---
title: Preface
summary: Introduces the overall background of Tuya's AI Application Development Solution, including core features (complete chain, scenario-based AI capabilities, rich templates, multi-type distribution) and On-App AI capabilities (lightweight and dynamic, real-time high efficiency, low-cost privacy protection).
questions:
  - What are the core features of Tuya's AI Application Development Solution?
  - What is On-App AI and what are its key advantages?
  - Do scenario-based AI capabilities require developers to have AI development skills?
  - What application scenarios does the AI solution support (device panels, smart apps, etc.)?
  - What does "lightweight and dynamic" mean specifically for On-App AI?
  - How does On-App AI achieve low cost and privacy protection?
  - What three layers does end-to-end AI collaboration refer to?
---

# Preface

As AIGC capabilities rapidly evolve and iterate, AI applications in the IoT domain are expanding at an accelerating pace. AI enhances interconnectivity among devices, enabling more intuitive interaction with users through technologies such as natural language processing (NLP), computer vision, and AIGC, while also enabling better coordination across devices. To meet the growing demand from customers and developers in AI applications, Tuya's AI Application Development Solution is built around **end-to-end collaboration** as its core architecture — connecting AI capabilities across the **cloud, device, and app layers** to provide developers with a complete-chain AI solution, enabling rapid deployment of AI hardware and applications without requiring deep AI development expertise.

## Core Features

- **Complete Link**: Provides end-to-end AI product solution capabilities, helping developers quickly understand and apply AI capabilities across categories.
- **Scenario-based AI Capabilities**: Offers pre-built, scenario-specific AI capabilities for various categories, allowing developers to integrate AI into devices and solutions without needing AI development skills.
- **Rich Templates**: Provides panel templates for various scenarios, enabling developers to quickly get started and complete application development.
- **Multi-type Distribution**: Supports not only device panels but also smart apps and other application scenarios, facilitating AI capability integration across devices and product solutions.

## End-to-End AI Collaboration

### Cloud AI

Provides dual-track large model and small model capabilities: **large models** target complex reasoning and generation needs by integrating mainstream LLMs and multimodal models; **small models** are fine-tuned for specific category scenarios, offering efficient inference at controlled cost, well-suited for high-frequency vertical use cases.

### Device-side AI

Deploys AI inference capabilities directly onto hardware devices, enabling **offline availability and real-time response**. Designed for network-constrained or privacy-sensitive scenarios, typical capabilities include local wake word recognition, on-device visual detection, and offline semantic understanding — significantly reducing reliance on the cloud and lowering data transmission costs.

### On-App AI

Runs lightweight inference models on the mobile app (On-App AI) to deliver local real-time AI capabilities, complementing the cloud and device layers at the mobile interaction level. Supports online dynamic model updates, low-latency local computation, and on-device data processing to protect user privacy, along with rich AI scenario templates for rapid application integration.

On-App AI capabilities include the following key advantages:

- **Lightweight and Dynamic**: Utilizes lightweight models optimized for mobile devices running on the app, supporting on-demand mechanisms such as online model downloading, updating, and deployment to improve operational efficiency and achieve a more flexible AI architecture.

- **Real-time and High Efficiency**: Relies on efficient local computation to ensure a smooth, latency-free user experience, enabling quick user responses and real-time processing for a superior user experience.

- **Low Cost and Privacy Protection**: Localized low-cost processing reduces unnecessary token consumption. Data is generated and processed within the app, protecting user privacy and ensuring secure computational costs.

- **Rich Templates**: Provides a wealth of AI scenario templates, allowing developers to quickly get started and complete application development, achieving a hardware-software integrated AI competitive solution.