--- title: Space AC Design Variations emoji: 🎨 colorFrom: purple colorTo: pink sdk: gradio sdk_version: 5.49.0 app_file: app.py pinned: false license: mit --- # 🎨 Space AC: Design Variations **AI-powered design system generation for Flutter apps** Part of the **AI Software Factory** - a complete system that generates Flutter applications from Notion wireframes to GitHub PRs. ## 🚀 Features This space contains specialized AI agents for design system generation: ### ⚡ Simple Generator (NEW!) ✅ LIVE - **Zero-configuration Flutter code generation** - Prompt → DeepSeek → Flutter Code → DartPad (auto-loaded!) - Automatic GitHub Gist creation for instant DartPad embedding - No themes, no presets - just describe what you want - Code automatically loaded in DartPad - just click "Run"! - Collapsible code viewer for clean UI **Example prompts:** - "Create a modern login screen with email and password" - "Generate a fitness app home page with statistics cards" - "Build a todo list with checkboxes and delete buttons" ### AC1: Moodboard Analyzer 🖼️ ✅ LIVE - Analyzes reference app screenshots - Extracts color palettes with visual previews - Identifies UI patterns and components - Detects typography trends - Generates actionable design insights ### AC2: Design Variation Generator 🎨 ✅ LIVE - Generates 4 complete design theme variations - Defines design tokens (colors, typography, spacing) - Creates comprehensive design system specifications - Interactive theme cards with visual preview ### AC3: Flutter Screen Generator 💻 ✅ LIVE - Generates runnable Flutter/Dart code for DartPad - AI-powered planning with Llama 3.1 70B + Jinja2 templates - Creates 3 screens per theme with realistic components - Supports 10 app types (fitness, banking, rider, ecommerce, social, etc.) - 400-500 lines of production-ready Dart code per screen ### Coming Soon: - **AC4:** Theme Comparator (pros/cons analysis and recommendations) ## 🎯 Use Cases - **Design System Creation:** Generate complete design systems from inspiration - **Theme Development:** Create multiple design variations quickly - **Competitor Analysis:** Analyze design patterns from reference apps - **Design Documentation:** Auto-generate design tokens and guidelines ## 🔧 API Endpoints ### POST /api/moodboard-analyzer Analyze a moodboard to extract design insights. **Request:** ```json { "reference_images": [ { "url": "https://example.com/screenshot.png", "description": "Instagram home feed - dark mode" } ], "mood_keywords": ["energetic", "gaming", "modern"] } ``` **Response:** ```json { "success": true, "output": { "analysis": { "color_palettes": [...], "typography_trends": [...], "ui_patterns": [...], "insights": "..." } } } ``` ### POST /api/design-variation-generator Generate 4 design theme variations from moodboard analysis. **Request:** ```json { "moodboard_analysis": { "color_palettes": [...], "typography_trends": [...], "ui_patterns": [...] }, "wireframe": "E-commerce app with product grid and checkout", "app_context": { "type": "mobile_app" } } ``` **Response:** ```json { "success": true, "output": { "themes": [ { "name": "Bold Energy", "style": "bold", "palette": {...}, "typography": {...}, "spacing": {...}, "component_style": {...}, "animations": {...} } ] } } ``` ### POST /api/flutter-screen-generator Generate Flutter/Dart code from design themes (AC3). **Request:** ```json { "theme": { "name": "Bold Energy", "palette": {...}, "typography": {...}, "spacing": {...} }, "app_type": "fitness", "screens": ["home", "profile", "settings"] } ``` **Response:** ```json { "success": true, "output": { "screens": [ { "name": "home", "code": "import 'package:flutter/material.dart';\n\n// Full Dart code...", "lines": 487, "components": ["StatsCard", "HeroBanner"], "status": "generated" } ], "theme_name": "Bold Energy", "app_type": "fitness", "total_lines": 1420 } } ``` **Note:** Requires `HF_TOKEN` environment variable for Llama 3.1 access. ## 🔑 Environment Variables To enable all features, you need to configure the following environment variables: ### HuggingFace Token (Required for AC1, AC2, AC3) All three agents use HuggingFace Inference API to access Llama models. 