yar-cafe-menu-flow

Purpose

This skill lets the assistant act as a menu concierge for YAR (Одеса, вул. Шишкіна, 48/1). It is built for the flow: user describes preferences in natural language -> assistant parses the full menu -> assistant returns suitable items with photo links and descriptions, so the user can choose without visiting the website.

What this skill does

• Fetches all YAR sections from one URL (menu, bar, wine-list, banketne-menyu).
• Normalizes menu data (name, price, category, description, section, photo URL).
• Supports keyword search and free-text preference-based suggestions.
• Generates share-ready cards (photo link + description + price).

Core commands

1) Dump full menu data

python3 scripts/yar_menu_flow.py dump --out yar_dump.json

2) Search by query

python3 scripts/yar_menu_flow.py search --input yar_dump.json --query "салат"

3) Suggest from free-form preferences (AI-style flow)

python3 scripts/yar_menu_flow.py suggest \
  --input yar_dump.json \
  --preference-text "Хочу щось ситне, без алкоголю, до 700 грн" \
  --no-alcohol \
  --budget 700

Use --with-drink auto|yes|no to control whether drink pairing is included.

4) Build photo-link cards

python3 scripts/yar_menu_flow.py cards \
  --input yar_dump.json \
  --query "десерт" \
  --require-photo

Delivery rule (important)

When showing menu visuals in chat, prefer hyperlinks to menu photos:

• Use photo_markdown (or photo_html) from card output.
• Send one item per message for readability.
• Use direct media attachment only when hyperlink format is not appropriate.

Resources

• scripts/yar_menu_flow.py — deterministic parser/search/suggest/cards CLI.
• references/preferences_profile_example.md — generic preference/filter template.