Neighborhood
“Beijing” or “Tokyo” is usually too broad. A useful agent asks for a neighborhood, meeting point, or travel limit if it could change the result.
Start with a decision-ready prompt
Beijing
北京找一家安静、适合约会、人均 300 元以内的餐厅。不要纯拍照网红店，最好方便聊天。
Tokyo
Find a relaxed date-night restaurant near Shibuya, around ¥8,000 per person. We care more about conversation and food than a flashy view.
Budget
Clarify whether the amount is per person, includes drinks, and is a hard cap or comfortable target.
Conversation
Noise, table spacing, pacing, and reservation pressure can matter more than décor when the goal is to talk.
Food constraints
Dietary needs and disliked cuisines are hard filters. Novelty, tasting menus, and sharing plates are preferences.
Evidence and uncertainty
Nomtiq searches current provider data, checks each candidate against the stated constraints, and labels uncertainty. It should not invent a private room, noise level, price, opening hour, or booking availability that the live evidence does not support.
Why the result is a 2+1 shortlist
A long ranked list transfers the decision back to the user. Nomtiq instead returns two strongly supported fits and one exploration choice:
- Fit 1: the safest overall match for the stated priorities.
- Fit 2: a credible alternative with a different strength or tradeoff.
- Explore: a potentially memorable option, clearly labeled when evidence or fit is less certain.
Each option should say why it fits, what could be a drawback, and which details need confirmation before going.
Make the next date-night search better
After the meal, a user can record concise feedback such as “food excellent, room too loud, service pace comfortable.” Nomtiq stores restaurant preferences locally and can apply them to later decisions without publishing a review.
Try it
Install the open-source skill, or give a compatible agent the GitHub repository URL.