Concise & Comprehensive

Developers will never have to learn an extra new concept or write code irrelevant to the business. The client supports development environments such as Arduino, IDF (under implementation), and PlatformIO.

Plug - in Architecture

Built - in IAT, TTS, and LLM from iFLYTEK, ByteDance, and Alibaba Cloud. Other platforms can be extended through plugins.

Multiple Interaction & Wake - up Methods

Customizable offline voice wake - up. Voice inference uses TensorFlow Lite models. An online wake - word generation service will be launched in the future. It supports continuous conversation, push - to - talk, or any other conversation mode.

Command Recognition

Developers can customize user commands or use built - in commands to enable robots to control home appliances, sing, etc.

C/S Architecture

The server and client are separated. Complex calculations and service scheduling are all handled by the server, greatly reducing the cost of the client and providing services to N clients simultaneously.

Identity Authentication

Identity verification of the client can be performed when each client connects and during each session.

Real - time Response

All service interfaces use the `WebScoket` protocol for streaming transmission, ensuring real - time server response. Every character inferred by the LLM can be immediately sent to the user.

Configurable

Except for plugins written for special businesses, the configurations provided by the framework or plugins can generally meet most scenarios.

Production - ready & Integratable

As a development library released for production environments, ESP - AI has strict requirements for scalability, flexibility, and stability. It can be easily integrated into existing projects.

🎉 ESP-AI 公告 🎉