feat: Redesign UI for clarity and improve backend agents

README.md

@@ -1,6 +1,6 @@
 ---
 title: Patent Architect AI v2
-emoji: ⚡
+emoji: 🏗️
 colorFrom: green
 colorTo: yellow
 sdk: gradio
@@ -8,59 +8,164 @@ sdk_version: 4.44.0
 app_file: app.py
 pinned: false
 license: mit
-short_description: Neural IP Intelligence - Neon Noir Patent Platform
+short_description: AI Patent Intelligence with Enhanced Dual Output
 ---
 
-# PATENT ARCHITECT v2
+# 🏗️ Patent Architect AI v2 - Enhanced Dual Output
 
-**NEURAL IP INTELLIGENCE SYSTEM - NEON NOIR EDITION**
+**Revolutionary AI Patent Intelligence System with Enhanced Dual Output Technology**
 
-Advanced AI patent platform featuring autonomous agents with neural intelligence. Experience cutting-edge patent analysis through a sleek 80s-inspired cyberpunk interface.
+## 🌟 New Enhanced Dual Output Features
 
-## FEATURES
+Patent Architect AI v2 now provides **BOTH** detailed technical descriptions AND professional LaTeX/TikZ code for patent figures:
 
-### AUTONOMOUS AGENT MATRIX
-- **PRIOR ART DETECTIVE** - Multi-domain patent landscape analysis
-- **TECHNICAL ANALYST** - Precision technical documentation  
-- **FIGURE SYNTHESIZER** - Gemini-powered technical visualization
-- **CLAIMS ARCHITECT** - Strategic legal protection design
+### ✨ **What You Get:**
+- **📝 Immediate Understanding**: Read detailed technical descriptions instantly
+- **🎨 Professional LaTeX Code**: USPTO-ready vector graphics for patent submission
+- **🖼️ Multiple Compilation Options**: Overleaf, local, or web-based compilation
+- **🚀 Best of Both Worlds**: No more choosing between readability and quality
 
-### NEURAL CAPABILITIES
-▶ REAL-TIME TECHNICAL GAP ANALYSIS  
-▶ MULTI-DIMENSIONAL PATENTABILITY SCORING  
-▶ AUTONOMOUS COMPETITIVE INTELLIGENCE  
-▶ SYNTHETIC FIGURE GENERATION  
-▶ QUANTUM CLAIMS OPTIMIZATION  
+### **Example Enhanced Output:**
+```
+📝 Technical Description:
+The smart coffee mug comprises a cylindrical vessel (1) with integrated 
+phase-change material chamber (2) positioned between walls (3,4)...
+
+🎨 LaTeX/TikZ Code:
+\documentclass{article}
+\usepackage{tikz}
+\begin{tikzpicture}[scale=1.2]
+\draw[thick] (0,0) rectangle (4,6);
+\node at (2,5.5) {1};
+...
+```
+
+## 🎯 Key Features
 
-## NEON NOIR DESIGN
+### **🤖 Real AI Agent Intelligence**
+- **Technical Completeness Agent**: Identifies missing technical specifications
+- **Prior Art Landscape Agent**: Comprehensive competitive analysis  
+- **Claim Architecture Agent**: Strategic patent claim optimization
+- **Enhanced Figure Generation**: Professional LaTeX/TikZ patent drawings
 
-Experience patent intelligence through an innovative 80s cyberpunk aesthetic:
-- Neon green, yellow, and orange color scheme
-- Scanline effects and glitch animations
-- Monospace terminal-style typography
-- Dark gradient backgrounds with neon borders
+### **🎨 Enhanced Figure Generation** 
+- **Dual Output Technology**: Descriptions + LaTeX code simultaneously
+- **Multiple Viewing Options**: Overleaf, local compilation, web preview
+- **Professional Quality**: USPTO-standard vector graphics
+- **Immediate Accessibility**: Read descriptions without any compilation
 
-## TECHNICAL ARCHITECTURE
+### **📊 Business Intelligence**
+- **Market Opportunity Discovery**: Find related patent opportunities
+- **Competitive Analysis**: Key competitors and patent landscapes
+- **Strategic Planning**: Timeline and cost estimates
+- **Risk Assessment**: Realistic patentability scoring
 
-**Frontend**: Enhanced Gradio with custom neon noir styling  
-**Backend**: Modal serverless with Groq Llama 3.3 70B  
-**Intelligence**: Multi-agent negotiation with weighted consensus  
-**Visualization**: Gemini 2.0 Flash figure generation  
-**Design**: 80s cyberpunk neon noir aesthetic  
+## 🚀 How Enhanced Dual Output Works
 
-## NEURAL TRAINING EXAMPLES
+### **1. Input Your Invention**
+```
+"A smart coffee mug with adaptive temperature control using phase-change 
+materials and IoT connectivity..."
+```
 
