NutrAI Searcher - AI that reveals a product's nutrients with a single photo click

This repository guides to develop a Multimodal FastAPI application that leverages the CLIP model for analyzing images or text. The app performs semantic search to identify and retrieve nutritional information by querying a vector database, making it a powerful tool for multimodal data processing. The fastapi app is connected with web frontend-end written in Next.js for a better user interaction.

Table of Contents

• Introduction
• Features
• Execution
• Front-End
• Contributors

Execution

In this project, we used OpenAI CLIP model to create images embeddings and compare them with the already created text embeddings present in the Qdrant database and give the respective nutirents. The FastAPI model takes the image file as request and after processing give the nutrients as the response.

Features

• Fast and efficient
• Best for Diet freaks
• Text and Image multimodal searching
• Easily intgrable with frontend with FastAPI

Installation

1. Clone the repository:

   git clone https://github.com/vansh-khaneja/NutrAI-Searcher-FastAPI
   cd NutrAI-Searcher-FastAPI

2. Set up the Python environment and install dependencies:

   python3 -m venv venv
   source venv/bin/activate
   pip install -r requirements.txt

3. Set up Qdrant:

   Login to Qdrant Cloud to get the api_key and url.

    qdrant_client = QdrantClient(
    url="https://xxxxxx-xxxxx-xxxxx-xxxx-xxxxxxxxx.us-east.aws.cloud.qdrant.io:6333",
    api_key="<your-api-key>",
    )

4. Run the data_upload.ipynb in jupyter notebook to setup and upload data in vector database.

5. Execute the main.py file by running this command to run the API.

   uvicorn main:app --reload

Front-End

1. For the execution of front-end refer this repository.

   https://github.com/shubhamchoudhary-2003/NutrAI-Frontend

Contributors

Vansh Khaneja (AI-ML/Back-end)

Shubham Choudhary (Front-end)