跪拜 Guibai
← All articles
Frontend · Programmer · AI Programming

Ink-Learner Puts Active Recall and Spelling Back Into a Desktop Vocabulary App

By 碎_浪 ·
Read original on juejin.cn ↗ Google Translate ↗ Alt translation

Most vocabulary tools train recognition, not production. Ink-Learner forces active recall through the keyboard, which matters for developers and anyone who needs to write English, not just read it. The local-first, AI-as-material-factory design also avoids the recurring token cost and latency of chat-wrapped memorization apps.

Summary

Ink-Learner is a macOS and Windows desktop application that treats vocabulary acquisition as a typing discipline. Instead of swiping flashcards or running speed tests, users move through a four-stage chain: spelling a target word, recalling it from a hidden prompt, full dictation, and typing it inside a short sentence. Eleven syllabus-aligned word lists are built in, and all practice records, wrong-word collections, and configurations live in a local SQLite database that supports backup and offline use.

Wrong words are never added automatically. After each chapter, users manually select which mistakes to keep, and those entries enter a spaced-repetition schedule. An optional AI pipeline can fill in missing definitions, phonetic symbols, and example sentences, but the results are written once into the local database rather than regenerated on every session. The tool deliberately borrows the typing-practice interaction from Qwerty Learner while diverging into a full desktop panel with a manual memory loop.

The project is open-source under GPLv3, with learning materials under CC BY-SA 4.0. Planned additions include a sentence-by-sentence intensive reading mode, a mini floating practice window, and mobile support.

Takeaways
Practice follows a four-stage chain: spelling, occluded recall, dictation, and short-sentence typing.
Wrong words are never auto-added; users manually curate their mistake list after each chapter.
Curated wrong words enter a spaced-repetition schedule for scheduled review.
An optional AI pipeline fills in missing definitions, phonetics, and example sentences, then writes results once into the local SQLite database.
All data stays local, with backup and restore support, and the app works fully offline.
Eleven built-in word lists cover middle school through GRE, GMAT, and BEC.
The desktop app is built with Tauri v2, React, and Vite, and is available for macOS and Windows.
Conclusions

Manual wrong-word curation is a deliberate design choice that treats the mistake list as a study plan rather than an error log, which contrasts sharply with apps that auto-pollute the review queue on a single typo.

Using AI as a one-time material generator that writes to a local database, rather than as a real-time tutor, sidesteps the token-cost and latency problems that make many AI vocabulary apps impractical for daily drill.

The project explicitly positions itself as a complement to Qwerty Learner, not a replacement, which is a rare and honest product differentiation in open-source tooling.

Concepts & terms
Spaced Repetition System (SRS)
A learning technique that schedules review of material at increasing intervals to improve long-term retention. Ink-Learner applies SRS only to words the user manually adds to the wrong-word list.
Tauri
A framework for building lightweight desktop applications using web frontend technologies (like React) with a Rust backend, offering smaller bundle sizes than Electron.
Source: juejin.cn ↗ Google Translate ↗ Backup ↗