AI Reading Assessment Platform
Pageflect
An AI-powered reading platform that assesses users on books they've read, built with Next.js, Prisma, and Gemini.
Problem
Readers wanted a way to verify and reinforce their understanding of books they'd read, with assessments generated from the actual book content instead of generic quizzes.
Solution
Built an 8-model Prisma/PostgreSQL schema (User, Book, Highlight, BookPage, BookTextSpan, PageStats, Assessment, ChatMessage) with automatic retry logic for transient DB connection failures. Integrated Gemini on Next.js API routes to reframe book content, with sub-2s measured response latency and model-fallback retry. Built a 5-question AI assessment engine that generates quizzes from book content and gives AI feedback on wrong answers.
System Architecture & Data Pipeline
Next.js 16 System: Multi-tenant reader backed by an 8-model Prisma/PostgreSQL schema with automatic connection retry logic, integrated with Gemini API for sub-2s book reframing and assessment generation.
Highlights
- Built an 8-model Prisma/PostgreSQL schema (User, Book, Highlight, BookPage, BookTextSpan, PageStats, Assessment, ChatMessage) with automatic retry logic for transient DB connection failures.
- Integrated Gemini on Next.js API routes to reframe book content, with sub-2s measured response latency and model-fallback retry.
- Built a 5-question AI assessment engine that generates quizzes from book content and gives AI feedback on wrong answers.
Impact
Built and deployed a Next.js/Prisma platform with an 8-model schema, Gemini-backed content reframing at sub-2s latency, and a working AI assessment engine generating book-specific quizzes with feedback.