Xiaomi Local Copilot is a future exploration solution for smart homes. Using Xiaomi Home cameras as the source of visual information and a self-developed LLM as its core, it connects all IoT devices throughout the house. Based on the development paradigm of LLM, it enables users to define various family needs and rules in natural language, achieving broader and more creative smart device integration.

• [2025-11] Xiaomi Miloco Framework Open Source

1. New Interaction Paradigm: Based on the development paradigm of LLM, rule-setting and complex device command control can be completed through natural language interaction.

2. New Use for Visual Data: Using camera data streams as a source of perceptual information, the LLM is used to analyze various home scene events contained in the visual data to respond to user queries.

3. On-Device LLM: The home scene tasks are split into two stages: planning and visual understanding. It provides Xiaomi's self-developed on-device model to realize on-device video understanding and ensure family privacy and security.

4. Xiaomi Home Ecosystem: It connects with the Xiaomi Home ecosystem, supports the retrieval and execution of Mi Home devices and scenes, and supports sending customized content for Xiao Home notifications.

   

• Hardware Requirements

CPU: x64 architecture
Graphics Card: NVIDIA 30 series and above, 8GB VRAM minimum (recommended 12GB and above)
Storage: Recommended 16GB or more available space (for local model storage)

• Software Requirements

Operating System:
  - Linux: x64 architecture, recommended Ubuntu 22.04 and above LTS versions
  - Windows: x64 architecture, recommended Windows 10 and above, requires WSL2 support
  - macOS: Not currently supported
Docker: Version 20.10 and above, requires docker compose support
NVIDIA Driver: NVIDIA driver with CUDA support
NVIDIA Container Toolkit: For Docker GPU support

Note: Please ensure your system meets the above hardware and software requirements. Windows systems need to enter the WSL environment.

Install with Docker
One-click installation via command line

bash -c "$(wget -qO- https://xiaomi-miloco.cnbj1.mi-fds.com/xiaomi-miloco/install.sh)"

Or download the source code first, then execute the one-click installation script:

git clone https://github.com/XiaoMi/xiaomi-miloco.git

bash scripts/install.sh

For detailed installation steps, please refer to the Docker Deployment Documentation.

Install with source code
For source code installation steps, please refer to the Development Guide.

Please refer to the Usage Documentation.

Please refer to the Contributing Guide.

For license details, please see LICENSE.md.

Important Notice: This project is limited to non-commercial use only. Without written authorization from Xiaomi Corporation, this project may not be used for developing applications, web services, or other forms of software.

If you discover potential security issues in this project, or believe you may have found a security issue, please notify the Miloco Team via our vulnerability reporting email. Please do not create public GitHub Issues.

For issue reporting, please participate through the following methods:

• Submit a GitHub Issue

• GitHub Discussions

• Project Discussion Group (WeChat):

  

The Xiaomi Miloco team is hiring. Send your resume to xiaomi-miloco@xiaomi.com, and it will be delivered directly to the project lead.

Thank you to the original team members who worked hard for Miloco：zhaoy、yangyongjie、xx、Changyu、yyk、junhui、郭兴宝、47、afei。

Your passion and talent are the fundamental driving force behind Miloco's continuous innovation and progress.

Special thanks to:

• The llama.cpp open source project for providing inference backend capabilities