#!/usr/bin/env python3 """ SUNO AI ML MUSIC SEARCH DEMO User: valkst Date: 2025-08-02 11:18:00 UTC Complete ML system demonstration """ import sqlite3 import numpy as np import pickle from sentence_transformers import SentenceTransformer from sklearn.metrics.pairwise import cosine_similarity import pandas as pd class SunoMLDemo: def __init__(self): self.db_path = "tracks.sqlite" self.model = SentenceTransformer('paraphrase-multilingual-MiniLM-L12-v2') print("🎵 SUNO AI ML MUSIC SEARCH SYSTEM DEMO") print("="*50) def semantic_search(self, query, limit=10): """Семантический поиск по запросу""" print(f"🔍 Поиск: '{query}'") # Создаем вектор запроса query_vector = self.model.encode([query], normalize_embeddings=True) # Получаем все векторы из базы conn = sqlite3.connect(self.db_path) # Получаем треки с векторами cursor = conn.execute(""" SELECT t.id, t.title, t.artist, emb.embedding_vector, ml.ml_emotion_dominant, ml.ml_emotion_confidence FROM tracks t JOIN ml_embeddings emb ON t.id = emb.track_id JOIN ml_analysis ml ON t.id = ml.track_id ORDER BY t.quality_score DESC LIMIT 1000 """) results = [] for row in cursor.fetchall(): track_id, title, artist, vector_blob, emotion, emotion_conf = row vector = pickle.loads(vector_blob) # Вычисляем similarity similarity = cosine_similarity(query_vector, vector.reshape(1, -1))[0][0] results.append({ 'id': track_id, 'title': title, 'artist': artist, 'similarity': similarity, 'emotion': emotion, 'emotion_confidence': emotion_conf }) conn.close() # Сортируем по similarity results.sort(key=lambda x: x['similarity'], reverse=True) print(f"📊 Найдено {len(results)} треков, показываем топ-{limit}:") print("-" * 80) for i, result in enumerate(results[:limit], 1): emotion_icon = {'joy': '😊', 'sadness': '😢', 'anger': '😡', 'fear': '😨', 'disgust': '🤢', 'surprise': '😮'}.get(result['emotion'], '🎵') print(f"{i:2d}. {result['title']}") print(f" 👤 {result['artist']}") print(f" 🎯 Similarity: {result['similarity']:.3f}") print(f" {emotion_icon} {result['emotion'].title()} ({result['emotion_confidence']:.2f})") print() return results[:limit] def emotion_filter(self, emotion, limit=15): """Фильтр по эмоции""" emotions_dict = { 'радость': 'joy', 'грусть': 'sadness', 'гнев': 'anger', 'страх': 'fear', 'отвращение': 'disgust', 'удивление': 'surprise', 'joy': 'joy', 'sadness': 'sadness', 'anger': 'anger', 'fear': 'fear', 'disgust': 'disgust', 'surprise': 'surprise' } emotion_en = emotions_dict.get(emotion.lower(), emotion.lower()) conn = sqlite3.connect(self.db_path) cursor = conn.execute(""" SELECT t.title, t.artist, ml.ml_emotion_confidence, t.quality_score FROM tracks t JOIN ml_analysis ml ON t.id = ml.track_id WHERE ml.ml_emotion_dominant = ? ORDER BY ml.ml_emotion_confidence DESC, t.quality_score DESC LIMIT ? """, (emotion_en, limit)) results = cursor.fetchall() conn.close() emotion_icon = {'joy': '😊', 'sadness': '😢', 'anger': '😡', 'fear': '😨', 'disgust': '🤢', 'surprise': '😮'}.get(emotion_en, '🎵') print(f"{emotion_icon} Треки с эмоцией '{emotion.title()}':") print("-" * 60) for i, (title, artist, conf, quality) in enumerate(results, 1): print(f"{i:2d}. {title}") print(f" 