migliorata visualizzazione BPO Dispatcher

.gitignore

@@ -1 +1,2 @@
 __pycache__/
+.venv

app.py

@@ -128,7 +128,7 @@ def bpo_dispatch_logic(text, request: gr.Request):
     start_time = time.time()
     
     try:
-        intent, urgency, entities = predict_bpo_ticket(text)
+        intent, urgency, entities, sentiment = predict_bpo_ticket(text)
         
         if intent is None:
              raise gr.Error("Errore nel modello BPO. Verifica i log.")
@@ -145,6 +145,11 @@ def bpo_dispatch_logic(text, request: gr.Request):
 
         html_output = utils.render_ner_html(entities)
 
+        sentiment_formatted = {
+            "Positivo": sentiment.get("positive", 0.0),
+            "Negativo": sentiment.get("negative", 0.0)
+        }
+
         elapsed_time = time.time() - start_time
         logger.log_interaction(
             request=request,
@@ -154,7 +159,7 @@ def bpo_dispatch_logic(text, request: gr.Request):
             execution_time=elapsed_time
         )
 
-        return intent, urgency, action, html_output
+        return intent, sentiment_formatted, urgency, action, html_output
 
     except Exception as e:
         raise gr.Error(f"Errore nell'analisi: {str(e)}")
@@ -190,8 +195,6 @@ with gr.Blocks(title="NGT AI Platform", theme=theme, css_paths="style.css") as d
                 gr.Image(value="data/icon.png", show_label=False, show_download_button=False, show_share_button=False, container=False, show_fullscreen_button=False, interactive=False, height=80, width=80)
             with gr.Column(scale=1, elem_classes="header-text-col"):
                 gr.Markdown("""<h1>AI Platform</h1><div class='subheader'>Advanced Machine Learning Solutions</div>""")
-        
-        # --- BPO INTELLIGENT DISPATCHER ---
         with gr.Tab("🧩 BPO Dispatcher") as tab_bpo:
             gr.Markdown("""
             # 🧩 Intelligent Ticket Routing & NER
@@ -205,24 +208,32 @@ with gr.Blocks(title="NGT AI Platform", theme=theme, css_paths="style.css") as d
                     analyze_btn_bpo = gr.Button("⚡ Analizza Richiesta", variant="primary")
                     gr.HTML("""
                     <div class='model-card'>
-                        <strong>🛠️ Model Architecture:</strong> NGT-BERT-Custom (DistilBERT)<br>
-                        <strong>📚 Training Data:</strong> Synthetic BPO Dataset (2025)<br>
-                        <strong>🎯 Tasks:</strong> Intent Classification (Multi-class), Entity Extraction (NER)
+                        <strong>🛠️ Model Architecture:</strong> NGT-BERT-Custom (DistilBERT) + NGT Sentiment Classifier<br>
+                        <strong>📚 Training Data:</strong> Synthetic BPO Dataset & Helpdesk Sentiment Dataset<br>
+                        <strong>🎯 Tasks:</strong> Intent Classification, Entity Extraction, & Sentiment-Driven Urgency Routing
                     </div>
                     """)
-
+ 
                 # OUTPUT
                 with gr.Column(scale=1):
                     with gr.Group():
                         gr.Markdown("#### 📋 Analisi Processata", elem_classes="h4-margin")
-                        bpo_intent_output = gr.Label(num_top_classes=3, label="Intento Rilevato")
+                        with gr.Row():
+                            bpo_intent_output = gr.Label(num_top_classes=3, label="Intento Rilevato", scale=1)
+                            bpo_sentiment_output = gr.Label(num_top_classes=2, label="Sentiment Rilevato (Sentiment Analysis)", scale=1)
                         with gr.Row():
                             bpo_urgency_output = gr.Textbox(label="Livello Urgenza", scale=1)
                             bpo_action_output = gr.Textbox(label="Azione Consigliata (Auto)", scale=1)
                         
