# Sentinel-AI Frontend Design ## 1. Product Vision Create a premium, high-impact web dashboard for the Sentinel-AI system that makes the backend’s face-analysis and semantic-search capabilities feel fast, intelligent, and trustworthy. The frontend should feel like a modern security operations console rather than a basic demo page. The experience must: - Present complex AI results clearly and professionally. - Make image-based analysis feel intuitive with strong visual previews. - Combine analytics, AI-generated insights, and searchable results in one polished workflow. - Feel suitable for a hackathon demo, investor pitch, or security operations showcase. --- ## 2. Project Understanding This project is a hybrid edge-cloud security agent that: - Detects faces from uploaded images. - Embeds the detected face and searches a FAISS index. - Extracts semantic attributes from natural-language queries. - Generates an AI-based security alert using Gemini. - Reports latency and estimated cost using cascadeflow metrics. The frontend should highlight these capabilities directly: - Image upload and analysis. - Semantic search by plain English. - Match confidence and attribute extraction. - AI-generated report and actionable insight. - System health and performance metrics. --- ## 3. Recommended Frontend Stack For a polished and fast implementation, use: - React + Vite for the app shell. - Tailwind CSS for styling and responsive layout. - Framer Motion for subtle animations. - Recharts or Chart.js for visual analytics. - Lucide React for icons. - Axios or Fetch for API calls. Suggested structure: - src/components - src/pages - src/services - src/hooks - src/styles - src/assets --- ## 4. Core User Experience Goals ### Primary goals 1. Make AI analysis feel instant and trustworthy. 2. Show visually rich result summaries. 3. Support both image upload and natural-language query modes. 4. Present technical performance metrics in a clean way. 5. Make the dashboard feel premium and modern. ### Secondary goals - Improve readability of attributes and match results. - Provide visual confidence indicators. - Enable quick comparison of results. - Offer a strong demo experience for presentation. --- ## 5. Main Page Layout The main experience should be a single polished dashboard with the following sections: ### A. Hero Section A high-end intro area at the top with: - A dark gradient background. - A bold headline such as: “AI-Powered Face Intelligence for Security Operations”. - A short supporting description. - Two primary CTAs: - Analyze Image - Run Semantic Search - A preview card or collage of facial images and analytic overlays. Suggested visual treatment: - Deep navy / black background. - Accent colors: electric blue, neon cyan, crimson, or amber. - Subtle animated grid overlays. - Floating metric cards for confidence, latency, and system status. ### B. Upload & Analysis Panel This should be the heart of the page. Features: - Drag-and-drop image uploader. - File preview after selection. - Optional camera capture placeholder for future extension. - Analyze button with loading state. - Clear error and empty states. Visuals: - Large upload card with a soft gradient border. - Preview image shown in a rounded content card. - Progress bar during analysis. - Result state card with confidence and attributes. ### C. Semantic Search Panel This panel should allow natural-language queries such as: - “young male with wavy hair and eyeglasses” - “older person with gray hair and a goatee” - “smiling woman with blond hair” Features: - Query input box. - Suggested search chips. - Top-k selector. - Search results grid. - Match count summary. Visuals: - Search result cards with thumbnails and attribute tags. - Compact metadata display for each result. ### D. Results Intelligence Panel This is where the core AI output becomes understandable. Include: - Confidence score gauge. - Detected attributes as pill chips. - AI-generated security report text. - System metrics cards for latency and cost. - Match status badge: success, no face, no match, or no attributes found. ### E. Visualization Panel A strong visual layer is essential for making the interface feel professional. Recommended visualizations: - Confidence gauge or radial chart. - Attribute distribution bar chart. - Latency and cost trend card. - Match result count chart. - Small status overview cards for index readiness, model availability, and API health. ### F. Gallery / Result Explorer A polished image gallery section should display matched media or sample references. Use thumbnails from the dataset directory such as: - Data/img_align_celeba/img_align_celeba/190538.jpg - Data/img_align_celeba/img_align_celeba/190539.jpg - Data/img_align_celeba/img_align_celeba/190540.jpg This section can show: - A large featured image. - Related matches below it. - Hover states and quick previews. - “View full match” interaction. --- ## 6. Key Interface Features ### 6.1 Upload and Analyze - Drag and drop image upload. - Image preview before submission. - Real-time validation for supported files. - Clear analysis progress feedback. - Result cards with confidence and attributes. ### 6.2 Semantic Search - Free-text query entry. - Smart suggested phrases. - Structured result list. - Filtered match cards. - Visual tags for extracted attributes. ### 6.3 AI Report Display - Show the Gemini-generated report in a visually distinct card. - Emphasize action-oriented recommendations. - Display confidence values clearly. ### 6.4 Analytics & Metrics - Latency in milliseconds. - Estimated cost in USD. - Match