This repository provides a Python-based sample application for building a Retrieval Augmented Generation (RAG) copilot, grounded in custom company data. It's designed for users of Azure AI Studio looking to create enterprise-grade copilots with customizable intelligence and capabilities.

How It Works

The sample implements a RAG pipeline using custom Python code orchestrated by Prompt Flow. It involves generating search queries from user input, embedding these queries using an Azure OpenAI embedding model, retrieving relevant documents from an Azure AI Search index, and finally, passing the retrieved context along with the user query to an Azure OpenAI chat completion model for response generation. This approach allows for grounding LLM responses in specific, proprietary data sources.

Quick Start & Requirements

• Install/Run: Primarily Python scripts. Requires Azure CLI (az login).
• Prerequisites: Azure subscription, Azure AI Studio project, Azure OpenAI resource, Azure AI Search resource. Python 3.x.
• Setup: Configuration involves updating a provision.yaml file with Azure resource details, followed by running provisioning scripts. Index creation and evaluation scripts are also provided.
• Links: Azure AI Studio, Prompt Flow

Highlighted Details

• End-to-end RAG implementation with Python and Azure AI services.
• Includes steps for data indexing, prompt testing, performance evaluation, and deployment.
• Supports custom evaluation metrics and safety risk assessments.
• Facilitates deployment to managed endpoints within Azure AI Studio.

Maintenance & Community

This is an Azure Samples repository, indicating official Microsoft backing. Specific contributor details or community channels (like Discord/Slack) are not explicitly mentioned in the README.

Licensing & Compatibility

The repository is licensed under the MIT License. This permissive license generally allows for commercial use and integration into closed-source projects.

Limitations & Caveats

The README explicitly states that response quality is not guaranteed and is subject to ongoing development. Users are responsible for validating the application's outputs for their specific scenarios. Deployment requires specific Azure resource configurations and quotas.