Record a patient consultation on Android and have AI draft the admission record automatically.
Open the HTML file in a Windows browser, type in patient data, and generate a progress note with no installation.
Connect either tool to any OpenAI-compatible AI service like Qwen or DeepSeek using your own API key.
Requires your own API key for an OpenAI-compatible AI service such as Alibaba Qwen or DeepSeek.
MedLift (规培-medlift) is a tool that uses AI to help Chinese residency training doctors write medical records faster. In China's hospital system, residency doctors (规培医生) rotate through multiple departments and are responsible for writing detailed documentation for every patient they handle, which can mean several hours of paperwork per shift on top of clinical duties. This project aims to reduce that documentation burden. The project has two separate tools. The first is an Android app where a doctor can record a spoken consultation with a patient, photograph printed lab results or test reports, and narrate updates about a patient's condition during their hospital stay. The app sends that audio and image content to an AI service, which generates the corresponding medical record sections: admission records and progress notes. The second tool is a simple HTML file that runs in any modern browser on Windows with no installation required. A doctor types in the patient's medical history, current lab results, orders, and key dates, and clicks a button to generate a progress note. Both tools work with any AI service that follows the OpenAI API standard, including Alibaba's Qwen models and DeepSeek. You supply your own API key in the app settings or in the browser tool. Audio recordings and photos are processed on your device and then sent to whatever AI provider you configure, no data is stored on third-party servers beyond your chosen provider. The README includes a clear reminder that AI-generated records are aids, not final documents. The responsible physician must review and sign off on everything before it becomes official clinical documentation. The README also notes that any patient identifying information (name, ID number, contact details) should be removed from photos before uploading. The Android app is built with Kotlin and Jetpack Compose. The Windows tool is pure HTML and JavaScript. The project is licensed under Apache 2.0.
← liuguangyihuanchang on gitmyhub — every repo by this author, as a profile.
Verify against the repo before relying on details.