Tuya AI Innovators Hackathon 2025 Submission > Track: AI Agent Application
Author: Ranajit Dhar
👉 CLICK HERE TO WATCH THE PROJECT DEMO VIDEO (Please click the link above to see the Autonomous AI Agent controlling the Tuya IoT Device in action!)
Traditional IoT systems are passive. They wait for human commands (e.g., "Alexa, turn on the lights"). They lack context, reasoning, and the ability to evaluate the quality of work. There is a disconnect between Generative AI (Creative Brain) and Smart Hardware (Physical Body).
YES Ai Studio is a fully autonomous Agentic AI System that bridges this gap. It doesn't just generate content; it evaluates its own output quality and autonomously controls the physical environment (Studio Lighting) based on that evaluation.
Target Users: Smart Studio Owners, AI Developers, Content Creators, and Smart Home Enthusiasts who want logic-based automation.
The system follows a Multi Agent Architecture where specialized AI agents collaborate to perform a task, culminating in a physical action via the Tuya Cloud Bridge.
Workflow:
- 🧭 Navigator Agent: Plans the project workflow & researches trends (using Google Gemini).
- 🎨 Curator Agent: Generates creative content (Social Media Posts, Image Prompts).
- ⚖️ Evaluator Agent: Critically reviews the content and assigns a Quality Score (1-10).
- 🔌 Tuya IoT Bridge: The Logic Layer that connects the AI Brain to Tuya Cloud.
- IF Score ≥ 8: ✅ High Quality -> Turn ON Studio Light (Green Signal).
- IF Score < 8: 📉 Low Quality -> Turn OFF Light (Power Save Mode).
Since this is a Cloud-Native AI Agent, it is designed to be hardware-agnostic. For this hackathon demonstration, we used:
- Device: Tuya Virtual Wi-Fi Switch (Simulating a Smart Studio Light).
- Connection: TuyaOpen API (Cloud-to-Cloud).
Note: This code works instantly with ANY physical Tuya-enabled Switch, Bulb, or Socket.
- AI Core: Google Gemini 2.5 Flash (via Kaggle).
- IoT Platform: Tuya Developer Platform (TuyaOpen API).
- SDK:
tuya-connector-python. - Validation: Pydantic (for Structured JSON Outputs).
Special thanks to Tuya Smart for the robust IoT API and Google for the Gemini AI models.
