Skip to main content
cloro API responses include multiple output formats. This guide covers how to use each one effectively — rendering HTML as proof of the scraped page, and handling special objects embedded in text and markdown fields.

​
HTML: proof of scraped page

When you request HTML from any endpoint using include.html: true, the response includes a URL to the full scraped page stored on our CDN. This URL expires after 24 hours. This raw HTML is the actual page the AI provider displayed. You can present it to your clients as proof of what was scraped.

​
Rendering HTML cleanly

The raw HTML includes browser chrome — headers, footers, sidebars, cookie banners, and scripts. To render a clean version, use the @cloro-dev/response-parser npm package. The library:
  • Auto-detects the AI provider from the HTML
  • Strips scripts, cookie banners, and other chrome
  • Returns sanitized HTML you can render in any framework (React, Vue, Svelte, vanilla JS)
See the response-parser README for installation, API reference, framework examples, and supported options.

​
Workflow

  1. Make an API request with include.html: true
  2. Fetch the HTML from the returned CDN URL
  3. Parse it with @cloro-dev/response-parser to get clean, sanitized HTML
  4. Render the result in your application
The HTML URL expires after 24 hours. Download and store the HTML if you need long-term access. There is no additional credit cost for including HTML.

​
Text and markdown: LLM analysis

For text analysis or LLM pipelines, use result.text or result.markdown (with include.markdown: true) from the API response directly — you don’t need the response-parser for this. However, be aware that special objects are embedded as text within these fields. When the AI provider displays shopping cards, places, ads, inline products, entities, or map entries, their content appears as text in result.text and result.markdown. Sources and footers are not included in these fields.

​
Handling special objects

The structured response fields (result.shoppingCards, result.places, result.ads, result.inlineProducts, result.entities, result.mapEntries) contain the same data in a clean, structured format. For LLM analysis:
  1. Use result.text or result.markdown as your primary text input
  2. Parse out special object content from the text using the patterns from the structured fields — you already have the exact data in those fields, so you know what to look for and remove
  3. Use the structured fields directly when you need object-specific data (prices, ratings, URLs, etc.)
If you pass result.text or result.markdown directly to an LLM without removing special object content, the model may treat product cards, ads, or place listings as part of the AI’s natural language response.

​
Common questions

​
Why do I see errors when opening the HTML file in my browser?

The scraped HTML may include JavaScript from the original AI provider page. When opened directly in a browser, this JavaScript executes and can cause errors, broken layouts, or unexpected behavior. To view the HTML correctly, disable JavaScript in your browser before opening the file. Alternatively, use the @cloro-dev/response-parser library which automatically strips all scripts during parsing.