Update app.py

app.py

@@ -8,6 +8,7 @@ import json
 import csv
 import os
 import tempfile
+import time
 from datetime import datetime
 from transformers import AutoTokenizer, AutoModelForSequenceClassification
 import torch.nn.functional as F
@@ -23,15 +24,40 @@ load_dotenv()
 aai.settings.api_key = os.getenv("ASSEMBLYAI_API_KEY")
 hf_token = os.getenv("HF_TOKEN")
 
+if not hf_token:
+    raise ValueError("❌ HF_TOKEN is missing. Add it to your .env")
+
 device = "cuda" if torch.cuda.is_available() else "cpu"
 
-tokenizer = AutoTokenizer.from_pretrained(
-    "nlptown/bert-base-multilingual-uncased-sentiment"
-)
+MODEL_NAME = "nlptown/bert-base-multilingual-uncased-sentiment"
+
+# =========================
+# LOAD MODEL (FIXED)
+# =========================
+def load_hf_model():
+    for attempt in range(3):
+        try:
+            tokenizer = AutoTokenizer.from_pretrained(
+                MODEL_NAME,
+                token=hf_token,
+                cache_dir="./models"
+            )
+
+            model = AutoModelForSequenceClassification.from_pretrained(
+                MODEL_NAME,
+                token=hf_token,
+                cache_dir="./models"
+            )
 
-model = AutoModelForSequenceClassification.from_pretrained(
-    "nlptown/bert-base-multilingual-uncased-sentiment"
-)
+            return tokenizer, model
+
+        except Exception as e:
+            print(f"⚠️ HF load failed (attempt {attempt+1}): {e}")
+            time.sleep(5)
+
+    raise RuntimeError("❌ Failed to load Hugging Face model after retries")
+
+tokenizer, model = load_hf_model()
 
 model.to(device)
 model.eval()
@@ -73,7 +99,12 @@ def build_segments(transcript):
 
 
 def analyze_text(text):
-    inputs = tokenizer(text[:512], return_tensors="pt", truncation=True).to(device)
+    inputs = tokenizer(
+        text,
+        return_tensors="pt",
+        truncation=True,
+        max_length=512
+    ).to(device)
 
     with torch.no_grad():
         logits = model(**inputs).logits
@@ -92,14 +123,13 @@ def process_audio(file, speakers=0, language="auto"):
         return "❌ No audio provided", "", ""
 
     path = file if isinstance(file, str) else file.name
-
     temp_path = None
 
     try:
         # Load audio
         audio, sr = librosa.load(path, sr=None, mono=True)
 
-        # 🔥 Create TEMP FILE (not saved permanently)
+        # Create TEMP FILE
         with tempfile.NamedTemporaryFile(delete=False, suffix=".wav") as tmp:
             sf.write(tmp.name, audio, sr)
             temp_path = tmp.name
@@ -116,7 +146,6 @@ def process_audio(file, speakers=0, language="auto"):
             return f"❌ {transcript.error}", "", ""
 
         global_segments = build_segments(transcript)
-
         speaker_count = len(set(s["speaker"] for s in global_segments))
 
         label_map = {
@@ -147,7 +176,6 @@ def process_audio(file, speakers=0, language="auto"):
         return f"❌ Error: {str(e)}", "", ""
 
     finally:
-        # 🔥 DELETE temp file ALWAYS
         if temp_path and os.path.exists(temp_path):
             os.remove(temp_path)
 
@@ -240,6 +268,5 @@ with gr.Blocks(title="AI Conversation Sentiment System") as app:
         outputs=[download]
     )
 
-
 if __name__ == "__main__":
     app.launch(theme=gr.themes.Soft())