confidence values. - Result count and responsiveness. ### 6.5 Health & System Status - API health indicator. - FAISS index readiness. - Model loading state. - Network/error states. ### 6.6 Professional Polish - Animated transitions. - Smooth hover effects. - Dark glassmorphism cards. - Responsive behavior on desktop and tablet. - Accessible contrast and keyboard support. --- ## 7. Suggested Page Components ### Main Dashboard Page - App shell with sidebar navigation. - Hero banner. - Upload analysis section. - Semantic search section. - Analytics widgets. - Result gallery. ### Reusable Components - HeaderBar - StatCard - UploadBox - ImagePreviewCard - ResultBadge - AttributeChip - ConfidenceGauge - MetricsChart - SearchResultCard - ReportPanel - EmptyState - ToastNotification --- ## 8. Visual Design Direction ### Style Theme Use a premium surveillance-inspired aesthetic: - Background: deep slate / midnight navy. - Surface cards: slightly lighter dark panels with glow borders. - Accent colors: cyan for intelligence, red/orange for alerts, green for healthy states. - Typography: clean sans-serif such as Inter, Sora, or Poppins. ### Visual Rules - Cards should have rounded corners and subtle shadow depth. - Important data should use strong contrast and spacing. - Results should be easy to scan at a glance. - Avoid clutter; use progressive disclosure. ### Motion Principles - Smooth transitions on upload and result loading. - Gentle floating effect on hero cards. - Subtle pulse animations for health indicators. - No unnecessary motion that distracts from the analysis. --- ## 9. Data Flow and API Integration ### Analyze Image flow 1. User uploads an image. 2. Frontend sends file to POST /api/v1/analyze-frame. 3. UI shows loading state. 4. Results display: - faces_detected - match_confidence - matched_attributes - agent_report - cascadeflow_metrics ### Semantic Search flow 1. User enters a natural-language query. 2. Frontend sends POST /api/v1/semantic-search. 3. UI renders extracted attributes and matching results. 4. Results include image IDs, attributes, and FAISS IDs. ### Health flow 1. Frontend calls GET /health. 2. Dashboard shows readiness, model, and index status. --- ## 10. Example UI Sections ### 10.1 Hero Section Content: - Headline: “Security intelligence from a single image.” - Subtext: “Upload a face, explore semantic attributes, and receive a professional AI-generated assessment.” - Two buttons: Analyze Image and Semantic Search. - Background image collage built from the CelebA dataset or a synthetic security-themed visual. ### 10.2 Analysis Results Card Content: - Uploaded image preview. - Match confidence indicator. - Detected attributes as chips. - AI-generated report. - Metrics summary for latency and cost. ### 10.3 Search Results View Content: - Search input field. - Result cards with face-like thumbnails. - Matched attribute tags. - Responsive grid layout. ### 10.4 Dashboard Analytics Strip Content: - System status cards. - Confidence speedometer. - Performance chart. - Recent search history. --- ## 11. Recommended Content and Image Strategy Use a mix of: - Real dataset faces from the local CelebA image folder. - Abstract security-themed backgrounds. - Minimal UI illustrations for empty states. - High-quality gradient overlays for visual depth. Suggested image usage: - Hero visual: collage or carousel of sample images from the dataset. - Result preview: uploaded or selected image. - Gallery: multiple dataset thumbnails for match exploration. Important note: - The interface should not feel overly technical or cluttered. - It should communicate confidence, intelligence, and control. --- ## 12. Implementation Plan ### Phase 1 – Layout and Shell - Build the dashboard shell. - Add navigation, hero section, and cards. - Create reusable UI components. ### Phase 2 – Core Analysis Flow - Implement image upload. - Connect to /api/v1/analyze-frame. - Display image preview and result card. ### Phase 3 – Search Experience - Add semantic search input and results grid. - Map extracted attributes into chips and cards. ### Phase 4 – Visual Intelligence - Add charts and gauges. - Create a polished analytics panel. - Add animations and transitions. ### Phase 5 – Polish and Hardening - Improve empty states, loading states, responsiveness, and accessibility. - Add toast notifications and error handling. --- ## 13. Suggested Mobile/Responsive Behavior - Desktop: full multi-panel dashboard. - Tablet: stacked panels with two-column sections. - Mobile: single-column layout with collapsible panels. The upload flow should remain primary and easy to access on every screen size. --- ## 14. Accessibility and Quality Standards - Good color contrast. - Keyboard-friendly controls. - Screen-reader-friendly labels. - Clear status messages. - Consistent spacing and typography. - Avoid overwhelming the user with too much information at once. --- ## 15. Final Design Direction The final frontend should feel like a serious AI security command center: - Modern - Premium - Fast - Visually rich - Professional - Intuitive - Clearly connected to the backend’s intelligence The page should make the system look not just functional, but impressive. --- ## 16. Recommended First-Version Wireframe Concept ### Top area - Brand/title - Navigation links - Status badge ### Main content - Left: upload and analysis card - Right: result intelligence card ### Lower section - Semantic search input and results - Analytics panel - Gallery of matched images This layout balances analysis, search, and visuals in a way that feels polished and demo-ready.