+                        gr.HTML("""
+                        <div class='sentiment-info-badge'>
+                            💡 <strong>Integrazione Sentiment Analysis:</strong> L'urgenza e l'assegnazione automatica del ticket sono calibrate in tempo reale integrando il modello di <em>Sentiment Analysis</em> per rilevare lo stato emotivo del cliente (es. minacce legali o toni irritati).
+                        </div>
+                        """)
+                        
                         gr.Markdown("#### 🔍 Dati Estratti (NER)", elem_classes="h4-margin")
                         bpo_ner_output = gr.HTML(label="Visualizzazione Entità")
-
+ 
             gr.Examples(
                 examples=[
                     ["Buongiorno, vi scrivo perché la fattura n. 99283 del mese scorso è sbagliata. Non ho consumato così tanto. Il mio codice cliente è 4599201. Attendo rettifica urgente."],
@@ -235,7 +246,7 @@ with gr.Blocks(title="NGT AI Platform", theme=theme, css_paths="style.css") as d
             analyze_btn_bpo.click(
                 bpo_dispatch_logic, 
                 inputs=bpo_input, 
-                outputs=[bpo_intent_output, bpo_urgency_output, bpo_action_output, bpo_ner_output]
+                outputs=[bpo_intent_output, bpo_sentiment_output, bpo_urgency_output, bpo_action_output, bpo_ner_output]
             )
 
         # --- AI FORECASTER ---
@@ -468,25 +479,26 @@ with gr.Blocks(title="NGT AI Platform", theme=theme, css_paths="style.css") as d
         return (
             None,   # 1. bpo_input
             None,   # 2. bpo_intent
-            None,   # 3. bpo_urgency
-            None,   # 4. bpo_action
-            None,   # 5. bpo_ner
-            None,   # 6. forecast_file
-            None,   # 7. forecast_plot
-            None,   # 8. forecast_stats
-            None,   # 9. image_input
-            None,   # 10. output_label
-            None,   # 11. image_input_dr
-            None,   # 12. output_dr_diag
-            None,   # 13. output_dr_prob
-            None,   # 14. multi_input
-            None,   # 15. multi_output
-            None,   # 16. sentiment_input
-            None    # 17. sentiment_output
+            None,   # 3. bpo_sentiment
+            None,   # 4. bpo_urgency
+            None,   # 5. bpo_action
+            None,   # 6. bpo_ner
+            None,   # 7. forecast_file
+            None,   # 8. forecast_plot
+            None,   # 9. forecast_stats
+            None,   # 10. image_input
+            None,   # 11. output_label
+            None,   # 12. image_input_dr
+            None,   # 13. output_dr_diag
+            None,   # 14. output_dr_prob
+            None,   # 15. multi_input
+            None,   # 16. multi_output
+            None,   # 17. sentiment_input
+            None    # 18. sentiment_output
         )
 
     reset_outputs = [
-        bpo_input, bpo_intent_output, bpo_urgency_output, bpo_action_output, bpo_ner_output,
+        bpo_input, bpo_intent_output, bpo_sentiment_output, bpo_urgency_output, bpo_action_output, bpo_ner_output,
         forecast_file, forecast_plot, forecast_stats,
         image_input, output_label,
         image_input_dr, output_dr_diagnosis, output_dr_prob,
@@ -545,6 +557,6 @@ with gr.Blocks(title="NGT AI Platform", theme=theme, css_paths="style.css") as d
 if __name__ == "__main__":
     demo.launch(
         server_name="0.0.0.0", 
-        server_port=7860, 
+        server_port=7862, 
         allowed_paths=["data"]
     )

data/logs/access_logs.csv

@@ -0,0 +1,2 @@
+timestamp,session_id,module,action,ip_address,user_agent,language,input_size,input_text,processing_time
+2026-06-05T12:50:51.155066+02:00,59c1bbc1,BPO Dispatcher,Prediction,127.0.0.1,"Mozilla/5.0 (X11; Linux x86_64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/148.0.0.0 Safari/537.36",it-IT,166 chars,"Buongiorno, vi scrivo perché la fattura n. 99283 del mese scorso è sbagliata. Non ho consumato così tanto. Il mio codice cliente è 4599201. Attendo rettifica urgente.",2.6220s

modules/bpo_dispatcher.py

@@ -121,7 +121,7 @@ class BPODispatcher:
                     add_entity(ent.text, "AZIENDA", ent.start_char, ent.end_char)
         return entities
     
