qingjian-team/qingjian

What it solves

It transforms the act of typing into a passive language learning experience. Instead of using a separate app for study, users can learn new vocabulary and translations naturally while performing their daily typing tasks without interrupting their workflow.

How it works

Developed in Rust, this cross-platform input method displays a concise translation of the candidate word in a target "learning language" next to the primary input. It uses a local small model to rerank candidates and provide translations and parts of speech. The system is designed to be lightweight, ensuring that the translation remains a secondary, non-intrusive piece of information.

Who it’s for

People who want to acquire a new language (such as English or Japanese) as a side effect of their normal typing habits, specifically those who prefer a low-friction, non-forced learning method.

Highlights

• Local AI Processing: Uses local small models for candidate reranking and translation, ensuring privacy.
• Privacy-First: No data is uploaded; all processing happens on the local machine.
• Cross-Platform: Designed with a platform-agnostic core to support macOS, Windows, and Linux.
• Single-Language Focus: Displays only one learning language at a time to keep the interface clean and efficient.