EOL: Final release

src/music_engine/llm_bridge.py

@@ -2,16 +2,16 @@ import os
 import json
 import re
 import requests
+from typing import List, Dict
 
 class LLMAcousticBridge:
-    def __init__(self, model_name="dolphin-llama3:8b"):
+    def __init__(self, model_name: str = "dolphin-llama3:8b"):
         self.model_name = model_name
         base_url = os.getenv("OLLAMA_API_URL", "http://emom_ollama:11434")
         self.api_url = f"{base_url}/api/generate"
 
-    def get_acoustic_profile(self, valence, arousal, semantics):
+    def get_acoustic_profile(self, valence: float, arousal: float, semantics: List[str]) -> Dict[str, float]:
         context_str = ", ".join(semantics) if semantics else "abstract scene"
-        
         prompt = f"""
         Analyze the visual context and emotions to determine the ideal background music properties. 
         Emotions: Valence {valence:.1f}/9.0 (Positivity), Arousal {arousal:.1f}/9.0 (Energy). 
@@ -34,32 +34,42 @@ class LLMAcousticBridge:
                 "model": self.model_name,
                 "prompt": prompt,
                 "stream": False,
-                "format": "json"  # Принудительный JSON-режим Ollama
+                "format": "json",
+                "options": {
+                    "temperature": 0.7,
+                    "top_p": 0.9
+                }
             }
             
             print(f"Запрос акустического профиля к Ollama...")
             response = requests.post(self.api_url, json=payload, timeout=120)
+            response.raise_for_status()
             
-            if response.status_code == 200:
+            data = response.json()
-                data = response.json()
+            response_text = data.get("response", "")
-                response_text = data.get("response", "")
+            
-                
+            profile = {}
-                try:
+            try:
-                    # 1. Попытка прямой десериализации
+                profile = json.loads(response_text)
-                    profile = json.loads(response_text)
+            except json.JSONDecodeError:
-                    return profile
+                match = re.search(r'\{.*\}', response_text, re.DOTALL)
-                except json.JSONDecodeError:
+                if match:
-                    # 2. Аварийное извлечение JSON из текста с помощью регулярного выражения
+                    profile = json.loads(match.group(0))
-                    match = re.search(r'\{.*\}', response_text, re.DOTALL)
+                else:
-                    if match:
+                    print(f"[ERROR] Ошибка парсинга LLM ответа: {response_text}")
-                        return json.loads(match.group(0))
-                    
-                    print(f"Ошибка парсинга LLM ответа: {response_text}")
                     return {}
+            
+            # 3. Жесткая валидация ключей (чтобы matcher.py не упал на демо)
+            required_keys = ['energy', 'flux', 'centroid', 'pitch', 'hnr', 'zcr']
+            if profile and all(k in profile for k in required_keys):
+                return profile
             else:
-                print(f"Ollama вернула ошибку HTTP: {response.status_code}")
+                print(f"[WARN] LLM вернула неполный JSON, активирован fallback: {profile}")
                 return {}
                 
+        except requests.exceptions.RequestException as e:
+            print(f"[ERROR] Ошибка соединения с Ollama: {str(e)}")
+            return {}
         except Exception as e:
-            print(f"Ошибка соединения с Ollama: {str(e)}")
+            print(f"[ERROR] Внутренняя ошибка семантического моста: {str(e)}")
             return {}

src/music_engine/matcher.py

@@ -78,6 +78,6 @@ class MusicMatcher:
         self.norm_db['acoustic_distance'] = acoustic_penalty / len(self.acoustic_features)
 
         # Вычисление интегральной метрики соответствия (мультимодальный скоринг)
-        self.norm_db['final_score'] = self.norm_db['emo_distance'] + (self.norm_db['acoustic_distance'] * 4.0)
+        self.norm_db['final_score'] = self.norm_db['emo_distance'] + (self.norm_db['acoustic_distance'] * 2.0)
 
         return self.norm_db.sort_values(by='final_score').head(top_k)

src/scripts/22_extract_embeddings.ipynb

@@ -264,9 +264,9 @@
  ],
  "metadata": {
   "kernelspec": {
-   "display_name": "Python (thesis)",
+   "display_name": "Python (my-python-project)",
    "language": "python",
-   "name": "thesis"
+   "name": "my-python-project"
   },
   "language_info": {
    "codemirror_mode": {