#!/usr/bin/env node
/**
* Test script for vector search functionality
* Demonstrates free local embeddings with Xenova + in-memory vector store
*/
import SimpleVectorStore from './lib/simple-vector-store.js'
import SemanticSplitter from './lib/semantic-splitter.js'
import { fileURLToPath } from 'url'
import { dirname } from 'path'
const __filename = fileURLToPath(import.meta.url)
const __dirname = dirname(__filename)
async function testVectorSearch() {
console.log('🚀 Testing Vector Search with Xenova + In-Memory Store\n')
try {
// Initialize the same components as the server
const semanticSplitter = new SemanticSplitter({
maxChunkSize: 2000,
includeContext: true,
groupRelated: true,
minFunctionSize: 50
})
const vectorStore = new SimpleVectorStore({
modelName: 'Xenova/all-MiniLM-L6-v2',
collectionName: 'test-code-chunks'
})
// Use the same patterns as the server
const patterns = ['**/*.{js,jsx,ts,tsx,mjs}']
console.log('📁 Analyzing project with SemanticSplitter (like the server)...')
const analysis = await semanticSplitter.extractSemanticChunks(__dirname, patterns)
console.log(`✅ Analysis complete!`)
console.log(`📊 Found ${analysis.chunks.length} semantic chunks`)
console.log(`📄 Total files: ${analysis.summary.totalFiles}`)
console.log(`⚡ Total functions: ${analysis.summary.totalFunctions}`)
console.log(`📦 Average chunk size: ${analysis.summary.averageChunkSize.toFixed(0)} chars\n`)
if (analysis.chunks.length > 0) {
console.log('🔍 Adding vector search capabilities...')
await vectorStore.storeChunks(analysis.chunks)
console.log('🧪 Testing Vector Search Features:\n')
// Test 1: Find similar code
console.log('1️⃣ Finding code similar to "semantic analysis":')
const similarChunks = await vectorStore.findSimilar('semantic analysis', { limit: 3 })
similarChunks.forEach((result, i) => {
console.log(` ${i + 1}. ${result.id} (${(result.similarity * 100).toFixed(1)}% similar)`)
console.log(` Type: ${result.metadata.semanticType || 'unknown'}`)
console.log(` Domain: ${result.metadata.businessDomain?.join(', ') || 'none'}`)
})
console.log()
// Test 2: Find by semantic type
console.log('2️⃣ Finding utility functions:')
const utilityFunctions = await vectorStore.findByType('utilityFunction', 3)
utilityFunctions.forEach((result, i) => {
console.log(` ${i + 1}. ${result.id} (${(result.similarity * 100).toFixed(1)}% similar)`)
})
console.log()
// Test 3: Find by domain
console.log('3️⃣ Finding code analysis related code:')
const analysisCode = await vectorStore.findByDomain('code-analysis', 3)
analysisCode.forEach((result, i) => {
console.log(` ${i + 1}. ${result.id} (${(result.similarity * 100).toFixed(1)}% similar)`)
})
console.log()
// Test 4: Find by pattern
console.log('4️⃣ Finding React-related code:')
const reactCode = await vectorStore.findByPattern('react', 3)
reactCode.forEach((result, i) => {
console.log(` ${i + 1}. ${result.id} (${(result.similarity * 100).toFixed(1)}% similar)`)
})
console.log()
// Test 5: Get vector database stats
const stats = await vectorStore.getStats()
console.log('📈 Vector Database Stats:')
console.log(` Total chunks: ${stats.totalChunks}`)
console.log(` Collection: ${stats.collectionName}`)
console.log(` Model: ${stats.modelName}`)
} else {
console.log('❌ No semantic chunks found - vector search not available')
}
} catch (error) {
console.error('❌ Test failed:', error.message)
console.error('Stack:', error.stack)
}
}
// Run the test
testVectorSearch()
