AvidiaExtract — Product Data Extraction

What AvidiaExtract Does

AvidiaExtract fetches a product URL — from a manufacturer, distributor, or supplier — and uses AI to parse the page into structured JSON fields. It handles HTML product pages, PDF datasheets, and JavaScript-rendered sites (with the browser extension).

The output is a normalized product record you can use as the input for all downstream modules: AvidiaDescribe for copy generation, AvidiaSEO for optimization, and your eCommerce platform for direct import.

Supported Input Types

• ✓Single URL — paste any product URL into the Extract UI and hit Extract. Ideal for testing and one-off products.
• ✓Batch URL list — paste up to 50 URLs at once using the bulk input field (newline-separated).
• ✓CSV import — upload a CSV with a url column to start a bulk extraction job. See the Import guide for column requirements.
• ✓API — submit extraction requests programmatically via POST /api/v1/ingest.
• ✓PDF datasheets — upload PDF files directly for extraction. Works best with machine-readable (non-scanned) PDFs.

Extracted Fields

AvidiaExtract attempts to populate the following fields from every product page. Fields that cannot be found are returned as null.

How Extraction Works

1. 1.Fetch: AvidiaTech's crawler fetches the HTML source of the URL, following redirects and handling standard HTTP headers. Pages that require JavaScript rendering fall back to a headless browser.
2. 2.AI Parsing: The raw HTML is sent to the extraction model, which identifies product data signals — spec tables, structured data (JSON-LD, microdata), image tags, pricing elements, and description blocks.
3. 3.Field Normalization: Raw values are normalized — units are standardized (e.g., "5 lbs" → {"value": 5, "unit": "lb"}), prices are stripped of currency symbols, and image URLs are resolved to absolute paths.
4. 4.JSON Output: The normalized record is stored as an ingestion and made available in the UI, via the API, and as input to downstream modules.

Handling Difficult Pages

Understanding Extraction Quality Scores

Every extraction is assigned a quality score from 0 to 100. This score reflects how completely the system was able to populate the standard field set.

Scores below 50 typically mean critical fields like product name, description, or specifications could not be found. Filter your product list by score using the Score < 50 filter to identify products that need manual review or reprocessing.

Reprocessing Failed Extractions

If an extraction fails or produces a low-quality result, you can reprocess it without consuming an additional extraction credit (first reprocess is free). From the product detail view, click Reprocess → choose whether to use the cached page or re-fetch the URL.

Via the API, set "reprocess": true in your ingest request body to force a fresh extraction even if a cached result exists.

Rate Limits and Quotas

API: Submit an Extraction

POST https://app.avidiatech.com/api/v1/ingest
Authorization: Bearer <your-api-key>
Content-Type: application/json

{
  "url": "https://www.boschtools.com/us/en/boschtools-ocs/cordless-drill-drivers-gsb-18v-755-06019H3110.html",
  "options": {
    "reprocess": false,
    "pipeline": ["extract", "describe", "seo"]
  }
}

// Response (202 Accepted)
{
  "id": "ing_01HX4K2QZRP7W9VMBGT3DENF8",
  "status": "queued",
  "created_at": "2024-03-15T14:22:00Z",
  "url": "https://www.boschtools.com/..."
}

Poll the result with GET /api/v1/ingest/:id or set up a webhook to be notified on completion. See the full API Reference.

Common Errors and Fixes