓' : '✗'}`);
console.log(`- Min time: ${minTime.toFixed(4)}ms`);
console.log(`- Max time: ${maxTime.toFixed(4)}ms`);
console.log(`- Target: <0.05ms`);
}
}
// =============================================================================
// Helper Functions
// =============================================================================
function createMockSteps(count: number, avgReward: number): TrajectoryStep[] {
const steps: TrajectoryStep[] = [];
const baseTime = Date.now();
for (let i = 0; i < count; i++) {
steps.push({
stepId: `step-${i}`,
timestamp: baseTime + i * 1000,
action: `action-${i}`,
stateBefore: new Float32Array(768).fill(0.1 + i * 0.05),
stateAfter: new Float32Array(768).fill(0.15 + i * 0.05),
reward: avgReward + (Math.random() - 0.5) * 0.1,
});
}
return steps;
}
// =============================================================================
// Run All Examples
// =============================================================================
async function main() {
console.log('\n╔═══════════════════════════════════════════════════════════╗');
console.log('║ SONA Integration Examples - V3 Neural Module ║');
console.log('║ @ruvector/sona v0.1.5 ║');
console.log('╚═══════════════════════════════════════════════════════════╝');
try {
await basicLearningExample();
await adaptationExample();
await patternDiscoveryExample();
await performanceMonitoringExample();
console.log('\n✓ All examples completed successfully!\n');
} catch (error) {
console.error('\n✗ Error running examples:', error);
process.exit(1);
}
}
// Run if executed directly
if (import.meta.url === `file://${process.argv[1]}`) {
main();
}
export { main };