AutoGenBook is a Python-based tool that automatically generates books using LLMs. It creates chapters, sections, and subsections recursively based on user-defined content and outputs the final book as a PDF using LaTeX.
The following modifications have been made to AutoGenBook created by hooked-on-mas:
- LLM:
- Modified to allow selection from Claude/Ollama in addition to OpenAI
- Execution Environment:
- Changed from Google Colab to execution via Docker container and command line
- Content:
- Modified to add O'REILLY-style covers
- Increased the inclusion of sample code for programming languages
-
Clone the repository:
git clone https://github.com/Tomatio13/AutoGenBook.git cd AutoGenBook -
Specify the API_KEY in the .env file: You must designate exactly one PROVIDER and MODEL.
cp -p env.example .env
PROVIDER=ANTHROPIC MODEL=claude-3-5-sonnet-20240620 ANTHROPIC_API_KEY=<your claude api key> # PROVIDER=OPENAI # MODEL=gpt-4o # OPENAI_API_KEY=<your openai api key> #PROVIDER=GEMINI #MODEL=gemini-1.5-pro #GEMINI_API_KEY=<your google api key> #GEMINI_BASE_URL=https://generativelanguage.googleapis.com/v1beta/ #PROVIDER=OLLAMA #MODEL=qwen2.5-coder:latest #OLLAMA_BASE_URL=http://localhost:11434/v1/ #OLLAMA_MAX_TOKENS=256
Note You can specify ollama, but the MODEL's max_tokens and num_ctx are small, which often results in generation failures, so it is not recommended.
-
Build the container.
docker compose build
-
Start the container.
docker compose up -d
-
Run the following command in the terminal:
docker exec autogenbook-autogenbook-1 python AutoGenBook.py "Next.jsに関する教科書" "Next.js初学者" 5 2024-10-23 01:23:11,686 - INFO - 1. 初期化しています 2024-10-23 01:23:11,687 - INFO - 2. 本のタイトルと概要を生成を開始します 2024-10-23 01:23:24,998 - INFO - HTTP Request: POST https://api.openai.com/v1/chat/completions "HTTP/1.1 200 OK" 2024-10-23 01:23:25,009 - INFO - 本のタイトル：Pythonプログラミング入門 2024-10-23 01:23:25,009 - INFO - 本の概要 ：この本は、Pythonを初めて学ぶ方から、基礎をしっかりと固めたい中級者までを対象とした教科書です。Pythonの基本的な文法から始め、データ構造、関数、モジュール、そしてオブジェクト指向プログラミングまでをカバーします。各章は実践的な例を通じて、読者が自分のペースで学習を進められるように設計されています。特に、プログラミングの初心者がつまずきやすいポイントを丁寧に解説し、理解を深めるための演習問題も豊富に用意しています。最終的には、読者がPythonを用いて簡単なプログラムを自ら作成できるようになることを目指しています。 2024-10-23 01:23:25,612 - INFO - HTTP Request: POST https://api.openai.com/v1/chat/completions "HTTP/1.1 200 OK" 2024-10-23 01:23:25,621 - INFO - ディレクトリ名:output/py_prog_getting_st 2024-10-23 01:23:25,621 - INFO - 3. 章・節の内容を生成しています 2024-10-23 01:23:34,826 - INFO - HTTP Request: POST https://api.openai.com/v1/chat/completions "HTTP/1.1 200 OK" 2024-10-23 01:23:43,232 - INFO - HTTP Request: POST https://api.openai.com/v1/chat/completions "HTTP/1.1 200 OK" 2024-10-23 01:23:51,013 - INFO - HTTP Request: POST https://api.openai.com/v1/chat/completions "HTTP/1.1 200 OK" 2024-10-23 01:24:03,703 - INFO - HTTP Request: POST https://api.openai.com/v1/chat/completions "HTTP/1.1 200 OK" 2024-10-23 01:24:10,871 - INFO - HTTP Request: POST https://api.openai.com/v1/chat/completions "HTTP/1.1 200 OK" 2024-10-23 01:24:10,874 - INFO - 4. PDFの生成を開始します 2024-10-23 01:24:11,934 - INFO - HTTP Request: POST https://api.openai.com/v1/chat/completions "HTTP/1.1 200 OK" 2024-10-23 01:24:12,007 - INFO - カバー画像:/app/cover_images/d906347b-53ed-4069-bb9e-bacce2aae797.eps 2024-10-23 01:24:12,614 - INFO - output/py_prog_getting_st/Pythonプログラミング入門.pdfの出力が完了しました
-
A PDF file is output under the output folder. In the example of step 5, it is output to the folder "output/py_prog_getting_st" with the file name "Introduction to Python Programming.pdf".
