Markdown / LLM-ready output #

Turn any page into clean Markdown instead of a wall of HTML — the feature Firecrawl and Crawl4AI are known for, now in the Laravel world. The result drops straight into an LLM context window or a RAG index with no HTML noise.

There are two ways to use it, and they pair naturally: extract an article body as a Markdown field, then export the whole crawl as a Markdown document.

1 · The markdown field type #

Render one element's content (an article body, a product description — any long-form block) as Markdown by setting the field's type to "markdown". The css selector locates the element; its content is converted:

{
  "name": "description",
  "css": ".product-description",
  "type": "markdown"
}

What the converter preserves and strips:

Use an empty css ("") in detail_fields[] to convert the whole detail page context, or point it at the main content container for article-only output.

Note: regex is not applied to markdown fields — it would corrupt multi-line output.

2 · The markdown output format #

Export a whole crawl as a single Markdown document, one section per item. Set the blueprint's output_config.format:

"output_config": {
  "format": "markdown"
}

php artisan datahelm:scrap:run example --output=storage/app/scrapes/example.md

Each item becomes a section:

• Heading — the most title-like field (title, name, heading, headline, label), falling back to Item N.
• Body — the first long-form field (markdown, content, body, description, text), rendered as-is.
• Metadata — every remaining scalar field as a bullet list (- **price:** 150000).
• Items are separated by --- horizontal rules.

Now the robot generates a .md file containing the scraped page's content, formatted as clean Markdown instead of raw JSON.

When a robot's blueprint has output_config.format = "markdown" (set via --output-format=markdown at generate time), running the robot produces a single Markdown document per crawl:

• A ## heading (from the page's title field)
• The main body text as readable prose
• Any remaining scraped fields listed as - **field:** value bullets below

This is meant to be dropped directly into an LLM's context window or a RAG index — no HTML tags, scripts, or site chrome to strip out first, unlike the default json format, which just dumps each field as a raw key/value pair.

Example output for a scraped page:

Next: Presets & item pipeline →