JSPM

  • ESM via JSPM
  • ES Module Entrypoint
  • Export Map
  • Keywords
  • License
  • Repository URL
  • TypeScript Types
  • README
  • Created
  • Published
  • Downloads 8
  • Score
    100M100P100Q58037F
  • License MIT

CLI tool to execute AI prompts with flexible output formatting

Package Exports

    This package does not declare an exports field, so the exports above have been automatically detected and optimized by JSPM instead. If any package subpath is missing, it is recommended to post an issue to the original package (heyi) to support the "exports" field. If that is not possible, create a JSPM override to customize the exports field for this package.

    Readme

    heyi

    CLI tool to execute AI prompts with flexible output formatting

    Execute AI prompts directly from your terminal with support for multiple models and structured output formats using OpenRouter and the Vercel AI SDK.

    Install

    npm install heyi -g

    Usage

    CLI

    heyi prompt [prompt] [options]
heyi preset [file] [options]

    Options

    • -m, --model <model> - AI model to use (default: openai/gpt-4o-mini)
    • -f, --format <format> - Output format: string, number, object, array (default: string)
    • -s, --schema <schema> - Zod schema for object/array format (required when format is object or array)
    • -c, --crawler <crawler> - Crawler to use for fetching URLs: fetch, chrome (default: fetch)
    • --file <path> - Read content from file and include as context (can be used multiple times)
    • --url <url> - Fetch content from URL and include as context (can be used multiple times)
    • --var <key=value> - Define variables for replacement in prompt using {{key}} syntax (can be used multiple times)
    • -h, --help - Display help information
    • -V, --version - Display version number

    Environment Variables

    • HEYI_API_KEY - OpenRouter API key (required, can be set via environment or .env file)
    • HEYI_MODEL - Default AI model to use (optional, can be overridden with --model flag)
    • HEYI_CRAWLER - Default crawler to use for fetching URLs (optional, can be overridden with --crawler flag)

    Examples

    # Simple text prompt
heyi prompt "What is the capital of France?"

# Use a different model
heyi prompt "Explain quantum computing" --model google/gemini-2.0-flash-exp

# Get structured output as array of strings
heyi prompt "List 5 programming languages" --format array --schema "z.string()"

# Get structured output as array of objects
heyi prompt "List 3 countries with their capitals" --format array --schema "z.object({name:z.string(),capital:z.string()})"

# Get structured output as single object
heyi prompt "Analyze: revenue 100k, costs 60k" --format object --schema "z.object({revenue:z.number(),costs:z.number()})"

# Complex nested schema
heyi prompt "Analyze top 3 tech companies" --format array --schema "z.object({name:z.string(),founded:z.number(),products:z.array(z.string())})"

# Variable replacement in prompts
heyi prompt "Preset in {{language}}" --var language="German"
heyi prompt "Preset in {{input}} and output in {{output}}" --var input="German" --var output="English"

# Interactive variable prompting (prompts for undefined variables)
heyi prompt "Translate {{text}} to {{language}}"
# Will interactively prompt: text: [user enters value]
#                            language: [user enters value]

# Variable with description (shows custom prompt text)
heyi prompt "Explain {{topic description='What to explain'}} in simple terms"
# Will interactively prompt: What to explain (topic): [user enters value]

# Variable replacement with stdin
echo "Translate to {{language}}" | heyi prompt --var language="Spanish"

# Set default model via environment variable
HEYI_MODEL=perplexity/sonar heyi prompt "Explain AI"

# Set API key via environment variable
HEYI_API_KEY=your-key heyi prompt "Hello, AI!"

# Input from file as context
heyi prompt "Summarize this content" --file input.txt

# Input from multiple files as context
heyi prompt "Compare these files" --file file1.txt --file file2.txt
heyi prompt "Analyze all these documents" --file doc1.md --file doc2.md --file doc3.md

# Input from URL as context
heyi prompt "Summarize this article" --url https://example.com/article.html

# Input from multiple URLs as context
heyi prompt "Compare these articles" --url https://example.com/article1.html --url https://example.com/article2.html

# Use Chrome crawler for JavaScript-heavy pages
heyi prompt "Summarize this SPA" --url https://example.com/spa --crawler chrome
HEYI_CRAWLER=chrome heyi prompt "Get content from dynamic page" --url https://example.com/dynamic

# Mix files and URLs as context
heyi prompt "Compare local and remote content" --file local.txt --url https://example.com/remote.txt

# Input from stdin
cat article.md | heyi prompt "Extract all URLs mentioned"
echo "Analyze this text" | heyi prompt

# Preset files
heyi preset file.json
heyi preset file.json --var language=german
heyi preset file.json --model openai/gpt-4o
heyi preset file.json --file additional.txt --url https://example.com

    Preset Files

    Preset files allow you to define reusable configurations with prompts, models, files, and URLs. Create a JSON file with the following structure:

    {
  "prompt": "Your prompt with {{variables}}",
  "model": "openai/gpt-4o-mini",
  "format": "array",
  "schema": "z.string()",
  "crawler": "fetch",
  "files": ["path/to/file1.txt", "path/to/file2.txt"],
  "urls": ["https://example.com/page.html"]
}

    Preset Configuration

    • prompt: The AI prompt to execute. Supports variable replacement using {{variable}} syntax.
    • model (optional): AI model to use (e.g., openai/gpt-4o-mini, google/gemini-2.0-flash-exp).
    • format (optional): Output format: string, number, object, array (default: string).
    • schema (optional): Zod schema for object/array format (required when format is object or array).
    • crawler (optional): Crawler to use for fetching URLs: fetch, chrome (default: fetch).
    • files (optional): Array of file paths to include as context.
    • urls (optional): Array of URLs to fetch and include as context.

