Clean Web HTML to LLM-Ready Markdown Converter

An ultra-fast, deterministic Apify Actor that strips away boilerplate, navigational noise, tracking scripts, and styling from messy HTML documents, outputting clean, token-efficient GitHub-flavored Markdown specifically formatted for Large Language Models (LLMs), RAG pipelines, and AI agent workflows.

---

🚀 Why Use This Actor?

Feeding raw web pages or messy HTML directly into LLMs wastes up to 70% of your prompt tokens on boilerplate headers, footers, stylesheets, SVG icons, and ads. This Actor provides a zero-latency, deterministic transformation pipeline that:

• Reduces LLM Token Usage: Cuts token counts drastically, saving significant API costs on OpenAI, Anthropic, Gemini, and local LLMs.
• Zero Hallucination / 100% Deterministic: Uses strict structural DOM parsing (BeautifulSoup4 + markdownify) rather than probabilistic AI models.
• Preserves Critical Content: Retains semantic headings, lists, inline formatting, code snippets, and Markdown tables.
• Extracts Key Meta Fields: Pulls document titles, SEO meta descriptions, and heading hierarchies into structured JSON attributes alongside the markdown body.
• Ideal for AI Automations: Seamlessly plugs into n8n, Make.com, LangChain, LlamaIndex, and Apify MCP servers.

---

📥 Input Schema

Field	Type	Default	Description
documents	Array of Objects	Sample Demo Document	Array of objects with source_id and raw html strings.
strip_links	Boolean	false	When true, strips hyperlinks while preserving anchor text to minimize token overhead.
heading_style	String	"ATX"	"ATX" uses # Heading, while "UNDERLINE" uses === and ---.
extract_tables	Boolean	true	Formats HTML tables into standard Markdown grid tables.

Sample Input JSON

{
  "documents": [
    {
      "source_id": "article-101",
      "html": "<html><head><title>AI Trends</title></head><body><header>Nav</header><main><h1>Top Trends</h1><p>Autonomous agents are scaling rapidly.</p></main><footer>Footer</footer></body></html>"
    }
  ],
  "strip_links": false,
  "heading_style": "ATX",
  "extract_tables": true
}

---

📤 Output Dataset Format

Each processed document yields a structured record in the default dataset:

{
  "source_id": "article-101",
  "title": "AI Trends",
  "meta_description": "",
  "markdown": "# Top Trends\n\nAutonomous agents are scaling rapidly.",
  "character_count": 52,
  "estimated_tokens": 13,
  "headings": [
    "H1: Top Trends"
  ],
  "status": "success"
}

---

💰 Pay-Per-Event (PPE) Pricing

This Actor utilizes Apify's modern Pay-Per-Event billing model:

• Event Name: item-processed
• Cost: Only $0.001 per document ($1.00 per 1,000 documents).
• No Monthly Minimums: Pay only for the documents you actually convert. Compute overhead is practically zero, providing the highest ROI for production ingestion pipelines.

---

🔌 API & Python Client Example

from apify_client import ApifyClient

client = ApifyClient("YOUR_APIFY_TOKEN")

run_input = {
    "documents": [
        {"source_id": "doc1", "html": "<h1>Hello World</h1><p>Content to clean.</p>"}
    ]
}

run = client.actor("YOUR_USERNAME/html-to-llm-markdown").call(run_input=run_input)
for item in client.dataset(run["defaultDatasetId"]).iterate_items():
    print(item["markdown"])