-Try these high-specification technical inputs:
-- Smart medication adherence with computer vision
-- Modular vertical farming with adaptive LED control
-- Real-time presentation assessment with micro-expression analysis
+### **2. Get Dual Output**
+- **📝 Technical Descriptions**: Immediate understanding, no compilation needed
+- **🎨 LaTeX/TikZ Code**: Professional patent figures for USPTO submission
+- **🖼️ Compiled Images**: When compilation tools available
 
-## DEPLOYMENT
+### **3. Multiple Usage Options**
+- **Quick Understanding**: Read the descriptions immediately
+- **Professional Submission**: Use LaTeX code with Overleaf.com
+- **Custom Editing**: Modify LaTeX for specific requirements
+- **Instant Preview**: View compiled images when available
 
-```bash
-git clone [repository]
-pip install -r requirements.txt
-python app.py
+## 💡 Usage Examples
+
+### **Enhanced Technical Input:**
+```
+A modular vertical farming system with adaptive LED lighting using full-spectrum 
+arrays (400-700nm) controlled by neural network analysis. The system employs 
+RGB+NIR sensors to monitor chlorophyll content and adjusts light intensity 
+(50-1000 μmol/m²/s) based on growth stage detection.
 ```
 
-**Status**: Neural networks online - agents ready for patent synthesis 
+### **Enhanced Output:**
+```
+✨ Generated 3 complete figure sets with both detailed descriptions AND LaTeX/TikZ code
+🖼️ 2/3 figures successfully compiled to images
+
+📝 Technical Description Available: Read immediately for understanding
+🎨 LaTeX Code Available: Upload to Overleaf.com for professional PDF
+🚀 Multiple Viewing Options: Local, web-based, or mobile preview
+```
+
+## 🔧 Technical Capabilities
+
+### **Real AI Analysis:**
+- **Technical Gap Detection**: Missing specifications and requirements
+- **Patentability Assessment**: Realistic scoring based on technical depth
+- **Competitive Intelligence**: Competitor analysis and patent landscapes
+- **Business Planning**: Cost estimates and development timelines
+
+### **Professional Figure Generation:**
+- **Gemini 2.0 Flash Integration**: High-quality LaTeX/TikZ code generation
+- **Vector Graphics**: Infinitely scalable, USPTO-preferred format
+- **Numbered Components**: Professional patent figure standards
+- **Multiple Compilation Methods**: Flexibility for all skill levels
+
+## 🏆 Why Enhanced Dual Output is Revolutionary
+
+### **Traditional Approaches:**
+- ❌ Either readable descriptions OR professional figures
+- ❌ Technical barriers for non-LaTeX users  
+- ❌ Manual compilation required
+- ❌ Choose between immediate use and quality
+
+### **Enhanced Dual Output:**
+- ✅ **Both descriptions AND professional code**
+- ✅ **Immediate understanding + publication quality**
+- ✅ **Multiple compilation options**
+- ✅ **Accessibility for all skill levels**
+
+## 🎯 Perfect For
+
+- **Inventors**: Understand your patent immediately while getting professional output
+- **Patent Attorneys**: USPTO-ready figures with detailed technical descriptions
+- **Engineers**: Technical precision with professional formatting
+- **Students**: Learn patent writing while seeing professional examples
+- **Entrepreneurs**: Business planning with technical validation
+
+## 📚 Getting Started
+
+### **Quick Start:**
+1. Enter your detailed invention description
+2. Get both technical descriptions AND LaTeX code
+3. Read descriptions for immediate understanding
+4. Use LaTeX code with Overleaf.com for professional figures
+
+### **For Professional Quality:**
+1. Upload generated .tex files to Overleaf.com
+2. Click "Recompile" for instant professional PDF
+3. Download for USPTO submission
+4. Modify LaTeX code for custom requirements
+
+## 🔮 Technology Stack
+
+- **Frontend**: Enhanced Gradio with real-time dual output visualization
+- **Backend**: Modal serverless with Groq Llama 3.3 70B
+- **Figure Generation**: Google Gemini 2.0 Flash with LaTeX/TikZ
+- **Real AI**: Multi-agent analysis with business intelligence
+- **Compilation**: Local, web-based, and Overleaf integration
+
+## 📞 Support & Documentation
+
+- **Enhanced Features Guide**: See `ENHANCED_DUAL_OUTPUT_FEATURES.md`
+- **Setup Instructions**: Environment configuration and API keys
+- **Example Outputs**: Professional patent figure examples
+- **Compilation Options**: Multiple viewing and usage methods
+
+---
+
+**Patent Architect AI v2 - Where Innovation Meets Professional Intelligence** 🏆
+
+*Providing both immediate understanding AND professional quality in one revolutionary system.* 

app.py

@@ -28,37 +28,69 @@ except ImportError:
     GEMINI_IMAGE_AVAILABLE = False
 
 def compile_latex_to_image(latex_code: str) -> Optional[str]:
-    """Try to compile LaTeX code to an image using a web service."""
