Promptbook

Supercharge your use of large language models

🤍 The Promptbook Whitepaper

If you have a simple, single prompt for ChatGPT, GPT-4, Anthropic Claude, Google Gemini, Llama 2, or whatever, it doesn't matter how you integrate it. Whether it's calling a REST API directly, using the SDK, hardcoding the prompt into the source code, or importing a text file, the process remains the same.

But often you will struggle with the limitations of LLMs, such as hallucinations, off-topic responses, poor quality output, language drift, word repetition repetition repetition repetition or misuse, lack of context, or just plain w𝒆𝐢rd responses. When this happens, you generally have three options:

1. Fine-tune the model to your specifications or even train your own.
2. Prompt-engineer the prompt to the best shape you can achieve.
3. Use multiple prompts in a pipeline to get the best result.

In all of these situations, but especially in 3., the Promptbook library can make your life easier.

• Separates concerns between prompt-engineer and programmer, between code files and prompt files, and between prompts and their execution logic.
• Establishes a common format .ptbk.md that can be used to describe your prompt business logic without having to write code or deal with the technicalities of LLMs.
• Forget about low-level details like choosing the right model, tokens, context size, temperature, top-k, top-p, or kernel sampling. Just write your intent and persona who should be responsible for the task and let the library do the rest.
• Has built-in orchestration of pipeline execution and many tools to make the process easier, more reliable, and more efficient, such as caching, compilation+preparation, just-in-time fine-tuning, expectation-aware generation, agent adversary expectations, and more.
• Sometimes even the best prompts with the best framework like Promptbook :) can't avoid the problems. In this case, the library has built-in anomaly detection and logging to help you find and fix the problems.
• Promptbook has built in versioning. You can test multiple A/B versions of pipelines and see which one works best.
• Promptbook is designed to do RAG (Retrieval-Augmented Generation) and other advanced techniques. You can use knowledge to improve the quality of the output.

🧔 Promptbook (for prompt-engeneers)

Prompt book markdown file (or .ptbk.md file) is document that describes a pipeline - a series of prompts that are chained together to form somewhat reciepe for transforming natural language input.

• Multiple pipelines forms a collection which will handle core know-how of your LLM application.
• Theese pipelines are designed such as they can be written by non-programmers.

Sample:

File write-website-content.ptbk.md:

🌍 Create website content

Instructions for creating web page content.

• PIPELINE URL https://promptbook.studio/webgpt/write-website-content.ptbk.md
• PROMPTBOOK VERSION 0.0.1
• INPUT  PARAM {rawTitle} Automatically suggested a site name or empty text
• INPUT  PARAM {rawAssigment} Automatically generated site entry from image recognition
• OUTPUT PARAM {websiteContent} Web content
• OUTPUT PARAM {keywords} Keywords

👤 Specifying the assigment

What is your web about?

• PROMPT DIALOG

{rawAssigment}

-> {assigment} Website assignment and specification

✨ Improving the title

• PERSONA Jane, Copywriter and Marketing Specialist.

As an experienced marketing specialist, you have been entrusted with improving the name of your client's business.

A suggested name from a client:
"{rawTitle}"

Assignment from customer:

> {assigment}

## Instructions:

-   Write only one name suggestion
-   The name will be used on the website, business cards, visuals, etc.

-> {enhancedTitle} Enhanced title

👤 Website title approval

Is the title for your website okay?

• PROMPT DIALOG

{enhancedTitle}

-> {title} Title for the website

🐰 Cunning subtitle

• PERSONA Josh, a copywriter, tasked with creating a claim for the website.

As an experienced copywriter, you have been entrusted with creating a claim for the "{title}" web page.

A website assignment from a customer:

> {assigment}

## Instructions:

-   Write only one name suggestion
-   Claim will be used on website, business cards, visuals, etc.
-   Claim should be punchy, funny, original

-> {claim} Claim for the web

🚦 Keyword analysis

• PERSONA Paul, extremely creative SEO specialist.

As an experienced SEO specialist, you have been entrusted with creating keywords for the website "{title}".