AutoGenBook.py is a script that assists in generating a book. This script creates a book based on the specified content, target readers, and the number of pages.
book_content(required): Specifies information about the content of the book.target_readers(required): Defines the target readers for the book.n_pages(required): Specifies the number of pages in the book.--level LEVEL(optional): Specifies the level of mathematical usage.
Below is a basic usage example of AutoGenBook.py:
docker exec autogenbook-autogenbook-1 python AutoGenBook.py "This is the content of the book" "young adults" 150 --level intermediate1: Use almost no mathematical expressions and explain all concepts in simple language. Use expressions only when absolutely necessary, and keep them to a minimum.
2: Use mathematical expressions sparingly and focus mainly on explanations in text. Use simple expressions only when necessary.
3: Combine explanations using both mathematical expressions and text in a balanced manner. Express important concepts with mathematical formulas and supplement with text for the rest.
4: Actively use mathematical expressions to accurately express concepts and relationships. However, provide important explanations in text as well.
5: Make maximum use of mathematical expressions. Express as many concepts and relationships as possible using formulas.
The service runs at http://localhost:8100.
Endpoint: POST /generate-book
Request Body:
{
"book_content": "Description of the book's content",
"target_readers": "Description of the target readers",
"n_pages": 50,
"level": 1 // Optional: frequency of mathematical expressions (1-5)
}Response:
{
"status": "accepted",
"message": "Book generation started",
"task_id": "Generated task ID",
"author": None
}Endpoint: GET /task/{task_id}
Response:
{
"status": "completed", // "processing", "completed", "failed"
"output_dir": "Path to the output directory",
"title": "Title of the generated book",
"author": "PROVIDER:MODEL_NAME" // Example: "OPENAI:gpt-4" or "ANTHROPIC:claude-3-sonnet"
}Endpoint: GET /download/{task_id}
Downloads the generated PDF book. This is available only if the task is completed.
Endpoint: GET /download-cover/{task_id}
Downloads the generated cover image in PNG format. This is available only if the task is completed.
Endpoint: GET /health
Response:
{
"status": "healthy"
}- Request book generation:
curl -X POST "http://localhost:8100/generate-book" \
-H "Content-Type: application/json" \
-d '{
"book_content": "Introduction to Python Programming",
"target_readers": "Beginner programmers",
"n_pages": 50,
"level": 1
}'- Check task status:
curl "http://localhost:8100/task/{task_id}"- Download PDF:
curl -O -J "http://localhost:8100/download/{task_id}"- Download Cover Image:
curl -O -J "http://localhost:8100/download-cover/{task_id}"API documentation using Swagger UI is available at:
While you can dive into the code for all the details, I realize that it might be a bit hard to follow, so let me explain the basic idea and workflow behind the tool.
Since ChatGPT has limitations on how much text it can generate at once, simply asking it to “write a textbook” results in only 1-2 pages of content. To overcome this, AutoGenBook recursively breaks down the structure of a book starting from the main topic or title. It goes from chapters → sections → subsections, and so on. This approach ensures that ChatGPT can generate meaningful, self-contained book for you without hitting its output limit.
Finally, the content for each subdivided section is generated using ChatGPT, and then output as a PDF.
Here’s an outline of the process. I've simplified it by only showing the flow up to the creation of subsections, but the same recursive structure continues for deeper levels.
graph TD
N[Input specifications] --> A[Generate book title and overview<br>Generate chapter titles and overviews]
A --> B[(Book title and overview)]
A --> B1[(Chapter 1 title and overview)]
A --> B2[(Chapter 2 title and overview)]
B1 --> BB1[Generate sections for Chapter 1]
B2 --> BB2[Generate sections for Chapter 2]
BB1 --> C11[(Section 1.1's title, overview<br>and necessity of division)]
BB1 --> C12[(Section 1.2's title, overview<br>and necessity of division)]
BB2 --> C21[(Section 2.1's title, overview<br>and necessity of division)]
BB2 --> C22[(Section 2.2's title, overview<br>and necessity of division)]
C11 --> D11{Should it be devided?}
C12 --> D12{Should it be devided?}
C21 --> D21{Should it be devided?}
C22 --> D22{Should it be devided?}
D11 --> |Yes| E11[Generate subsections of Section 1.1]
D11 --> |No| F11[Generate content of Section 1.1]
E11 --> E111[(Subsection 1.1.1's title, overview<br>and necessity of division)]
E11 --> E112[(Subsection 1.1.2's title, overview<br>and necessity of division)]
E111 --> F111{Should it be devided?}
E112 --> F112{Should it be devided?}
F11 --> FF[(Content of Section 1.1)]
The O'Reilly-style cover image was created based on the source code from O-RLY-Book-GeneratorPublic.