# Boojee AI Voice Ordering - Built on AsqVox

## What it is

**Boojee** is a voice-ordering assistant for dine-in guests, built by **AsqVox**. A guest at a
café table scans the QR code on their table, the menu opens on their phone - no app install, no
login, no sign-up - and they simply *talk*. "I want something healthy and high-protein for
dinner." "Add the Hummus & Chicken." "Give me a high-protein drink to go with it." The assistant
listens, understands, and builds the order in the cart on their screen in real time. The guest
reviews it, tweaks it, and taps **Place Order** themselves - nothing is submitted on their behalf,
and there is no in-app payment; the order goes to the kitchen and is settled at the table, exactly
as café service already works.

It is a focused, honest product: **table-based dine-in only**. No delivery, takeaway,
reservations, or accounts to manage - just a faster, more natural way to order what is in front
of you.

## How quickly it came together on AsqVox's infrastructure

The reason Boojee could be stood up fast is that **almost none of the hard parts were new**.
AsqVox already runs the difficult machinery of production voice AI - live speech sessions, a tuned
voice agent, anti-abuse gates on every call, a real-time server-to-browser channel, a shared
design system, and a repeatable multi-environment deploy. Boojee is a *vertical* assembled on top
of that foundation, not a rebuild.

Concretely, the whole ordering experience rides on a **one-way command relay**: the voice agent
calls a function ("add to cart," "show me the desserts"), AsqVox's backend receives it over a
secured per-widget key, and streams it to the guest's screen over a live channel - where the
widget runs it against the on-screen cart. The backend is a **deliberately "dumb" relay**: it
holds *no menu of its own*. The menu lives in the café's own app and in the assistant's knowledge
base. That single design decision is what makes the platform scalable - onboarding a new café is a
matter of **menu data + a knowledge document + flipping one toggle**, not new engineering.

The database footprint tells the same story: enabling the entire feature took **two new columns**
(an on/off switch and a private per-widget key), mirroring capabilities AsqVox had already shipped.
The ordering app itself is a lightweight static site served on its own subdomain, bundling the
voice widget same-origin so there is nothing extra for a café to wire up. This is the leverage of
a platform: a whole new product surface, built mostly from parts that were already battle-tested
in production.

## What it can do

- **Natural voice ordering** - guests order in plain language; the assistant maps what they say to
  real menu items, even loosely worded ("a berry smoothie with protein" → *Berry Smoothie + whey
  isolate boost*).
- **Handles the messy real world** - it collects **required choices** on its own (bread, bun,
  salad size, tofu-vs-egg, protein) and can add **several items in a single request**
  ("add all three").
- **Mood & suggestion engine** - "something light and fresh," "spicy and bold," "sharing for the
  table" → the assistant surfaces 2–3 fitting picks as a *Suggested for you* strip on screen,
  leading with house signatures.
- **Nutrition intelligence** - every dish carries **protein / carbs / fat / calorie** tags shown
  right on the menu cards, and the assistant can answer real dietary questions: "a 50-gram-protein
  meal with a drink," "high-protein but low-fat," "healthy and clean." It reasons over pre-built
  macro leaderboards and best-combination tables, so the answers are fast and grounded.
- **Show-me-the-menu** - "what coffees do you have?" pulls the right section onto the guest's
  screen with a short spoken summary, instead of reading a wall of items aloud.
- **Diet-aware and honest** - full veg / egg / non-veg awareness, a veg-only mode, and
  allergen/dietary flags (Vegan, GF, Jain, Nut, Dairy-free). It **never invents items** - if
  something is not on the menu, it says so plainly and offers a real alternative.
- **On-screen, in control** - the assistant *fills* the cart; the guest always reviews and places
  the order. Voice and screen stay in lockstep in real time.
- **Secure and abuse-resistant by default** - it inherits AsqVox's front-door protections
  (trusted-gesture and challenge gates, per-widget authentication) so café minutes are not burned
  by bots.

## How customizable it is

Boojee is one *instance* of a reusable template - the same stack re-skins to any café:

- **The menu is pure data.** Items, prices, sections, dietary flags, mood mappings, and the macro
  tags are all data-driven - change the data, and both the on-screen menu and the assistant update
  together.
- **The assistant's knowledge is a single document.** One knowledge base defines everything the
  voice agent knows and how it suggests - swap it to change the café's entire personality and
  catalogue.
- **The voice agent is tunable** - language and accent handling (e.g. Indian-English), interruption
  sensitivity, response pacing, and boosted keywords are all adjustable per outlet without touching
  code.
- **Look and feel** follow the AsqVox design system, so branding, colour, and layout stay
  consistent and on-brand.
- **Per-table, per-outlet** context comes straight from the QR link, so the same deployment serves
  every table and location.
- **Toggle-controlled** - voice ordering is a per-widget switch; a café turns it on when ready,
  with its own private key.

## Why it matters

Boojee shows what AsqVox's platform is really for: taking a genuinely hard capability - a reliable,
real-time voice assistant that *does things* on screen - and making it a configuration exercise for
the next business. The heavy engineering was solved once, in the platform. Boojee is the proof that
a new voice-commerce experience can be launched on top of it quickly, safely, and with deep
customization - and that the next café, retailer, or clinic can be next.