name	description	languages	products	page_type	urlFragment
RAG using Semantic Kernel with Azure OpenAI and Azure Cosmos DB for MongoDB vCore	A Python sample for implementing retrieval augmented generation using Azure Open AI to generate embeddings, Azure Cosmos DB for MongoDB vCore to perform vector search and semantic kernel. Deployed to Azure App service using Azure Developer CLI (azd).	azdeveloper python bicep html	azure azure-app-service azure-openai cosmos-db mongodb-vcore	sample	rag-semantic-kernel-mongodb-vcore

RAG using Semantic Kernel with Azure OpenAI and Azure Cosmos DB for MongoDB vCore

A Python sample for implementing retrieval augmented generation using Azure Open AI to generate embeddings, Azure Cosmos DB for MongoDB vCore to perform vector search and semantic kernel. Deployed to Azure App service using Azure Developer CLI (azd).

🎥 Click this image to watch the recorded reactor workshop

How to use?

1. Create the following resources on Microsoft Azure:

   • Azure Cosmos DB for MongoDB vCore cluster. See the Quick Start guide here.
   • Azure OpenAI resource with:
     • Embedding model deployment. (ex. text-embedding-ada-002) See the guide here.
     • Chat model deployment. (ex. gpt-35-turbo)

2. 📝 Start here 👉 rag-azure-openai-cosmosdb-notebook.ipynb

notebook-full.mp4

Test it inside codespaces 👇

Running the web app locally

To run the Quart application, follow these steps:

1. Download the project starter code locally

   git clone https://github.com/john0isaac/rag-semantic-kernel-mongodb-vcore.git
   cd rag-semantic-kernel-mongodb-vcore

2. Install, initialize and activate a virtualenv using:

   pip install virtualenv
   python -m virtualenv .venv
   source .venv/bin/activate

   Note - In Windows, the .venv does not have a bin directory. Therefore, you'd use the analogous command shown below:

   source .venv/Scripts/activate

3. Install the dependencies:

   pip install -r requirements-dev.txt

4. Run the notebook to generate the .env file and test out everything first

5. Install the app as an editable package:

   pip install -e src

6. Execute the following command in your terminal to start the quart app

   export QUART_APP=src.quartapp
   export QUART_ENV=development
   export QUART_DEBUG=true
   quart run --reload

   For Windows, use setx command shown below:

    setx QUART_APP src.quartapp
    setx QUART_ENV development
    setx QUART_DEBUG true
    quart run --reload

7. Verify on the Browser

Navigate to project homepage http://127.0.0.1:5000/ or http://localhost:5000

website-full.mp4

Step-by-Step Deployment

Follow this guide 👉 Build RAG Chat App using Azure Cosmos DB for MongoDB vCore and Azure OpenAI: Step-by-Step Guide

azd Deployment

This repository is set up for deployment on Azure App Service (w/Azure Cosmos DB for MongoDB vCore) using the configuration files in the infra folder.

To deploy your own instance, follow these steps:

1. Sign up for a free Azure account

2. Install the Azure Dev CLI.

3. Login to your Azure account:

   azd auth login

4. Initialize a new azd environment:

   azd init

   It will prompt you to provide a name (like "quart-app") that will later be used in the name of the deployed resources.

5. Provision and deploy all the resources:

   azd up

   It will prompt you to login, pick a subscription, and provide a location (like "eastus"). Then it will provision the resources in your account and deploy the latest code. If you get an error with deployment, changing the location (like to "centralus") can help, as there may be availability constraints for some of the resources.

When azd has finished deploying, you'll see an endpoint URI in the command output. Visit that URI to browse the app! 🎉

Note

If you make any changes to the app code, you can just run this command to redeploy it:

azd deploy

Add the Data

1. Open the Azure portal and sign in.

2. Navigate to your App Service page.

   

3. Select SSH from the left menu then, select Go.

4. In the SSH terminal, run python ./scripts/add_data.py.

Add your Own Data

The Python scrips that adds the data is configured to accept any JSON file with your data but you need to specify the following parameters when you run it:

• Data file path: Path to the JSON file that contains your data. --file="./data/text-sample.json" or -f "./data/text-sample.json"

• ID field: This is the name of the field that cosmos uses to identify your database records. --id-field=id or -id id

• Text field: This is the name of the field that will be used to generate the vector embeddings from and stored in the database. --text-field=content or -txt content

• Description field: This is the name of the description field that cosmos will store along with the embeddings. --description-field=title or -desc title

  python ./scripts/add_data.py --file="./data/text-sample.json" --id-field=id --text-field=content --description-field=title

Example for Step-by-step Manual Deployment

1. Add your JSON data to the data folder.

2. The workflow will trigger automatically and push your data to the Azure App service.

3. Open the Azure portal and sign in.

4. Navigate to your App Service page.

   

5. Select SSH from the left menu then, select Go.

6. In the SSH terminal, run the following command with the changed values to suit your data:

   python ./scripts/add_data.py --file="./data/text-sample.json" --id-field=id --text-field=content --description-field=title

Example for azd Deployment

1. Add your JSON data to the data folder.

2. Run azd deploy to upload the data to Azure App Service.

3. Open the Azure portal and sign in.

4. Navigate to your App Service page.

   

5. Select SSH from the left menu then, select Go.

6. In the SSH terminal, run the following command with the changed values to suit your data:

   python ./scripts/add_data.py --file="./data/text-sample.json" --id-field=id --text-field=content --description-field=title