← Back to list feature

Local-First Session Engine & Targeted Review Loop

Client-side architecture with Fisher-Yates shuffling, targeted review queue, and defensive localStorage persistence without backend dependencies.

Four-step cognitive learning loop diagram for Pocket English. — Think of the situation response, reveal multi-tone cards, evaluate confidence, and targeted review.

Problem

Learners need focused practice without registration or backend overhead, while retaining card shuffling and refresh-safe session persistence.

Decision

Built an isolated state machine in useFlashcardSession with Fisher-Yates shuffle, 3-tier confidence rating (Again, Unsure, Got it), and automated review queue filters persisting to localStorage.

Result

A 50-card session can be resumed seamlessly (Continue Learning), and learning progress stays locally in the browser without an account or backend.

Implementation

Local persistence wraps schema parsing in defensive guards preventing runtime crashes from corrupted storage.
The review queue filters Again and Unsure cards into a targeted mini-deck with zero duplicate items.
Learning progress stays locally in the browser without an account or external tracking.

Related screenshots

Session result evaluation page showing confidence breakdown and review queue. — Summary of 50-card session results and targeted review queue trigger.

Flashcard back face displaying Formal, Polite, and Casual expressions with TTS audio. — Multi-tone expression variants with pronunciation audio buttons and rating controls.