📚 TrialReads: Full-Stack AI Book Assistant & Library Manager
A modern full-stack app that helps you decide what to read and manage your library — a FastAPI backend orchestrating an agentic ReAct AI, paired with a responsive Next.js + TypeScript frontend featuring a chat interface and an editable library data grid.
A Look at the App
Screenshots from the live TrialReads application.
SituationThe context
Readers struggle to judge whether a book is worth their time and money from limited previews, while tracking a personal reading history across scattered spreadsheets is tedious and error-prone. TrialReads set out to solve this with a polished, production-grade product that combines an intelligent AI assistant with hands-on library management.
TaskThe objective
Design and ship a modern, decoupled full-stack application: an agentic AI backend that intelligently routes a user's intent to the right tool, and a responsive frontend that unifies conversation with hands-on library management — architected to be independently deployable and to run end-to-end in production.
ActionWhat I built
- Designed a monorepo with a clean separation of concerns — a FastAPI backend for LLM orchestration and business logic, and a Next.js frontend for the UI — communicating over an API-first REST interface.
- Built an agentic ReAct agent (OpenAI gpt-4o-mini) that performs multi-tool orchestration, automatically routing each message to the right capability: summarization, recommendations, or library queries.
- Implemented core services as tools: book summarization (chapter-by-chapter), a text-to-SQL library manager via LlamaIndex that turns natural-language questions into exact SQL queries, and an LLM recommendation engine returning similar titles with Amazon links.
- Integrated the Google Books API to fetch book cover images, giving the library grid and recommendations an attractive, visually rich interface.
- Crafted a responsive Next.js 14 (App Router) + TypeScript frontend styled with Tailwind CSS and shadcn/ui, featuring a real-time chat tab (with typing indicators and history) and an editable library data grid supporting inline add/edit/delete, validation, and save/refresh.
- Deployed the full stack end-to-end: Vercel for the Next.js frontend, Render for the FastAPI backend, and Supabase (PostgreSQL) for the database — demonstrating the ability to build and ship a complete application from development to production.
ResultThe outcome
- Shipped a production-ready, fully deployed full-stack product — live across Vercel, Render, and Supabase — with a polished, responsive UI and a maintainable, scalable architecture.
- Delivered a genuinely agentic experience — the ReAct agent picks the right tool on its own, so users just type naturally with no modes or tabs to manage.
- Made library answers exact and verifiable through real text-to-SQL queries, and enriched the UI with Google Books cover art for an attractive, browsable library.