Skip to content
Masab Farooque
Contact

A-01RotiShoti AI

RotiShoti AI

Food discovery for Pakistani cuisine where you describe what you feel like eating and it answers properly.

Client
Independent build
Year
2025
Role
Sole builder. Data model, retrieval, AI layer and interface.
RotiShoti AI, food discovery for Pakistani cuisine where you describe what you feel like eating and it answers properly.

Fig. 1 · Architecture

How it works

Every part is drawn from what the project does. The traffic moving through it is illustrative.

Drag to turn · point to inspect

Overview

Most restaurant search is a filter list. RotiShoti lets people ask the way they actually talk, something like cheap karahi near me that is open late, and returns places that fit.

Semantic search runs over the restaurant and dish data, with budget and location narrowing the result set. Conversation context carries between turns so a follow up question does not start from zero.

It runs on Groq LLaMA models for fast responses with Supabase behind it, which keeps the experience quick without an expensive per query bill.

What it does
  • Natural language search across restaurants and dishes
  • Budget aware filtering
  • Location aware recommendations
  • Conversation memory across turns
  • Live data from Supabase
The hard parts
  • Semantic search that stays fast as the dataset grows
  • Holding context without the token bill climbing
  • Location queries combined with several other filters at once

Built with

Next.js 15FastAPISupabaseGroqZustandTailwind CSS

Next project

Javea Denia Rentals