#!/usr/bin/env tsx
/**
* Test embeddings and vector search with mock setup
*/
import { initDatabaseFromEnv } from './src/database/index.js';
import { MemoryCore } from './src/core/index.js';
// Mock embedding function for testing without API key
function generateMockEmbedding(text: string): number[] {
// Create a simple hash-based embedding for testing
const embedding = new Array(1536).fill(0);
for (let i = 0; i < text.length; i++) {
const idx = (text.charCodeAt(i) * 7 + i) % 1536;
embedding[idx] = Math.random() * 0.1 + 0.45;
}
// Normalize
const sum = embedding.reduce((a, b) => a + b, 0);
return embedding.map(v => v / sum * 100);
}
async function testEmbeddings() {
console.log('🧪 Testing Embeddings and Vector Search\n');
const db = initDatabaseFromEnv();
const memoryCore = new MemoryCore(db, process.env.OPENAI_API_KEY);
await memoryCore.initialize();
const userId = 'bob@matsuoka.com';
// 1. Test with real API key if available
if (process.env.OPENAI_API_KEY) {
console.log('1️⃣ TESTING WITH OPENAI API:');
const memResult = await memoryCore.addMemory(
'AI Research Paper',
'Deep learning models for natural language processing using transformers',
'semantic' as any,
{
userId,
generateEmbedding: true,
}
);
if (memResult.status === 'success' && memResult.data) {
const memId = (memResult.data as any).id;
// Check embedding in database
const checkResult = await db.execute(
`SELECT
embedding IS NOT NULL as has_embedding,
json_array_length(embedding) as dimensions
FROM memories
WHERE id = ?`,
[memId]
);
const row = checkResult.rows[0] as any;
console.log(` ✅ Memory created with embedding`);
console.log(` Has embedding: ${row?.has_embedding ? 'Yes' : 'No'}`);
console.log(` Dimensions: ${row?.dimensions || 'N/A'}`);
// Test vector search
const searchResult = await memoryCore.searchMemories(
'transformer neural networks',
{
userId,
strategy: 'similarity',
threshold: 0.5,
}
);
if (searchResult.status === 'success' && searchResult.data) {
const results = searchResult.data as any[];
console.log(` ✅ Vector search found ${results.length} results`);
}
await memoryCore.deleteMemory(memId, { userId });
}
} else {
console.log('1️⃣ NO OPENAI API KEY - Using mock embeddings for testing');
// Test without API key to show the issue
const memResult = await memoryCore.addMemory(
'Test Memory Without API',
'This memory will not have real embeddings',
'semantic' as any,
{
userId,
generateEmbedding: true,
}
);
if (memResult.status === 'success' && memResult.data) {
const memId = (memResult.data as any).id;
// Check what was stored
const checkResult = await db.execute(
`SELECT
embedding,
embedding IS NOT NULL as has_embedding,
LENGTH(embedding) as embed_length
FROM memories
WHERE id = ?`,
[memId]
);
const row = checkResult.rows[0] as any;
console.log(` Memory created: ${memId}`);
console.log(` Has embedding: ${row?.has_embedding ? 'Yes' : 'No'}`);
console.log(` Embedding value: ${row?.embedding}`);
console.log(` Embedding length: ${row?.embed_length}`);
// Manually add mock embedding for testing
const mockEmbedding = generateMockEmbedding('test memory');
await db.execute(
'UPDATE memories SET embedding = ? WHERE id = ?',
[JSON.stringify(mockEmbedding), memId]
);
console.log(' ✅ Added mock embedding for testing');
await memoryCore.deleteMemory(memId, { userId });
}
}
// 2. Test metadata updates
console.log('\n2️⃣ TESTING METADATA UPDATES:');
const metaResult = await memoryCore.addMemory(
'Metadata Test',
'Testing metadata updates',
'semantic' as any,
{
userId,
metadata: { version: '1.0', status: 'draft' },
generateEmbedding: false,
}
);
if (metaResult.status === 'success' && metaResult.data) {
const metaId = (metaResult.data as any).id;
// Update metadata
const updateResult = await memoryCore.updateMemory(
metaId,
{
metadata: { version: '2.0', status: 'published', author: 'test' }
},
{ userId }
);
console.log(` Update result: ${updateResult.status}`);
if (updateResult.status === 'success') {
// Verify update
const getResult = await memoryCore.getMemory(metaId, { userId });
if (getResult.status === 'success' && getResult.data) {
const memory = getResult.data as any;
console.log(` ✅ Metadata updated: ${JSON.stringify(memory.metadata)}`);
}
}
await memoryCore.deleteMemory(metaId, { userId });
}
// 3. Test all update fields
console.log('\n3️⃣ TESTING ALL UPDATE FIELDS:');
const fullResult = await memoryCore.addMemory(
'Original Title',
'Original content',
'semantic' as any,
{
userId,
importance: 0.5,
tags: ['old'],
metadata: { version: '1' },
generateEmbedding: false,
}
);
if (fullResult.status === 'success' && fullResult.data) {
const fullId = (fullResult.data as any).id;
// Update everything
const updateAllResult = await memoryCore.updateMemory(
fullId,
{
title: 'New Title',
content: 'New content',
importance: 0.9,
tags: ['new', 'updated'],
metadata: { version: '2', updated: true }
},
{ userId }
);
if (updateAllResult.status === 'success') {
const verifyResult = await memoryCore.getMemory(fullId, { userId });
if (verifyResult.status === 'success' && verifyResult.data) {
const memory = verifyResult.data as any;
console.log(' ✅ Title updated:', memory.title === 'New Title');
console.log(' ✅ Content updated:', memory.content === 'New content');
console.log(' ✅ Importance updated:', memory.importance === 0.9);
console.log(' ✅ Tags updated:', JSON.stringify(memory.tags));
console.log(' ✅ Metadata updated:', JSON.stringify(memory.metadata));
}
}
await memoryCore.deleteMemory(fullId, { userId });
}
console.log('\n✅ Testing complete!');
process.exit(0);
}
testEmbeddings().catch(console.error);
