No. The system prompt holds a firm boundary on diagnosis, treatment, dosage, and weight-loss recommendations. The bot can describe the products you carry, their macros, ingredients, and certifications, and refers customers to registered dietitians, pharmacists, or doctors for anything that crosses into medical territory. This is the right behaviour for any retailer not licensed to give medical advice.

Only against tagged data. The bot confirms gluten-free, dairy-free, soy-free, nut-free, sesame-free status against the allergen fields you set on each product. For unflagged products, the bot says it cannot verify and asks the customer to check the label. The bot always reminds customers with severe allergies to double-check packaging, because labels can change between batches and the bot is not a substitute for label reading.

Yes. Vegan + gluten-free + nut-free + low-sugar all combine into one filtered shortlist. The bot returns picks that meet every constraint, or honestly says 'nothing in stock meets all four' if that is the truth. This is more useful than a generic 'check our health-food aisle' deflection because the customer gets a concrete answer with shelf locations.

Yes. 'No added sugar' vs 'sugar-free,' 'gluten-free' vs 'wheat-free,' 'plant-based' vs 'vegan,' 'natural' vs 'organic': the bot explains these distinctions against your product descriptions and the regulatory definitions you publish on category pages. The explanations stay factual and avoid marketing language, which is what health-conscious customers are listening for.

Yes. Multibot plus display conditions can put a supplements-info bot on /supplements/ with a tighter boundary: it describes products you stock, their ingredients, and the manufacturer's stated use, without making efficacy claims or dosage recommendations. The medical-advice referral language stays prominent, and the bot defers to pharmacists or doctors for any therapeutic question.

Yours. SleekAI is bring-your-own-key for OpenAI, Anthropic, Google, or OpenRouter. Health food retailers often prefer a model with strong refusal behaviour for the medical-advice line. You can route allergen-flagged conversations to a higher-end model while a smaller model handles in-stock checks and routine diet filtering at lower cost-per-message.

If those products have stock fields, yes. Plant-based meats, dairy alternatives, kefir, kombucha, frozen acai packs, and prepared bowls all surface in recommendations. The bot can also respect expiry-window fields, so a customer planning a Friday dinner on Tuesday gets a steer to the kefir arriving Wednesday rather than the one ending its window Thursday.

It can describe the product (strain, CFU count, capsule vs powder, with-meal or empty-stomach storage) from your data. It does not recommend a probiotic for a specific health condition or suggest dosage outside what the manufacturer prints on the bottle. For anything beyond product description (which strain for which use), the bot refers to pharmacists, dietitians, or doctors. The same boundary applies to vitamins, minerals, herbs, and adaptogens.