Yes. SleekAI reads your live WooCommerce stock so recommendations stay inside what is on the shelf today. The bot can mention adjacent brands for context (the Lee Kum Kee version when the Koon Chun is out) but the recommendation, the brand the shopper should grab, stays in stock. Out-of-stock items get filtered before the prompt is built.

Yes. The underlying LLM handles dozens of languages. The system prompt instructs the bot to match the customer's language and switch mid-conversation if the cook flips between English and their first language for ingredient questions. For mixed-script names (pinyin vs Chinese characters), you can also load a glossary into the knowledge base so transliteration matches your house style.

Yes. Add region (Sichuan, Cantonese, Hunanese, Shandong) and category (sauces, noodles, dried goods) as taxonomy or custom fields and the bot reasons across them. A customer asking 'what is the most authentic Cantonese soy sauce you carry' gets a real answer, not a generic 'try our soy sauce' line. The same applies to North vs South Indian, Levantine vs Maghreb, regional Chinese, and so on.

Yes. Multibot plus display conditions lets a Latin-American bot live on /latino/, an East Asian bot on /asia/, a Middle Eastern bot on /middleeast/, each with its own knowledge base and prompt. Display conditions can also tune the bot by user role (wholesale buyers vs retail) or by URL pattern (recipe pages vs product pages).

Yes. Drop your house recipes into the OpenAI Files vector store and the bot retrieves the right card mid-conversation. A cook asking 'I want to make mapo tofu' gets a list of the items you stock plus a note on what they need elsewhere. The same flow works for biryani, mole, kibbeh, paella, and any dish with a defined ingredient list.

Yours. SleekAI is bring-your-own-key for OpenAI, Anthropic, Google, or OpenRouter. For multilingual conversations a higher-end model often handles register and idiom better, so you can route language-flagged conversations to a stronger model while a smaller model handles English in-stock checks at a lower cost-per-message.

Yes, when packaging color, brand, and shelf location live as fields on each product. A customer asking 'which one is the red-label Pixian' gets a real answer with shelf aisle and brand color, which is the kind of detail that turns a 10-minute wander into a 90-second find. For Asian groceries especially, packaging-color cues are how regulars navigate, and the bot uses them when you tag them.

If those live as separate products with stock fields, yes. Fresh herbs, tofu, dumpling skins, fresh noodles, and refrigerated sauces all surface in recommendations. The bot also respects expiry-window fields if you publish them, so a customer planning Sunday dumplings on Wednesday gets a steer to the fresh skins arriving Friday rather than the ones that will need to move before then.