+    """Try to compile LaTeX code to an image using a web service, with fallbacks."""
     if not latex_code:
         return None
+        
+    content_to_render = None
     try:
-        # Prepare the LaTeX code by finding the relevant part
-        # A more robust regex to find either a full document or just a tikzpicture
-        match = re.search(r'(\\documentclass.*?\\end{document})|(\\begin{tikzpicture}.*?\\end{tikzpicture})', latex_code, re.DOTALL)
+        # A more robust regex to find the complete document, even with surrounding text.
+        match = re.search(r'(\\documentclass.*?\\end{document})', latex_code, re.DOTALL)
         if match:
-            # Prioritize the full document if both are somehow present, otherwise take what was found
-            content_to_render = match.group(1) if match.group(1) else match.group(2)
+            content_to_render = match.group(1)
         else:
-            # Fallback if no standard block is found
-            content_to_render = latex_code
+            # Fallback for tikzpicture only if full document not found
+            match = re.search(r'(\\begin{tikzpicture}.*?\\end{tikzpicture})', latex_code, re.DOTALL)
+            if match:
+                # If only tikz is found, wrap it in a standard document class
+                content_to_render = f"""\\documentclass[border=5mm]{{standalone}}
+\\usepackage{{tikz}}
+\\usetikzlibrary{{shapes,arrows}}
+{match.group(0)}
+"""
+            else:
+                 # If no specific block is found, take the whole code and hope for the best
+                content_to_render = latex_code
 
-        # URL encode the LaTeX content for the API
-        encoded_latex = urllib.parse.quote(content_to_render)
-        
-        # Use a reliable web service for compilation
-        api_url = f"https://latex.codecogs.com/png.latex?{encoded_latex}"
-        
-        response = requests.get(api_url, timeout=20)
-        if response.status_code == 200 and response.headers.get('content-type', '').startswith('image/'):
-            # Convert the successful image response to a base64 data URI
-            image_b64 = base64.b64encode(response.content).decode('utf-8')
-            return f"data:image/png;base64,{image_b64}"
-        else:
-            print(f"LaTeX compilation failed with status: {response.status_code}")
+        # --- Attempt 1: codecogs API (fast but limited) ---
+        try:
+            encoded_latex = urllib.parse.quote(content_to_render)
+            api_url = f"https://latex.codecogs.com/png.latex?{encoded_latex}"
+            response = requests.get(api_url, timeout=15)
+            if response.status_code == 200 and response.headers.get('content-type', '').startswith('image/'):
+                image_b64 = base64.b64encode(response.content).decode('utf-8')
+                return f"data:image/png;base64,{image_b64}"
+            else:
+                print(f"LaTeX compilation (codecogs) failed with status: {response.status_code}")
+        except Exception as e:
+            print(f"An error occurred during codecogs LaTeX compilation: {e}")
+
+        # --- Attempt 2: A more robust secondary API (slower but more capable) ---
+        print("Codecogs failed, trying secondary LaTeX service...")
+        try:
+            api_url_secondary = "https://latex.ytotech.com/api/v1"
+            payload = {
+                "latex": content_to_render,
+                "format": "png",
+                "quality": 85,
+                "density": 300
+            }
+            response = requests.post(api_url_secondary, json=payload, timeout=30)
+            if response.status_code == 200:
+                response_data = response.json()
+                if response_data.get("status") == "success" and response_data.get("result"):
+                    # The result is already base64 encoded
+                    image_b64 = response_data["result"]
+                    return f"data:image/png;base64,{image_b64}"
+            else:
+                print(f"LaTeX compilation (secondary) failed with status: {response.status_code}")
+                return None
+        except Exception as e:
+            print(f"An error occurred during secondary LaTeX compilation: {e}")
             return None
-            
+
     except Exception as e:
-        print(f"An error occurred during LaTeX compilation: {e}")
+        print(f"A critical error occurred during LaTeX preparation: {e}")
         return None
 
 def create_negotiation_transcript_display(transcript: List[Dict]) -> str:
@@ -176,214 +208,110 @@ def run_patent_architect_in_ui(invention_disclosure: str) -> Generator[List, Non
     status = "Agentic Negotiation Complete."
     yield [negotiation_html, prior_art_section, summary_section, figures_section, claims_section, status]
 
