You can import your product catalog using a CSV file, reusing an existing file, whether it's in Google Shopping format or another format you are already using.

If you need to create a new CSV file specifically for the catalog, the data feed description below outlines the essential information to include.

The product catalog provides the Assistant with the necessary data to assist customers in their search and answer questions regarding product characteristics. Therefore, an optimal structure of this CSV file is crucial to ensure effective functionality.

1. Data feed description

1.1. File format

The CSV file must include a header row indicating the content of each column, followed by a row for each product. The columns should list the product characteristics such as Title, Description, Price, etc.

An example of the CSV file is available via this link. Ensure to use the CSV format with UTF-8 encoding to avoid any errors during import.

1.2. Expected data

Required fields:

id: Unique Product ID
title: Product Title
description: Product Description
link: Product Page URL of your brand's website

Other available and strongly recommended fields:

item_group_id: Group ID (used to group variants of the same product: sizes, colors, materials, conditions)
availability: Availability
Standard values: in_stock, out_of_stock, preorder, backorder

The system also accepts:
- Writing variations: uppercase/lowercase, with or without hyphens/spaces (e.g., in stock, IN-STOCK)
- Numeric values: positive number = in stock | 0 or negative = out of stock
- Multilingual synonyms: yes, true, available = in stock | no, false = out of stock
The system automatically normalizes these values to one of the 4 standard statuses.
availability_date: Availability Date
price: Price
sale_price: Sale Price
image_link: URL of the main product image

Required for your AI Shopping Assistant's product discovery feature
additional_image_link: URL(s) of additional product images

Required for your AI Shopping Assistant's product discovery feature
product_types: Product Category

Example value: Home > Men > Clothing > Shirts
gender: Gender

Supported values: male, female, unisex
brand: Brand
color: Color

The primary color must be listed first
size: Size
material: Material

Can include a primary material followed by secondary ones separated by a comma, slash, etc.
condition: Condition

Supported values: new, refurbished, used
crosssellproducts: cross selling (learn more)
compatiblewith: compatibility (learn more)

These pieces of information are essential to provide an optimal experience for your visitors. If they are not correctly provided for the majority of products, the catalog cannot be activated or used by your Assistant.

The CSV file may contain additional columns to enrich the Assistant. You have two options:

Use the mapping feature to define the content of additional columns, such as color or size.
Let our system store these details as complementary data. In this case, ensure the column names are explicit, such as "Manufacturing Lead Time" or "Power Supply".

Once your product catalog is created, use the knowledge explorer to ensure all the provided information is correctly integrated into the knowledge base.

1.3. Data update

You can import a CSV file either by selecting it from your computer or by specifying the URL of a file hosted on your website. Using a URL offers the advantage of automatic daily synchronization of your data.

To ensure the Assistant always provides up-to-date information to your customers, it is essential to regularly update your CSV file, adding new products and adjusting existing information.

1.4. Product variant management

When a product exists in multiple variations (size, color, material, etc.), each variant must be represented as a separate row in the CSV file. All variants of the same product are linked together through the item_group_id field, which must carry the same value across the entire group.

Structuring rules to follow:

Each variant has a unique id that identifies it individually.
All variants of the same product share the same item_group_id.
The item_group_id field must be filled in on every variant row.

Example:

A pair of jeans available in three colours is structured as follows in the catalog:

id	title	color	item_group_id
79052-RAW	Straight-cut jeans – Raw	Raw	79052
79052-BLK	Straight-cut jeans – Black	Black	79052
79052-FAD	Straight-cut jeans – Faded	Faded	79052

Each variant has its own id and all share the same item_group_id: 79052.

⚠ Do not include any "parent" or "group" row in the catalog.

Some catalog formats include an additional row whose id matches the item_group_id value of the variants, without that row itself being a real purchasable variant (no individual colour, size or stock level).

This type of row is not compliant with the structure expected by the Assistant and may cause unexpected behaviour:

The Assistant may identify this row during product matching, but the Add-To-Cart feature will not work correctly, as this row does not correspond to any real variant.
The variant selection panel shown to the visitor will be incorrect or absent.

If your catalog contains this type of row, remove it. Only rows corresponding to real, purchasable variants should be present.

NB The Add-To-Cart feature identifies products via the id field (unique variant identifier) and not via the item_group_id field. It is therefore essential that every row in the catalog represents a real variant, with its own id and a populated item_group_id.

2. Assistance in creating the product catalog through CSV file

Use a URL to import your CSV file and benefit from automatic daily synchronization. If authentication is required, include the necessary information in the headers.

If your file does not use the standard CSV format (comma, semicolon, tab, etc.), adjust the field separator to ensure correct reading.

When the columns in your file follow standard labels, they are automatically recognized and mapped to the corresponding fields. Additional columns that are not configured in the mapping will be added as complementary information, as long as their name is clear enough to be interpreted correctly.

Finally, be sure to regularly update your CSV file to ensure the Assistant always has the most up-to-date information about your products.

The SFTP (Secure File Transfer Protocol) is not supported as an ingestion method. If your files are hosted on an SFTP server, you will need to export them to a compatible source first (for example, a public URL or a locally imported CSV file) before ingesting them into your knowledge base.

NB When importing your CSV file (via URL or local file), a preliminary check is performed on the first 10,000 rows of the file.

This allows potential errors to be detected immediately before starting the full ingestion. This makes the process smoother, especially for large files.

3. Error management

In case of an error, an explanatory message will specify the reasons for the issue.

The most common errors include:

Incorrect separator: The separator specified does not match the file format.
CSV format violation: Some rows do not contain the same number of columns. Each row must have a fixed number of columns, even if some are empty.
Non-compliant encoding: The file is not encoded in UTF-8.
Field format violation: The content of one or more fields does not follow the expected specifications (see the product catalog field requirements).

These error messages help quickly identify the source of the problem to make the necessary corrections.