Case Study

Redictionary (WordBrain)

An AI-powered English-Korean vocabulary learning service combining structured dictionary generation, Redis caching, TTS audio, and quizzes.

Project Overview

Redictionary turns English word searches into structured learning content. A NestJS API generates dictionary entries through structured OpenAI output, caches results in RedisJSON, creates pronunciation audio with Google Cloud Text-to-Speech, and serves quiz sessions, search history, and a React/Vite web app from an Nx monorepo.

Key Challenges

Stabilizing LLM output as product data with a strict dictionary schema
Reducing generation cost and latency through RedisJSON caching
Avoiding duplicate pronunciation audio generation with hashed storage keys
Managing web, backend, mobile, and shared types in one Nx workspace

Key Outcomes

Built a full-stack vocabulary service with React, NestJS, Redis, OpenAI, and Google Cloud
Implemented structured dictionary generation, regeneration, and UI rendering flows
Added TTS pronunciation audio through Cloud Storage signed URLs
Created quiz lists, quiz sessions, answer tracking, and guest learning flows

Technologies

NxTypeScriptReactViteNestJSRedisJSONOpenAI APIGoogle Cloud TTSExpoDocker