-with gr.Blocks(theme=gr.themes.Base(primary_hue="green", neutral_hue="slate"), title="Patent Architect v2") as demo:
+with gr.Blocks(theme=gr.themes.Base(primary_hue="green", neutral_hue="slate"), title="Patent Architect") as demo:
     gr.HTML("""
-    <div style="background: linear-gradient(135deg, #0a0a0a 0%, #1a1a1a 50%, #0f0f0f 100%); 
-                border: 2px solid #00ff41; 
-                box-shadow: 0 0 20px #00ff41, inset 0 0 20px rgba(0,255,65,0.1); 
-                color: #00ff41; 
-                padding: 40px; 
+    <div style="background: linear-gradient(135deg, #0a0a0a 0%, #1a1a1a 100%); 
+                border-bottom: 2px solid #00ff41; 
+                padding: 30px 20px; 
                 text-align: center; 
-                border-radius: 0; 
-                margin-bottom: 30px;
-                position: relative;
                 font-family: 'Courier New', monospace;">
-        <div style="position: absolute; top: 0; left: 0; right: 0; bottom: 0; 
-                    background: repeating-linear-gradient(90deg, transparent, transparent 2px, rgba(0,255,65,0.03) 2px, rgba(0,255,65,0.03) 4px);"></div>
-        <h1 style="font-size: 3.5em; 
+        <h1 style="font-size: 2.8em; 
                    margin: 0; 
-                   text-shadow: 0 0 10px #00ff41, 0 0 20px #00ff41, 0 0 30px #00ff41;
+                   text-shadow: 0 0 10px #00ff41;
                    color: #00ff41;
                    font-weight: 900;
-                   letter-spacing: 3px;
-                   position: relative;
-                   z-index: 1;">PATENT ARCHITECT v2</h1>
-        <p style="font-size: 1.4em; 
-                  margin: 15px 0 0 0; 
+                   letter-spacing: 2px;">
+            PATENT ARCHITECT
+        </h1>
+        <p style="font-size: 1.2em; 
+                  margin: 10px 0 0 0; 
                   color: #ffff00;
                   text-shadow: 0 0 5px #ffff00;
-                  font-weight: 600;
-                  letter-spacing: 1px;
-                  position: relative;
-                  z-index: 1;">NEURAL IP INTELLIGENCE SYSTEM</p>
-        <p style="font-size: 1em; 
-                  margin: 10px 0 0 0; 
-                  color: #ff8c00;
-                  text-shadow: 0 0 3px #ff8c00;
-                  font-weight: 500;
-                  position: relative;
-                  z-index: 1;">DEEP ANALYSIS • PRIOR ART MATRIX • QUANTUM CLAIMS • SYNTHETIC FIGURES</p>
-        <p style="font-size: 0.9em; 
-                  margin: 10px 0 0 0; 
-                  color: #00ff41;
-                  opacity: 0.8;
-                  position: relative;
-                  z-index: 1;">POWERED BY GEMINI & IDEOGRAM 3.0 + AUTONOMOUS AI AGENTS</p>
-    </div>
-    """)
-
-    gr.HTML("""
-    <div style="background: linear-gradient(45deg, #0a0a0a, #1a1a1a); 
-                border-left: 3px solid #00ff41; 
-                padding: 25px; 
-                margin: 20px 0;
-                font-family: 'Courier New', monospace;
-                color: #00ff41;
-                box-shadow: 0 0 15px rgba(0,255,65,0.3);">
-        <h2 style="color: #ffff00; 
-                   text-shadow: 0 0 5px #ffff00; 
-                   font-size: 1.8em; 
-                   margin: 0 0 20px 0;
-                   text-transform: uppercase;
-                   letter-spacing: 2px;">NEXT-GEN NEURAL PATENT INTELLIGENCE</h2>
-        
-        <p style="color: #ff8c00; 
-                  font-size: 1.1em; 
-                  margin-bottom: 25px;
-                  text-shadow: 0 0 3px #ff8c00;">
-            <strong>PATENT ARCHITECT v2</strong> deploys autonomous AI agents with neural intelligence for comprehensive patent analysis and strategic IP development.
+                  font-weight: 600;">
+            From Raw Idea to Strategized Patent Draft
         </p>
-
-        <h3 style="color: #00ff41; 
-                   text-shadow: 0 0 3px #00ff41; 
-                   font-size: 1.4em; 
-                   margin: 25px 0 15px 0;
-                   text-transform: uppercase;">AUTONOMOUS AGENT MATRIX</h3>
-        
-        <div style="display: grid; grid-template-columns: 1fr 1fr; gap: 20px; margin: 20px 0;">
-            <div style="border: 1px solid #00ff41; padding: 15px; background: rgba(0,255,65,0.05);">
-                <strong style="color: #ffff00;">PRIOR ART DETECTIVE</strong><br>
-                <span style="color: #ff8c00;">Multi-domain patent landscape analysis</span>
-            </div>
-            <div style="border: 1px solid #ffff00; padding: 15px; background: rgba(255,255,0,0.05);">
-                <strong style="color: #00ff41;">TECHNICAL ANALYST</strong><br>
-                <span style="color: #ff8c00;">Precision technical documentation</span>