1. Get your token from [HuggingFace Settings](https://huggingface.co/settings/tokens) 2. Create a `.env` file in the project root: ```bash cp .env.example .env ``` 3. Add your token to `.env`: ```env HF_TOKEN=your_huggingface_token_here ``` **On HuggingFace Spaces:** - Go to Space Settings → Variables - Add `HF_TOKEN` as a secret **Note:** Without the token, agents will use fallback/mock data. ### GitHub Token (Required for Simple Generator Auto-load DartPad) The **Simple Generator** feature automatically loads generated Flutter code into DartPad by creating GitHub Gists. **Required for:** - ⚡ Simple Generator → Auto-load code in DartPad **Setup Instructions:** 1. Create a GitHub Personal Access Token (classic): - Go to [GitHub Settings → Tokens](https://github.com/settings/tokens) - Click "Generate new token (classic)" - Give it a name like "HF Space Gist Creator" - Select scopes: **`gist`** (only this scope is needed) - Click "Generate token" - **Copy the token** (you won't see it again!) 2. Add to `.env` file: ```env GITHUB_TOKEN=ghp_yourGitHubTokenHere ``` **On HuggingFace Spaces:** - Go to Space Settings → Variables - Add `GITHUB_TOKEN` as a **secret** (not a public variable!) **Note:** Without this token, the Simple Generator will still work, but users will need to manually copy-paste code into DartPad instead of automatic loading. ## 💻 Local Development ```bash # Clone the repository git clone https://huggingface.co/spaces/OneWeekApp/space-ac-design cd space-ac-design # Install dependencies pip install -r requirements.txt # Configure environment variables cp .env.example .env # Edit .env and add your HuggingFace token # Run locally python app.py ``` The Gradio interface will be available at `http://localhost:7860` ## 🧠 AI Models Used - **DeepSeek Coder** - Direct Flutter code generation (Simple Generator) - **Llama 3.2 Vision (11B)** - Image analysis (AC1) - **Llama 3.1 (70B)** - Text generation and synthesis (AC2) - **Llama 3.1 (70B)** - Flutter code planning (AC3) - **Jinja2 Templates** - Dart code generation (AC3) ## 📊 Example Usage ### Python ```python import requests response = requests.post( "https://OneWeekApp-space-ac-design.hf.space/api/moodboard-analyzer", json={ "reference_images": [ {"description": "Instagram dark mode", "url": ""} ], "mood_keywords": ["energetic", "gaming"] } ) result = response.json() palettes = result["output"]["analysis"]["color_palettes"] print(palettes) ``` ### cURL ```bash curl -X POST https://OneWeekApp-space-ac-design.hf.space/api/moodboard-analyzer \ -H "Content-Type: application/json" \ -d '{ "reference_images": [{"description": "Uber driver app", "url": ""}], "mood_keywords": ["professional", "modern"] }' ``` ## 🏗️ Architecture This space is part of the **AI Software Factory** system: ``` Notion Wireframe ↓ Space AC (Design) ← You are here ↓ Space A (Architecture) ↓ Space B (Code Generation) ↓ Spaces C-N (Testing, DB, Platform, etc.) ↓ GitHub PR with complete Flutter app ``` ## 📝 Input Examples ### Gaming/Energetic Theme ``` Reference Images: - PUBG Mobile home screen - dark mode, neon accents - Fortnite battle pass - vibrant colors, bold typography - Uber driver stats - clean cards, green highlights Mood: gaming, energetic, bold ``` ### Professional Theme ``` Reference Images: - LinkedIn feed - white cards, blue accents - Banking app - clean layout, trust signals - Notion workspace - minimal, organized Mood: professional, trustworthy, modern ``` ## 🔗 Related Spaces Part of the **AI Software Factory** ecosystem: - **space-a-architecture** - System architecture design - **space-b-codegen** - Flutter code generation - **space-c-testing-white** - Unit & widget tests - **space-d-testing-black** - E2E tests - **space-e-database** - Database & backend ## 📄 License MIT License ## 🤝 Contributing This space is part of a research project. Contributions welcome! ## 📧 Contact - **GitHub:** [AI Software Factory](https://github.com/sabatino81/AISoftwareFactory) - **Space:** [OneWeekApp/space-ac-design](https://huggingface.co/spaces/OneWeekApp/space-ac-design) --- **Built with ❤️ using Gradio, FastAPI, and HuggingFace Inference API**