Emotion-Based Book Recommendations with AI

This Agentic AI project is an European initiative that combines a digital platform with physical libraries to help people choose novels based on the emotions they evoke (joy, sadness, fear, love, anger, surprise).

Wizards Reply supports an AI-driven Agent to automate emotional classification, enrich it with metadata/summaries/reviews/user signals, and provide a multilingual web back office and APIs to connect partner libraries, promoting well-being and accessible reading across multiple countries. 

Challenges

Here are the main challenges we faced : 

  • Replace subjective, manually maintained emotion labels with a reliable automated scoring. 

  • Design models that work across multiple languages and data sources. 

  • Integrate with partner libraries with very different levels of digital maturity and existing tools. 

  • Deliver a large functional scope safely via incremental releases while keeping quality, security and operational usability for both our client teams and partner staff. 

Solutions

We are delivering an iterative MVP to establish the product foundations: 
An AI Agent that analyzes books and generates emotion profiles; 
A secure multilingual web platform with a back office to manage books, libraries and users; 

Benchmarking and on-demand AI evaluations to monitor model quality over time; 

Bulk import via ISBN/EAN; and a partner API exposing our client's data in JSON plus an embeddable iFrame for fast integration. 
The API was co-designed with a pilot library to ensure interoperability.

Results and value added

This project shows how we can define and implement an MVP scope and roadmap enabling a fast operational launch within a controlled budget. 
The MVP provides: a secure back office for our client's operations (libraries/users/books/indicators management), AI-driven emotional book evaluation with benchmark tracking, multilingual access to book pages, and integration assets for partner libraries (documented JSON APIs and an iFrame widget). 
The approach supports progressive rollout across partner countries while keeping the option to extend with higher-value features in later phases.


Countries: Belgium Slovenia Lithuania Tunisia Poland 
Topics: Library accessibility, AI-Driven Insights, MVP Agentic AI 
Methodology Reply: Agile 
Technologies: MongoDB, Radix UI, Tailwind CSS, ESLint 9, Payload CMS, TypeScript, Next.js