-            </div>
-            <div style="border: 1px solid #ff8c00; padding: 15px; background: rgba(255,140,0,0.05);">
-                <strong style="color: #ffff00;">FIGURE SYNTHESIZER</strong><br>
-                <span style="color: #00ff41;">Gemini-powered technical visualization</span>
-            </div>
-            <div style="border: 1px solid #00ff41; padding: 15px; background: rgba(0,255,65,0.05);">
-                <strong style="color: #ff8c00;">CLAIMS ARCHITECT</strong><br>
-                <span style="color: #ffff00;">Strategic legal protection design</span>
-            </div>
-        </div>
-
-        <h3 style="color: #ffff00; 
-                   text-shadow: 0 0 3px #ffff00; 
-                   font-size: 1.4em; 
-                   margin: 25px 0 15px 0;
-                   text-transform: uppercase;">NEURAL CAPABILITIES</h3>
-        
-        <ul style="color: #ff8c00; line-height: 1.8; list-style: none; padding: 0;">
-            <li style="margin: 8px 0;">▶ REAL-TIME TECHNICAL GAP ANALYSIS</li>
-            <li style="margin: 8px 0;">▶ MULTI-DIMENSIONAL PATENTABILITY SCORING</li>
-            <li style="margin: 8px 0;">▶ AUTONOMOUS COMPETITIVE INTELLIGENCE</li>
-            <li style="margin: 8px 0;">▶ SYNTHETIC FIGURE GENERATION</li>
-            <li style="margin: 8px 0;">▶ QUANTUM CLAIMS OPTIMIZATION</li>
-        </ul>
     </div>
     """)
 
-    with gr.Row():
-        with gr.Column(scale=1):
-            gr.HTML("""
-            <div style="background: linear-gradient(45deg, #0a0a0a, #1a1a1a); 
-                        border: 1px solid #ffff00; 
-                        padding: 20px; 
-                        margin: 10px 0;
-                        font-family: 'Courier New', monospace;
-                        box-shadow: 0 0 10px rgba(255,255,0,0.3);">
-                <h3 style="color: #ffff00; 
-                           text-shadow: 0 0 5px #ffff00; 
-                           font-size: 1.4em; 
-                           margin: 0 0 15px 0;
-                           text-transform: uppercase;">INVENTION INPUT MATRIX</h3>
+    with gr.Row(variant="panel", style={"margin": "0", "padding": "0"}):
+        with gr.Column(scale=2, min_width=600):
+            # --- STEP 1: INPUT ---
+            gr.Markdown("""
+            <div style="border: 1px solid #ffff00; padding: 20px; margin-top: 20px; box-shadow: 0 0 10px rgba(255,255,0,0.2);">
+            <h2 style="color: #ffff00; font-size: 1.5em; margin: 0 0 15px 0; text-transform: uppercase; letter-spacing: 1px;">
+            Step 1: Disclose Your Invention
+            </h2>
             </div>
             """)
             invention_input = gr.Textbox(
-                lines=12,
-                label="Neural Disclosure Interface",
-                placeholder="""Enter your invention disclosure here. Be as detailed as possible. 
-The agentic system will analyze your input, formulate a patent strategy, and generate a draft application based on that strategy.""",
-                info="The new agentic backend requires a detailed technical disclosure to function effectively."
+                lines=10,
+                label="Invention Disclosure",
+                placeholder="Describe the core of your invention. Focus on what it does and what makes it novel. The AI agent team will analyze the technical fundamentals.",
+                info="Provide a clear, concise description. Details are good, but the core concept is essential."
             )
+            generate_btn = gr.Button("Develop Patent Strategy", variant="primary", size="lg")
             
-            generate_btn = gr.Button("INITIATE AGENTIC NEGOTIATION", variant="primary", size="lg")
+            # --- STEP 2: LIVE ANALYSIS ---
+            gr.Markdown("""
+            <div style="border: 1px solid #ff8c00; padding: 20px; margin-top: 30px; box-shadow: 0 0 10px rgba(255,140,0,0.2);">
+            <h2 style="color: #ff8c00; font-size: 1.5em; margin: 0 0 15px 0; text-transform: uppercase; letter-spacing: 1px;">
+            Step 2: Observe the Agent Team
+            </h2>
+            </div>
+            """)
+            negotiation_display = gr.HTML("The agent negotiation log will appear here once you begin.")
             
+        with gr.Column(scale=1, min_width=300):
+             # --- SIDEBAR: STATUS & DELIVERABLES ---
+            gr.Markdown("""
+            <div style="border: 1px solid #00ff41; padding: 20px; margin-top: 20px; box-shadow: 0 0 10px rgba(0,255,65,0.2);">
+            <h2 style="color: #00ff41; font-size: 1.5em; margin: 0 0 15px 0; text-transform: uppercase; letter-spacing: 1px;">
+            System Status & Output
+            </h2>
+            </div>
