---
title: Semantic Book Search (2.4M)
emoji: 📚
colorFrom: blue
colorTo: indigo
sdk: gradio
sdk_version: 6.1.0
app_file: app.py
pinned: false
license: mit
---

# 📚 Semantic Book Search Engine

Welcome to the **AI-powered Book Search Engine**.
Stop searching by exact keywords. This tool allows you to search for books by **describing the plot, the atmosphere, or the emotions** you are looking for.

The system indexes over **2.4 million books**, allowing you to uncover hidden gems using state-of-the-art Natural Language Processing.

## 🚀 How to use it

### 1. 🔎 Search by Plot (Semantic Search)
Can't remember the title? Looking for a specific vibe?
* Try: *"A dystopian novel where books are banned and burned by firemen"*
* Try: *"A psychological thriller set in Victorian London with a plot twist"*
* The model understands the **concept** and retrieves the most semantically similar books.

### 2. 📖 I liked... (Recommendation)
Did you love a specific book?
* Switch to the second tab.
* Search for a title (e.g., *"Harry Potter"*).
* The system retrieves the existing vector from the database and recommends books that are mathematically closest in the latent space (similar style, genre, and plot).

---

## 🛠️ Under the Hood (Technical Architecture)

This project is a showcase of **End-to-End AI Engineering**, designed to handle large-scale datasets in a **Low-Resource Environment**.

* **Dataset:** ~2.4 Million books processed and indexed.
* **AI Embedding Model:** `paraphrase-multilingual-MiniLM-L12-v2`.
* **Hybrid Retrieval Architecture:**
    * 🧠 **Qdrant (Vector DB):** Handles semantic similarity search. Vectors are compressed using **INT8 Scalar Quantization**.
    * 🗄️ **Turso (LibSQL):** Relational database for low-latency metadata retrieval (Title, Author, Year, Rating), keeping the vector payload lightweight.

### 👨‍💻 Author
**Antonio Gagliostro**
* [GitHub Profile](https://github.com/ninooo96)
* [LinkedIn](https://www.linkedin.com/in/antonio-gagliostro-1b4751121)