Update app.py

app.py

@@ -1,203 +1,176 @@
-from smolagents import CodeAgent, HfApiModel, load_tool, tool
+from smolagents import CodeAgent, HfApiModel, tool
 import yaml
-import sys
+import requests
+import base64
+import io
 from tools.final_answer import FinalAnswerTool
 from Gradio_UI import GradioUI
 
-# Force logging to stdout
-import logging
-logging.basicConfig(
-    level=logging.DEBUG,
-    format='%(asctime)s - %(name)s - %(levelname)s - %(message)s',
-    handlers=[
-        logging.StreamHandler(sys.stdout),
-        logging.FileHandler('debug.log')
-    ]
-)
-logger = logging.getLogger(__name__)
-
-# Print function for immediate visibility
-def debug_print(message):
-    print(f"DEBUG: {message}")
-    logger.info(message)
-
-# Alternative image generation approach
+# Create a custom image generation tool that returns a string instead of AgentImage
 @tool
-def manual_image_generator(prompt: str) -> str:
-    """Generate an image using direct API calls with debugging.
+def text_to_image_generator(prompt: str) -> str:
+    """Generate an image from text and return the result as a string with image details.
     Args:
-        prompt: A detailed text description of the image to generate
+        prompt: A detailed description of the image to generate
     """
-    debug_print(f"=== MANUAL IMAGE GENERATION START ===")
-    debug_print(f"Prompt: {prompt}")
-    
     try:
-        import requests
-        import base64
-        from PIL import Image
-        import io
+        print(f"🎨 Starting image generation for: {prompt}")
         
-        # Try Hugging Face Inference API directly
-        API_URL = "https://api-inference.huggingface.co/models/stabilityai/stable-diffusion-2-1"
-        headers = {"Authorization": "Bearer hf_your_token_here"}  # You'll need to add your token
+        # Method 1: Try Hugging Face Inference API
+        api_urls = [
+            "https://api-inference.huggingface.co/models/stabilityai/stable-diffusion-2-1",
+            "https://api-inference.huggingface.co/models/runwayml/stable-diffusion-v1-5",
+            "https://api-inference.huggingface.co/models/stabilityai/stable-diffusion-xl-base-1.0"
+        ]
         
-        debug_print("Attempting direct API call...")
+        for api_url in api_urls:
+            try:
+                print(f"🔄 Trying API: {api_url}")
+                
+                # Make request without auth token first (some models allow public access)
+                response = requests.post(
+                    api_url,
+                    json={"inputs": prompt},
+                    timeout=30
+                )
+                
+                print(f"📡 API Response: {response.status_code}")
+                
+                if response.status_code == 200:
+                    print("✅ Image generated successfully!")
+                    
+                    # Save the image
+                    image_filename = f"generated_image_{hash(prompt) % 10000}.png"
+                    with open(image_filename, 'wb') as f:
+                        f.write(response.content)
+                    
+                    image_size = len(response.content)
+                    print(f"💾 Image saved as: {image_filename} (Size: {image_size} bytes)")
+                    
+                    return f"""✅ IMAGE GENERATED SUCCESSFULLY!
+
+📝 Prompt: {prompt}
+🖼️ File: {image_filename}
+📦 Size: {image_size} bytes
+🌐 API: {api_url.split('/')[-1]}
+⏰ Status: Ready for viewing
+
+The image has been generated and saved successfully. You can view it in your file system."""
+                    
+                elif response.status_code == 503:
+                    print("⏳ Model is loading, trying next option...")
+                    continue
+                elif response.status_code == 429:
+                    print("⚠️ Rate limit exceeded, trying next option...")
+                    continue
+                else:
+                    print(f"❌ API Error: {response.status_code} - {response.text}")
+                    continue
+                    
+            except requests.exceptions.RequestException as e:
+                print(f"🔗 Connection error: {e}")
+                continue
+            except Exception as e:
+                print(f"🚫 Unexpected error: {e}")
+                continue
         
-        # Make the API call
-        response = requests.post(API_URL, headers=headers, json={"inputs": prompt})
-        debug_print(f"API Response status: {response.status_code}")
+        # Method 2: If all APIs fail, create a detailed description
+        print("🎭 Falling back to detailed description mode...")
         
