app.py

import spaces
import gradio as gr
import torch
from PIL import Image
import numpy as np
from clip_interrogator import Config, Interrogator
import logging
import os
import warnings
from datetime import datetime
import gc
import re

warnings.filterwarnings("ignore", category=FutureWarning)
warnings.filterwarnings("ignore", category=UserWarning)
os.environ["TOKENIZERS_PARALLELISM"] = "false"

logging.basicConfig(level=logging.INFO)
logger = logging.getLogger(__name__)

def get_device():
    if torch.cuda.is_available():
        return "cuda"
    elif torch.backends.mps.is_available():
        return "mps"
    else:
        return "cpu"

DEVICE = get_device()

class DeepFluxAnalyzer:
    """
    Deep analysis engine that understands image content and applies Flux rules intelligently
    """
    
    def __init__(self):
        self.forbidden_elements = ["++", "weights", "white background [en dev]"]
        
        # Deep vocabulary for intelligent analysis
        self.age_descriptors = {
            "young": ["young", "youthful", "fresh-faced"],
            "middle": ["middle-aged", "mature"], 
            "elderly": ["elderly", "aged", "distinguished", "weathered"]
        }
        
        self.facial_features = {
            "beard": ["bearded", "with a full beard", "with facial hair", "with a silver beard", "with a gray beard"],
            "glasses": ["wearing glasses", "with wire-frame glasses", "with spectacles", "with eyeglasses"],
            "eyes": ["intense gaze", "piercing eyes", "contemplative expression", "focused stare"]
        }
        
        self.clothing_religious = {
            "hat": ["black hat", "traditional hat", "religious headwear", "Orthodox hat"],
            "clothing": ["traditional clothing", "religious attire", "formal wear", "dark clothing"]
        }
        
        self.settings_detailed = {
            "indoor": ["indoor setting", "interior space", "indoor environment"],
            "outdoor": ["outdoor setting", "natural environment", "exterior location"],
            "studio": ["studio setting", "controlled environment", "professional backdrop"]
        }
        
        self.lighting_advanced = {
            "portrait": ["dramatic portrait lighting", "studio portrait lighting", "professional portrait setup"],
            "natural": ["natural lighting", "window light", "ambient illumination"],
            "dramatic": ["dramatic lighting", "high contrast lighting", "chiaroscuro lighting"]
        }
        
        self.technical_professional = {
            "portrait_lens": ["85mm lens", "135mm lens", "medium telephoto"],
            "standard_lens": ["50mm lens", "35mm lens", "standard focal length"],
            "aperture": ["f/1.4 aperture", "f/2.8 aperture", "f/4 aperture"],
            "camera": ["Shot on Phase One XF", "Shot on Hasselblad", "Shot on Canon EOS R5"]
        }
    
    def analyze_clip_deeply(self, clip_result):
        """Extract detailed information from CLIP analysis"""
        clip_lower = clip_result.lower()
        analysis = {
            "subjects": [],
            "age": None,
            "features": [],
            "clothing": [],
            "setting": None,
            "mood": None,
            "composition": None
        }
        
        # Subject and age detection
        if any(word in clip_lower for word in ["man", "person", "male"]):
            if any(word in clip_lower for word in ["old", "elderly", "aged", "gray", "grey", "silver"]):
                analysis["subjects"].append("elderly man")
                analysis["age"] = "elderly"
            elif any(word in clip_lower for word in ["young", "youth", "boy"]):
                analysis["subjects"].append("young man")  
                analysis["age"] = "young"
            else:
                analysis["subjects"].append("man")
                analysis["age"] = "middle"
        
        if any(word in clip_lower for word in ["woman", "female", "lady"]):
            if any(word in clip_lower for word in ["old", "elderly", "aged"]):
                analysis["subjects"].append("elderly woman")
                analysis["age"] = "elderly"
            else:
                analysis["subjects"].append("woman")
        
