BookBot

A Streamlit book recommendation and Q&A chatbot. Search for a book by title or description, get similar recommendations retrieved from a Pinecone vector index, and chat with the bot about any book using GPT-3.5-turbo.

How it works

1. Search query is embedded with sentence-transformers/all-MiniLM-L6-v2 (runs locally via the transformers library).
2. The embedding is sent to a Pinecone index that stores metadata for a large book catalogue (title, author, year, publisher, language, category, summary).
3. Top-1 nearest neighbor returns the matched book's metadata; top-10 returns similar books.
4. Summaries that are short or non-English are expanded into 150-300 word English paragraphs via GPT-3.5-turbo.
5. A persistent chat session backed by GPT-3.5-turbo lets you ask questions about any book.

Built with

• Streamlit -- UI and session state management
• Pinecone -- vector index for book metadata and similarity search
• OpenAI GPT-3.5-turbo -- summary expansion and conversational Q&A
• sentence-transformers/all-MiniLM-L6-v2 -- local text embedding model (via HuggingFace transformers)

UI

Two tabs:

Tab	What it shows
Recommended Books	Top-10 most similar books to the search query, each with cover image, metadata, and an expanded English summary
Chat	Persistent multi-turn chat with BookBot; context resets when a new book is searched

The sidebar displays the matched book's cover image, title, author, year, publisher, language, category, and an expanded summary.

Setup

Prerequisites

• Python 3.8+
• OpenAI API key
• Pinecone API key (with a pre-populated index -- see Data below)

Install

git clone https://github.com/Animesh352/BookBot.git
cd BookBot
pip install -r requirements.txt

Set environment variables

macOS / Linux:

export OPENAI_API_KEY=your_openai_api_key_here
export PINECONE_API_KEY=your_pinecone_api_key_here

Windows:

set OPENAI_API_KEY=your_openai_api_key_here
set PINECONE_API_KEY=your_pinecone_api_key_here

Run

streamlit run app.py

Data

Book metadata is stored in chunked CSV files under Data/ (df_chunk_1.csv through df_chunk_14.csv). Each row contains at minimum: book_title, book_author, year_of_publication, publisher, Language, Category, Summary, img_l (cover image URL).

You need to load this data into a Pinecone index with embeddings generated from utils.py's text_to_embedding() before the app can run. The embedding model is sentence-transformers/all-MiniLM-L6-v2.

Language support

The app recognizes and displays language names for 30+ languages including Arabic, Chinese (Simplified/Traditional), French, German, Hindi, Japanese, Korean, Russian, Spanish, and more.

Author

Animesh Giri -- Animesh352