ConfiDict: A Private AI Vocabulary Coach That Grades Your Own Definitions
ConfiDict is a local-first vocabulary trainer built for the Hacktoberfest Open-Source AI Challenge, designed to help learners focus on the words they understand least. Instead of self-grading flashcards, users write definitions in their own words, and a locally run open-weight Gemma model rates each answer as strong, correct, partial, or incorrect with brief feedback. Each word carries a visible 0-100 confidence score, and a daily five-card session is drawn from the 15 lowest-scoring words using weighted randomness. Intended users include exam candidates preparing for IELTS, TOEFL, or GRE, adults improving workplace vocabulary, and tutors sharing curated word lists. The project is MIT-licensed, stores all data in the browser via IndexedDB without accounts or backend storage, and uses React, Vite, Ollama, and Vitest.
This is an AI-generated summary. ShortSingh links to the original source for the complete article.
Discussion (0)
Log in to join the discussion and vote.
Log in