-    def _calculate_smart_urgency(self, text, intent_label):
+    def _calculate_smart_urgency(self, text, intent_label, sentiment_score_neg, sentiment_score_pos):
         """
         MATRICE DI URGENZA (Intent + Sentiment)
         Combina la gravità del problema con lo stato d'animo del cliente.
@@ -129,18 +129,6 @@ class BPODispatcher:
         urgency = "Bassa"
         text_lower = text.lower()
         
-        # Analisi Sentiment
-        sentiment_score_neg = 0.0
-        sentiment_score_pos = 0.0
-        
-        if binary_classification:
-            try:
-                sent_result = binary_classification(text)
-                sentiment_score_neg = float(sent_result.get('negative', 0.0))
-                sentiment_score_pos = float(sent_result.get('positive', 0.0))
-            except Exception:
-                sentiment_score_neg = 0.5  # Fallback neutro
-        
         # CASO CHURN (Disdetta) -> Sempre Critico
         # Indipendentemente dal tono, se uno vuole andare via è priorità assoluta.
         if intent_label == "Retention / Churn Risk":
@@ -187,8 +175,8 @@ class BPODispatcher:
         return urgency
 
     def predict(self, text):
-        if self.model is None: return None, "Errore", []
-        if not text.strip(): return None, "Vuoto", []
+        if self.model is None: return None, "Errore", [], {}
+        if not text.strip(): return None, "Vuoto", [], {}
 
         # Intent Classification (BERT)
         inputs = self.tokenizer(text, return_tensors="pt", truncation=True, max_length=128, padding=True)
@@ -202,13 +190,29 @@ class BPODispatcher:
         top_idx = torch.max(probs, dim=-1)[1].item()
         predicted_label = LABELS_MAP[top_idx]
         
+        # Analisi Sentiment
+        sentiment_score_neg = 0.5
+        sentiment_score_pos = 0.5
+        if binary_classification:
+            try:
+                sent_result = binary_classification(text)
+                sentiment_score_neg = float(sent_result.get('negative', 0.5))
+                sentiment_score_pos = float(sent_result.get('positive', 0.5))
+            except Exception:
+                pass
+        
         # Urgenza (AI + Sentiment + Rules)
-        urgency = self._calculate_smart_urgency(text, predicted_label)
+        urgency = self._calculate_smart_urgency(text, predicted_label, sentiment_score_neg, sentiment_score_pos)
         
         # NER Extraction
         entities = self._extract_smart_entities(text)
         
-        return label_output, urgency, entities
+        sentiment_output = {
+            "negative": sentiment_score_neg,
+            "positive": sentiment_score_pos
+        }
+        
+        return label_output, urgency, entities, sentiment_output
 
 dispatcher = BPODispatcher()
 def predict_bpo_ticket(text): return dispatcher.predict(text)

modules/image_classification.py

@@ -30,13 +30,13 @@ def load_resources():
         except Exception as e:
             print(f"❌ Errore caricamento modello X-Ray: {e}")
 
-# Inizializziamo subito
-load_resources()
-
 def image_classification(image_array):
     """
     Analizza un'immagine radiografica (numpy array) e restituisce le probabilità.
     """
+    global model
+    if model is None:
+        load_resources()
     if model is None:
         return {"Errore": "Modello non disponibile (Verifica il percorso file)"}
     if image_array is None:

modules/multilabel_classification.py

@@ -2,7 +2,7 @@ import os
 import json
 from keras.models import load_model
 from keras_preprocessing.sequence import pad_sequences
-from keras.preprocessing.text import tokenizer_from_json
+from keras_preprocessing.text import tokenizer_from_json
 
