// Test final pour vérifier l'optimisation des scores
import { setEmbeddingProvider, embedAndStore, semanticSearch } from './build/rag/vector-store.js';
async function runTest() {
console.log('=== TEST D\'OPTIMISATION DES SCORES ===\n');
// Configurer Ollama avec nomic-embed-text
setEmbeddingProvider('ollama', 'nomic-embed-text');
// Indexer des documents variés
const documents = [
{ path: '/tmp/doc1.txt', content: 'Le machine learning est une branche de l\'intelligence artificielle' },
{ path: '/tmp/doc2.txt', content: 'PostgreSQL est un système de gestion de base de données relationnelle' },
{ path: '/tmp/doc3.txt', content: 'React est une bibliothèque JavaScript pour construire des interfaces utilisateur' },
{ path: '/tmp/doc4.txt', content: 'Python est un langage de programmation interprété et de haut niveau' },
{ path: '/tmp/doc5.txt', content: 'Docker est une plateforme de conteneurisation d\'applications' }
];
console.log('1. Indexation des documents...');
for (const doc of documents) {
await embedAndStore('/tmp/test_final', doc.path, doc.content);
}
console.log(' ✓ 5 documents indexés\n');
// Tests de recherche
const testQueries = [
{ query: 'intelligence artificielle', expected: 'doc1.txt' },
{ query: 'base de données', expected: 'doc2.txt' },
{ query: 'JavaScript', expected: 'doc3.txt' },
{ query: 'programmation', expected: 'doc4.txt' },
{ query: 'conteneur', expected: 'doc5.txt' }
];
console.log('2. Tests de recherche sémantique:');
for (const test of testQueries) {
const results = await semanticSearch(test.query, {
projectFilter: '/tmp/test_final',
limit: 3,
threshold: 0
});
console.log(`\n Requête: "${test.query}"`);
console.log(' Résultats:');
results.forEach((r, i) => {
const fileName = r.filePath.split('/').pop();
const match = fileName === test.expected ? '✓' : ' ';
console.log(` ${match} ${i+1}. ${fileName}: ${r.score.toFixed(4)}`);
});
}
// Analyse statistique
console.log('\n3. Analyse statistique des scores:');
const { Pool } = await import('pg');
const pool = new Pool({
host: "localhost",
port: 16432,
database: "rag_mcp_dedicated",
user: "rag_user",
password: "secure_rag_password",
});
const stats = await pool.query(`
SELECT
MIN(score) as min_score,
MAX(score) as max_score,
AVG(score) as avg_score,
STDDEV(score) as stddev_score
FROM (
SELECT a.vector <=> b.vector as score
FROM rag_store a, rag_store b
WHERE a.id < b.id AND a.project_path = '/tmp/test_final' AND b.project_path = '/tmp/test_final'
) distances;
`);
const row = stats.rows[0];
console.log(` Score minimum: ${row.min_score?.toFixed(4) || 'N/A'}`);
console.log(` Score maximum: ${row.max_score?.toFixed(4) || 'N/A'}`);
console.log(` Score moyen: ${row.avg_score?.toFixed(4) || 'N/A'}`);
console.log(` Écart-type: ${row.stddev_score?.toFixed(4) || 'N/A'}`);
// Vérification du problème des scores élevés
console.log('\n4. Vérification de l\'optimisation:');
if (row.max_score && row.max_score < 0.9) {
console.log(' ✓ PROBLÈME RÉSOLU: Les scores ne sont plus uniformément élevés');
console.log(` ✓ Score maximum: ${row.max_score.toFixed(4)} (acceptable)`);
} else {
console.log(' ✗ PROBLÈME PERSISTANT: Scores trop élevés');
}
await pool.end();
console.log('\n=== TEST TERMINÉ ===');
}
runTest().catch(console.error);
