Products
A product catalog turns your assistant from something that answers questions into something that sells. Connect your existing product API once and the bot can search it, filter it, sort it, and show real products as cards in the chat — with your prices, your stock status, and links straight to your product pages.
Requires the Product Catalog module. Everything in this section is unavailable without it; see Modules.
Open it at Dashboard → Product Catalogs.
How it fits together
Promptly doesn't query your API on every visitor question. It imports your products, embeds them for search, and generates a search tool the assistant can call — so a product question is answered against Promptly's copy in a fraction of a second rather than waiting on your store. A sync keeps that copy current.
That's the whole model, and the four setup steps map onto it directly: connect the API, tell Promptly what your fields mean, import, then teach the AI your business's vocabulary.
In this section
- Connecting a catalog — create a catalog and point it at your product API. Platform presets for Shopify, WooCommerce, Magento 2, BigCommerce, PrestaShop, and generic JSON, plus auth, headers, and manual import for tiny catalogs.
- Schema detection & field mapping — the AI reads a sample product and infers your fields; you confirm which are searchable, which are filterable, and how they map onto title, price, image, and URL. Also pagination and the final test-and-activate step.
- Sync & scheduling — run the first import, then keep it fresh. Incremental versus full sync, the five sync phases, history, and automatic schedules that suspend themselves when your API breaks.
- AI context & examples — where you teach the bot your domain: business rules, the mapping from customer words to filter values, and few-shot examples for tricky queries. This is where a mediocre product bot becomes a good one.
- Product cards — what a card shows, which mapped fields feed it, and where cards appear in the dashboard and the widget.
Suggested order
Work the wizard front to back — connect, map, test, activate, sync — then stop and actually talk to your bot. Ask it the five questions your customers ask most.
The gaps you find will almost all be AI context problems rather than mapping problems: the bot doesn't know that "kuća" means one thing in your propertyType field, or that "cheapest" should sort ascending, or that your city names are uppercase. That page is where you fix them, and each fix takes effect immediately.
A note on data quality
The most common cause of a disappointing product bot isn't the AI — it's the catalog. Duplicated entries, two different things sharing one price field (sale price and rental price, say), HTML markup left in descriptions, and inconsistent category spellings all survive the import and all surface in answers.
If your bot is confidently wrong about products, look at the raw data on the Products tab before you rewrite any prompts.