genaiops-azureaisdk-template

Local Environment Setup with Conda

Prerequisites

• Python 3.9 or higher
• Conda package manager
• Azure CLI installed
• Git installed

Step 1: Using Conda Environment Manager

You can use Conda for managing your Python environment. This guide will focus on Conda setup.

Conda Environment Setup

# Create a new conda environment with Python 3.9
conda create -n myenv python=3.9

# Activate the environment
conda activate myenv

Step 2: Azure Authentication

# Login to Azure
az login

# Verify your login
az account show

Step 3: Install Required Packages

# Install execution requirements
python -m pip install -r ./.github/requirements/execute_job_requirements.txt

# Install validation requirements
python -m pip install -r ./.github/requirements/build_validation_requirements.txt

# Install the package in editable mode
python -m pip install -e .

Step 4: Environment Configuration

Create and configure your environment file:

# Create .env file from template
cp .env.sample .env

Environment Variables Configuration

Variable	Description	Where to Find
AZURE_OPENAI_API_VERSION	Azure OpenAI API version	Use the latest stable version from Azure OpenAI documentation e.g. 2024-02-15-preview
AZURE_AI_CHAT_ENDPOINT	Azure OpenAI Chat API endpoint	Found in Azure Portal → Azure OpenAI resource → Keys and Endpoint. e.g. https://your-resource.openai.azure.com/openai/deployments/your-deployed-model
AZURE_AI_CHAT_KEY	Azure OpenAI Chat API key	Found in Azure Portal → Azure OpenAI resource → Keys and Endpoint e.g. xxxxxx
AOAI_API_KEY	Azure OpenAI API key	Found in Azure Portal → Azure OpenAI resource → Keys and Endpoint e.g. xxxxxx
GPT4O_DEPLOYMENT_NAME	GPT-4 model deployment name	Found in Azure Portal → Azure OpenAI resource → Model Deployments
GPT4O_API_KEY	GPT-4 API key	Found in Azure Portal → Azure OpenAI resource → Keys and Endpoint
SUBSCRIPTION_ID	Azure subscription ID	Found in Azure Portal → Subscriptions
RESOURCE_GROUP_NAME	Resource group name	Name of your Azure resource group
PROJECT_NAME	AI Foundry project name	Name of your AI Foundry project in Azure
USER_CLIENT_ID	User-defined managed identity (Optional)	Used only for online evaluations. Create in Azure Portal → Managed Identities
CONNECTION_STRING	AI Foundry project connection string	Found in AI Foundry project settings (Must be enclosed in quotes)
PROMPTY_FILE	Path to prompt template file	Local path to your prompt template file

Example .env File

AZURE_OPENAI_API_VERSION=2024-02-01
AZURE_AI_CHAT_ENDPOINT=https://your-resource.openai.azure.com/openai/deployments/your-deployed-model
AZURE_AI_CHAT_KEY=your-chat-key
AOAI_API_KEY=your-openai-key
GPT4O_DEPLOYMENT_NAME=your-deployment-name
GPT4O_API_KEY=your-gpt4-key
SUBSCRIPTION_ID=your-subscription-id
RESOURCE_GROUP_NAME=your-resource-group
PROJECT_NAME=your-project-name
USER_CLIENT_ID=your-client-id
CONNECTION_STRING="your-connection-string"
PROMPTY_FILE=xxxxx.prompty

Additional configuration for Semantic Kernel agent with AI Foundry

AZURE_AI_AGENT_PROJECT_CONNECTION_STRING=”your-connection-string” AZURE_AI_AGENT_MODEL_DEPLOYMENT_NAME=your-deployment-name AZURE_AI_AGENT_ENDPOINT=https://xxxdomain.openai.azure.com/ AZURE_AI_AGENT_SUBSCRIPTION_ID==your-subscription-id AZURE_AI_AGENT_RESOURCE_GROUP_NAME=your-resource-group AZURE_AI_AGENT_PROJECT_NAME=your-project-name

AZURE_AI_AGENT_ENDPOINT value is available from Azure AI Foundry Overview page under Endpoints and keys section associated with Azure OpenAI Service key

Important Notes

• The CONNECTION_STRING value must be enclosed in quotes in the .env file
• The CONNECTION_STRING value is available from project overview page of Azure AI Foundry under Project connection string
• Ensure the PROMPTY_FILE value is correct and points to a valid file name. prompty files are available under math_coding/flows/math_code_generation folder
• USER_CLIENT_ID is optional and only required for online evaluations
• Keep your .env file secure and never commit it to version control
• Regularly update your API versions to ensure compatibility
• Multiple examples are shown by means of using different endpoint types.

Verification

To verify your setup:

1. Ensure your conda environment is activated
2. Verify Azure login status
3. Test package installation by running a simple command
4. Check environment variables are loaded correctly

Troubleshooting

• If packages fail to install, try updating pip: python -m pip install --upgrade pip
• For Azure login issues, clear your credentials and login again
• Check file permissions if you have issues creating the .env file
• Verify your Python version matches the requirements

This site is open source. Improve this page.