 BASE_DIR = os.path.dirname(os.path.dirname(os.path.abspath(__file__))) # Risale alla root
 MODEL_PATH = os.path.join(BASE_DIR, 'data', 'model', 'multi-classification.h5')
@@ -12,9 +12,10 @@ CLASS_NAMES = ['Economia', 'Politica', 'Scienza_e_tecnica', 'Sport', 'Storia']
 
 # Caricamento Singleton (lo carichiamo una volta sola)
 model = None
+tokenizer = None
 
 def load_resources():
-    global model
+    global model, tokenizer
     if model is None and os.path.exists(MODEL_PATH):
         try:
             # Carica Tokenizer
@@ -29,10 +30,10 @@ def load_resources():
             print(f"Errore caricamento risorse MultiLabel: {e}")
             return None, None
 
-# Carichiamo una volta sola all'avvio (Singleton) per velocità
-model, tokenizer = load_resources()
-
 def multi_classification(text):
+    global model, tokenizer
+    if model is None or tokenizer is None:
+        load_resources()
     if model is None or tokenizer is None:
         return {"Errore": "Modello non caricato"}
     try:

modules/retina.py

@@ -19,11 +19,11 @@ def load_resources():
         except Exception as e:
             print(f"❌ Errore caricamento modello Retina: {e}")
 
-# Caricamento all'avvio
-load_resources()
-
 def predict_diabetic_retinopathy(image_array):
     # Restituiamo sempre DUE valori (Diagnosi, Percentuale) anche in caso di errore
+    global model
+    if model is None:
+        load_resources()
     if model is None:
         return "❌ Errore: Modello non trovato", "0%"
     

modules/utilities/logger.py

@@ -42,14 +42,27 @@ if not LOG_FILE.exists() or LOG_FILE.stat().st_size == 0:
             "ip_address", "user_agent", "language", "input_size", "input_text" ,"processing_time"
         ])
 
-scheduler = CommitScheduler(
-    repo_id=DATASET_REPO_ID,
-    repo_type="dataset",
-    folder_path=LOG_DIR,
-    path_in_repo="logs",
-    every=TIME,
-    token=HF_TOKEN
-)
+class DummyScheduler:
+    def __init__(self):
+        class Lock:
+            def __enter__(self): pass
+            def __exit__(self, exc_type, exc_val, exc_tb): pass
+        self.lock = Lock()
+
+try:
+    scheduler = CommitScheduler(
+        repo_id=DATASET_REPO_ID,
+        repo_type="dataset",
+        folder_path=LOG_DIR,
+        path_in_repo="logs",
+        every=TIME,
+        token=HF_TOKEN
+    )
+    print("✅ CommitScheduler caricato con successo.")
+except Exception as e:
+    print(f"⚠️ Impossibile inizializzare CommitScheduler: {e}")
+    print("ℹ️ I log verranno salvati solo in locale.")
+    scheduler = DummyScheduler()
 
 def log_interaction(request, module_name, action, input_data=None, execution_time=0.0):
     try:

modules/utilities/utils.py

@@ -1,8 +1,8 @@
 import json
 import gradio as gr
 import os
-from keras.preprocessing.text import Tokenizer
-from keras.preprocessing.text import tokenizer_from_json
+from keras_preprocessing.text import Tokenizer
+from keras_preprocessing.text import tokenizer_from_json
 
 def load_doc(filename):
 	# open the file as read only 

style.css

@@ -180,4 +180,21 @@ button.primary:hover { filter: brightness(1.1); box-shadow: 0 4px 15px rgba(139,
     }
 
 }
-footer {visibility: hidden}
+footer {visibility: hidden}
+
+/* Sentiment info badge for BPO Dispatcher */
+.sentiment-info-badge {
+    background: #f0fdf4;
+    border: 1px solid #bbf7d0;
+    border-radius: 8px;
+    padding: 12px;
+    font-size: 0.9em;
+    color: #166534;
+    margin-top: 15px;
+    margin-bottom: 15px;
+    line-height: 1.4;
+}
+
+.sentiment-info-badge strong {
+    color: #14532d !important;
+}