AI mapping

AI mapping helps create the first version of a source-to-nopCommerce mapping. It is not a replacement for administrator review. The best result comes from AI suggestions plus a short manual validation pass.

Where AI mapping fits

Use AI mapping after the source schema is detected and before the first dry run.

The AI mapper uses:

1. Source field names.
2. Field types.
3. Sample values.
4. Target entity type, such as Products or Customers.
5. Existing nopCommerce target fields.

Recommended workflow

1. Create or test the source.
2. Review the typed source schema.
3. Open the mapping workflow.
4. Select the target entity.
5. Generate AI suggestions.
6. Review confidence and reasoning.
7. Save the mapping.
8. Run dry run.
9. Adjust any low-confidence fields.

What AI usually maps well

AI mapping is strongest for common fields:

Source example	Likely target
sku, itemNo, productCode	Product SKU
title, name, product_name	Product name
description, body_html	Full description
price, sale_price	Price
brand, manufacturer	Manufacturer
category, department, taxonomy	Category
image, imageUrl, images[0]	Product pictures
color, size, material	Product attributes
weight, warranty, rating	Specification attributes

What must be reviewed manually

Always review:

1. Fields that look similar but mean different things, such as cost, price, list price, and sale price.
2. Category hierarchy rules.
3. Attribute versus specification attribute decisions.
4. Image source rules.
5. Duplicate matching fields.
6. Required fields.
7. Any field with low AI confidence.

AI result modes

NopBridge can display which engine produced the suggestion:

Mode	Meaning
Bitlogika AI Gateway	The request was processed through the Bitlogika API gateway.
Direct OpenAI	A direct configured OpenAI key was used.
Heuristic fallback	No AI gateway result was available, so deterministic local rules created a fallback mapping.

If the UI shows heuristic fallback, treat the mapping as a useful starter only. It can still be helpful, but it is not a real AI response.

Quality checklist

Before real import:

1. Every required nopCommerce field has a source or default value.
2. SKU or source identity is stable.
3. Category mapping creates the expected hierarchy.
4. Attributes are separated from specification attributes.
5. Image mapping points to the right source: URL, file system, FTP, Drive, or object storage.
6. Dry run shows the expected create, update, and skip counts.
7. A second dry run does not produce unexpected duplicates.