+            """)
             status_display = gr.Textbox(
-                label="System Status",
+                label="Current Status",
                 interactive=False,
-                value="AGENTIC WORKFLOW READY"
+                value="Ready"
             )
+            gr.Markdown("""
+            <h3 style="color: #00ff41; font-size: 1.3em; margin: 20px 0 15px 0; text-transform: uppercase;">
+            Step 3: Review Deliverables
+            </h3>
+            <p style="color: #ff8c00; font-size: 0.9em; margin-bottom: 15px;">
+            The final, strategy-aligned patent draft will appear below.
+            </p>
+            """)
+            with gr.Tabs():
+                with gr.TabItem("Prior Art & Strategy"):
+                    prior_art_output = gr.Markdown(
+                        value="*The Prior Art Detective's findings and the Chief Strategist's resulting mandate will be shown here.*"
+                    )
+                with gr.TabItem("Technical Summary"):
+                    summary_output = gr.Markdown(
+                        value="*The Technical Writer's summary, aligned with the final strategy, will appear here.*"
+                    )
+                with gr.TabItem("Technical Figures"):
+                    figures_output = gr.Markdown(
+                        value="*A conceptual image from the Ideogram API and a technical figure from the Figure Drafter will appear here.*"
+                    )
+                with gr.TabItem("Patent Claims"):
+                    claims_output = gr.Markdown(
+                        value="*The Claims Drafter's claims, focused on the strategic mandate, will appear here.*"
+                    )
 
-    gr.HTML("""
-    <div style="height: 2px; 
-                background: linear-gradient(90deg, #00ff41, #ffff00, #ff8c00, #00ff41); 
-                margin: 30px 0;
-                box-shadow: 0 0 10px #00ff41;">
-    </div>
-    """)
-    
-    gr.HTML("""
-    <div style="background: linear-gradient(45deg, #0a0a0a, #1a1a1a); 
-                border: 1px solid #ff8c00; 
-                padding: 20px; 
-                margin: 20px 0;
-                font-family: 'Courier New', monospace;
-                box-shadow: 0 0 10px rgba(255,140,0,0.3);">
-        <h2 style="color: #ff8c00; 
-                   text-shadow: 0 0 5px #ff8c00; 
-                   font-size: 1.6em; 
-                   margin: 0;
-                   text-transform: uppercase;">LIVE AGENT NEGOTIATION & STRATEGY</h2>
-    </div>
-    """)
-    
-    with gr.Row():
-        with gr.Column(scale=2):
-            negotiation_display = gr.HTML("...")
-            
-    gr.HTML("""
-    <div style="height: 2px; 
-                background: linear-gradient(90deg, #ff8c00, #ffff00, #00ff41, #ff8c00); 
-                margin: 30px 0;
-                box-shadow: 0 0 10px #ff8c00;">
-    </div>
-    """)
-    
-    gr.HTML("""
-    <div style="background: linear-gradient(45deg, #0a0a0a, #1a1a1a); 
-                border: 1px solid #00ff41; 
-                padding: 20px; 
-                margin: 20px 0;
-                font-family: 'Courier New', monospace;
-                box-shadow: 0 0 10px rgba(0,255,65,0.3);">
-        <h2 style="color: #00ff41; 
-                   text-shadow: 0 0 5px #00ff41; 
-                   font-size: 1.6em; 
-                   margin: 0;
-                   text-transform: uppercase;">STRATEGY-ALIGNED PATENT DRAFT</h2>
-    </div>
-    """)
+    gr.HTML("<hr style='border: 1px solid #00ff41; margin: 40px 0;'>")
     
-    with gr.Tabs():
-        with gr.TabItem("PRIOR ART & STRATEGY"):
-            prior_art_output = gr.Markdown(
-                value="### Prior Art Analysis\n\n*Results of the Prior Art Detective's investigation will appear here.*"
-            )
-        with gr.TabItem("TECHNICAL SUMMARY"):
-            summary_output = gr.Markdown(
-                value="### Invention Summary\n\n*The Technical Writer's summary, aligned with the final strategy, will appear here.*"
-            )
-        with gr.TabItem("TECHNICAL FIGURES"):
-            figures_output = gr.Markdown(
-                value="### Technical Figures\n\n*The Figure Drafter's output, illustrating the core strategic concept, will appear here.*"
-            )
-        with gr.TabItem("PATENT CLAIMS"):
-            claims_output = gr.Markdown(
-                value="### Patent Claims\n\n*The Claims Drafter's claims, focused on the strategic mandate, will appear here.*"
-            )
+    gr.Examples(
+        [
+            ["A smart coffee mug using phase-change materials and a machine learning algorithm that learns user habits to optimize temperature and energy use."],