        # Facial features detection
        if any(word in clip_lower for word in ["beard", "facial hair", "mustache"]):
            if any(word in clip_lower for word in ["gray", "grey", "silver", "white"]):
                analysis["features"].append("silver beard")
            else:
                analysis["features"].append("beard")
        
        if any(word in clip_lower for word in ["glasses", "spectacles", "eyeglasses"]):
            analysis["features"].append("glasses")
        
        # Clothing and accessories
        if any(word in clip_lower for word in ["hat", "cap", "headwear"]):
            analysis["clothing"].append("hat")
        
        if any(word in clip_lower for word in ["suit", "formal", "dress", "shirt"]):
            analysis["clothing"].append("formal wear")
            
        # Setting detection  
        if any(word in clip_lower for word in ["indoor", "inside", "interior", "room"]):
            analysis["setting"] = "indoor"
        elif any(word in clip_lower for word in ["outdoor", "outside", "landscape", "street"]):
            analysis["setting"] = "outdoor"
        elif any(word in clip_lower for word in ["studio", "backdrop"]):
            analysis["setting"] = "studio"
        
        # Mood and composition
        if any(word in clip_lower for word in ["portrait", "headshot", "face", "close-up"]):
            analysis["composition"] = "portrait"
        elif any(word in clip_lower for word in ["sitting", "seated", "chair"]):
            analysis["composition"] = "seated"
        elif any(word in clip_lower for word in ["standing", "upright"]):
            analysis["composition"] = "standing"
            
        return analysis
    
    def build_flux_prompt(self, analysis, clip_base):
        """Build optimized Flux prompt using deep analysis"""
        components = []
        
        # 1. Article (intelligent selection)
        if analysis["subjects"]:
            subject = analysis["subjects"][0]
            article = "An" if subject[0] in 'aeiou' else "A"
        else:
            article = "A"
        components.append(article)
        
        # 2. Descriptive adjectives (context-aware)
        adjectives = []
        if analysis["age"] == "elderly":
            adjectives.extend(["distinguished", "weathered"])
        elif analysis["age"] == "young":
            adjectives.extend(["young", "fresh-faced"])
        else:
            adjectives.extend(["professional", "elegant"])
        
        # Add up to 2-3 adjectives as per Flux rules
        components.extend(adjectives[:2])
        
        # 3. Main subject (enhanced with details)
        if analysis["subjects"]:
            main_subject = analysis["subjects"][0]
            # Add religious/cultural context if detected
            if "hat" in analysis["clothing"] and "beard" in [f.split()[0] for f in analysis["features"]]:
                main_subject = "Orthodox Jewish " + main_subject
        else:
            main_subject = "subject"
        components.append(main_subject)
        
        # 4. Features integration (intelligent placement)
        feature_descriptions = []
        if "glasses" in analysis["features"]:
            feature_descriptions.append("with distinctive wire-frame glasses")
        if any("beard" in f for f in analysis["features"]):
            if "silver beard" in analysis["features"]:
                feature_descriptions.append("with a distinguished silver beard")
            else:
                feature_descriptions.append("with a full beard")
        
        if feature_descriptions:
            components.extend(feature_descriptions)
        
        # 5. Clothing and accessories
        clothing_desc = []
        if "hat" in analysis["clothing"]:
            clothing_desc.append("wearing a traditional black hat")
        if "formal wear" in analysis["clothing"]:
            clothing_desc.append("in formal attire")
            
        if clothing_desc:
            components.extend(clothing_desc)
        
        # 6. Verb/Action (based on composition analysis)
        if analysis["composition"] == "seated":
            action = "seated contemplatively"
        elif analysis["composition"] == "standing":
            action = "standing with dignity"
        else:
            action = "positioned thoughtfully"
        components.append(action)
        
        # 7. Context/Location (enhanced setting)
        setting_map = {
            "indoor": "in an intimate indoor setting",
            "outdoor": "in a natural outdoor environment",
            "studio": "in a professional studio environment"
        }
        
        if analysis["setting"]:
            context = setting_map.get(analysis["setting"], "in a carefully composed environment")
        else:
            context = "in a thoughtfully arranged scene"
        components.append(context)
        
