reader3

A lightweight, self-hosted EPUB reader with a built-in AI chat window. Read any EPUB one chapter at a time, and ask a local LLM (Ollama) or the OpenAI API questions about the current chapter—no copy/paste required.

Inspired by Karpathy’s idea of “reading books with LLMs” can be found in this repository "karpath/reader3". This project is intentionally simple, hackable, and easy to modify with the help of any LLM.

Final UI

Below is how the reader appears with integrated chat window....

Response with questions in the chapter.....

Response with questions not in the chapter....

---

Features

• 📖 Clean, chapter-based EPUB reader
• 💬 Integrated chat window on the right
• 🤖 Supports local LLMs (Ollama) or OpenAI GPT models
• 📘 Automatically sends the current chapter text as context
• 🔒 LLM is restricted to chapter content only
• 🧹 .env support with python-dotenv
• 🗂️ Simple library system (*_data folders)

---

Usage

1. Install dependencies

Install all the required dependencies of the file

pip install -r requirements.txt

2. Process an EPUB

Download an EPUB (e.g., Dracula from Project Gutenberg) and process it. Place this file in the project folder

python reader3.py dracula.epub

This creates a folder:

dracula_data/
  ├── book.pkl
  └── images/

Each *_data folder becomes a book in your library.

3. Run the server

python server.py

visit: http://localhost:8123/ You'll see local libary and can open any book

4. Enable OpenAI

Create a .env file in your project folder with below line defining your open ai key.

OPENAI_API_KEY=your_key_here

If omitted, only local LLM model is available