“I’m pregnant,” “there are four of us,” “no more than €60 including drinks”: guests describe a situation. iAlacarte understands it, applies it to your menu and responds with dishes you serve.
Guests can select their allergens, diets and unwanted foods in a selector, or simply state them. A health or wellness situation (“I’m diabetic”, “I’m tired”) is stated to the assistant. In all cases, they apply throughout the conversation: recommendations, menus, pairings and suggestions.
A dislike (“I don’t like anchovies”) excludes the ingredient, not an entire food family. A declared allergy (“no peanuts”) excludes the whole family, as a precaution.
| The guest | The assistant |
|---|---|
| “I’m pregnant, what do you recommend?” | Only suggests dishes compatible with pregnancy according to its guidelines (alcohol, including cooked alcohol, raw foods, cured meats, raw-milk cheeses…). If she asks for wine, it reminds her that alcohol is not recommended and suggests non-alcoholic drinks. |
| “I don’t eat tomatoes. Suggest a salad.” | The ingredient becomes a restriction chip for the whole session. If no salads remain, it says so and suggests something else. |
| “I don’t want to go over 1,500 calories for this meal, including drinks.” | Builds complete menus below this limit; the estimated total is displayed beneath each menu. |
| “€60 including drinks, what can I eat?” | Menus within the budget, with the drink and glass of wine included in the total. |
| “What’s in the Caesar salad? Does the toast contain gluten?” | An answer read from the dish record, without using a language model. |
| “Which wine goes with pata negra?” | A pairing taken from your wine list; the pairing printed on your menu comes first. |
| “There are four of us, what can we share?” | The sharing dishes on the menu: boards, tapas and plates for the table. |
| “What time are you open until?” | Your opening hours for the day, exactly as you entered them. |
| “And in English for my friend?” | The same answer, replayed in English. |
Each answer fits into a few sentences. Mentioned dishes appear as cards: one tap opens the full record (ingredients, allergens, nutrition, pairing). Follow-up suggestions are offered, also filtered by the guest’s restrictions.
Factual questions (ingredients, prices, calories, opening hours) receive an answer written from the data, without a language model. The language model writes recommendations and can only mention dishes selected for that answer.
Fifteen allergens (the fourteen covered by European regulations and buckwheat), thirteen diets, including pregnancy, and an “unwanted foods” section with free-text entry. Each selection marks the relevant dishes on the menu and filters answers. A health situation or wellness goal is not selected: guests state it to the assistant, which applies it and displays it as a chip, and the menu highlights dishes that meet it. A chip can be removed with one tap; guests can also remove it in their own words (“remove the restriction”).
Guests type or dictate in their language; the assistant recognises it and answers in that language. Changing language during a conversation keeps the context: the latest answer is replayed in the new language, and restriction chips remain. See the menu in 27 languages.
Health guidance and nutritional values are estimates. For an allergy, the assistant names excluded dishes and reminds guests, once per conversation, that preparation may involve traces: confirmation must be obtained from the team. A question about medication is referred to a doctor or pharmacist.
Import your menu and ask it your guests’ questions: allergies, budget, pairings and opening hours.