Sift: A Local-First Intelligent Medical Record Processing Tool Built with Rust + Tauri

Sift is a high-performance Windows desktop application developed with Rust and Tauri, focusing on automated medical record processing. It adopts a local-first architecture, converting unstructured clinical data into structured longitudinal summaries via local LLM inference—enhancing medical data processing efficiency while protecting patient privacy.

In modern healthcare systems, a large amount of unstructured medical data (e.g., PDFs, handwritten notes, scattered electronic records) relies on manual organization, which is time-consuming, labor-intensive, and error-prone. More critically, medical data is privacy-sensitive—cloud processing carries risks of leakage and compliance issues. How to efficiently convert unstructured data under security premises is a major challenge in healthcare informatization.

Developed by the fleXRPL team, Sift is an open-source Windows desktop application. It uses Rust as the core backend (for memory safety, high performance, and strong concurrent processing capabilities) and combines the Tauri framework to build a lightweight interface (small size, fast startup, low memory usage), balancing performance and user experience. Its core concept is local-first: all processing is done on local devices, eliminating data leakage risks and being free from network restrictions.

Local LLM Inference Pipeline: Accepts PDF, scanned images, or text input → OCR text extraction → Local LLM analyzes and identifies key data such as diagnoses, medications, and test results → Organizes into a structured longitudinal timeline. Privacy Protection Design: All processing is done locally—no usage data or telemetry is collected; Rust's memory safety features prevent vulnerabilities like buffer overflows, reducing data leakage risks.

• Clinical Doctors: Automatically generates structured patient timelines, reducing medical record organization time—quickly access key information during emergencies or consultations.
• Medical Research: Automates extraction of specific information, compressing manual organization time (from weeks to hours), meeting ethical review and informed consent requirements.
• Small Clinics: Low-cost digital solution—no need for expensive HIS systems or cloud subscriptions; ordinary Windows computers can achieve intelligent medical record management.

Technical Implementation: Modular architecture separates document parsing, OCR, LLM inference, and UI components for easy maintenance and expansion; supports multiple local model backends—users can choose model sizes based on hardware (large models for high-performance GPUs, lightweight models for ordinary computers). Open-Source Ecosystem: Hosted on GitHub with a permissive license to encourage community participation; README documentation lowers entry barriers, and the development team actively responds to feedback; open-source transparency allows users to review code and confirm no hidden data collection.

Limitations: Currently only supports Windows platform, limiting cross-system popularity; local LLM performance is hardware-dependent—processing speed on low-end devices needs improvement. Future Outlook: Based on Tauri architecture, it can be extended to macOS and Linux; with the advancement of edge AI technology and hardware upgrades, performance issues are expected to be alleviated.

Sift represents the development direction of healthcare software—leveraging AI capabilities while upholding privacy and security bottom lines. The local-first architecture provides a paradigm for sensitive data applications. For developers, it demonstrates the potential of Rust+Tauri in high-performance desktop applications; for healthcare practitioners, it offers a practical and secure tool; for privacy-conscious users, it proves that convenience and security can coexist. Amid the trend of healthcare informatization, such tools will help institutions in digital transformation, enhancing efficiency while safeguarding privacy.