Insurance Image Vector Search Demo - MongoDB

Overview

This project showcases a vector-based image search application tailored for the insurance industry. Using MongoDB’s vector search capabilities, the demo allows for efficient querying and retrieval of images by leveraging image embeddings (numerical representations of image features) to find visually similar insurance claim images. The backend API facilitates this search, enabling comparisons based on image content rather than traditional keywords.

Key Features

• Similarity Search: Uses vector embeddings to find similar images based on content rather than text tags or metadata.
• MongoDB Vector Indexing: Efficiently manages and queries large sets of image embeddings.

Prerequisites:

• MongoDB Atlas account: A cloud-based MongoDB account for creating the database and managing data.
• Dataset File: A sample dataset located at data/demo_rag_insurance.claims.json to populate the MongoDB collection with sample insurance claim data, including image embeddings.

Setup Instructions

Step 0: Set Up MongoDB Database and Collection

1. Log in to MongoDB Atlas and create a new database named demo_rag_insurance.
2. Inside this database, create a collection called claims_final.
3. Import the dataset file data/demo_rag_insurance.claims.json into the claims_final collection.

MongoDB Vector Search Index

To enable efficient similarity search using vector embeddings, create a vector index for the photoEmbedding field in your collection. This index will use cosine similarity to compare image embeddings.

{
  "fields": [
    {
      "type": "vector",
      "path": "photoEmbedding",
      "numDimensions": 1000,
      "similarity": "cosine"
    }
  ]
}

Ensure that numDimensions matches the output dimensions of your embedding model.

Step 1: Configure the Environment Variables for the backend

1. In the backend folder, create a .env file.
2. Add your MongoDB connection string using the following format and also specify the frontend URL that is permitted to interact with the backend API:

MONGO_URI="mongodb+srv://<username>:<password>@<cluster-name>.mongodb.net/"
ORIGINS=http://localhost:3000

Replace username, password, and cluster-name with your MongoDB credentials.

Step 2. Configure the Environment Variables for the frontend

1. In the frontend folder, create a .env.local file.
2. Add the URL for the API using the following format:

NEXT_PUBLIC_IMAGE_SEARCH_API_URL="http://localhost:8000/imageSearch"

Step 3: Run the backend

1. Navigate to the backend folder.
2. Install dependencies using Poetry by running:

poetry install

3. Start the backend server with the following command:

poetry run uvicorn main:app --host 0.0.0.0 --port 8000

The backend will now be accessible at http://localhost:8000, ready to handle API requests for image vector search.

Step 4 Run the frontend

1. Navigate to the frontend folder.
2. Install dependencies by running:

npm install

3. Start the frontend development server with:

npm run dev

The frontend will now be accessible at http://localhost:3000 by default, providing a user interface to interact with the image vector search demo.

Docker Setup Instructions

To run the application using Docker, follow these setup steps:

Build the Application

NOTE: If you don’t have make installed, you can install it using sudo apt install make or brew install make

To build the Docker images and start the services, run the following command:

make build

NOTE: Depending on the version of Docker Compose you have installed, you might need to modify the Makefile to use docker-compose (with a hyphen) instead of docker compose (without a hyphen), as the command format can vary between versions.

Stopping the Application

To stop all running services, use the command:

make stop

Cleaning Up

To remove all images and containers associated with the application, execute:

make clean

Deploy on AWS EC2 Instance

In this guide, we'll deploy a t2.micro instance running Ubuntu Server 24.04 LTS with approximately 20 GB of storage.

Step 1: Create the EC2 Instance

• Launch a t2.micro EC2 instance with Ubuntu Server 24.04 LTS from the AWS Console.

NOTE: Ensure that you open port 3000 for the frontend and port 8000 for the backend in your security group settings. Additionally, allow outbound traffic to port 27017, which is the default port for MongoDB.

Step 2: SSH into the Instance

Once the instance is up and running, SSH into the machine using the following command:

ssh ubuntu@<your-ec2-ip-address>

Step 3: Update the Package Index

Before installing any packages, it's good practice to update the package index:

sudo apt update

Step 4: Install Docker

Install Docker on your EC2 instance by running the following command:

sudo apt install docker.io -y

Verify the installation by checking the Docker version:

docker --version

Step 5: Install Docker Compose

Download Docker Compose:

sudo curl -L "https://github.com/docker/compose/releases/download/1.29.2/docker-compose-$(uname -s)-$(uname -m)" -o /usr/local/bin/docker-compose

Make Docker Compose executable:

sudo chmod +x /usr/local/bin/docker-compose

Check the Docker Compose version:

docker-compose --version

Step 6: Start and Enable Docker Service

Start the Docker service:

sudo systemctl start docker

Enable Docker to start on boot:

sudo systemctl enable docker