This repository provides the Information Assistant (IA) agent template, an end-to-end solution for building custom generative AI capabilities using Azure OpenAI Service. It targets organizations and developers looking to implement "chat with your data" scenarios via the Retrieval Augmented Generation (RAG) pattern, enabling natural language interaction with proprietary datasets. The benefit is a comprehensive, customizable reference sample for rapid development and deployment of AI-powered information retrieval systems.

How It Works

The IA leverages the Retrieval Augmented Generation (RAG) pattern, combining Azure OpenAI's LLMs with Azure AI Search for data retrieval. This approach grounds LLM responses in specific, user-provided data, ensuring accuracy and providing verifiable citations. It also offers "Ungrounded" generation using the LLM's broad knowledge and "Work and Web" for responses grounded in web content. A preview "Assistants" feature allows LLMs to autonomously reason, decompose tasks, and select tools without predefined workflows.

Quick Start & Requirements

• Deployment: Follow the instructions in the deployment guide.
• Prerequisites: Azure account with Azure OpenAI service access, Azure subscription with specific RBAC permissions (Role Based Access Control Administrator, User Access Administrator, Owner, Application Administrator), acceptance of Azure AI Services Responsible AI Notices.
• Models: Requires access to Azure OpenAI models like gpt-35-turbo, gpt-4, or gpt-4o (recommended), and optionally text-embedding-ada-002.
• Environment: Visual Studio Code is recommended for development, especially if conditional access policies are in place.

Highlighted Details

• Supports RAG, ungrounded, and web-grounded response generation.
• Includes features like explainable thought processes and referenceable citations.
• Offers preview access to autonomous agents for dynamic task execution.
• Integrates Azure AI Search for hybrid text and image search capabilities.

Maintenance & Community

This project is maintained by Microsoft. Further community engagement and support can be found via the Microsoft Open Source Code of Conduct.

Licensing & Compatibility

The IA is provided "AS-IS" and is excluded from Azure Service Level Agreements and limited warranty. It is not designed for processing Personal Data or data subject to legal/regulatory compliance. Terms in the DPA related to Personal Data, GDPR, Data Security, and HIPAA do not apply.

Limitations & Caveats

The IA may employ lesser or different privacy and security measures than typical Azure Services. It is not intended as legal services or a substitute for professional legal counsel. Microsoft makes no warranty that the IA is accurate, up-to-date, or complete. Users are responsible for ensuring compliance with all applicable laws and regulations.