#!/usr/bin/env bun
/**
* Fix vector dimensions from 512 to 768
* This script updates the embeddings table to use 768-dimensional vectors
*/
import { database } from '../src/backend/services/database.js';
async function fixVectorDimensions() {
try {
console.log('đ§ Fixing vector dimensions from 512 to 768...');
// Connect to database
const config = {
host: process.env.DB_HOST || 'localhost',
port: parseInt(process.env.DB_PORT || '5432'),
database: process.env.DB_NAME || 'article_manager_test',
user: process.env.DB_USER || 'postgres',
password: process.env.DB_PASSWORD || 'postgres',
};
await database.connect(config);
console.log('â
Connected to database');
// Check if embeddings table exists and has vector column
const tableCheck = await database.query(`
SELECT column_name, data_type, character_maximum_length
FROM information_schema.columns
WHERE table_name = 'embeddings' AND column_name = 'vector'
`);
if (tableCheck.rows.length === 0) {
console.log('âšī¸ No vector column found in embeddings table');
return;
}
console.log('đ Current vector column info:', tableCheck.rows[0]);
// Check if we have any embeddings data
const dataCheck = await database.query('SELECT COUNT(*) as count FROM embeddings');
const embeddingCount = parseInt(dataCheck.rows[0].count);
console.log(`đ Found ${embeddingCount} existing embeddings`);
if (embeddingCount > 0) {
console.log('â ī¸ Backing up existing embeddings...');
// Clear existing embeddings since they have wrong dimensions
await database.query('DELETE FROM embeddings');
console.log('đī¸ Cleared existing embeddings (they will be regenerated)');
}
// Drop and recreate the vector column with correct dimensions
console.log('đ Updating vector column to 768 dimensions...');
await database.query('ALTER TABLE embeddings DROP COLUMN IF EXISTS vector');
await database.query('ALTER TABLE embeddings ADD COLUMN vector VECTOR(768)');
// Recreate the vector index
console.log('đ Recreating vector index...');
await database.query('DROP INDEX IF EXISTS idx_embeddings_vector');
await database.query('CREATE INDEX idx_embeddings_vector ON embeddings USING ivfflat (vector vector_cosine_ops)');
console.log('â
Vector dimensions updated successfully!');
console.log('âšī¸ Existing articles will need to have their embeddings regenerated');
} catch (error) {
console.error('â Error fixing vector dimensions:', error);
throw error;
} finally {
await database.disconnect();
console.log('đĄ Database disconnected');
}
}
// Run the fix
fixVectorDimensions().catch(console.error);
