Azure-Samples/rag-with-azure-ai-search-notebooks

Jupyter notebooks that demonstrate vector search, hybrid search, image search, RAG, and evaluation, all with Azure AI Search.

Azure AI Search Demos

This repository contains many notebooks that explain how Azure AI Search works, including several showcasing how vector search works.

Environment setup

1. Run azd up on azure-search-openai-demo with GPT-4-vision enabled. This will create the necessary resources for the Azure OpenAI, Azure AI Search services, and the Computer Vision service.

2. Create a .env with these variables, with the values taken from .azure/ENV-NAME/.env in the azure-search-openai-demo repository.

   AZURE_OPENAI_SERVICE=YOUR-SERVICE-NAME
   AZURE_OPENAI_DEPLOYMENT_NAME=YOUR-OPENAI-DEPLOYMENT-NAME
   AZURE_OPENAI_ADA_DEPLOYMENT=YOUR-EMBED-DEPLOYMENT-NAME
   AZURE_SEARCH_SERVICE=YOUR-SEARCH-SERVICE-NAME
   AZURE_COMPUTERVISION_SERVICE=YOUR-COMPUTERVISION-SERVICE-NAME
   AZURE_TENANT_ID=YOUR-TENANT-ID

3. Login to your Azure account using the Azure CLI. Specify --tenant-id if you deployed that repo to a non-default tenant.

   azd auth login

4. Create a Python virtual environment or open the project in a container.

5. Install the requirements:

   pip install -r requirements.txt

Notebooks

These are the available notebooks, in suggested order:

• Vector Embeddings Notebook
• Azure AI Search Notebook
• Image Search Notebook
• Azure AI Search Relevance Notebook
• RAG with Azure AI Search
• RAG Evaluation

You can find video recordings going through the notebooks here.