👤 {artist}") print(f" 🎯 Уверенность: {conf:.3f}, Качество: {quality:.3f}") print() return results def cluster_analysis(self): """Анализ кластеров""" conn = sqlite3.connect(self.db_path) cursor = conn.execute(""" SELECT ml.semantic_cluster, COUNT(*) as track_count, AVG(ml.cluster_confidence) as avg_confidence, ml.ml_emotion_dominant as dominant_emotion, COUNT(CASE WHEN ml.ml_emotion_dominant = 'joy' THEN 1 END) as joy_count, COUNT(CASE WHEN ml.ml_emotion_dominant = 'sadness' THEN 1 END) as sad_count FROM ml_analysis ml WHERE ml.semantic_cluster >= 0 GROUP BY ml.semantic_cluster, ml.ml_emotion_dominant ORDER BY track_count DESC """) clusters_data = {} for row in cursor.fetchall(): cluster_id, count, conf, emotion, joy, sad = row if cluster_id not in clusters_data: clusters_data[cluster_id] = { 'total_tracks': 0, 'avg_confidence': conf, 'emotions': {} } clusters_data[cluster_id]['total_tracks'] += count clusters_data[cluster_id]['emotions'][emotion] = count conn.close() print("🔍 АНАЛИЗ СЕМАНТИЧЕСКИХ КЛАСТЕРОВ:") print("=" * 50) for cluster_id, data in sorted(clusters_data.items(), key=lambda x: x[1]['total_tracks'], reverse=True): dominant_emotion = max(data['emotions'].items(), key=lambda x: x[1]) emotion_icon = {'joy': '😊', 'sadness': '😢', 'anger': '😡', 'fear': '😨', 'disgust': '🤢', 'surprise': '😮'}.get(dominant_emotion[0], '🎵') print(f"🔸 Кластер {cluster_id}: {data['total_tracks']} треков") print(f" {emotion_icon} Доминирующая эмоция: {dominant_emotion[0]} ({dominant_emotion[1]} треков)") print(f" 🎯 Уверенность: {data['avg_confidence']:.3f}") print() def statistics_overview(self): """Общая статистика""" conn = sqlite3.connect(self.db_path) # Основная статистика cursor = conn.execute(""" SELECT COUNT(*) as total_tracks, AVG(ml_emotion_confidence) as avg_emotion_conf, AVG(cluster_confidence) as avg_cluster_conf, COUNT(CASE WHEN ml_emotion_confidence > 0.8 THEN 1 END) as high_conf_tracks FROM ml_analysis """) stats = cursor.fetchone() total, avg_emotion, avg_cluster, high_conf = stats print("📈 ОБЩАЯ СТАТИСТИКА ML СИСТЕМЫ:") print("=" * 40) print(f"🎵 Всего треков: {total:,}") print(f"🎭 Средняя эмоциональная уверенность: {avg_emotion:.3f}") print(f"🔍 Средняя кластерная уверенность: {avg_cluster:.3f}") print(f"🏆 Треков с высокой уверенностью: {high_conf:,} ({high_conf/total*100:.1f}%)") print() # Топ исполнители по количеству треков cursor = conn.execute(""" SELECT t.artist, COUNT(*) as track_count FROM tracks t JOIN ml_analysis ml ON t.id = ml.track_id GROUP BY t.artist ORDER BY track_count DESC LIMIT 10 """) print("🎤 ТОП-10 ИСПОЛНИТЕЛЕЙ:") print("-" * 30) for i, (artist, count) in enumerate(cursor.fetchall(), 1): print(f"{i:2d}. {artist}: {count} треков") conn.close() def main(): demo = SunoMLDemo() print(""" Доступные команды: 1. demo.semantic_search("запрос") - семантический поиск 2. demo.emotion_filter("радость") - фильтр по эмоции 3. demo.cluster_analysis() - анализ кластеров 4. demo.statistics_overview() - общая статистика Примеры: demo.semantic_search("love song about heartbreak") demo.emotion_filter("грусть") """) return demo if __name__ == "__main__": demo = main()