Haymaker Azure Workloads

Azure infrastructure workloads with goal-seeking agents for Agent Haymaker platform

View the Project on GitHub rysweet/haymaker-azure-workloads

Configuration

This page documents all configuration options for Haymaker Azure Workloads, including deployment settings, environment variables, and the workload manifest format.

Deployment Configuration

Configuration is passed via the --config flag when deploying, or through a YAML configuration file.

Options

Option Type Default Required Description
scenario string – Yes Name of the scenario to execute (e.g., linux-vm-web-server)
duration_hours integer 8 No Duration of the operations phase in hours
region string eastus No Azure region for resource deployment
enable_llm boolean false No Enable LLM-powered adaptive agent behavior

Via CLI Flags

haymaker deploy azure-infrastructure \
  --config scenario=linux-vm-web-server \
  --config duration_hours=4 \
  --config region=westus2 \
  --config enable_llm=true

Via YAML File

workload_name: azure-infrastructure
scenario: linux-vm-web-server
duration_hours: 4
region: westus2
enable_llm: true

Environment Variables

Azure Credentials

The workload uses Azure CLI for all resource management. You must be authenticated before deploying.

# Interactive login
az login

# Service principal login (CI/CD)
az login --service-principal \
  --username $AZURE_CLIENT_ID \
  --password $AZURE_CLIENT_SECRET \
  --tenant $AZURE_TENANT_ID

# Set the target subscription
az account set --subscription $AZURE_SUBSCRIPTION_ID
Variable Required Description
AZURE_SUBSCRIPTION_ID Yes Target Azure subscription
AZURE_CLIENT_ID For SP auth Service principal application ID
AZURE_CLIENT_SECRET For SP auth Service principal secret
AZURE_TENANT_ID For SP auth Azure AD tenant ID

LLM Provider

When enable_llm=true, the workload requires LLM provider credentials. The agent uses the agent-haymaker LLM client, which supports multiple providers.

# Azure OpenAI
export LLM_PROVIDER=azure_openai
export AZURE_OPENAI_ENDPOINT=https://my-resource.openai.azure.com/
export AZURE_OPENAI_API_KEY=your-key-here
export AZURE_OPENAI_DEPLOYMENT=gpt-4

# Anthropic
export LLM_PROVIDER=anthropic
export ANTHROPIC_API_KEY=sk-ant-...

# OpenAI
export LLM_PROVIDER=openai
export OPENAI_API_KEY=sk-...
Variable Required When Description
LLM_PROVIDER enable_llm=true LLM provider name (azure_openai, anthropic, openai)
AZURE_OPENAI_ENDPOINT Azure OpenAI Azure OpenAI resource endpoint URL
AZURE_OPENAI_API_KEY Azure OpenAI Azure OpenAI API key
AZURE_OPENAI_DEPLOYMENT Azure OpenAI Model deployment name
ANTHROPIC_API_KEY Anthropic Anthropic API key
OPENAI_API_KEY OpenAI OpenAI API key

Install the AI extras to enable LLM support:

pip install "haymaker-azure-workloads[ai]"

See the agent-haymaker LLM docs for full provider configuration details.

Workload Manifest (workload.yaml)

The workload.yaml file in the repository root defines the workload metadata for the Agent Haymaker platform. This file is read by the haymaker workload install command.

name: azure-infrastructure
version: "0.1.0"
type: runtime
description: "Azure infrastructure workloads with goal-seeking agents"

# Python package info
package:
  name: haymaker-azure-workloads
  entrypoint: haymaker_azure_workloads:AzureInfrastructureWorkload

# Target requirements
targets:
  - type: azure_subscription
    required_roles:
      - Contributor
      - User Access Administrator
    description: "Azure subscription for deploying infrastructure scenarios"

# Configuration schema
config_schema:
  scenario:
    type: string
    required: true
    description: "Name of the scenario to execute"
  duration_hours:
    type: integer
    default: 8
    description: "Duration of operations phase in hours"
  region:
    type: string
    default: "eastus"
    description: "Azure region for deployment"
  enable_llm:
    type: boolean
    default: false
    description: "Enable LLM-powered adaptive agent behavior"

# Available scenarios (populated from scenarios/ directory)
scenarios:
  compute:
    - linux-vm-web-server
    - windows-vm-iis
    - app-service-python
    - azure-functions-http
    - vm-scale-set
  databases:
    - mysql-wordpress
    - postgresql-django
    - cosmos-db-api
  networking:
    - virtual-network
    - load-balancer
    - application-gateway
  security:
    - key-vault-secrets
    - managed-identity
  ai-ml:
    - cognitive-services
    - azure-openai
    - ml-workspace

Manifest Fields

Field Description
name Workload identifier used in haymaker deploy <name>
version Semantic version of the workload package
type Must be runtime for workloads that execute infrastructure
description Human-readable workload description
package.name PyPI package name
package.entrypoint Python import path to the WorkloadBase subclass
targets List of target environments with required Azure roles
config_schema Schema for workload configuration options
scenarios Registry of available scenarios organized by category

Azure Requirements

The target Azure subscription must have the following: