Azure Vector Database Samples

As the need for customers to build copilots over their data grows, Vector Databases are becoming crucial in the architecture of production-grade copilot applications. This repository is a collection of samples that demonstrates how to use different vector database tools in Azure to store and query embeddings from text, documents and images.

The samples focus on -

• Working with text, documents and images
• Ingesting embeddings and constructing complex queries
• IaC scripts to spin up vector storage in Azure
• Common best practices

What this repository is not - This repository doesn't offer any guidance on how to build LLM apps (for example RAG pattern). Please check the following repositories for LLM app development guidance.

• LLMOps with Prompt flow (preview)
• Chat with your data - Solution accelerator

Repository Structure

• Code Samples
  • Azure AI Search
  • Azure Database for PostgreSQL
  • Azure Cache for Redis
  • Azure CosmosDb for PostgreSQL
  • Azure Cosmos DB for MongoDB vCore
  • Fabric Real-Time Analytics(Kusto)
• Evaluation Samples
• Best Practices

Run the Code Locally

• To run the code locally, install the Jupyter extension in Visual Studio Code. Please check Jupyter Notebooks in VS Code to understand how to use this extension.

• The samples uses conda to manage python dependencies. Each sample comes with a conda environment (yml) file. Use the following command to create the conda environment.

  conda env create -f environment.yml

  Please check Python environments in VS Code how to use conda with VS Code.