ai-tools

name	ai-tools
description	AI/ML utilities including embeddings generation, vision analysis, text-to-speech, and structured output extraction. Supports Anthropic Claude, OpenAI, and compatible APIs.

AI Tools

Modern AI/ML capabilities for embeddings, vision, speech, and structured data extraction.

Prerequisites

Requires one of these environment variables:

• ANTHROPIC_API_KEY - For Claude models
• OPENAI_API_KEY - For OpenAI models
• AI_GATEWAY_API_KEY + AI_GATEWAY_BASE_URL - For Cloudflare AI Gateway

Quick Start

Generate Embeddings

node /path/to/skills/ai-tools/scripts/embeddings.js "Your text to embed"

Analyze Image

node /path/to/skills/ai-tools/scripts/vision.js /path/to/image.png "Describe this image"

Extract Structured Data

node /path/to/skills/ai-tools/scripts/extract.js "John Doe, 30 years old, lives in NYC" --schema '{"name":"string","age":"number","city":"string"}'

Scripts

embeddings.js

Generate vector embeddings for semantic search and RAG.

Usage:

node embeddings.js <text> [OPTIONS]

Options:

• --model <model> - Embedding model (default: text-embedding-3-small)
• --dimensions <n> - Output dimensions (default: 1536)
• --output <file> - Save embeddings to file

vision.js

Analyze images using multimodal AI models.

Usage:

node vision.js <image> <prompt> [OPTIONS]

Arguments:

• <image> - Path to image file or URL
• <prompt> - Question or instruction about the image

Options:

• --model <model> - Vision model (default: claude-3-5-sonnet-20241022)
• --detail <level> - Image detail: auto, low, high (default: auto)

extract.js

Extract structured data from unstructured text.

Usage:

node extract.js <text> --schema <json_schema>

Options:

• --schema <json> - JSON schema defining expected output structure
• --model <model> - Model to use (default: claude-3-5-sonnet-20241022)

summarize.js

Intelligent text summarization with configurable length.

Usage:

node summarize.js <text|file> [OPTIONS]

Options:

• --length <words> - Target summary length (default: 100)
• --style <style> - Style: brief, detailed, bullets (default: brief)
• --file - Treat input as file path

sentiment.js

Analyze sentiment and emotional tone.

Usage:

node sentiment.js <text>

Examples

Semantic Search Embeddings

node embeddings.js "Machine learning is transforming healthcare" --output embedding.json

Image Description

node vision.js screenshot.png "What UI elements are visible? List any errors."

Extract Contact Info

node extract.js "Contact Sarah at [e-mail] or call 555-0123" \
  --schema '{"name":"string","email":"string","phone":"string"}'

Summarize Article

node summarize.js article.txt --file --length 50 --style bullets

Output Formats

embeddings.js

{
  "model": "text-embedding-3-small",
  "dimensions": 1536,
  "embedding": [0.123, -0.456, ...],
  "usage": { "tokens": 8 }
}

vision.js

{
  "model": "claude-3-5-sonnet-20241022",
  "analysis": "The image shows...",
  "usage": { "input_tokens": 1200, "output_tokens": 150 }
}

extract.js

{
  "extracted": { "name": "John", "age": 30, "city": "NYC" },
  "confidence": 0.95
}

Supported Models

Embeddings

• text-embedding-3-small (OpenAI)
• text-embedding-3-large (OpenAI)
• @cf/baai/bge-base-en-v1.5 (Cloudflare)

Vision

• claude-3-5-sonnet-20241022 (Anthropic)
• claude-3-opus-20240229 (Anthropic)
• gpt-4o (OpenAI)
• gpt-4o-mini (OpenAI)

Text

• claude-3-5-sonnet-20241022 (Anthropic)
• claude-3-5-haiku-20241022 (Anthropic)
• gpt-4o (OpenAI)
• gpt-4o-mini (OpenAI)