Overview

This exercise demonstrates how to implement Retrieval Augmented Generation (RAG) using Azure OpenAI Service and Azure AI Search. The goal is to enhance AI-generated responses by grounding them in custom data sources.

Steps & Configuration Details

1. Provision Azure Resources

To complete this exercise, you need:

  • Azure OpenAI resource
  • Azure AI Search resource
  • Azure Storage Account resource

Configuration Items:

  • Azure OpenAI Resource:

    • Subscription: Select an approved Azure subscription.
    • Resource Group: Choose or create a resource group.
    • Region: Choose from:
      • East US
      • East US 2
      • North Central US
      • South Central US
      • Sweden Central
      • West US
      • West US 3
    • Name: A unique name.
    • Pricing Tier: Standard S0

  • Azure AI Search Resource:

    • Service Name: A unique name.
    • Location: Same as OpenAI resource.
    • Pricing Tier: Basic

  • Azure Storage Account Resource:

    • Storage Account Name: A unique name.
    • Region: Same as OpenAI resource.
    • Primary Service: Azure Blob Storage or Azure Data Lake Storage Gen 2
    • Performance: Standard
    • Redundancy: Locally Redundant Storage (LRS)

After provisioning, gather:

  • Azure OpenAI Endpoint & Key
  • Azure AI Search Endpoint & Admin Key

2. Upload Your Data

  • Download the dataset:
    https://aka.ms/own-data-brochures
  • Extract the brochures and upload them to Azure Blob Storage:
    • Create a Blob Container named margies-travel.
    • Upload .pdf brochures to the root folder.

3. Deploy AI Models

Two models are required:

  1. Text Embedding Model → Converts text into vector format.
  2. GPT Model → Generates responses based on indexed data.

Deployment Commands (Azure Cloud Shell - Bash):

  • Text Embedding Model:
    az cognitiveservices account deployment create \
  -g <your_resource_group> \
  -n <your_OpenAI_resource> \
  --deployment-name text-embedding-ada-002 \
  --model-name text-embedding-ada-002 \
  --model-version "2" \
  --model-format OpenAI \
  --sku-name "Standard" \
  --sku-capacity 5
  • GPT Model:
    az cognitiveservices account deployment create \
  -g <your_resource_group> \
  -n <your_OpenAI_resource> \
  --deployment-name gpt-4o \
  --model-name gpt-4o \
  --model-version "2024-05-13" \
  --model-format OpenAI \
  --sku-name "Standard" \
  --sku-capacity 5
  • Navigate to Azure AI Search → Import and Vectorize Data.
  • Configuration Items:
    • Data Source: Azure Blob Storage
    • Blob Container: margies-travel
    • Vectorization Model: text-embedding-ada-002
    • Authentication Type: API Key
    • Index Name: margies-index
    • Enable Semantic Ranking: Yes
    • Schedule Indexer: Run Once

5. Configure Your Application

  • Clone the repository:
    git clone https://github.com/MicrosoftLearning/mslearn-openai
  • Open the folder in Visual Studio Code.
  • Install dependencies:
    • C#:
      dotnet add package Azure.AI.OpenAI --version 2.1.0
dotnet add package Azure.Search.Documents --version 11.6.0
    • Python:
      pip install openai==1.65.2
  • Open the configuration file:
    • C#: appsettings.json
    • Python: .env
  • Update Configuration Values:
    • Azure OpenAI Endpoint
    • API Key
    • Deployment Name (gpt-4o)
    • Azure AI Search Endpoint
    • Search API Key
    • Search Index Name (margies-index)
  • Save the configuration file.

6. Add Code to Use Azure OpenAI

  • Open the code file:
    • C#: ownData.cs
    • Python: ownData.py
  • Configure the Azure OpenAI client:
    • C#:
      ChatCompletionOptions chatCompletionsOptions = new ChatCompletionOptions() {
    MaxOutputTokenCount = 600,
    Temperature = 0.9f,
};
chatCompletionsOptions.AddDataSource(
    new AzureSearchChatDataSource() {
        Endpoint = new Uri(azureSearchEndpoint),
        IndexName = azureSearchIndex,
        Authentication = DataSourceAuthentication.FromApiKey(azureSearchKey),
    }
);
    • Python:
      text = input('\\nEnter a question:\\n')
completion = client.chat.completions.create(
    model=deployment,
    messages=[{"role": "user", "content": text}],
    extra_body={
        "data_sources": [
            {
                "type": "azure_search",
                "parameters": {
                    "endpoint": os.environ["AZURE_SEARCH_ENDPOINT"],
                    "index_name": os.environ["AZURE_SEARCH_INDEX"],
                    "authentication": {
                        "type": "api_key",
                        "key": os.environ["AZURE_SEARCH_KEY"],
                    }
                }
            }
        ]
    }
)

7. Run Your Application

  • Open Visual Studio Code.
  • Run the application:
    • C#:
      dotnet run
    • Python:
      python ownData.py
  • Example prompt:
    Tell me about London.
  • The response should include AI-generated text grounded in indexed data.

8. Clean Up

  • Delete Azure resources to avoid unnecessary costs:

Related Posts