Update app.py

app.py

@@ -6,23 +6,20 @@ import time
 from PIL import Image as PILImage
 import io
 import os
-import base64
+from tempfile import NamedTemporaryFile
+from pathlib import Path
 
 def create_monitor_interface():
     api_key = os.getenv("GROQ_API_KEY")
     
     class SafetyMonitor:
-        def __init__(self, model_name: str = "llama-3.2-90b-vision-preview"):
+        def __init__(self):
             self.client = Groq(api_key=api_key)
-            self.model_name = model_name
-            self.max_image_size = (128, 128)  # Drastically reduced size
-            self.jpeg_quality = 20  # Very low quality
-
+            self.model_name = "llama-3.2-90b-vision-preview"
+            self.max_image_size = (128, 128)
+            
         def resize_image(self, image):
-            """Resize image while maintaining aspect ratio"""
             height, width = image.shape[:2]
-            
-            # Calculate aspect ratio
             aspect = width / height
             
             if width > height:
@@ -32,8 +29,7 @@ def create_monitor_interface():
                 new_height = min(self.max_image_size[1], height)
                 new_width = int(new_height * aspect)
                 
-            resized = cv2.resize(image, (new_width, new_height), interpolation=cv2.INTER_AREA)
-            return resized
+            return cv2.resize(image, (new_width, new_height), interpolation=cv2.INTER_AREA)
 
         def analyze_frame(self, frame: np.ndarray) -> str:
             if frame is None:
@@ -48,23 +44,37 @@ def create_monitor_interface():
             frame = self.resize_image(frame)
             frame_pil = PILImage.fromarray(frame)
             
-            # Convert to base64 with minimal size
+            # Convert to base64 with minimal quality
             buffered = io.BytesIO()
             frame_pil.save(buffered, 
                          format="JPEG", 
-                         quality=self.jpeg_quality, 
+                         quality=20, 
                          optimize=True)
             img_base64 = base64.b64encode(buffered.getvalue()).decode('utf-8')
             
             try:
-                # Minimal prompt
-                prompt = f"""List safety issues: <image>data:image/jpeg;base64,{img_base64}</image>"""
-
                 completion = self.client.chat.completions.create(
                     messages=[
                         {
                             "role": "user",
-                            "content": prompt
+                            "content": "You are a workplace safety expert. Analyze the following image for safety concerns."
+                        },
+                        {
+                            "role": "assistant",
+                            "content": "I'll analyze the image for workplace safety concerns and provide specific observations."
+                        },
+                        {
+                            "role": "user",
+                            "content": [
+                                {
+                                    "type": "text",
+                                    "text": "What safety issues do you see?"
+                                },
+                                {
+                                    "type": "image_url",
+                                    "url": f"data:image/jpeg;base64,{img_base64}"
+                                }
+                            ]
                         }
                     ],
                     model=self.model_name,