microsoft/Build-your-own-copilot-Solution-Accelerator

What it solves

This project provides a reusable architecture and set of code snippets to help organizations build custom AI copilots using their own enterprise data. It specifically targets the challenge of streamlining daily tasks and meeting preparation for customer-facing roles, such as client advisors, by automating the synthesis of client profiles and conversation history.

How it works

The solution integrates several Azure services to create a RAG-based assistant:

• Azure OpenAI Service: Handles text generation, summarization of conversations, and autonomous function calling via Semantic Kernel to interact with structured and unstructured data.
• Azure AI Search: Provides semantic search capabilities to ground the AI's responses in processed enterprise datasets using vector embeddings.
• Data Processing: Scales the processing of past conversations to populate the AI search index.
• Infrastructure: Deployed via the Azure Developer CLI (azd) and hosted using Azure Container Apps, with data stored in Azure SQL DB and Azure Cosmos DB.

Who it’s for

It is designed for customers and enterprise software engineers who need a production-ready template to build custom copilots for customer-facing employees to improve client retention and productivity.

Highlights

• RAG Implementation: Uses Azure AI Search for grounding and semantic search.
• Autonomous Function Calling: Leverages Semantic Kernel to chat with both structured and unstructured data.
• Enterprise-Ready Architecture: Includes integrated security via Azure Key Vault and Managed Identity.
• Quick Deployment: Supports rapid setup through the Azure Developer CLI.