memento-mcp

import { describe, it, expect, beforeAll } from 'vitest'; import { EmbeddingServiceFactory } from '../EmbeddingServiceFactory.js'; import type { EmbeddingService } from '../EmbeddingService'; // This is a live test that connects to OpenAI and generates real embeddings // If OpenAI API key isn't available, it will fall back to the default provider describe('OpenAI Embedding Live Example', () => { let embeddingService: EmbeddingService; let hasApiKey = false; // Check for mock mode const useMockEmbeddings = process.env.MOCK_EMBEDDINGS === 'true'; console.log( `OpenAI API example tests WILL run (using API key: ${process.env.OPENAI_API_KEY !== undefined}, mock=${useMockEmbeddings})` ); beforeAll(() => { // Initialize the embedding services hasApiKey = process.env.OPENAI_API_KEY !== undefined; // Create a service with the OpenAI provider if available, otherwise fallback to default if (hasApiKey && !useMockEmbeddings) { embeddingService = EmbeddingServiceFactory.createService({ provider: 'openai', apiKey: process.env.OPENAI_API_KEY!, model: 'text-embedding-3-small', }); console.log('Using OpenAI embedding service with real API key'); } else { embeddingService = EmbeddingServiceFactory.createService({ provider: 'default', dimensions: 1536, // Match OpenAI dimensions for testing }); console.log('Using default embedding service fallback'); } }); it('generates embeddings from the API', async () => { // Generate an embedding for a text sample const text = 'The quick brown fox jumps over the lazy dog'; const embedding = await embeddingService.generateEmbedding(text); // Verify the embedding properties expect(Array.isArray(embedding)).toBe(true); expect(embedding.length).toBe(1536); // text-embedding-3-small has 1536 dimensions // Check that the embedding is normalized const magnitude = Math.sqrt(embedding.reduce((sum, val) => sum + val * val, 0)); expect(magnitude).toBeCloseTo(1.0, 1); }); it('generates embeddings for multiple texts in a batch', async () => { // Generate embeddings for multiple text samples const texts = [ 'Machine learning is a subset of artificial intelligence', 'Natural language processing is used for text understanding', 'Vector embeddings represent semantic meaning in a high-dimensional space', ]; const embeddings = await embeddingService.generateEmbeddings(texts); // Verify the embeddings properties expect(Array.isArray(embeddings)).toBe(true); expect(embeddings.length).toBe(3); // Check each embedding embeddings.forEach((embedding, index) => { expect(Array.isArray(embedding)).toBe(true); expect(embedding.length).toBe(1536); // Check that each embedding is normalized const magnitude = Math.sqrt(embedding.reduce((sum, val) => sum + val * val, 0)); expect(magnitude).toBeCloseTo(1.0, 1); }); // Calculate cosine similarities between embeddings to demonstrate semantic relationships const similarity12 = calculateCosineSimilarity(embeddings[0], embeddings[1]); const similarity13 = calculateCosineSimilarity(embeddings[0], embeddings[2]); const similarity23 = calculateCosineSimilarity(embeddings[1], embeddings[2]); // Check that all similarities are reasonable values // (actual semantic relationships can vary based on the embedding model's understanding) expect(similarity12).toBeGreaterThan(0.2); expect(similarity13).toBeGreaterThan(0.2); expect(similarity23).toBeGreaterThan(0.2); // Verify that the similarities are relatively close to each other (within 0.3) // This is a more robust test than assuming specific relative magnitudes expect(Math.abs(similarity12 - similarity13)).toBeLessThan(0.3); expect(Math.abs(similarity12 - similarity23)).toBeLessThan(0.3); expect(Math.abs(similarity13 - similarity23)).toBeLessThan(0.3); }); }); /** * Calculate cosine similarity between two vectors * @param vecA - First vector * @param vecB - Second vector * @returns Cosine similarity (between -1 and 1) */ function calculateCosineSimilarity(vecA: number[], vecB: number[]): number { if (vecA.length !== vecB.length) { throw new Error('Vectors must have the same dimensions'); } // For normalized vectors, the dot product equals cosine similarity let dotProduct = 0; for (let i = 0; i < vecA.length; i++) { dotProduct += vecA[i] * vecB[i]; } return dotProduct; }