Skip to content

yushi-apps/yushi

Repository files navigation

English | 简体中文

What is Yushi

Yushi is a no-code on-device AI agent development platform.

Total control. Infinite extensibility.

Get your own agent running locally in minutes.

Distinctive features

1. AI-powered, intelligently assembled

2. Generalized reversible computing, delta evolution.

3. Everything is a plugin, designed for extension.

4. What you see is what you get, incredibly easy to use

Features

  • Deploy SmallThinker locally with support for RK3588 and RK3576.

Quick Start

Install from Git Source

Prerequisites

  • Git and Rust is installed
  • If it’s not installed yet, refer to the prerequisites to set it up.

1. Download Yushi to your local machine

git clone https://github.com/yushi-apps/yushi.git

2. Change directory

cd yushi

3. Install Yushi

cargo install --path cli

4. Create an agent

cargo yushi new my_agent

The directory structure of an agent is as follows:

.
├── Cargo.toml
├── src/
│   ├── main.js
├── .yushi/
│   ├── main_prompt.md
│   ├── model.toml
│   ├── agents/

An agent is simply a regular Cargo project with a .yushi configuration file.

  • main_prompt.md is Agent Prompt file,write in Markdown format.
  • model. tomlis the model configuration file for the agent
  • agents/ directory contains sub-agents, each written in Markdown—one file per sub-agent.

5. Configure the agent

cd my_agent

Navigate to the agent directory and edit .yushi/model.toml to set the model parameters. By default, DeepSeek is used—just fill in your api_key and you're ready to go.

Edit .yushi/main_prompt.md to define the agent; if the file is empty, the LLM model will be used directly.

6. Build and run

cargo yushi build --release
./my_agent/target/release/my_agent

After the build finishes, copy the executable my_agent/target/release/my_agent to your target device and run it.

For cross-compilation, please refer to the Cross-compilation Guide .

7. Use it

curl -X POST http://127.0.0.1:22786/chat -d "What can you do?"

Community

License

This project is offered under a dual-license model.

  • AGPL for community edition
  • For commercial licensing, please contact yushi_app@163.com for details.

Examples

Contact us

  • Email:yushi_app@163.com

About

No-code On-device Agent Development Platform

Resources

License

Unknown, AGPL-3.0 licenses found

Licenses found

Unknown
LICENSE.txt
AGPL-3.0
LICENSE-AGPL.txt

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Contributors

Languages