Website assignment from the customer:

> {assigment}

## Instructions:

-   Write a list of keywords
-   Keywords are in basic form

## Example:

-   Ice cream
-   Olomouc
-   Quality
-   Family
-   Tradition
-   Italy
-   Craft

-> {keywords} Keywords

🔗 Combine the beginning

• SIMPLE TEMPLATE

# {title}

> {claim}

-> {contentBeginning} Beginning of web content

🖋 Write the content

• PERSONA Jane

As an experienced copywriter and web designer, you have been entrusted with creating text for a new website {title}.

A website assignment from a customer:

> {assigment}

## Instructions:

-   Text formatting is in Markdown
-   Be concise and to the point
-   Use keywords, but they should be naturally in the text
-   This is the complete content of the page, so don't forget all the important information and elements the page should contain
-   Use headings, bullets, text formatting

## Keywords:

{keywords}

## Web Content:

{contentBeginning}

-> {contentBody} Middle of the web content

🔗 Combine the content

• SIMPLE TEMPLATE

{contentBeginning}

{contentBody}

-> {websiteContent}

Following is the scheme how the promptbook above is executed:

%% 🔮 Tip: Open this on GitHub or in the VSCode website to see the Mermaid graph visually

flowchart LR
  subgraph "🌍 Create website content"

      direction TB

      input((Input)):::input
      templateSpecifyingTheAssigment(👤 Specifying the assigment)
      input--"{rawAssigment}"-->templateSpecifyingTheAssigment
      templateImprovingTheTitle(✨ Improving the title)
      input--"{rawTitle}"-->templateImprovingTheTitle
      templateSpecifyingTheAssigment--"{assigment}"-->templateImprovingTheTitle
      templateWebsiteTitleApproval(👤 Website title approval)
      templateImprovingTheTitle--"{enhancedTitle}"-->templateWebsiteTitleApproval
      templateCunningSubtitle(🐰 Cunning subtitle)
      templateWebsiteTitleApproval--"{title}"-->templateCunningSubtitle
      templateSpecifyingTheAssigment--"{assigment}"-->templateCunningSubtitle
      templateKeywordAnalysis(🚦 Keyword analysis)
      templateWebsiteTitleApproval--"{title}"-->templateKeywordAnalysis
      templateSpecifyingTheAssigment--"{assigment}"-->templateKeywordAnalysis
      templateCombineTheBeginning(🔗 Combine the beginning)
      templateWebsiteTitleApproval--"{title}"-->templateCombineTheBeginning
      templateCunningSubtitle--"{claim}"-->templateCombineTheBeginning
      templateWriteTheContent(🖋 Write the content)
      templateWebsiteTitleApproval--"{title}"-->templateWriteTheContent
      templateSpecifyingTheAssigment--"{assigment}"-->templateWriteTheContent
      templateKeywordAnalysis--"{keywords}"-->templateWriteTheContent
      templateCombineTheBeginning--"{contentBeginning}"-->templateWriteTheContent
      templateCombineTheContent(🔗 Combine the content)
      templateCombineTheBeginning--"{contentBeginning}"-->templateCombineTheContent
      templateWriteTheContent--"{contentBody}"-->templateCombineTheContent

      templateCombineTheContent--"{websiteContent}"-->output
      output((Output)):::output

      classDef input color: grey;
      classDef output color: grey;

  end;

Loading

• More template samples
• Read more about .ptbk.md file format here

Note: We are using postprocessing functions like unwrapResult that can be used to postprocess the result.