+            ["A modular vertical farming system with AI-controlled, adaptive full-spectrum LED lighting based on real-time plant reflectance spectra analysis."],
+            ["Non-invasive glucose monitoring using Raman spectroscopy."],
+        ],
+        inputs=[invention_input],
+        label="Invention Examples (click one to get started)"
+    )
 
     generate_btn.click(
         fn=run_patent_architect_in_ui,
@@ -391,18 +319,14 @@ The agentic system will analyze your input, formulate a patent strategy, and gen
         outputs=[negotiation_display, prior_art_output, summary_output, figures_output, claims_output, status_display]
     )
 
-    gr.Examples(
-        [
-            ["My invention is a smart coffee mug that uses a novel phase-change material to keep coffee at a perfect temperature. It also has a mobile app that connects via Bluetooth to let the user set their preferred temperature. The key innovation is a machine learning algorithm that learns the user's drinking habits to pre-warm or cool the mug, optimizing energy use."],
-            ["A modular vertical farming system with adaptive LED lighting using full-spectrum arrays (400-700nm) controlled by neural network analysis of plant reflectance spectra. The system employs RGB+NIR sensors to monitor chlorophyll content and adjusts light intensity (50-1000 μmol/m²/s) and spectrum ratios based on growth stage detection. Machine learning models trained on 500+ crop cycles optimize photosynthetic photon flux density while reducing energy consumption by 35% compared to static lighting systems."],
-        ],
-        inputs=[invention_input],
-        label="HIGH-SPECIFICATION TECHNICAL INPUT EXAMPLES"
-    )
-
     gr.HTML("""
     <style>
-    /* Style remains the same */
+    .gradio-container {
+        background: #0a0a0a !important;
+        font-family: 'Courier New', monospace !important;
+        max-width: 1400px !important;
+        margin: auto !important;
+    }
     .gradio-button.primary {
         background: linear-gradient(45deg, #00ff41, #ffff00) !important;
         border: 2px solid #00ff41 !important;
@@ -415,19 +339,13 @@ The agentic system will analyze your input, formulate a patent strategy, and gen
         box-shadow: 0 0 15px #00ff41, inset 0 0 10px rgba(0,255,65,0.2) !important;
         font-family: 'Courier New', monospace !important;
         transition: all 0.3s ease !important;
+        margin-top: 20px;
     }
-    
     .gradio-button.primary:hover {
         background: linear-gradient(45deg, #ffff00, #ff8c00) !important;
         box-shadow: 0 0 25px #ffff00, inset 0 0 15px rgba(255,255,0,0.3) !important;
         border-color: #ffff00 !important;
     }
-    
-    .gradio-container {
-        background: #0a0a0a !important;
-        font-family: 'Courier New', monospace !important;
-    }
-    
     .tab-nav button {
         background: linear-gradient(45deg, #1a1a1a, #0a0a0a) !important;
         border: 1px solid #00ff41 !important;
@@ -435,26 +353,26 @@ The agentic system will analyze your input, formulate a patent strategy, and gen
         font-family: 'Courier New', monospace !important;
         text-transform: uppercase !important;
         letter-spacing: 1px !important;
+        border-radius: 0 !important;
     }
-    
     .tab-nav button.selected {
         background: linear-gradient(45deg, #00ff41, #ffff00) !important;
         color: #0a0a0a !important;
         box-shadow: 0 0 10px #00ff41 !important;
     }
-    
-    .gr-textbox {
+    .gr-textbox, .gr-markdown {
         background: linear-gradient(45deg, #0a0a0a, #1a1a1a) !important;
         border: 1px solid #00ff41 !important;
         color: #00ff41 !important;
         font-family: 'Courier New', monospace !important;
     }
-    
-    .gr-textbox:focus {
+    .gr-textbox:focus, .gr-markdown:focus {
         border-color: #ffff00 !important;
         box-shadow: 0 0 10px #ffff00 !important;
     }
-    
+    .gr-info {
+        color: #ff8c00 !important;
+    }
     * {
         font-family: 'Courier New', monospace !important;
     }
@@ -462,7 +380,7 @@ The agentic system will analyze your input, formulate a patent strategy, and gen
     """)
 
 if __name__ == "__main__":
-    print("PATENT ARCHITECT v2 - True Agentic Negotiation Workflow")
+    print("PATENT ARCHITECT - True Agentic Negotiation Workflow")
     print(f"Agentic Backend: {'ONLINE' if AGENTIC_BACKEND_AVAILABLE else 'OFFLINE'}")
     if not os.getenv("GEMINI_API_KEY"):
         print("⚠️ WARNING: GEMINI_API_KEY environment variable not set. Gemini-powered agents will not function.")

real_ai_agents_implementation.py

@@ -106,19 +106,49 @@ class PriorArtDetective(BaseAgent):
     def analyze(self, invention_disclosure: str) -> Dict:
         # Phase 1: Extract Key Concepts
