BusyBox client adds LLM fallback to Xiaomi AI speaker
The project author released an MIT-licensed BusyBox shell client for the Xiaomi AI Speaker MDZ-25-DA, also identified as S12A. It keeps native XiaoAI handling for requests such as lighting, volume, alarms and weather, while routing unsupported questions or explicit DeepSeek requests to a configurable large-model API. Answers are synthesized with EdgeTTS and played sentence by sentence, without requiring a relay server or an always-on computer. The author reports testing only the MDZ-25-DA/S12A. Installation requires opening the device, connecting a serial interface and possibly modifying system partitions, creating risks of bricking, warranty problems and instability. Voice interruption during large-model playback is not supported, and all-day stability remains under validation.
The project offers a way to add configurable large-model responses to an existing Xiaomi speaker while retaining its native controls and avoiding an always-on relay computer. However, the evidence comes from the project author, and testing is limited to one model.

TL;DR:在老款小米 AI 音箱(MDZ-25-DA / S12A)上跑一个 shell 客户端。开灯、天气这类请求仍由小爱原生处理;小爱答不上来、或者你点名“问问 DeepSeek”时,音箱直接调大模型 API,用 EdgeTTS 逐句播报,播完绿灯亮起,可以免唤醒继续追问。不需要中转服务器,也不用电脑常驻。MIT 开源:github.com/open-speake…
一、背景与痛点
我家有一台 2019 年买的小米 AI 音箱。控制家电它很好用:开灯、定闹钟、问天气,反应快,也很稳定。但只要问点开放性问题,比如“介绍一下西湖”“最近三天有什么 AI 新闻”,它多半回一句“这个我还不会”。
想给它接大模型,需求其实很明确:
- 原有能力一个都不能丢:家电控制是刚需,不能因为接了大模型就变慢、变不稳定。
- 不要再养一台常驻设备:不想为了一个音箱让 NAS 或电脑 7×24 小时开着。
- 对话要自然:答完能接着问,不用每句都喊“小爱同学”。
#xiaomi_ai_llm#Xiaomi AI Speaker MDZ-25-DA#Xiaomi AI Speaker S12A#XiaoAI#DeepSeek#EdgeTTS#OpenAI-compatible API#GLM#Kimi#MiniMax