📦 Packages

This library is divided into several packages, all are published from single monorepo. You can install all of them at once:

npm i ptbk

Or you can install them separately:

⭐ Marked packages are worth to try first

• ⭐ ptbk - Bundle of all packages, when you want to install everything and you don't care about the size
• promptbook - Same as ptbk
• @promptbook/core - Core of the library, it contains the main logic for promptbooks
• @promptbook/node - Core of the library for Node.js environment
• @promptbook/browser - Core of the library for browser environment
• ⭐ @promptbook/utils - Utility functions used in the library but also useful for individual use in preprocessing and postprocessing LLM inputs and outputs
• @promptbook/markdown-utils - Utility functions used for processing markdown
• (Not finished) @promptbook/wizzard - Wizard for creating+running promptbooks in single line
• @promptbook/execute-javascript - Execution tools for javascript inside promptbooks
• @promptbook/openai - Execution tools for OpenAI API, wrapper around OpenAI SDK
• @promptbook/anthropic-claude - Execution tools for Anthropic Claude API, wrapper around Anthropic Claude SDK
• @promptbook/azure-openai - Execution tools for Azure OpenAI API
• @promptbook/langtail - Execution tools for Langtail API, wrapper around Langtail SDK
• @promptbook/fake-llm - Mocked execution tools for testing the library and saving the tokens
• @promptbook/remote-client - Remote client for remote execution of promptbooks
• @promptbook/remote-server - Remote server for remote execution of promptbooks
• @promptbook/types - Just typescript types used in the library
• @promptbook/cli - Command line interface utilities for promptbooks

📚 Dictionary

The following glossary is used to clarify certain concepts:

Core concepts

• 📚 Collection of pipelines
• 📯 Pipeline
• 🎺 Pipeline templates
• 🤼 Personas
• ⭕ Parameters
• 🚀 Pipeline execution
• 🧪 Expectations
• ✂️ Postprocessing
• 🔣 Words not tokens
• ☯ Separation of concerns

Advanced concepts

• 📚 Knowledge (Retrieval-augmented generation)
• 🌏 Remote server
• 🃏 Jokers (conditions)
• 🔳 Metaprompting
• 🌏 Linguistically typed languages
• 🌍 Auto-Translations
• 📽 Images, audio, video, spreadsheets
• 🔙 Expectation-aware generation
• ⏳ Just-in-time fine-tuning
• 🔴 Anomaly detection
• 👮 Agent adversary expectations
• view more

🔌 Usage in Typescript / Javascript

• Simple usage
• Usage with client and remote server

➕➖ When to use Promptbook?

➕ When to use

• When you are writing app that generates complex things via LLM - like websites, articles, presentations, code, stories, songs,...
• When you want to separate code from text prompts
• When you want to describe complex prompt pipelines and don't want to do it in the code
• When you want to orchestrate multiple prompts together
• When you want to reuse parts of prompts in multiple places
• When you want to version your prompts and test multiple versions
• When you want to log the execution of prompts and backtrace the issues

➖ When not to use

• When you have already implemented single simple prompt and it works fine for your job
• When OpenAI Assistant (GPTs) is enough for you
• When you need streaming (this may be implemented in the future, see discussion).
• When you need to use something other than JavaScript or TypeScript (other languages are on the way, see the discussion)
• When your main focus is on something other than text - like images, audio, video, spreadsheets (other media types may be added in the future, see discussion)
• When you need to use recursion (see the discussion)

🐜 Known issues

• 🤸‍♂️ Iterations not working yet
• ⤵️ Imports not working yet

🧼 Intentionally not implemented features

• ➿ No recursion
• 🏳 There are no types, just strings

❔ FAQ

If you have a question start a discussion, open an issue or write me an email.

Why not just use the OpenAI SDK / Anthropic Claude SDK / ...?

Different levels of abstraction. OpenAI library is for direct use of OpenAI API. This library is for a higher level of abstraction. It is for creating prompt templates and promptbooks that are independent of the underlying library, LLM model, or even LLM provider.

How is it different from the Langchain library?

Langchain is primarily aimed at ML developers working in Python. This library is for developers working in javascript/typescript and creating applications for end users.

We are considering creating a bridge/converter between these two libraries.

Promptbooks vs. OpenAI`s GPTs

GPTs are chat assistants that can be assigned to specific tasks and materials. But they are still chat assistants. Promptbooks are a way to orchestrate many more predefined tasks to have much tighter control over the process. Promptbooks are not a good technology for creating human-like chatbots, GPTs are not a good technology for creating outputs with specific requirements.