Prompt-keeper NPM

prompt-keeper 🚀

A CLI tool for managing AI prompts linked to code files, making your AI-assisted development traceable and maintainable.

Why prompt-keeper? 🤔

📝 Track Your AI Conversations: Keep a history of all AI prompts used in your codebase
🔍 Code Traceability: Easily find which AI prompts generated specific code sections
🤖 Multi-Model Support: Works with popular AI models (Claude, GPT-4, Gemini, etc.)
🎯 Developer Friendly: Simple CLI interface with interactive mode

Quick Demo

terminal

Quick Start 🚀

# Install globally
npm install -g prompt-keeper

# Start using with interactive mode
pk i your-file.ts

View full documentation

Features

Manage prompt history for each code file
Version control for prompts
AI model specification support
Link prompts to code files with comments
Flexible comment insertion position
Interactive prompt input support

Installation

npm install -g prompt-keeper

Usage

The tool provides two command aliases for easier use:

pk: Short for "prompt-keeper"
prompt-keeper: Full command name

Both commands work exactly the same way. Use whichever you prefer.

Interactive Mode

pk i <file-path>
# or
pk interactive <file-path>
# or
prompt-keeper interactive <file-path>

The interactive mode guides you through the following steps:

Enter your prompt
Select an AI model from the available options:
- Claude 3.5 Sonnet v2
- O3-mini
- O3-mini High
- Gemini 2.0 Flash
- Gemini 1.5 Pro
- Gemini 2.0 Experimental Advanced
- GPT-4o
- GPT-4o Mini
- Custom
Choose where to add the prompt ID comment:
- Beginning of file
- End of file
- Specific line number

Command Line Mode

pk s <file-path> "your prompt" --model <model-name>
# or
pk set <file-path> "your prompt" --model <model-name>
# or
prompt-keeper set <file-path> "your prompt" --model <model-name>

Options

-m, --model <model>: Specify AI model (e.g., claude-3.5-sonnet-v2)

Generated Files

The tool generates a prompt file in the same directory as your code file with the format {filename}.prompt.json.

Example: For MyComponent.tsx → MyComponent.prompt.json

{
  "file": "test/MyComponent.tsx",
  "prompts": [
    {
      "id": "8ffd78c0-d00c-433d-a8d5-af111cc79020",
      "prompt": "This is a test prompt, and it's a library that can manage what kind of prompts were used to implement code in the past. This is a sample saved prompt.",
      "model": "o3-mini",
      "createdAt": "2025-02-15T02:00:22.206Z",
      "version": 1
    },
    {
      "id": "b19e27f8-9b25-4c2f-a011-da3898206f82",
      "prompt": "Implement memoization for this component",
      "model": "o3-mini-high",
      "createdAt": "2025-02-15T12:10:18.760Z",
      "version": 2
    }
  ]
}

It also adds a prompt ID comment to your code file:

import React, { memo } from "react";

// prompt-id: 8ffd78c0-d00c-433d-a8d5-af111cc79020
const MyComponent: React.FC = () => {
  return (
    <div>
      <h1>Test Component</h1>
    </div>
  );
};

// prompt-id: b19e27f8-9b25-4c2f-a011-da3898206f82
export default memo(MyComponent);

Development

Local Setup

# Clone the repository
git clone https://github.com/kazuya-sakamoto/prompt-keeper.git
cd prompt-keeper

# Install dependencies
npm install

# Link package globally (Option 1)
npm link

# Now you can use the pk command directly
pk i <file-path>

# Or use npm script without linking (Option 2)
npm run pk -- i <file-path>