This sample application demonstrates building a Generative AI chat application using Azure Cosmos DB's vector search capabilities and Azure OpenAI Service. It targets developers looking to create scalable AI-powered applications with natural conversational interactions, improved performance, and cost efficiency through semantic caching.

How It Works

The application leverages Azure OpenAI Service for generating text completions and embeddings, and Semantic Kernel SDK for orchestrating these AI interactions. It utilizes Azure Cosmos DB for NoSQL, specifically its preview vector search feature, to manage chat history, optimize token consumption, and implement a semantic cache for enhanced performance and cost savings.

Quick Start & Requirements

• Install/Run: Use Azure Developer CLI (azd auth login, azd init --template cosmosdb-chatgpt, azd up).
• Prerequisites: Azure Subscription, Azure Developer CLI, Azure OpenAI Service access, Azure Cosmos DB for NoSQL Vector Search preview enrollment.
• Setup: Requires Azure resource deployment and configuration.
• Links: Latest Version, Quickstarts, Vector Search Announcement, Azure OpenAI Docs, Semantic Kernel

Highlighted Details

• Demonstrates building a highly scalable Generative AI chat application.
• Covers managing context windows and token consumption for LLMs.
• Implements a semantic cache using Cosmos DB vector search.
• Utilizes Semantic Kernel SDK for AI task orchestration.

Maintenance & Community

This sample is now maintained in a new repository: AzureCosmosDB/cosmosdb-nosql-copilot.

Licensing & Compatibility

The repository does not explicitly state a license in the provided README. Compatibility for commercial use or closed-source linking would depend on the final license of the maintained repository.

Limitations & Caveats

The core feature, Vector Search for Azure Cosmos DB for NoSQL, is in preview, which may imply potential breaking changes or API instability. The README also notes that this sample is being maintained in a new repository, suggesting potential divergence or deprecation of this specific version.