Basketball Coach AI Agent Demo Repository

This repository contains multiple basketball coaching AI agent implementations demonstrating different Azure AI technologies and frameworks.

Applications Overview

App1 - Azure AI Foundry Agent Service

Technology: Azure AI Foundry
Framework: Azure AI Agent Service
Description: Demo code for Azure AI Agent Service intro post. Basic agent implementation using Azure AI Foundry's native agent capabilities.
Blog Post: https://beyondelastic.github.io/posts/agent/

App2 - Semantic Kernel Multi-Agent System

Technology: Azure AI Foundry
Framework: Semantic Kernel
Description: Multi-agent basketball coaching system using Semantic Kernel's agent orchestration capabilities.
Blog Post: https://beyondelastic.github.io/posts/multi-agent/

App2-Modular - Modular Semantic Kernel Implementation

Technology: Azure AI Foundry
Framework: Semantic Kernel
Description: Refactored version of App2 with improved modularity and separation of concerns using Semantic Kernel agents.

App3 - Semantic Kernel Function Calling

Technology: Azure OpenAI
Framework: Semantic Kernel
Description: Demonstrates Semantic Kernel function calling and plugins for basketball coaching scenarios.
Blog Post: https://beyondelastic.github.io/posts/skfc/

App4 - Advanced Azure AI Foundry Features

Technology: Azure AI Foundry
Framework: Azure AI Agent Service
Description: Advanced implementation showcasing Azure AI Foundry features like Bing Grounding Tool, Connected Agent Tool, and File Search capabilities.

App5 - Microsoft Agent Framework Implementation

Technology: Azure AI Foundry
Framework: Microsoft Agent Framework
Description: Clean, modular multi-agent basketball coaching system using the new Microsoft Agent Framework with proper resource management and workflow orchestration.

Clean-up - Azure AI Foundry Resource Management

Technology: Azure AI Foundry
Framework: Azure AI Projects SDK
Description: Utility script for cleaning up Azure AI Foundry resources (agents, threads, files, vector stores) to manage costs and maintain clean environments.

---

Setup Instructions

1. Clone the repository:

   git clone https://github.com/yourusername/basketball-ai-agent.git
   cd basketball-ai-agent

2. Create and activate a virtual environment (optional but recommended):

   python3 -m venv .venv
   source .venv/bin/activate

3. Install dependencies:

   • Each app has its own requirements.txt file. Install dependencies for the app you want to run:
     • For App 1:

       pip install -r src/app1/requirements.txt

     • For App 2:

       pip install -r src/app2/requirements.txt

     • For App 2-Modular:

       pip install -r src/app2-modular/requirements.txt

     • For App 3:

       pip install -r src/app3/requirements.txt

     • For App 4:

       pip install -r src/app4/requirements.txt

     • For App 5:

       pip install -r src/app5/requirements.txt

     • For Clean-up:

       pip install -r src/clean-up/requirements.txt

   • (If a requirements.txt is missing, install dependencies as needed for that app, e.g. semantic-kernel, openai, etc.)

4. Set up environment variables:

   • Set the following variables in your shell or .env file as needed for each app:
     • App 1 (Azure AI Foundry):
       • AZURE_AI_AGENT_PROJECT_CONNECTION_STRING=""
       • AZURE_AI_AGENT_MODEL_DEPLOYMENT_NAME=""
       • NBA_CSV_FILE_PATH="../data/nba3p.csv"
     • App 2 (Azure AI Foundry + Semantic Kernel):
       • AZURE_AI_AGENT_PROJECT_CONNECTION_STRING=""
       • AZURE_AI_AGENT_MODEL_DEPLOYMENT_NAME=""
     • App 2-Modular (Azure AI Foundry + Semantic Kernel):
       • AZURE_AI_AGENT_PROJECT_CONNECTION_STRING=""
       • AZURE_AI_AGENT_MODEL_DEPLOYMENT_NAME=""
     • App 3 (Azure OpenAI + Semantic Kernel):
       • AZURE_OPENAI_ENDPOINT=""
       • AZURE_OPENAI_API_KEY=""
       • AZURE_OPENAI_DEPLOYMENT=""
     • App 4 (Azure AI Foundry):
       • AZURE_AI_AGENT_PROJECT_CONNECTION_STRING=""
       • AZURE_AI_AGENT_MODEL_DEPLOYMENT_NAME=""
     • App 5 (Azure AI Foundry + Agent Framework):
       • AZURE_AI_AGENT_PROJECT_CONNECTION_STRING=""
       • AZURE_AI_AGENT_MODEL_DEPLOYMENT_NAME=""
     • Clean-up (Azure AI Foundry):
       • AZURE_AI_AGENT_PROJECT_CONNECTION_STRING=""

Usage

• App 1 (Azure AI Foundry Agent Service):

  • Navigate to src/app1/ and run the main script:

    cd src/app1
    python app1.py

• App 2 (Semantic Kernel Multi-Agent):

  • Navigate to src/app2/ and run the app:

    cd src/app2
    python app2.py

• App 2-Modular (Modular Semantic Kernel):

  • Navigate to src/app2-modular/ and run the main app:

    cd src/app2-modular
    python main.py

• App 3 (Semantic Kernel Function Calling):

  • Navigate to src/app3/ and run the app:

    cd src/app3
    python app3.py

• App 4 (Advanced Azure AI Foundry):

  • Navigate to src/app4/ and run the app:

    cd src/app4
    python app4.py

• App 5 (Microsoft Agent Framework):

  • Navigate to src/app5/ and run the app:

    cd src/app5
    python app5.py

  • Or with custom task:

    cd src/app5
    python app5.py "We need help with our defense against pick and roll plays"

• Clean-up (Resource Management):

  • Navigate to src/clean-up/ and run the cleanup script:

    cd src/clean-up
    python clean_up.py

  • Use dry-run mode to preview deletions:

    python clean_up.py --dry-run

Data

• The data/ folder contains basketball datasets (e.g., nba3p.csv) used by app1.

Technology Stack Summary

App	Technology	Framework	Use Case
App1	Azure AI Foundry	Azure AI Agent Service	Basic agent implementation
App2	Azure AI Foundry	Semantic Kernel	Multi-agent orchestration
App2-Modular	Azure AI Foundry	Semantic Kernel	Modular multi-agent system
App3	Azure OpenAI	Semantic Kernel	Function calling & plugins
App4	Azure AI Foundry	Azure AI Agent Service	Advanced features (Bing, File Search)
App5	Azure AI Foundry	Microsoft Agent Framework	Clean workflow orchestration
Clean-up	Azure AI Foundry	Azure AI Projects SDK	Resource management utility

Notes

• Make sure you have the necessary Azure AI Foundry or Azure OpenAI access and correct API keys/connection strings.
• For Azure AI Foundry apps, you'll need a project connection string.
• For Azure OpenAI apps (App3), you'll need direct Azure OpenAI service credentials.
• Each app demonstrates different approaches to building basketball coaching AI agents.
• For more details, refer to the code comments and the referenced blog posts above.