mission3

Mission 3: Orchestrate SQL + AI workflows

This mission advances your skills by teaching you how to orchestrate complex workflows that combine SQL database operations with AI capabilities. You'll learn to build intelligent, multi-step processes that seamlessly integrate data retrieval, transformation, and AI-powered decision making.

Overview

Learn how to design and implement sophisticated AI workflows that coordinate multiple database queries, embedding operations, and language model calls to solve complex business problems requiring multiple steps and decision points.

Learning Objectives

• Build Multi-Step Workflows: Create orchestrated processes that chain together SQL queries, vector searches, and AI model calls
• Implement Conditional Logic: Use AI responses to determine next steps in your workflow
• Handle Data Transformations: Process and prepare data between workflow stages

Prerequisites

1. Mission 1 completed (embeddings and vector search with AI_GENERATE_EMBEDDINGS() and External Model MyEmbeddingModel)
2. Mission 2 completed (RAG implementation with stored procedures get_embedding and get_similar_items)
3. Embedding model access (Azure OpenAI text-embedding-3-small)
4. Chat model access (Azure OpenAI gpt-5-mini)

Walkthrough

Choose your preferred language and open the corresponding notebook:

Python

Open the Python notebook (missions/mission3/py/mission3.ipynb) and follow the step-by-step instructions to:

1. Set up your environment and install dependencies
2. Connect to your SQL database using mssql-python
3. Build an AI agent using LangChain that orchestrates SQL queries
4. Use the get_similar_items stored procedure for vector search
5. Run the agent with sample queries

Requirements: See requirements.txt in the py folder.

Key Components:

• AzureAIChatCompletionsModel for chat completions (gpt-5-mini)
• Custom find_products tool that calls the stored procedure

.NET (C#)

Open the .NET notebook (missions/mission3/dotnet/mission3.ipynb) and follow the step-by-step instructions to:

1. Set up your environment and install NuGet packages
2. Connect to your SQL database using Microsoft.Data.SqlClient
3. Build an AI agent using Semantic Kernel that orchestrates SQL queries
4. Register a find_products function as a kernel plugin
5. Run the agent with sample queries

Key Components:

• Semantic Kernel with Azure OpenAI Chat Completion (gpt-5-mini)
• ChatCompletionAgent for agent orchestration
• KernelFunctionFactory for registering SQL query functions

Environment Configuration

Before running either notebook, copy .env.sample to .env and fill in your credentials:

# Database connection
SERVER_CONNECTION_STRING=Server=YOUR_SERVER;Database=YOUR_DATABASE;Trusted_Connection=yes;TrustServerCertificate=yes;

# Azure OpenAI
AZURE_OPENAI_ENDPOINT=https://your-resource.openai.azure.com/
MODEL_ENDPOINT_URL=https://your-resource.cognitiveservices.azure.com/
MODEL_API_KEY=your-api-key-here

Next Steps

After completing this mission, you will have built a robust AI orchestration layer that can handle complex scenarios by integrating SQL data operations with AI capabilities.

Proceed to Mission 4: Building a Full-Stack AI Application to learn how to create a complete application that leverages the workflows you've built in this mission.