Documentation

Products

Load a CSV product feed into a brand and your AI can answer questions about what you sell — recommending items in conversation, showing product cards in chat, and quoting prices and stock read live at the moment of answering.

Overview

A brand's product catalogue lives at Brand → (your brand) → Products. You point the screen at a CSV feed; we download it, turn each product into something the AI can search, and keep the catalogue in step with the feed.

The catalogue belongs to the brand, not to the AI agent. Delete or replace the agent and your products stay exactly where they are.

Product feeds are re-processed automatically every 24 hours, so the catalogue keeps up with the feed. Prices and stock do not wait for that — they are read live every time the AI answers.

Adding a product feed

  1. Open Brand → (your brand) → Products
  2. Click Add Feed to open the feed configuration modal
  3. Enter the URL of your CSV product feed
  4. Click Save to start processing
Your CSV file must be accessible via a public URL. The system will automatically download and process the feed.
A brand needs an AI agent before it can hold a feed. If the brand has no Vex yet, the screen says so up front — "Add an AI agent first — Product feeds are still tied to an AI agent for now" — with a Go to Vex button, rather than letting you fill in a feed and fail on save. This is a known interim state, not how it will stay.Once the brand has Vex but no feed, you get the other empty state: "No product catalogue — add a product feed and your AI can answer questions about what you sell, with live prices and stock — and show product cards in chat."

CSV format requirements

Your CSV file should contain these standard product fields:

id,title,description,link,image_link,price,sale_price,brand,condition,availability,product_type,google_product_category
PROD-001,"Wireless Headphones","High-quality wireless headphones with noise cancellation","https://example.com/products/headphones","https://example.com/images/headphones.jpg","$199.99","$149.99","AudioTech","new","in_stock","Electronics > Audio","Electronics > Audio > Headphones"
id
string required
Unique identifier for the product (used for deduplication)
title
string required
Product name/title
description
string
Detailed product description
link
string required
URL to the product page
image_link
string
URL to the product image
price
string
Regular product price (e.g., "$99.99")
sale_price
string
Discounted/sale price if applicable
brand
string
Product brand/manufacturer
availability
string
Stock status (e.g., "in_stock", "out_of_stock")
condition
string
Product condition (e.g., "new", "used", "refurbished")

Managing product feeds

The feeds table shows each feed's Name, when it was Last checked, a status badge under Auto re-scan, and its Created At date.

Feed status

  • ACTIVE: Feed processed successfully and products are available for recommendations
  • PROCESSING: Feed is currently being processed
  • WARNING: Feed processed with warnings (some products may have issues). Automatic re-scanning continues
  • INACTIVE: Feed processing failed — automatic re-scanning is switched off for it until you run Rescrape on it yourself. Hover the information icon for the reason

Feed actions

  • View Chunks: See exactly what the AI stored for this feed's products
  • Rescrape: Manually trigger a reprocessing of the feed (5-minute cooldown)
  • Delete: Remove the feed and all associated products
Deleting a product feed removes every product that came from it, together with everything the AI learned about them. The AI will no longer be able to recommend those products.

Managing individual products

Below the feeds, the products table lists every product from your feeds, with its Title, Price, Availability and whether it is Active:

  • Enable/Disable: Toggle whether an individual product appears in recommendations
  • Delete: Remove products one at a time, or select several and delete them together
Products are automatically enabled when first imported. You can disable specific products if you don't want them recommended.

How product recommendations work

When visitors chat with your AI:

  1. The AI analyzes the conversation context
  2. If product-related intent is detected, relevant products are suggested
  3. Products appear as an interactive carousel in the chat widget
  4. Visitors can click products to view details or make purchases

The same carousel goes out to Facebook Messenger and Instagram as Meta's own product cards, and your agents now see it in the Yaplet inbox too, instead of the bare text the visitor's message thread used to show them.

The AI uses semantic search to match conversation topics with your product catalog, recommending the most relevant items based on product descriptions and metadata.

Prices and stock are read live

Price, sale price and availability are not part of what the AI learned about a product. They are read from your catalogue at the moment the answer is written, so the number in the AI's sentence and the number on the product card beside it can never disagree.

Two practical consequences:

  • Changing a price costs nothing. There is no re-processing to wait for and no re-processing to pay for.
  • You refresh a feed to add and remove products, not to correct prices. Prices were already right.

Automatic updates

Product feeds are re-processed every 24 hours for accounts on a paid plan or a trial. Each run:

  • Downloads the latest CSV data
  • Stores every row, which is what keeps price and stock current
  • Re-processes only the products whose durable details changed — the title, brand, condition, description or link — plus any product that has no AI record yet
  • Deletes products that have disappeared from the feed, along with what the AI learned about them
If you make urgent product changes, run Rescrape on the feed's row to update products immediately (subject to the 5-minute cooldown).

When some products seem missing

A feed can look perfectly healthy while only part of the catalogue is actually findable, and nothing on the screen tells you.

That repairs itself. Because any product with no AI record is always picked up, an interrupted or half-failed run is finished off by the next one — whether that is your own Rescrape or the nightly one. If a product is still missing after a full run, check that it is Active in the products table and that its row in the CSV has a title and a stable id.

Ownership and permissions

The permission is Vex AI (key Vex) — the same one that gates the Vex screens and the AI sources on Brand → Knowledge.

A product feed deliberately does not appear in the Knowledge sources list, so the same thing cannot be deleted from two different screens. Its refresh status, its last error and its product count live only on this page.