AI Trends

Deploying the Mirobody Local AI Health System and Using It to Track Fitness

Ayi AI NotesUpdated Nov 20 18:59

English text is being prepared.

AI summary

The author runs Mirobody on a personal computer, integrating fitness records, Apple Health data, and medical examination records into a single local AI health system. The system can run offline on a laptop with 16GB of RAM and no dedicated graphics card, and can connect to Claude Code or Cursor via MCP. The author also notes that it cannot provide diagnoses or prescribe medication, some terminology requires manual verification, processing blurry images on a CPU is slow, and once connected to a cloud-based AI client, relevant readings are still sent to the model provider.

Why read this

AI score 61

The original article also outlines local deployment requirements, the health data organization workflow, and privacy boundaries, providing a basis for evaluating whether the personal health record workflow is suitable for individual use.

Topics

Tags

Quoted post

AYi

@AYi_AInotes

When you get your annual checkup at a different hospital, last year's report ends up buried at the bottom of a drawer; switch phones and you cannot find it at all. Even the mmol/L and mg/dL units do not match up. Then, when a doctor asks how your results have changed over the past few years, you are completely stumped! This step-by-step guide shows you how to build a local AI health system for yourself and your parents: it runs offline with 16GB of RAM and no graphics card, connects to Claude Code / Cursor, and thoroughly explains the four current limitations 👇 https://t.co/h9TKShKv3p

View original

每年换医院体检,去年的单子压在抽屉底、换个手机找不着,连 mmol/L 和 mg/dL 单位都对不上,看病时医生问前几年指标怎么变的,当场抓瞎! 手把手教你给自己和爸妈搭一套本地 AI 健康系统:16GB 无显卡就能离线跑、接进 Claude Code / Cursor,连目前做不到的 4 个坑全讲透👇 https://t.co/h9TKShKv3p

View original post

Body

看到这个我不会再做传统俯卧撑了, 也建议所有铁汁,不管是小登,中登,还是老登, 不说每天200个,多大多是人来说要求太高了,

迈克泰森俯卧撑,同样的时间,能同时锻炼到全身大部分肌群,包括三头肌、肩,胸,核心力量,心肺耐力、腿部, 以前每天死磕两百个传统俯卧撑,练半天只酸胸肌,换成迈克泰森俯卧撑后,同样的时间能把三头肌、肩膀、胸、核心、心肺和腿部全带一遍,每天做五十到一百个就够了, 但练完最头疼的是身体账本一团乱,今天做了多少组、Apple Health 里的静息心率降没降、去年体检单上的 mmol/L 和今年 mg/dL 血脂怎么对比,全散落在手机相册和抽屉底, 最近我把 GitHub 排名第一的开源健康引擎 Mirobody 跑在了自己电脑上,直接把日常健身记录和体检档案接在了一起, 👉 16GB 内存普通笔记本就能纯本地离线运行,不用独立显卡,还能通过 MCP 直接接进 Claude Code 或 Cursor, 👉 练完像发微信一样输入一句刚做完六十个迈克泰森俯卧撑、静息心率降到六十、左肩有点紧,系统自动拆解编码记成标准日志, 👉 内置 63,416 个国际医学标准编码和 102 万条多语言别名,把 Apple Health 导出包和跨医院体检照片往里一丢,本机词表 0.4 秒自动换算单位拉出多年趋势线,每个数字点开都能跳回原报告核对, 不过动手跟练和部署前,我先把这事做不到的几个坑全摊开,免得你白折腾: 迈克泰森俯卧撑讲究爆发力和控制力,对胸肌的孤立刺激不如传统动作,姿势不对很容易伤肩膀,新手别硬凑五十个,做十个力竭就先停, Mirobody 只负责数据整理和原件溯源,绝对不看病开处方,指标正不正常依然得听医生的, 遇到冷门医院缩写或血脂这类大类统称,它会老实留空标成待确认,需要你手动点一次确认,不会硬凑, 无显卡纯 CPU 跑本地模型,读手机拍的糊图要等几分钟,存自己和爸妈真实报告前,务必先关掉配置里的演示数据开关,一旦连上云端 AI 客户端,查到的读数依然会发给模型厂商, 大家每天练完汗流浃背、吃了一堆补剂,到底有哪一项在你的真实体检单和心率曲线上留下过证据? Mirobody 2 分钟零代码体验 + 本地 Docker 部署避坑参数评论区自取呀~ https://t.co/KlfjFlILPl

Source

Ayi AI Notes