Design Patterns MCP Server

#!/usr/bin/env node /** * Generate Embeddings for Design Patterns using Strategy Pattern * Creates vector embeddings for all patterns in the database using best available strategy */ import { initializeDatabaseManager } from '../services/database-manager.js'; import { EmbeddingServiceAdapter } from '../adapters/embedding-service-adapter.js'; import { getAvailableEmbeddingStrategies } from '../factories/embedding-factory.js'; import { logger } from '../services/logger.js'; export async function main(): Promise<void> { try { logger.info('generate-embeddings', 'Starting embedding generation with strategy pattern...'); // Check available strategies logger.info('generate-embeddings', 'Checking available embedding strategies...'); const availableStrategies = await getAvailableEmbeddingStrategies(); for (const strategy of availableStrategies) { logger.info( 'generate-embeddings', `Strategy: ${strategy.name} (${strategy.model}) - Available: ${strategy.available}` ); } // Initialize database const dbConfig = { filename: './data/design-patterns.db', options: { readonly: false, fileMustExist: false, timeout: 5000, verbose: (message: string) => logger.debug('generate-embeddings', message), }, }; const dbManager = await initializeDatabaseManager(dbConfig); // Initialize embedding adapter with SAME strategy as SemanticSearch (simple-hash) // This ensures consistency between database embeddings and query embeddings const embeddingAdapter = new EmbeddingServiceAdapter({ cacheEnabled: false, // Disable cache for bulk generation batchSize: 20, // Process in larger batches retryAttempts: 3, retryDelay: 1000, preferredStrategy: 'simple-hash', // Match SemanticSearch strategy to fix MCP search issue fallbackToSimple: true, }); await embeddingAdapter.initialize(); const strategyInfo = embeddingAdapter.getStrategyInfo(); logger.info( 'generate-embeddings', `Using strategy: ${strategyInfo?.name} (${strategyInfo?.model})` ); // Get all patterns const patterns = dbManager.query<{ id: string; name: string; description: string }>( 'SELECT id, name, description FROM patterns' ); logger.info('generate-embeddings', `Found ${patterns.length} patterns to process`); // Prepare texts for batch processing const texts = patterns.map(pattern => `${pattern.name} ${pattern.description}`); const patternIds = patterns.map(pattern => pattern.id); // Generate embeddings using batch processing logger.info('generate-embeddings', 'Generating embeddings using strategy pattern...'); const embeddings = await embeddingAdapter.generateEmbeddings(texts); logger.info('generate-embeddings', `Generated ${embeddings.length} embeddings`); // Store embeddings in database logger.info('generate-embeddings', 'Storing embeddings...'); // Create embeddings table if it doesn't exist dbManager.execute(` CREATE TABLE IF NOT EXISTS pattern_embeddings ( pattern_id TEXT PRIMARY KEY, embedding TEXT NOT NULL, model TEXT NOT NULL, strategy TEXT NOT NULL, dimensions INTEGER NOT NULL, created_at DATETIME DEFAULT CURRENT_TIMESTAMP ) `); // Validate schema integrity (Fail-Fast Pattern) const schemaInfo = dbManager.query<{ name: string }>(`PRAGMA table_info(pattern_embeddings)`); const expectedColumns = new Set([ 'pattern_id', 'embedding', 'model', 'strategy', 'dimensions', 'created_at', ]); const actualColumns = new Set(schemaInfo.map(col => col.name)); if ( expectedColumns.size !== actualColumns.size || ![...expectedColumns].every(col => actualColumns.has(col)) ) { const missing = [...expectedColumns].filter(col => !actualColumns.has(col)); const extra = [...actualColumns].filter(col => !expectedColumns.has(col)); throw new Error( `Schema mismatch for pattern_embeddings table!\n` + `Expected columns: ${[...expectedColumns].join(', ')}\n` + `Actual columns: ${[...actualColumns].join(', ')}\n` + `Missing columns: ${missing.join(', ') || 'none'}\n` + `Extra columns: ${extra.join(', ') || 'none'}\n` + `This usually means migrations need to be re-run. Try: npm run migrate` ); } logger.info('generate-embeddings', '✓ Schema validation passed - all required columns present'); // Store embeddings with strategy information for (let i = 0; i < patterns.length; i++) { const patternId = patternIds[i]; const embedding = embeddings[i]; dbManager.execute( 'INSERT OR REPLACE INTO pattern_embeddings (pattern_id, embedding, model, strategy, dimensions) VALUES (?, ?, ?, ?, ?)', [ patternId, JSON.stringify(embedding), strategyInfo?.model ?? 'unknown', strategyInfo?.name ?? 'unknown', embedding.length, ] ); if ((i + 1) % 50 === 0) { logger.info('generate-embeddings', `Stored ${i + 1}/${patterns.length} embeddings`); } } logger.info( 'generate-embeddings', `Successfully generated and stored embeddings for ${embeddings.length} patterns` ); // Verify embeddings const storedCount = dbManager.queryOne<{ count: number }>( 'SELECT COUNT(*) as count FROM pattern_embeddings' ); const strategyStats = dbManager.query<{ strategy: string; model: string; count: number }>( 'SELECT strategy, model, COUNT(*) as count FROM pattern_embeddings GROUP BY strategy, model' ); logger.info( 'generate-embeddings', `Verification: ${storedCount?.count ?? 0} embeddings stored` ); for (const stat of strategyStats) { logger.info( 'generate-embeddings', ` ${stat.strategy} (${stat.model}): ${stat.count} embeddings` ); } // Close database await dbManager.close(); logger.info('generate-embeddings', 'Embedding generation completed successfully!'); } catch (error) { console.error('❌ Embedding generation failed:', error); throw error; } } // Run if executed directly if (import.meta.url === `file://${process.argv[1]}`) { main().catch(error => { console.error('Error:', error); process.exit(1); }); }