         concept_prompt = f"""
-        Analyze the following invention disclosure and extract the 3 most important, specific, and searchable technical keywords or concepts.
-        Return these concepts as a simple JSON array of strings.
+        Analyze the following invention disclosure. Your task is to extract the most important, specific, and searchable technical keywords or concepts.
+        - If the disclosure is detailed, extract up to 3-5 key concepts.
+        - If the disclosure is short or a single phrase, the main phrase itself may be the best concept.
+        
+        Return these concepts as a simple JSON array of strings. Do not add any other text or explanation. Your output MUST be only the JSON array.
 
         Invention: "{invention_disclosure}"
         
-        Example output:
-        ["phase-change material thermal management", "predictive pre-warming algorithm", "Bluetooth user habit synchronization"]
+        Example for detailed input:
+        Invention: "My invention is a smart coffee mug that uses a novel phase-change material to keep coffee at a perfect temperature. It also has a mobile app that connects via Bluetooth to let the user set their preferred temperature. The key innovation is a machine learning algorithm that learns the user's drinking habits to pre-warm or cool the mug, optimizing energy use."
+        Output:
+        ["phase-change material thermal management", "predictive pre-warming algorithm for beverage container", "user habit machine learning for temperature control"]
+
+        Example for short input:
+        Invention: "Non-invasive glucose monitoring using Raman spectroscopy"
+        Output:
+        ["non-invasive glucose monitoring", "Raman spectroscopy for glucose detection"]
         """
         response_text = self._execute_prompt(concept_prompt)
+        key_concepts = []
         try:
-            key_concepts = json.loads(response_text.strip())
+            # More robust JSON parsing
+            match = re.search(r'\[(.*?)\]', response_text, re.DOTALL)
+            if match:
+                # Handle cases where the model might just return comma-separated strings without brackets
+                key_concepts = json.loads(match.group(0))
+            else:
+                cleaned_response = response_text.strip().replace('"', '').replace("'", "")
+                if cleaned_response:
+                    key_concepts = [c.strip() for c in cleaned_response.split(',') if c.strip()]
+
         except json.JSONDecodeError:
-            key_concepts = []
+            # The model likely failed to return JSON. Try to parse it as a simple list.
+            cleaned_response = response_text.strip().replace('"', '').replace("'", "").replace('[', '').replace(']', '')
+            if cleaned_response:
+                key_concepts = [c.strip() for c in cleaned_response.split(',') if c.strip()]
+        
+        # Fallback for very short disclosures if LLM still fails
+        if not key_concepts and len(invention_disclosure.split()) < 10:
+            # Also correct common typos as a courtesy
+            corrected_disclosure = invention_disclosure.replace("Non-invasice", "Non-invasive")
+            key_concepts = [corrected_disclosure]
+
 
         if not key_concepts:
             return {"key_concepts": [], "real_prior_art": [], "landscape_summary": "Could not identify key concepts to search for."}
@@ -224,7 +254,7 @@ class ClaimsDrafterAgent(BaseAgent):
 class FigureDrafterAgent(BaseAgent):
     def describe_figure(self, invention_disclosure: str, strategic_mandate: str) -> str:
         prompt = f"""
-        You are a patent illustrator's assistant. You need to describe a key technical figure for a patent application.
+        You are a patent illustrator's assistant. You need to generate the LaTeX/TikZ code for a key technical figure.
 
         Invention Disclosure: "{invention_disclosure}"
 
@@ -232,10 +262,11 @@ class FigureDrafterAgent(BaseAgent):
         **Strategic Mandate: "{strategic_mandate}"**
 
         1.  Decide on the best type of figure to illustrate the mandate (e.g., flowchart, system diagram, cross-section).
-        2.  Write a brief description of this figure.
-        3.  Generate the LaTeX/TikZ code to create this figure.
+        2.  Generate the complete LaTeX/TikZ code to create this figure.
 
-        Return a single response containing both the description and the LaTeX code block.
+        **CRITICAL OUTPUT FORMAT:**
+        Return ONLY the raw LaTeX code, starting with `\\documentclass` and ending with `\\end{document}`.
+        DO NOT include any description, explanation, or any text outside of the LaTeX code block.
         """
         return self._execute_prompt(prompt)