Skip to main content
Mizen automatically extracts recipe information from URLs using a two-layer parsing approach that prioritizes speed and accuracy.

​
How It Works

Mizen uses a unified parsing system that tries multiple extraction methods:
1

Enter a recipe URL

Paste any recipe URL into the search bar. You can enter partial URLs like allrecipes.com/recipe/123 - Mizen automatically adds the protocol.
2

URL validation

Mizen performs a quick validation to ensure the URL contains recipe-related keywords before making a full parsing request.
3

Structured data extraction (JSON-LD)

Mizen first attempts to extract recipe data from JSON-LD structured data embedded in the page. This method is fast and doesn’t consume AI tokens.Most major recipe sites (AllRecipes, Serious Eats, Bon Appétit, etc.) include JSON-LD data following the schema.org Recipe format.
4

AI-powered fallback

If JSON-LD data is unavailable or incomplete, Mizen falls back to AI-based parsing. The system:
  • Fetches and cleans the HTML content
  • Uses AI vision models to extract recipe structure
  • Returns consistently formatted data

​
Extracted Data

Mizen extracts the following information when available:
  • Title - Recipe name
  • Ingredients - Organized by groups (e.g., “For the dough”, “For the filling”)
  • Instructions - Step-by-step cooking directions
  • Author - Recipe creator or source
  • Servings - Number of servings or yield
  • Timing - Prep time, cook time, and total time
  • Cuisine - Cuisine type tags (Italian, Thai, etc.)
  • Summary - AI-generated recipe overview
  • Storage - Storage instructions and shelf life
  • Plating - Serving suggestions, vessel recommendations, and temperature

​
API Endpoints

The parsing functionality is available at /home/daytona/workspace/source/src/app/api/parseRecipe/route.ts:49.

​
Request Format

{
  "url": "https://example.com/recipe"
}

​
Response Format

{
  "success": true,
  "title": "Recipe Title",
  "ingredients": [
    {
      "groupName": "Main",
      "ingredients": [
        {
          "amount": "1",
          "units": "cup",
          "ingredient": "flour"
        }
      ]
    }
  ],
  "instructions": ["Step 1", "Step 2"],
  "method": "json-ld"
}
The method field indicates which parsing method was used: json-ld for structured data extraction, or ai for AI-based parsing.

​
Error Handling

Mizen provides specific error codes for different failure scenarios:
  • ERR_INVALID_URL - Malformed or missing URL
  • ERR_NO_RECIPE_FOUND - No recipe content detected on the page
  • ERR_RATE_LIMIT - Too many requests (includes retry-after timestamp)
  • ERR_TIMEOUT - Request took too long
  • ERR_FETCH_FAILED - Could not connect to the recipe site
  • ERR_API_UNAVAILABLE - Parsing service temporarily unavailable

​
Supported Websites

Mizen works with any website that includes recipe content. Sites with JSON-LD structured data parse faster and don’t consume AI tokens:
  • AllRecipes
  • Serious Eats
  • Bon Appétit
  • Food Network
  • NYT Cooking
  • BBC Good Food
  • And thousands more
For sites without structured data, Mizen automatically uses AI parsing to extract the recipe. This works with blog posts, PDFs converted to HTML, and other non-standard formats.

​
Performance

  • JSON-LD parsing: ~1-2 seconds
  • AI parsing: ~3-5 seconds
  • Max file size: No limit (processes HTML only)
  • Timeout: 2 minutes per request

Build docs developers (and LLMs) love

Get started for freeTalk to us