This repository provides an accelerator for building a private, secure AI multi-agent system using Azure Cognitive Search and Azure OpenAI. It's designed for organizations to create ChatGPT-like experiences tailored to their data, enabling intelligent agents to answer questions with detailed explanations and source references across multiple communication channels.

How It Works

The system leverages a Retrieval-Augmented Generation (RAG) multi-agent architecture. An agent determines the appropriate data source (Azure SQL, APIs, Bing Search, Azure AI Search with enriched documents, or CSV files) based on user input. It retrieves information, crafts an answer, and saves conversation state to CosmosDB for persistence. LangChain is used for interacting with Azure OpenAI, vector stores, prompt construction, and agent creation, while LangGraph facilitates the multi-agentic design.

Quick Start & Requirements

• Install: Clone the repository and install dependencies using pip install -r ./common/requirements.txt within a Python 3.12 environment.
• Prerequisites: Azure subscription, Azure OpenAI Service with deployed models (gpt-4o, gpt-4o-mini, text-embedding-3-large, tts-whisper), Azure Machine Learning Workspace (preferred), or Visual Studio Code. Customer data must be uploaded to blob storage two weeks prior.
• Setup: Requires Azure infrastructure deployment via a provided button. Configuration involves editing credentials.env.
• Links: Demo, Troubleshooting Private Repo

Highlighted Details

• Utilizes Azure AI Search's hybrid search for semantic answers.
• Supports multi-modal input/output and multi-lingual capabilities.
• Integrates with various data sources including APIs and tabular data.
• Employs Azure AI Document Intelligence for complex PDF parsing.
• Offers multi-channel deployment via Bot Framework and Bot Service.

Maintenance & Community

This project is an official Microsoft accelerator, indicated by its association with Microsoft architects and customer-funded Value-Based Delivery (VBD) assets. Contributions are welcomed under a Microsoft CLA.

Licensing & Compatibility

The repository is licensed under the MIT License, allowing for commercial use and integration with closed-source projects.

Limitations & Caveats

The project is presented as a workshop accelerator and Proof of Concept (POC), implying it may require further development for production readiness. Specific Azure OpenAI model versions are required, and initial setup involves significant Azure resource provisioning.