        # 8. Environmental details (lighting-aware)
        if analysis["composition"] == "portrait":
            env_detail = "with dramatic portrait lighting that emphasizes facial features and texture"
        else:
            env_detail = "captured with sophisticated atmospheric lighting"
        components.append(env_detail)
        
        # 9. Technical specifications (composition-appropriate)
        if analysis["composition"] == "portrait":
            tech_spec = "Shot on Phase One XF, 85mm lens, f/2.8 aperture"
        else:
            tech_spec = "Shot on Phase One, 50mm lens, f/4 aperture"
        components.append(tech_spec)
        
        # 10. Quality marker (always professional)
        components.append("professional photography")
        
        # Join with proper punctuation
        prompt = ", ".join(components)
        
        # Clean up and optimize
        prompt = re.sub(r'\s+', ' ', prompt)  # Remove extra spaces
        prompt = prompt.replace(", ,", ",")   # Remove double commas
        
        return prompt
    
    def calculate_intelligence_score(self, prompt, analysis):
        """Calculate how well the prompt reflects intelligent analysis"""
        score = 0
        
        # Structure compliance (Flux rules 1-10)
        if prompt.startswith(("A", "An")):
            score += 10
            
        # Feature recognition accuracy
        if len(analysis["features"]) > 0:
            score += 15
            
        # Context understanding
        if analysis["setting"]:
            score += 15
            
        # Subject detail depth
        if len(analysis["subjects"]) > 0:
            score += 15
            
        # Technical specs presence
        if "Phase One" in prompt and "lens" in prompt:
            score += 15
            
        # Lighting specification
        if "lighting" in prompt:
            score += 10
            
        # Composition awareness
        if analysis["composition"]:
            score += 10
            
        # Forbidden elements check
        if not any(forbidden in prompt for forbidden in self.forbidden_elements):
            score += 10
            
        return min(score, 100)

class FluxPromptOptimizer:
    def __init__(self):
        self.interrogator = None
        self.analyzer = DeepFluxAnalyzer()
        self.usage_count = 0
        self.device = DEVICE
        self.is_initialized = False
    
    def initialize_model(self):
        if self.is_initialized:
            return True
            
        try:
            config = Config(
                clip_model_name="ViT-L-14/openai",
                download_cache=True,
                chunk_size=2048,
                quiet=True,
                device=self.device
            )
            
            self.interrogator = Interrogator(config)
            self.is_initialized = True
            
            if self.device == "cpu":
                gc.collect()
            else:
                torch.cuda.empty_cache()
                
            return True
            
        except Exception as e:
            logger.error(f"Initialization error: {e}")
            return False
    
    def optimize_image(self, image):
        if image is None:
            return None
            
        if isinstance(image, np.ndarray):
            image = Image.fromarray(image)
        elif not isinstance(image, Image.Image):
            image = Image.open(image)
        
        if image.mode != 'RGB':
            image = image.convert('RGB')
        
        max_size = 768 if self.device != "cpu" else 512
        if image.size[0] > max_size or image.size[1] > max_size:
            image.thumbnail((max_size, max_size), Image.Resampling.LANCZOS)
        
        return image
    
    @spaces.GPU
    def generate_optimized_prompt(self, image):
        try:
            if not self.is_initialized:
                if not self.initialize_model():
                    return "❌ Model initialization failed.", "Please refresh and try again.", 0
            
            if image is None:
                return "❌ Please upload an image.", "No image provided.", 0
            
            self.usage_count += 1
            
            image = self.optimize_image(image)
            if image is None:
                return "❌ Image processing failed.", "Invalid image format.", 0
            
            start_time = datetime.now()
            
