title	app_file	sdk	sdk_version
Book-Recommender	gradio-dashboard.py	gradio	5.27.0

📚 Book Recommender Application

Welcome to the Book Recommender!
An intelligent system built using LangChain, OpenAI embeddings, and a dataset of 7K+ books to help users find books similar to their interests.
This project uses semantic search powered by vector databases for smarter, meaning-based recommendations.

---

📂 Dataset Information

Dataset sourced from Kaggle: 7k+ Books with Metadata

Main columns:

Column Name	Description
isbn13	ISBN 13 Identifier
isbn10	ISBN 10 Identifier
title	Title of the book
subtitle	Subtitle of the book
authors	Authors (separated by ;)
categories	Categories/Genres (separated by ;)
thumbnail	URL of the book thumbnail
description	Book description
published_year	Year the book was published
average_rating	Average user rating (where available)

---

🔥 Features Implemented

• Data Exploration: Cleaning, understanding, and preparing book metadata.
• Vector Search Engine:
  • Uses LangChain for document loading, text splitting, and embedding generation.
  • OpenAI's Embedding Models used to embed book descriptions into vector space.
  • ChromaDB as a lightweight, fast vector database for storage and retrieval.
• Semantic Search: Find books not just by keywords, but based on meaning!
• Batch Processing: Batching documents for efficient embedding within OpenAI’s rate limits.
• Text Classification using Zero-shot Learning:
  • Broadly categorizing books into higher-level groups (Fiction, Nonfiction, Children's categories).
  • Using Hugging Face Transformers (e.g., facebook/bart-large-mnli) without additional model training.
• Sentiment and Emotion Analysis:
  • Predicting emotional tone (anger, joy, sadness, etc.) from book descriptions.
  • Fine-tuned transformer models used (e.g., j-hartmann/emotion-english-distilroberta-base).
  • Allows filtering books based on emotional tone for better personalized recommendations.
• Environment Management:
  • API keys and other secrets managed securely using .env files.
  • Requires OPENAI_API_KEY to be set.

---

⚙️ Setup Instructions

1. Clone the repository

git clone https://github.com/your-username/book-recommender.git
cd book-recommender

2. Set up a virtual environment (recommended)

# Create a virtual environment
python -m venv venv

# Activate the virtual environment
# On Windows:
venv\Scripts\activate
# On macOS/Linux:
source venv/bin/activate

3. Install dependencies

pip install -r requirements.txt

Make sure you have Python 3.10+ installed.

4. Setup your API Key

Create a .env file in the project root directory and add your OpenAI API Key:

OPENAI_API_KEY=your_openai_api_key_here

You can find or create your key at https://platform.openai.com/account/api-keys.

5. Additional Setup for Hugging Face and Mac MPS (Apple Silicon) Support

If you plan to use Hugging Face Transformers for zero-shot classification and want to enable hardware acceleration on your Mac (using MPS — Metal Performance Shaders), you need to install PyTorch correctly.

(I) Install PyTorch with MPS support (for macOS)

Follow the official installation instructions here, or run the following for macOS:

# Install PyTorch for macOS with MPS backend
pip install torch torchvision torchaudio

Note: MPS support is available on macOS 12.3+ with Apple Silicon (M1, M2, M3 chips) or newer Intel Macs.

---

(II) Check if MPS is available

After installing, you can check MPS support with this snippet:

import torch

if torch.backends.mps.is_available():
    print("✅ MPS is available. Using hardware acceleration!")
else:
    print("⚠️ MPS is not available. Running on CPU.")

(III) Why is PyTorch Required?

The Hugging Face transformers library needs either PyTorch, TensorFlow, or Flax as a backend to load models like facebook/bart-large-mnli.
Without a deep learning backend, you will only be able to use tokenizers and configs — not the actual models.

(IV) Additional Requirements for Hugging Face

Make sure you have installed:

pip install transformers

If not already present, install huggingface_hub and other related libraries too:

pip install huggingface_hub

---

✨ How It Works

• Load book descriptions into documents
• Split text into manageable chunks
• Embed documents using OpenAI’s text-embedding-ada-002 model
• Store embeddings in Chroma (persisted locally in SQLite and binary files)
• Query the database semantically to find similar books
• Classify books into higher-level groups to support better filtering
• Analyze emotional tones of descriptions for sentiment-based recommendation enhancements

The embeddings and metadata are stored in the datasets/chroma_db/ directory.

---

🔮 Upcoming Features

• Gradle Dashboard:
  An interactive dashboard to visually explore books, categories, and search recommendations.

---

🛠️ Tech Stack

• Python (Jupyter Notebooks)
• LangChain
• OpenAI API
• Hugging Face Transformers
• ChromaDB
• dotenv
• Pandas, NumPy
• Matplotlib, Seaborn (optional)

---

📜 License

This project is licensed under the MIT License — feel free to use, modify, and distribute!