# Extraction

Source: https://docs.evomi.com/scraping-products-instructions/scraper-api/parameters/extraction/

Reliable structured extraction is achieved through Css, Xpath, Regex, & Llms.

Css, Xpath, & Regex extraction is communicated though the `extract_scheme` or `scheme_id` parameter which is a extraction scheme where each object defines a data/area "bucket." These objects act as either a **Rule** (to pull data) or a **Container** (to scope data).

**Note:** `extract_scheme` can be defined with each scraper request or using a presaved configuration using `scheme_id`. View "Scheme" Tab.

## Defining an `extract_scheme`

1 - Select data/container with CSS/xpath/regex locators

2 - Nest to target extraction into subsections if needed

3 - Configure extract field types (content/attribute/exists/count)

## Extraction Parameters

| Parameter | Type | Description |
| :--- | :--- | :--- |
| **`label`** | String | The key name for the field in the resulting JSON. |
| **`type`** | Enum | `content` (text), `attribute` (links/src), `exists` (bool), `count` (int), or `nest`. |
| **`selector`** | CSS | A standard CSS selector to target the element. |
| **`xpath`** | XPath | A standard XPath expression for complex navigation. |
| **`regex`** | Pattern | A Regular Expression to extract text from the raw HTML/text. |
| **`attribute`** | String | **Required if type="attribute"**. Specify the target (e.g., `href`, `src`). |
| **`fields`** | Array | **Required if type="nest"**. A recursive list of extraction objects. |

---

Below is a standard configuration for extracting a list of blog posts from a page:

```json
{
  "extract_scheme": [
    {
      "label": "blog_posts",
      "type": "nest",
      "selector": "article.post-card",
      "fields": [
        {
          "label": "title",
          "type": "content",
          "selector": "h2"
        },
        {
          "label": "link",
          "type": "attribute",
          "selector": "a.main-link",
          "attribute": "href"
        },
        {
          "label": "author_via_xpath",
          "type": "content",
          "xpath": ".//span[contains(@class, 'author-name')]" 
        }
      ]
    }
  ]
}
```
Set `delivery=json` to receive results, View "Output Formats" Tab.

* **Hybrid Targeting:** You can **mix and match** locator types. For example, you can use a `selector` to find a parent container and then use `regex` or `xpath` to find specific data inside it.
* **Use `regex` for raw strings:** Best when data isn't wrapped in its own tag (e.g., pulling a price out of a string like "Price: $45.00 only today").
* **Use `type: "nest"` for lists:** Use this for repeated elements like product grids to keep your JSON structured.
* **Use `type: "exists"` for validation:** Perfect for checking if a "Sold Out" badge is present without needing the actual text.
---

## Request Mode with Filtering then Extracting (2 credits)
```json
{
  "url": "https://spa-website.com",
  "mode": "request",
  "excluded_tags": ["aside", "form", "nav"],
  "excluded_selectors": [".ads"],
  "proxy_type": "residential",
  "proxy_country": "US",
  "extract_scheme": [
    {
      "label": "main_content",
      "type": "nest",
      "selector": "div",
      "fields": [
        {
          "label": "headline",
          "type": "content",
          "xpath": "//h1"
        },
        {
          "label": "body_text",
          "type": "content",
          "selector": "p"
        },
        {
          "label": "external_link",
          "type": "attribute",
          "selector": "a",
          "attribute": "href"
        }
      ]
    }
  ]
}
```

## AI Extraction

| Parameter | Type | Default | Required When | Description |
|-----------|------|---------|---------------|-------------|
| `ai_enhance` | boolean | `false` | - | Enable AI processing |
| `ai_source` | string | - | ai_enhance=true | Source: `markdown` or `screenshot` |
| `ai_prompt` | string | - | No | Custom prompt for AI processing |
| `ai_force_json` | boolean | `true` | No | Force JSON output format |

**AI Model:** Google Gemini 2.0 Flash  
**Cost:** +30 credits (additive)

**Example:**
```json
{
  "ai_enhance": true,
  "ai_source": "markdown",
  "ai_prompt": "Extract product name, price, and availability as JSON"
}
```

See [AI Enhancement](/scraping-products-instructions/scraper-api/parameters/ai-enhancement/) for detailed examples.

## Pages in this section

- [Fix](https://docs.evomi.com/scraping-products-instructions/scraper-api/parameters/extraction/schema/fix/index.md): With test and fix both true, the API writes or repairs an extractscheme from your prompt against the live page, and saves it only if it works.
- [Test](https://docs.evomi.com/scraping-products-instructions/scraper-api/parameters/extraction/schema/test/index.md): test=true validates an extractscheme against a live URL before it is saved, and shows what each selector actually returned.