            # Get comprehensive CLIP analysis
            clip_result = self.interrogator.interrogate(image)
            
            # Deep analysis of the CLIP result
            deep_analysis = self.analyzer.analyze_clip_deeply(clip_result)
            
            # Build optimized Flux prompt
            optimized_prompt = self.analyzer.build_flux_prompt(deep_analysis, clip_result)
            
            # Calculate intelligence score
            score = self.analyzer.calculate_intelligence_score(optimized_prompt, deep_analysis)
            
            end_time = datetime.now()
            duration = (end_time - start_time).total_seconds()
            
            # Memory cleanup
            if self.device == "cpu":
                gc.collect()
            else:
                torch.cuda.empty_cache()
            
            # Generate detailed analysis info
            gpu_status = "⚡ ZeroGPU" if torch.cuda.is_available() else "💻 CPU"
            
            features_detected = ", ".join(deep_analysis["features"]) if deep_analysis["features"] else "None"
            subjects_detected = ", ".join(deep_analysis["subjects"]) if deep_analysis["subjects"] else "Generic"
            
            analysis_info = f"""**Deep Analysis Complete**

**Processing:** {gpu_status} • {duration:.1f}s  
**Intelligence Score:** {score}/100  
**Generation:** #{self.usage_count}  

**Detected Elements:**
• **Subjects:** {subjects_detected}
• **Features:** {features_detected}  
• **Setting:** {deep_analysis["setting"] or "Unspecified"}
• **Composition:** {deep_analysis["composition"] or "Standard"}

**CLIP Base:** {clip_result[:80]}...  
**Flux Enhancement:** Applied deep analysis with Pariente AI rules"""
            
            return optimized_prompt, analysis_info, score
            
        except Exception as e:
            logger.error(f"Generation error: {e}")
            return f"❌ Error: {str(e)}", "Please try with a different image.", 0

optimizer = FluxPromptOptimizer()

def process_image_wrapper(image):
    """Simplified wrapper - no unnecessary options"""
    try:
        prompt, info, score = optimizer.generate_optimized_prompt(image)
        
        # Create score HTML
        color = "#22c55e" if score >= 80 else "#f59e0b" if score >= 60 else "#ef4444"
        score_html = f'''
        <div style="text-align: center; padding: 1rem; background: linear-gradient(135deg, #f0fdf4 0%, #dcfce7 100%); border: 2px solid {color}; border-radius: 12px; margin: 1rem 0;">
            <div style="font-size: 2rem; font-weight: 700; color: {color}; margin: 0;">{score}</div>
            <div style="font-size: 0.875rem; color: #15803d; margin: 0; text-transform: uppercase; letter-spacing: 0.05em;">Intelligence Score</div>
        </div>
        '''
        
        return prompt, info, score_html
        
    except Exception as e:
        logger.error(f"Wrapper error: {e}")
        return "❌ Processing failed", f"Error: {str(e)}", '<div style="text-align: center; color: red;">Error</div>'

def clear_outputs():
    gc.collect()
    if torch.cuda.is_available():
        torch.cuda.empty_cache()
    return "", "", '<div style="text-align: center; padding: 1rem;"><div style="font-size: 2rem; color: #ccc;">--</div><div style="font-size: 0.875rem; color: #999;">Intelligence Score</div></div>'

def create_interface():
    css = """
    @import url('https://fonts.googleapis.com/css2?family=Inter:wght@300;400;500;600;700&display=swap');
    
    .gradio-container {
        max-width: 1200px !important;
        margin: 0 auto !important;
        font-family: 'Inter', -apple-system, BlinkMacSystemFont, sans-serif !important;
        background: linear-gradient(135deg, #f8fafc 0%, #f1f5f9 100%) !important;
    }
    
    .main-header {
        text-align: center;
        padding: 2rem 0 3rem 0;
        background: linear-gradient(135deg, #1e293b 0%, #334155 100%);
        color: white;
        margin: -2rem -2rem 2rem -2rem;
        border-radius: 0 0 24px 24px;
    }
    