    Preset Examples

    Basic preset with variables:

    {
  "prompt": "Explain {{topic}} in {{language}}"
}
    heyi preset explain.json --var topic="quantum computing" --var language="simple terms"

    Preset with files and URLs:

    {
  "prompt": "Analyze and compare the following documents",
  "model": "google/gemini-2.0-flash-exp",
  "files": ["report1.txt", "report2.txt"],
  "urls": ["https://example.com/data.html"]
}
    heyi preset analyze.json

    Preset with structured output:

    {
  "prompt": "List programming languages mentioned in these files",
  "format": "array",
  "schema": "z.string()",
  "files": ["code1.js", "code2.py"]
}
    heyi preset languages.json

    CLI Override Behavior

    • Model override: Using --model flag overrides the model specified in the preset file.
    • Format override: Using --format flag overrides the format specified in the preset file.
    • Schema override: Using --schema flag overrides the schema specified in the preset file.
    • Crawler override: Using --crawler flag overrides the crawler specified in the preset file.
    • Files and URLs append: Using --file or --url flags adds additional context to the preset's files and URLs.
    • Variables: Use --var to replace variables in the preset's prompt.
    # Override model from preset
heyi preset file.json --model openai/gpt-4o

# Override format from preset
heyi preset file.json --format object --schema "z.object({name:z.string()})"

# Override crawler from preset
heyi preset file.json --crawler chrome

# Add additional files to preset's files
heyi preset file.json --file extra.txt

# Replace variables in preset prompt
heyi preset file.json --var name="Alice" --var role="developer"

    Output Formats

    • string (default): Plain text response from the AI model
    • number: Numeric response from the AI model
    • object: Single JSON object with structured data (requires --schema flag)
    • array: JSON array with structured data (requires --schema flag)

    The tool uses Zod schemas to ensure the AI model returns data in the requested format. When using object or array formats, you must provide a Zod schema string via the --schema flag.

    Schema Examples

    • String array: --format array --schema "z.string()"
    • URL array: --format array --schema "z.url()" (not supported by all models)
    • Object array: --format array --schema "z.object({name:z.string(),age:z.number()})"
    • Single object: --format object --schema "z.object({total:z.number(),items:z.array(z.string())})"

    Variables

    The tool supports variable replacement in prompts using {{variable}} syntax. Variables can be provided via the --var flag or through interactive prompting.

    Variable Syntax

    Basic variable:

    {{variableName}}

    Variable with description (for interactive prompting):

    {{variableName description="Description shown to user"}}

    Variable Behavior

    1. Provided via --var flag: Variables are directly replaced with the provided values
    2. Not provided: The tool will interactively prompt the user to enter the value
    3. With description: When prompting, the description is shown to help the user understand what to enter

    Variable Examples

    # Provide variables via flag
heyi prompt "Translate {{text}} to {{language}}" --var text="Hello" --var language="Spanish"

# Interactive prompting for undefined variables
heyi prompt "Translate {{text}} to {{language}}"
# Prompts:
#   text: [user enters value]
#   language: [user enters value]

# Mix provided and interactive variables
heyi prompt "Translate {{text}} to {{language}}" --var language="French"
# Only prompts for 'text' since 'language' is provided

# Use descriptions for better user experience
heyi prompt "Explain {{topic description='Enter a topic to explain'}} in simple terms"
# Prompts:
#   Enter a topic to explain (topic): [user enters value]

# Variables work in preset files too
heyi preset translate.json
# Prompts for any undefined variables in the preset's prompt

    Crawlers

    The tool supports two crawlers for fetching content from URLs:

    • fetch (default): Uses the native fetch API to retrieve HTML content. Fast and lightweight, but may not work well with JavaScript-heavy or dynamically rendered pages.
    • chrome: Uses Puppeteer to launch a headless Chrome browser and retrieve content after the page has fully loaded. Ideal for single-page applications (SPAs) and JavaScript-heavy websites, but slower and requires more resources.

    When to Use Chrome Crawler

    Use the chrome crawler when:

    • The target website relies heavily on JavaScript for rendering content
    • Content is loaded dynamically after the initial page load
    • You need to interact with a single-page application (SPA)
    • The fetch crawler returns incomplete or missing content

    Crawler Examples

    # Use default fetch crawler
heyi prompt "Summarize this page" --url https://example.com

# Use Chrome crawler for JS-heavy page
heyi prompt "Extract data from SPA" --url https://app.example.com --crawler chrome

# Set Chrome as default crawler via environment
HEYI_CRAWLER=chrome heyi prompt "Get content" --url https://dynamic-site.com

    Development

    # Install dependencies
npm install

# Run tests
npm test

# Lint and format code
npm run format

# Run the CLI in development
npm start -- prompt "Your prompt here"

# Or run directly
./bin/index.js prompt "Your prompt here"