Update app.py

app.py

@@ -31,7 +31,11 @@ SESSION_TOKEN_LENGTH = 32
 HF_TOKEN = os.getenv("HF_TOKEN")
 
 # Initialize logging
-logging.basicConfig(filename='app.log', level=logging.INFO)
+logging.basicConfig(
+    filename='app.log', 
+    level=logging.INFO,
+    format='%(asctime)s - %(levelname)s - %(message)s'
+)
 
 # Model configuration
 MODEL_CHOICES = {
@@ -43,8 +47,11 @@ DEFAULT_MODEL = "TinyLlama (Fastest)"
 
 # Initialize Hugging Face API
 if HF_TOKEN:
-    hf_api = HfApi(token=HF_TOKEN)
-    HfFolder.save_token(HF_TOKEN)
+    try:
+        hf_api = HfApi(token=HF_TOKEN)
+        HfFolder.save_token(HF_TOKEN)
+    except Exception as e:
+        logging.error(f"Failed to initialize Hugging Face API: {str(e)}")
 
 # ========== OPTIMIZED MODEL LOADING ==========
 class ModelLoader:
@@ -56,15 +63,17 @@ class ModelLoader:
         self.error = None
         self.current_model = None
     
-    def load_model(self, model_name, progress=gr.Progress()):
+    def load_model(self, model_name: str, progress: gr.Progress = None) -> Tuple[Optional[AutoModelForCausalLM], Optional[AutoTokenizer]]:
         """Lazy load the model with progress feedback"""
         if self.loaded and self.current_model == model_name:
             return self.model, self.tokenizer
             
         self.loading = True
         self.error = None
+        
         try:
-            progress(0.1, desc="Initializing...")
+            if progress:
+                progress(0.1, desc="Initializing...")
             
             # Clear previous model if any
             if self.model:
@@ -84,25 +93,29 @@ class ModelLoader:
             if "TinyLlama" in model_name:
                 model_kwargs["attn_implementation"] = "flash_attention_2"
             
-            progress(0.3, desc="Loading tokenizer...")
+            if progress:
+                progress(0.3, desc="Loading tokenizer...")
             self.tokenizer = AutoTokenizer.from_pretrained(
                 MODEL_CHOICES[model_name],
                 trust_remote_code=True
             )
             
-            progress(0.6, desc="Loading model...")
+            if progress:
+                progress(0.6, desc="Loading model...")
             self.model = AutoModelForCausalLM.from_pretrained(
                 MODEL_CHOICES[model_name],
                 **model_kwargs
             )
             
             # Verify model responsiveness
-            progress(0.8, desc="Verifying model...")
+            if progress:
+                progress(0.8, desc="Verifying model...")
             test_input = self.tokenizer("Test", return_tensors="pt").to(self.model.device)
             _ = self.model.generate(**test_input, max_new_tokens=1)
             
             self.model.eval()  # Disable dropout
-            progress(0.9, desc="Finalizing...")
+            if progress:
+                progress(0.9, desc="Finalizing...")
             self.loaded = True
             self.current_model = model_name
             return self.model, self.tokenizer
@@ -113,7 +126,7 @@ class ModelLoader:
             return None, None
         except Exception as e:
             self.error = str(e)
-            logging.error(f"Model loading error: {self.error}")
+            logging.error(f"Model loading error: {str(e)}")
             return None, None
         finally:
             self.loading = False
@@ -394,10 +407,12 @@ def parse_transcript_with_ai(text: str, progress=gr.Progress()) -> Dict:
     
     # First try the structured parser
     try:
-        progress(0.1, desc="Parsing transcript structure...")
+        if progress:
+            progress(0.1, desc="Parsing transcript structure...")
         parser = TranscriptParser()
         parsed_data = parser.parse_transcript(text)
-        progress(0.9, desc="Formatting results...")
+        if progress:
+            progress(0.9, desc="Formatting results...")
         
         # Convert to expected format
         formatted_data = {
@@ -420,7 +435,8 @@ def parse_transcript_with_ai(text: str, progress=gr.Progress()) -> Dict:
                 "grade_level": course["grade_level"]
             })
         
-        progress(1.0)
+        if progress:
+            progress(1.0)
         return validate_parsed_data(formatted_data)
     
     except Exception as e:
@@ -452,11 +468,13 @@ def parse_transcript_with_ai_fallback(text: str, progress=gr.Progress()) -> Dict
     """
     
     try:
-        progress(0.1, desc="Processing transcript with AI...")
+        if progress:
+            progress(0.1, desc="Processing transcript with AI...")
         
         # Tokenize and generate response
         inputs = model_loader.tokenizer(prompt, return_tensors="pt").to(model_loader.model.device)
-        progress(0.4)
+        if progress:
+            progress(0.4)
         
         outputs = model_loader.model.generate(
             **inputs,
@@ -464,7 +482,8 @@ def parse_transcript_with_ai_fallback(text: str, progress=gr.Progress()) -> Dict
             temperature=0.1,
             do_sample=True
         )
-        progress(0.8)
+        if progress:
+            progress(0.8)
         
         # Decode the response
         response = model_loader.tokenizer.decode(outputs[0], skip_special_tokens=True)
@@ -478,7 +497,8 @@ def parse_transcript_with_ai_fallback(text: str, progress=gr.Progress()) -> Dict
             json_str = re.search(r'\{.*\}', response, re.DOTALL).group()
             parsed_data = json.loads(json_str)
         
-        progress(1.0)
+        if progress:
+            progress(1.0)
         return validate_parsed_data(parsed_data)
     
     except torch.cuda.OutOfMemoryError:
@@ -828,7 +848,7 @@ class ProfileManager:
                 json.dump(data, f, indent=2, ensure_ascii=False)
             
             # Upload to HF Hub if token is available
-            if HF_TOKEN:
+            if HF_TOKEN and hf_api:
                 try:
                     hf_api.upload_file(
                         path_or_fileobj=filepath,
@@ -867,7 +887,7 @@ class ProfileManager:
                 
                 if not profile_file.exists():
                     # Try loading from HF Hub
-                    if HF_TOKEN:
+                    if HF_TOKEN and hf_api:
                         try:
                             hf_api.download_file(
                                 path_in_repo=f"profiles/{profile_file.name}",
@@ -1379,7 +1399,7 @@ def create_interface():
                     except Exception as e:
                         logging.error(f"Upload error: {str(e)}")
                     return (
-                        "Error processing transcript. Please try again.", 
+                        f"Error processing transcript: {str(e)}", 
                         None, 
                         current_tab_status,
                         gr.update(),