This repository provides materials for a Microsoft AI Tour 2026 workshop focused on building a conversational AI agent for a retail company. It addresses the challenge of integrating external data and securing sensitive information by leveraging Microsoft Foundry Agent Service, the Model Context Protocol (MCP), and PostgreSQL with Row Level Security (RLS) and the pgvector extension for semantic search. The project targets developers and researchers seeking to create intelligent, secure, and data-rich AI agents.

How It Works

The core approach involves constructing an AI agent using Microsoft Foundry Agent Service, which facilitates agent deployment and observability. External data and tool integration are managed via the Model Context Protocol (MCP), enabling agents to access diverse information sources. PostgreSQL serves as the data backend, enhanced with pgvector for efficient semantic search and RLS to enforce role-based data access controls, ensuring data security and privacy. This architecture allows for sophisticated data analysis and customer assistance.

Quick Start & Requirements

Project setup involves compiling Python dependencies using pip-tools from requirements.in. Specific hardware or software version prerequisites are not detailed, though the project leverages enterprise-grade Microsoft Foundry and PostgreSQL. Key resources are available via the AI Tour 2026 Resource Center (https://aka.ms/AITour26-Resource-center) and Learn at AI Tour (https://aka.ms/LearnAtAITour).

Highlighted Details

• Develops a conversational AI agent for Zava, a retail DIY company, to analyze sales data and assist customers with product discovery.
• Employs the Model Context Protocol (MCP) for robust external data and tool connectivity.
• Utilizes PostgreSQL with pgvector for semantic search capabilities and RLS for granular, role-based data protection.
• Showcases Microsoft Foundry as an enterprise AI platform offering unified model access, monitoring, tracing, and governance.
• Integrates Responsible AI principles and tools like Azure AI Content Safety and the Azure AI Evaluation SDK.

Maintenance & Community

Key content owners include Dave Glover, Marlene Mhangami, and Aaron Powell. Community engagement is encouraged via the Microsoft Foundry Community Discord. Questions can be directed to Mike Kinsman and Erik Weis.

Licensing & Compatibility

The repository's README does not specify a software license, which may impact commercial use or integration into closed-source projects.

Limitations & Caveats

This repository appears to be primarily workshop material, not a production-ready library. Multi-language support is noted as "coming soon," indicating current limitations. Specific performance benchmarks or detailed scalability assessments are not provided within the README.