✨ What is Promptbook?
Promptbook is an innovative ecosystem that lets you write AI applications (chatbots, knowledge bases, agents, translators, automation, etc.) in human-readable language.
"A paradigm shift is needed! The future of software lies in plain languages like English, French, and Latin."
🚀 New Features
- VSCode Plugin: A VSCode plugin supporting the
.bookextension - Docker Image: Also available as a Docker image
- OpenAI o3-mini model support: Supports the o3-mini model
- DeepSeek model support: Supports the DeepSeek model
⚠️ Warning: This library is a pre-release version. It is not yet recommended for production use. See the latest stable version for details.
📦 What is the @promptbook/markitdown Package?
- Promptbook is split into multiple packages, all managed in one monorepo.
@promptbook/markitdownis one of them — a package that integrates Microsoft's Markitdown into Promptbook.
Installation
# Install the full Promptbook ecosystem
npm i ptbk
# Install only this package (to save space)
npm install @promptbook/markitdown
🤍 The Book Abstract
"We have already experienced multiple generations of computer languages. And now, we stand on the threshold of the next revolution!"
- The Book language proposes a new paradigm where software is written in human-understandable language — something both AI and humans can understand and execute.
- Just as advances in microprocessors and Moore's Law drove the evolution of languages, large language models (LLMs) will lead the innovation of tomorrow.
"This change will happen whether we are ready or not. Our goal is not merely to be 'good' but to be 'excellent'."
🚀 Getting Started
- Try Book immediately with a Hello World example!
💜 The Promptbook Project Ecosystem
Promptbook is an ecosystem made up of various projects and tools.
| Project | Description |
|---|---|
| Book language | A human-friendly markup language for AI apps |
| Promptbook Engine | The engine that runs apps written in the Book language |
| Promptbook Studio | Web-based editor and runner (experimental MVP stage) |
- VSCode Plugin: Supports
.bookfiles - Community: Discord, Landing page, Github Discussions, LinkedIn, Facebook
- Sub-brands:
- Promptujeme (for Czech clients)
- Promptbook.city (for image/graphic generation)
💙 What is the Book Language?
The Book language is a language that lets you write AI apps, agents, workflows, automation, knowledge bases, translators, sheet processors, email automation, and more — without needing to know the complex internals of AI models — as easily as speaking naturally.
📄 Example
# 🌟 My first Book
- BOOK VERSION 1.0.0
- URL https://promptbook.studio/hello.book
- INPUT PARAMETER {topic}
- OUTPUT PARAMETER {article}
# Write an article
- PERSONA Jane, marketing expert (experienced in writing tech and AI articles)
- KNOWLEDGE https://wikipedia.org/
- KNOWLEDGE ./journalist-ethics.pdf
- EXPECT MIN 1 sentence
- EXPECT MAX 5 pages
> Write an article about {topic}
> -> {article}
The 3 Core Principles of the Book Language
1. What: Workflows, Tasks, Parameters
- What will be done? Each Book defines a workflow (scenario/pipeline). Example: a workflow that receives a topic and generates an article.
2. Who: Personas
- Who will do it?
Each task is performed by a persona (a virtual employee).
You can describe them in natural language, like "Jane, a creative writer with a sharp sense of humor."
3. How: Knowledge, Tools, Actions
- How will it be done? You can specify the knowledge (documents, websites, etc.), tools (real-time data, calculations, etc.), and actions (sending emails, creating files, etc.) the persona will use.
📌 General Principles of the Book Language
- Markdown-based (easy to read and write)
- Uses the .book extension and UTF-8 encoding
- Two forms: Flat (unstructured) / Full (structured)
- Compatible with version control (e.g., git)
- Understandable even to non-experts
🔒 Security
- To report security vulnerabilities, see the Security Policy
📦 Packages for Developers
Promptbook is divided into packages for various purposes.
- ptbk: Bundle of all packages (when size doesn't matter)
- @promptbook/wizzard: Easily run Books from Node
- @promptbook/core: Core logic
- @promptbook/node / browser: Core for each environment
- @promptbook/utils: Utilities for preprocessing/postprocessing LLM input and output
- @promptbook/markdown-utils: Markdown processing utilities
- @promptbook/openai / anthropic-claude / google / deepseek / azure-openai: Integrations for various AI APIs
- @promptbook/pdf / documents / legacy-documents / website-crawler: Support for various document types and web crawling
- @promptbook/cli: Command-line tool
- Docker image: Can be used as a server
⭐️ Recommended packages are marked with a star!
📚 Glossary
General LLM/AI Terms
- Prompt drift: The phenomenon where AI produces results that deviate from the original intent
- Pipeline/Workflow: A structure that executes multiple tasks in sequence
- Fine-tuning: Additional training of a pre-trained model on specific data
- Zero-shot/Few-shot learning: Learning with no examples or very few examples
- Meta-learning: The ability to quickly learn across a variety of tasks
- Retrieval-augmented generation: Finding information from a database and using it in generation
- Longtail: Rare or unusual data
💯 Core Concepts
- Pipeline collection, pipeline, task, persona, parameter, execution, expectation, postprocessing, word-based, separation of concerns, etc.
Advanced Concepts
- Knowledge (RAG), remote server, conditions (Jokers), meta-prompts, language types, automatic translation, images/audio/video, expectation-aware generation, JIT fine-tuning, anomaly detection, adversarial agent expectations, etc.
🚂 The Promptbook Engine
- This is the engine that actually runs apps written in the Book language.
- View the engine architecture diagram
➕➖ When Should You Use Promptbook?
➕ Great for these cases!
- When building complex outputs with LLMs (websites, articles, code, stories, etc.)
- When you want to separate code from prompts
- When you want to design complex prompt pipelines without writing code
- When you want to orchestrate multiple prompts
- When you need prompt reuse, version control, testing, and execution logs
"Instead of writing complex prompts directly in code, you can design and manage them in human-readable language."
➖ Not recommended for these cases
- When you only need a single simple prompt
- When OpenAI Assistant (GPTs) alone is sufficient
- When streaming is required (planned for future support)
- When you specifically need languages other than JavaScript/TypeScript (planned for future support)
- When the focus is non-text content like images, audio, or video (planned for future support)
- When recursion is required
🐜 Known Issues
- Iteration not supported
- Import not supported
🧼 Intentionally Unimplemented Features
- No recursion
- No types (all values are strings)
❔ Frequently Asked Questions (FAQ)
- How is this different from the OpenAI SDK / Anthropic Claude SDK?
- How is this different from OpenAI GPTs?
- How is this different from Langchain and DSPy?
- Is Promptbook a RAG (retrieval-augmented generation) system?
- Is function calling supported?
For questions, you can start a discussion, file an issue, or contact by email.
⌚ Changelog
- See CHANGELOG.md
📜 License
- BUSL 1.1 license (learn more)
🎯 Roadmap
- See TODO.md
🤝 Partners
🖋️ Contributing
- ⭐️ Star the project, follow on GitHub, or see the Contributing Guide
- Pull requests, feedback, and suggestions are all welcome!
📞 Support
💡 Summary Keywords
- Promptbook
- Book language
- Workflow/Pipeline
- Persona
- Knowledge/Tools/Actions
- Markdown-based
- AI applications
- Prompt management/reuse/version control
- LLM (Large Language Model)
- Open-source ecosystem
"Join us on this journey!" 🚀