    .main-title {
        font-size: 2.5rem !important;
        font-weight: 700 !important;
        margin: 0 0 0.5rem 0 !important;
        letter-spacing: -0.025em !important;
        background: linear-gradient(135deg, #60a5fa 0%, #3b82f6 100%);
        -webkit-background-clip: text;
        -webkit-text-fill-color: transparent;
        background-clip: text;
    }
    
    .subtitle {
        font-size: 1.125rem !important;
        font-weight: 400 !important;
        opacity: 0.8 !important;
        margin: 0 !important;
    }
    
    .prompt-output {
        font-family: 'SF Mono', 'Monaco', 'Inconsolata', 'Roboto Mono', monospace !important;
        font-size: 14px !important;
        line-height: 1.6 !important;
        background: linear-gradient(135deg, #ffffff 0%, #f8fafc 100%) !important;
        border: 1px solid #e2e8f0 !important;
        border-radius: 12px !important;
        padding: 1.5rem !important;
        box-shadow: 0 4px 6px -1px rgba(0, 0, 0, 0.1) !important;
    }
    """
    
    with gr.Blocks(
        theme=gr.themes.Soft(),
        title="Flux Prompt Optimizer",
        css=css
    ) as interface:
        
        gr.HTML("""
        <div class="main-header">
            <div class="main-title">⚡ Flux Prompt Optimizer</div>
            <div class="subtitle">Deep AI analysis • Intelligent prompt generation • Research-based optimization</div>
        </div>
        """)
        
        with gr.Row():
            with gr.Column(scale=1):
                gr.Markdown("## 📷 Image Analysis")
                
                image_input = gr.Image(
                    label="Upload your image",
                    type="pil",
                    height=400
                )
                
                optimize_btn = gr.Button(
                    "🧠 Analyze & Optimize",
                    variant="primary",
                    size="lg"
                )
                
                gr.Markdown("""
                ### Deep Analysis Engine
                
                This system performs comprehensive image analysis:
                
                • **Subject Recognition** - Identifies people, objects, context
                • **Feature Detection** - Facial features, clothing, accessories  
                • **Composition Analysis** - Lighting, setting, mood
                • **Flux Optimization** - Applies research-validated rules
                
                No options needed - the AI decides what's optimal.
                """)
            
            with gr.Column(scale=1):
                gr.Markdown("## 🎯 Optimized Result")
                
                prompt_output = gr.Textbox(
                    label="Flux-Optimized Prompt",
                    placeholder="Upload an image to see the intelligent analysis and optimization...",
                    lines=8,
                    max_lines=12,
                    elem_classes=["prompt-output"],
                    show_copy_button=True
                )
                
                score_output = gr.HTML(
                    value='<div style="text-align: center; padding: 1rem;"><div style="font-size: 2rem; color: #ccc;">--</div><div style="font-size: 0.875rem; color: #999;">Intelligence Score</div></div>'
                )
                
                info_output = gr.Markdown(value="")
                
                clear_btn = gr.Button("🗑️ Clear", size="sm")
        
        gr.Markdown("""
        ---
        ### 🔬 Pariente AI Research Foundation
        
        This optimizer implements deep computer vision analysis combined with validated Flux prompt engineering rules. 
        The system intelligently recognizes image content and applies structured optimization without requiring user configuration.
        
        **Research-based • Intelligence-driven • Zero configuration needed**
        """)
        
        # Simple event handlers
        optimize_btn.click(
            fn=process_image_wrapper,
            inputs=[image_input],
            outputs=[prompt_output, info_output, score_output]
        )
        
        clear_btn.click(
            fn=clear_outputs,
            outputs=[prompt_output, info_output, score_output]
        )
    
    return interface

if __name__ == "__main__":
    logger.info("🚀 Starting Deep Flux Prompt Optimizer")
    interface = create_interface()
    interface.launch(
        server_name="0.0.0.0",
        server_port=7860,
        show_error=True
    )