Smart Infra: AI-Driven Cloud Infrastructure Design and IaC Auto-Generation System

Smart Infra is an open-source project that leverages AI Agent Skill workflows to enable natural language interaction for cloud infrastructure design. It covers the entire process from requirement clarification to Terraform code generation, supports AWS/GCP/Azure multi-cloud environments, and automatically generates architecture visualization documents, significantly lowering the entry barrier for Infrastructure as Code (IaC).

Traditional cloud architecture design involves multiple steps: requirement collection (with business teams), architecture design (selecting cloud services/patterns), document writing (architecture diagrams/specs), code implementation (Terraform/CloudFormation), and review/validation. This process requires collaboration across roles (architects, ops, security engineers) and can take weeks, becoming a bottleneck for startups or small enterprises without dedicated cloud experts.

Smart Infra encapsulates professional skills as AI Agent Skills instead of building complex low-code platforms. Its core concepts include:

• Dialogue as design: Clarify requirements via natural language interaction.
• Automated generation: Directly output production-grade Terraform code from requirements.
• Visualization documents: Auto-generate architecture diagrams and explanatory docs.
• Multi-cloud support: Compatible with AWS, GCP, and Azure.

Smart Infra consists of four core Agent Skills:

1. cloud-architect: Guides users to clarify requirements (availability, traffic, budget, security, tech preferences) and outputs structured requirements.md.
2. iac-generate: Reads requirements to generate modular Terraform code (modular design, multi-cloud abstraction, built-in best practices, custom templates) with a standard directory structure.
3. arch-diagram: Parses Terraform configs to generate Mermaid diagrams (quick preview) and Python diagrams (high-quality PNG) for architecture visualization.
4. infra-flow: Orchestrates the end-to-end workflow from user queries to final outputs.

Usage:

1. One-stop workflow (recommended): Use infra-flow Skill to input natural language (e.g., "design a high-availability employee management system") for automated requirement clarification, code generation, and visualization.
2. Step-by-step: Use cloud-architect → iac-generate → arch-diagram for more control.

Technical Details:

• Skill structure: Stored in .agents/skills/ with metadata, prompts, and tools/templates.
• Lazy dependency management: Auto-installs missing tools (e.g., diagrams library) on first execution.
• Template extension: Custom Terraform templates can be added to templates/terraform/common-patterns/.

Multi-cloud Support: Achieved via conditional logic and abstraction layers (e.g., network/compute resource mappings across AWS/GCP/Azure).

Output Management: Results are stored in project-isolated directories under output/ (e.g., output/ecommerce-platform/), which is excluded from version control via .gitignore to avoid sensitive config leaks.

Scenarios:

• Startups: Fast production environment setup without cloud experts.
• Enterprise IT: Standardize infrastructure templates for consistency.
• Consulting: Quick proposal generation (architecture diagrams, cost estimates).
• Education: Interactive learning tool for Terraform and cloud architecture.

Limitations:

• Complex compliance requires manual review.
• No direct cost estimation (use tools like Infracost).
• Does not manage Terraform state files.
• Templates need updates for cloud API/provider changes.

Smart Infra represents a new direction for AI-assisted infrastructure management:

• Agent Skill architecture is more flexible than low-code platforms.
• Natural language interaction lowers IaC barriers but should not replace underlying knowledge.
• Modular design ensures maintainable code.
• Visualization improves communication efficiency.
• Multi-cloud abstraction balances generality and platform-specific features.

For teams lacking IaC expertise, Smart Infra provides a practical starting point. The "dialogue as infrastructure" model may become a standard for cloud resource management as AI evolves.