Yes. Brewing recipes live in postmeta as a structured list of dose, yield, temperature, grind, and time. SleekAI reads those fields and replies with a complete recipe, like 15 g coffee with 250 g water at 92 degrees over 3 minutes on a V60. Different coffees get different recipes based on their actual data, not a generic filter rule.

Map origin country, region, farm or co-op, process, and varietal as product attributes. The bot can confirm that a Yirgacheffe Kochere is washed or that a Sidamo is natural. For blends it can quote the component lots and roughly how much of each is in the blend without overstating sourcing claims.

Yes. Roast level and intended use map from product attributes. The bot can recommend an espresso roast for an espresso machine owner and a medium-light filter roast for a V60 customer, complete with their respective $16 and $18 250 g prices. Omni roasts are flagged as suitable for both.

Yes. Each coffee usually has 250 g, 500 g, and 1 kg variations. SleekAI reads variation postmeta and quotes the right row directly. A 500 g Yirgacheffe quotes as $32, with the saving over two 250 g bags called out, while a 1 kg wholesale bag is priced and packaged differently for cafe customers.

Yes. Map Woo Subscriptions coffee club postmeta and the bot can quote a 500 g monthly subscription at $32 with a 10 percent subscriber saving. It can explain how to swap coffees between cycles, skip a month, change grind from whole bean to V60 fineness, or pause via the account page.

Yes. Display conditions support post type, product category, tag, user role, and URL pattern. A retail bot focuses on origin and home brewing recipes, while a wholesale bot specialises in 1 kg pricing, weekly delivery, and cafe friendly grind options. Both read the same WooCommerce data with no duplication.

Decaf and limited release coffees use the same WooCommerce schema. Map their decaffeination process and lot number as attributes. The bot can confirm whether a decaf is sugarcane EA processed or Swiss Water, and reflect stock status on every turn so a sold-out limited release does not get promised to a customer who is about to be disappointed.

Yes. Every chat is logged with the model name, token usage, and the page URL it ran on. Patterns like repeated questions about missing brewing recipes or unclear process fields feed back into product attributes and recipe postmeta, so the roastery's online voice tightens to match the bar.