-        if response.status_code == 200:
-            debug_print("✅ API call successful")
-            # Convert response to image
-            image_bytes = response.content
-            image = Image.open(io.BytesIO(image_bytes))
-            debug_print(f"Generated image size: {image.size}")
-            
-            # Save the image temporarily
-            image_path = f"generated_image.png"
-            image.save(image_path)
-            debug_print(f"Image saved to: {image_path}")
-            
-            return f"Image generated successfully! Size: {image.size}"
-            
-        else:
-            debug_print(f"❌ API call failed: {response.status_code}")
-            debug_print(f"Response: {response.text}")
-            return f"API Error: {response.status_code} - {response.text}"
-            
-    except ImportError as e:
-        debug_print(f"Missing dependencies: {e}")
-        return "Error: Missing required libraries (requests, PIL)"
-    except Exception as e:
-        debug_print(f"Manual generation failed: {e}")
-        return f"Error: {str(e)}"
+        description = f"""🎨 IMAGE CONCEPT GENERATED
 
-# Simplified approach - try to bypass the AgentImage issue
-@tool
-def simple_image_describer(prompt: str) -> str:
-    """Describe what an image would look like instead of generating it.
-    Args:
-        prompt: A text description of the desired image
-    """
-    debug_print(f"=== SIMPLE IMAGE DESCRIPTION ===")
-    debug_print(f"Input: {prompt}")
-    
-    description = f"""
-    IMAGE DESCRIPTION for: "{prompt}"
-    
-    This would be a detailed visual representation showing:
-    - Main subject: {prompt.split()[0] if prompt else 'unknown'}
-    - Style: Realistic digital art
-    - Composition: Well-balanced and visually appealing
-    - Colors: Vibrant and harmonious
-    - Quality: High resolution, professional quality
-    
-    [Note: This is a text description. In a working system, this would be an actual generated image]
-    """
-    
-    debug_print("Description created successfully")
-    return description
+📝 Original Prompt: "{prompt}"
+
+🖼️ Detailed Visual Description:
+This image would feature {prompt.lower()}, rendered in high quality with:
+
+🎯 Main Elements:
+- Primary subject matter as described in the prompt
+- Professional composition and framing
+- Balanced lighting and shadows
+- Rich, vibrant colors
+
+🎨 Style Characteristics:
+- Digital art quality
+- Realistic rendering
+- High resolution output
+- Professional photography aesthetic
+
+📐 Technical Specs:
+- Dimensions: 512x512 pixels (standard)
+- Format: PNG with transparency support
+- Quality: Professional grade
+- Style: Photorealistic
+
+💡 Note: This is a detailed concept description. In a fully functional system, this would be accompanied by the actual generated image file."""
+
+        print("📋 Description generated successfully!")
+        return description
+        
+    except Exception as e:
+        error_msg = f"❌ Image generation failed: {str(e)}"
+        print(error_msg)
+        return error_msg
 
 # Initialize components
-debug_print("=== INITIALIZING COMPONENTS ===")
+print("🚀 Initializing Image Generation Agent...")
+
 final_answer = FinalAnswerTool()
 
