RotiShoti AI: AI-Powered Islamabad Restaurant Recommender using LangChain and Vector Embeddings

Описание: RotiShoti AI is an AI-first food recommendation engine built for smarter and faster restaurant discovery.
Developed by Masab Farooque and Saad Ilyas, it uses a multi-LLM architecture powered by LangChain, OpenAI, and Supabase Vector DB to make food discovery conversational and personalized.

Unlike basic filter-based apps, RotiShoti understands natural language queries like:

“Find me beef burgers under 2000”
“Quiet rooftop for coffee”
“Good biryani spots near F-7 Islamabad”

It embeds restaurant data, including menus, reviews, cuisines, deals, and locations, using OpenAI’s text-embedding-3-small model for semantic search. A hybrid retrieval chain then ranks results by user intent, budget, and preferences in real time.

Tech Stack:
LangChain, FastAPI, OpenAI GPT, Supabase, PostgreSQL, React 19, Next.js 15, Tailwind CSS, Zustand, Vector Search, React Leaflet.

Currently in active development for Islamabad-based foodies.