MCP Quoting System

/** * Embedding Service * * Generates vector embeddings using OpenAI's text-embedding-ada-002 model. * Used for semantic similarity search of RFPs and historical quotes. */ import OpenAI from 'openai'; export class EmbeddingService { private client: OpenAI; private static instance: EmbeddingService; private enabled: boolean; private constructor() { const apiKey = process.env.OPENAI_API_KEY; this.enabled = !!(apiKey && apiKey !== 'sk-your-openai-api-key-here'); if (this.enabled) { this.client = new OpenAI({ apiKey }); console.log('✓ OpenAI Embedding Service initialized'); } else { console.log('⚠ OpenAI Embedding Service disabled (no API key)'); } } public static getInstance(): EmbeddingService { if (!EmbeddingService.instance) { EmbeddingService.instance = new EmbeddingService(); } return EmbeddingService.instance; } public isEnabled(): boolean { return this.enabled; } /** * Generate embedding for a text string * Returns a 1536-dimensional vector */ async generateEmbedding(text: string): Promise<number[]> { if (!this.enabled) { throw new Error('Embedding service is not enabled. Please set OPENAI_API_KEY.'); } try { const response = await this.client.embeddings.create({ model: 'text-embedding-ada-002', input: text.substring(0, 8000) // API limit is 8191 tokens, ~8000 chars to be safe }); return response.data[0].embedding; } catch (error: any) { console.error('Error generating embedding:', error.message); throw new Error(`Failed to generate embedding: ${error.message}`); } } /** * Generate embeddings for multiple texts in batch * More efficient than calling generateEmbedding multiple times */ async generateEmbeddingsBatch(texts: string[]): Promise<number[][]> { if (!this.enabled) { throw new Error('Embedding service is not enabled. Please set OPENAI_API_KEY.'); } if (texts.length === 0) { return []; } try { // Truncate long texts const truncatedTexts = texts.map(t => t.substring(0, 8000)); const response = await this.client.embeddings.create({ model: 'text-embedding-ada-002', input: truncatedTexts }); return response.data.map(d => d.embedding); } catch (error: any) { console.error('Error generating embeddings batch:', error.message); throw new Error(`Failed to generate embeddings: ${error.message}`); } } /** * Create a text representation of an RFP for embedding * Combines key fields into a searchable string */ createRfpText(rfp: { rawText?: string; material?: string; processes?: string[]; qty?: number; tolerances?: string; finish?: string; notes?: string; }): string { const parts: string[] = []; if (rfp.rawText) { parts.push(rfp.rawText); } if (rfp.material) { parts.push(`Material: ${rfp.material}`); } if (rfp.processes && rfp.processes.length > 0) { parts.push(`Processes: ${rfp.processes.join(', ')}`); } if (rfp.qty) { parts.push(`Quantity: ${rfp.qty}`); } if (rfp.tolerances) { parts.push(`Tolerances: ${rfp.tolerances}`); } if (rfp.finish) { parts.push(`Finish: ${rfp.finish}`); } if (rfp.notes) { parts.push(`Notes: ${rfp.notes}`); } return parts.join(' | '); } /** * Calculate cosine similarity between two embeddings * Returns a value between -1 and 1 (typically 0 to 1 for our use case) */ cosineSimilarity(embedding1: number[], embedding2: number[]): number { if (embedding1.length !== embedding2.length) { throw new Error('Embeddings must have the same length'); } let dotProduct = 0; let norm1 = 0; let norm2 = 0; for (let i = 0; i < embedding1.length; i++) { dotProduct += embedding1[i] * embedding2[i]; norm1 += embedding1[i] * embedding1[i]; norm2 += embedding2[i] * embedding2[i]; } const denominator = Math.sqrt(norm1) * Math.sqrt(norm2); if (denominator === 0) { return 0; } return dotProduct / denominator; } } export const embeddingService = EmbeddingService.getInstance();