🥗 MatKompis – SmartRecipe AI Assistant

MatKompis (“Food Buddy”) is a conversational AI assistant that provides personalized meal recommendations and nutritional insights using real Swedish supermarket data and LLM-powered reasoning.

It combines intent detection, retrieval-augmented generation (RAG), and local dataset lookups to deliver realistic and context-aware food suggestions.

---

🧭 Overview

The system allows users to chat about what they want to eat (“I want a high-protein lunch under 600 kcal”) and receive meal ideas, ingredient lists, and nutritional breakdowns — all grounded in real product data.

💬 “Your AI friend for healthier, smarter, and more sustainable eating.”

---

🧩 Architecture

🔹 Flow Summary

1. User Input via Gradio Chat UI
2. FastAPI Backend receives /ask requests
3. Gemini API (Intent Classification) identifies query type
   • Meal recommendation / recipe query
   • Nutrition or price query
4. SentenceTransformer Models
   • all-MiniLM-L6-v2: for recipe semantic matching
   • KBLab/sentence-bert-swedish-cased: for Swedish text embedding
5. String Lookup Modules
   • Recipe dataset lookup
   • Product dataset lookup
6. Gemini API (RAG Answer Generation) synthesizes final response
7. Response Displayed in Gradio interface

---

⚙️ Features

• 🍽️ Personalized Meal Recommendations
• 🧮 Nutritional Estimation (kcal, protein, carbs, fat)
• 🛒 Integration with ICA & Hemköp Datasets
• 🧠 LLM-powered Understanding (Gemini / OpenAI)
• 🧾 RAG Architecture with Sentence Embeddings
• 🧑‍🍳 Local Recipe & Product Lookup

---

🚀 Getting Started

1️⃣ Clone the repository

git clone https://github.com/alireza1420/MatKompis.git
cd MatKompi

2️⃣ Create and activate a virtual environment

python -m venv venv
source venv/bin/activate  # or on Windows: venv\Scripts\activate

3️⃣ Start the FastAPI backend AND launch Gradio

uvicorn app:main --port 8000
gradio client.py