-# Create model
-debug_print("Creating model...")
+# Create model with optimized settings
 model = HfApiModel(
     max_tokens=1024,
     temperature=0.7,
     model_id='Qwen/Qwen2.5-Coder-32B-Instruct',
     custom_role_conversions=None,
 )
-debug_print("✅ Model created")
-
-# Try original tool loading with detailed debugging
-debug_print("=== ATTEMPTING ORIGINAL TOOL LOADING ===")
-original_tool = None
-try:
-    debug_print("Loading agents-course/text-to-image...")
-    original_tool = load_tool("agents-course/text-to-image", trust_remote_code=True)
-    debug_print(f"✅ Original tool loaded: {type(original_tool)}")
-    
-    # Test the original tool
-    debug_print("Testing original tool...")
-    test_result = original_tool("test red circle")
-    debug_print(f"Original tool test result: {type(test_result)}")
-    debug_print(f"Test result details: {test_result}")
-    
-    if hasattr(test_result, 'size'):
-        debug_print(f"Test result size: {test_result.size}")
-        if test_result.size == (0, 0):
-            debug_print("⚠️ FOUND THE ISSUE: Original tool returns size 0x0")
-            debug_print("This indicates the Hugging Face model is not responding properly")
-        else:
-            debug_print("✅ Original tool working correctly")
-    
-except Exception as e:
-    debug_print(f"❌ Original tool failed: {e}")
-    import traceback
-    debug_print(f"Full traceback: {traceback.format_exc()}")
 
 # Load prompts
-debug_print("Loading prompts...")
 try:
     with open("prompts.yaml", 'r') as stream:
         prompt_templates = yaml.safe_load(stream)
-    debug_print("✅ Prompts loaded from file")
+    print("✅ Loaded custom prompts")
 except:
     prompt_templates = {
-        "system": "You are an AI assistant. Use the available tools to help users with their requests.",
-        "user": "{input}"
+        "system": """You are an AI assistant specialized in image generation. 
+        When users request images, use the text_to_image_generator tool with detailed, descriptive prompts.
+        Always provide clear feedback about the image generation process.""",
+        "user": "Request: {input}"
     }
-    debug_print("✅ Using default prompts")
+    print("✅ Using default prompts")
 
-# Create tools list with multiple options
-debug_print("=== CREATING TOOLS LIST ===")
-tools_list = [final_answer]
-
-# Add image tools in order of preference
-if original_tool:
-    tools_list.append(original_tool)
-    debug_print("✅ Added original image tool")
-
-tools_list.append(simple_image_describer)
-debug_print("✅ Added simple image describer")
-
-tools_list.append(manual_image_generator)
-debug_print("✅ Added manual image generator")
-
-debug_print(f"Total tools: {len(tools_list)}")
-
-# Create agent
-debug_print("=== CREATING AGENT ===")
+# Create agent with simplified tool set
 agent = CodeAgent(
     model=model,
-    tools=tools_list,
+    tools=[final_answer, text_to_image_generator],
     max_steps=3,
-    verbosity_level=2,
+    verbosity_level=1,
     grammar=None,
     planning_interval=None,
-    name="MultiMethodImageAgent",
-    description="AI agent with multiple image generation methods and comprehensive debugging",
+    name="StringBasedImageAgent",
+    description="AI agent that generates images and returns detailed text descriptions",
     prompt_templates=prompt_templates
 )
-debug_print("✅ Agent created")
+
+print("✅ Agent initialized successfully")
+print("🎯 Available tools: FinalAnswer, TextToImageGenerator")
 
 # Launch function
-def launch_debug_agent():
-    debug_print("=== LAUNCHING DEBUG AGENT ===")
-    debug_print(f"Tools available: {[getattr(t, '__name__', str(t)) for t in tools_list]}")
-    
+def launch_agent():
     try:
-        debug_print("Starting Gradio UI...")
+        print("🌐 Starting Gradio interface...")
         GradioUI(agent).launch()
     except Exception as e:
-        debug_print(f"Launch failed: {e}")
-        import traceback
-        debug_print(f"Full traceback: {traceback.format_exc()}")
+        print(f"❌ Launch failed: {e}")
+        # Try basic launch
+        try:
+            print("🔄 Trying basic launch...")
+            GradioUI(agent).launch()
+        except Exception as e2:
+            print(f"💥 Basic launch also failed: {e2}")
 
 if __name__ == "__main__":
-    debug_print("=== STARTING APPLICATION ===")
-    launch_debug_agent()
+    print("=" * 50)
+    print("🎨 STRING-BASED IMAGE GENERATION AGENT")
+    print("=" * 50)
+    launch_agent()