-
Notifications
You must be signed in to change notification settings - Fork 2
Installation
Before installing az prototype, ensure the following tools are available:
| Tool | Minimum Version | Required For | Install |
|---|---|---|---|
| Python | 3.9+ (3.10-3.12 tested in CI) | Extension runtime | python.org |
| Azure CLI | 2.50.0+ | All commands | Install Azure CLI |
GitHub CLI (gh) |
2.30.0+ | Copilot/GitHub Models auth, backlog push | cli.github.com |
| Terraform | 1.14.0+ | IaC generation and deployment (if --iac-tool terraform) |
Terraform Install |
| Bicep | (bundled with Azure CLI) | IaC generation and deployment (if --iac-tool bicep) |
Included with Azure CLI |
The governor agent uses pre-computed neural embeddings for built-in policy retrieval. These are generated at build time and shipped with the wheel -- no extra install needed, works on all platforms including Windows 32-bit.
For custom policies (added to .prototype/policies/), TF-IDF is used by default. Non-Windows users who want neural-quality embeddings for custom policies can optionally install:
pip install sentence-transformersNote: sentence-transformers requires torch, which has no 32-bit Windows wheels. This optional install is only for macOS/Linux or 64-bit Windows Python environments.
You also need access to at least one AI provider. See the AI Provider Setup section below.
Download the latest .whl file from the Releases page, then install it as an Azure CLI extension:
az extension add --source azext_prototype-0.2.1b6-py3-none-any.whlIf you have a previous version installed, remove it first (see Upgrade).
az prototype --helpYou should see the available subcommands: init, design, build, deploy, status, and backlog.
The extension supports three AI providers. Configure one during az prototype init with --ai-provider.
Requires an active GitHub Copilot subscription (Individual, Business, or Enterprise). The extension authenticates through the GitHub CLI and validates your Copilot license during init. Default model: claude-sonnet-4.5.
See AI Providers for detailed setup.
Requires a GitHub Personal Access Token with the models scope. Set up via gh auth login or by providing the token in your project secrets. Default model: gpt-4o.
See AI Providers for detailed setup.
Requires an Azure OpenAI resource with a deployed model, plus the endpoint URL and API key. Configure the endpoint and deployment name during init or in prototype.yaml. Default model: gpt-4o.
# prototype.yaml
ai:
provider: azure-openai
azure_openai:
endpoint: https://your-resource.openai.azure.com/
deployment: gpt-4oAPI keys are stored in prototype.secrets.yaml (git-ignored).
See AI Providers for detailed setup.
Remove the existing extension and install the new version:
az extension remove --name prototype
az extension add --source azext_prototype-<new-version>-py3-none-any.whlYour project files (prototype.yaml, generated code) are not affected by extension upgrades.
az extension remove --name prototypeThis removes the CLI extension only. Project directories and configuration files created by az prototype init remain on disk.
Getting Started
Stages
Interfaces
Configuration
Agent System
Features
- Backlog Generation
- Cost Analysis
- Error Analysis
- Docs & Spec Kit
- MCP Integration
- Knowledge System
- Escalation
Quality
Help
Policies — Azure
AI Services
Compute
Data Services
- Azure SQL
- Backup Vault
- Cosmos Db
- Data Factory
- Databricks
- Event Grid
- Event Hubs
- Fabric
- IoT Hub
- Mysql Flexible
- Postgresql Flexible
- Recovery Services
- Redis Cache
- Service Bus
- Stream Analytics
- Synapse Workspace
Identity
Management
Messaging
Monitoring
Networking
- Application Gateway
- Bastion
- CDN
- DDoS Protection
- DNS Zones
- Expressroute
- Firewall
- Load Balancer
- Nat Gateway
- Network Interface
- Private Endpoints
- Public Ip
- Route Tables
- Traffic Manager
- Virtual Network
- Vpn Gateway
- WAF Policy
Security
Storage
Web & App
Policies — Well-Architected
Reliability
Security
Cost Optimization
Operational Excellence
Performance Efficiency
Integration