👉 Click here to follow our complete step-by-step tutorial and learn how to build this from scratch with detailed code walkthroughs, explanations, and best practices.
A multi-agent system built with Google ADK that analyzes photos of your space, creates personalized renovation plans, and generates photorealistic renderings using Gemini 3 Flash and Gemini 3 Pro's multimodal capabilities.
- 🔍 Smart Image Analysis: Upload room photos and inspiration images - agent automatically detects and analyzes them
- 🎨 Photorealistic Rendering: Generates professional-quality images of your renovated space using Gemini 3 Pro
- 💰 Budget-Aware Planning: Tailors recommendations to your budget constraints
- 📊 Complete Roadmap: Provides timeline, budget breakdown, contractor list, and action checklist
- 🤖 Multi-Agent Orchestration: Demonstrates Coordinator/Dispatcher + Sequential Pipeline patterns
- ✏️ Iterative Refinement: Edit generated renderings based on feedback
The system uses a Coordinator/Dispatcher pattern with three specialized agents:
-
Visual Assessor 📸
- Analyzes uploaded room photos (layout, condition, dimensions)
- Extracts style from inspiration images
- Estimates costs and identifies improvement opportunities
-
Design Planner 🎨
- Creates budget-appropriate design plans
- Specifies exact materials, colors, and fixtures
- Prioritizes high-impact changes
-
Project Coordinator 🏗️
- Generates comprehensive renovation roadmap
- Creates photorealistic rendering of renovated space
- Provides budget breakdown, timeline, and action steps
-
Clone the repository
git clone https://github.com/Shubhamsaboo/awesome-llm-apps.git cd awesome-llm-apps/advanced_ai_agents/multi_agent_apps/ai_home_renovation_agent -
Install dependencies
pip install -r requirements.txt
-
Set up your API key
export GOOGLE_API_KEY="your_gemini_api_key"
Or create a
.envfile:GOOGLE_API_KEY=your_gemini_api_key -
Launch ADK Web
cd multi_agent_apps adk web -
Open browser and select "ai_home_renovation_agent"
[Upload photo of your kitchen]
"What can I improve here with a $5k budget?"
→ Agent analyzes your space, suggests budget-friendly improvements, generates rendering
[Upload photo 1: your kitchen]
[Upload photo 2: Pinterest inspiration]
"Transform my kitchen to look like this. What's the cost?"
→ Agent extracts style from inspiration, applies to your room, provides budget + rendering
"Renovate my 10x12 kitchen with oak cabinets and laminate counters.
Want modern farmhouse style with white shaker cabinets. Budget: $30k"
→ Agent creates design plan and generates rendering from description
[After initial rendering]
"Make the cabinets cream instead of white"
"Add pendant lights over the island"
"Change flooring to lighter oak"
→ Agent refines the rendering with your feedback
- "I want to renovate my small galley kitchen. It's 8x12 feet, has oak cabinets from the 90s. I love modern farmhouse style. Budget: $25k"
- "My master bathroom is tiny (5x8) with a cramped tub. I want a spa-like retreat with walk-in shower. Budget: $15k"
- "Transform my boring bedroom into a cozy retreat. Thinking accent wall, new flooring. Budget: $12k"
- google_search: Finds renovation costs, materials, and trends
- estimate_renovation_cost: Calculates costs by room type and scope
- calculate_timeline: Estimates project duration
- generate_renovation_rendering: Creates photorealistic renderings
- edit_renovation_rendering: Refines renderings based on feedback
- Versioned artifacts: Automatic version tracking for all renderings
Demonstrates Coordinator/Dispatcher + Sequential Pipeline:
Coordinator (Root Agent)
├── Info Agent (quick Q&A)
└── Planning Pipeline (Sequential)
├── Visual Assessor (image analysis)
├── Design Planner (specifications)
└── Project Coordinator (rendering + roadmap)
Why this pattern?
- Efficient: Only runs workflows that are needed
- Modular: Each agent has clear responsibilities
- Scalable: Easy to add new features
- Production-ready: Real-world agentic system pattern