AI-powered Infrastructure-as-Code Generator

AIAC implements a Backend interface abstraction layer that enables seamless switching between OpenAI, AWS Bedrock, and Ollama LLM providers through a single unified API. Each backend implementation handles provider-specific authentication, request formatting, and response parsing, allowing the core library to remain agnostic to the underlying LLM provider. This layered architecture decouples the code generation logic from provider-specific details, enabling users to swap backends via configuration without code changes.

AIAC accepts plain English descriptions of infrastructure requirements and translates them into production-ready IaC templates through LLM prompting. The system constructs context-aware prompts that guide the LLM toward generating syntactically correct, idiomatic code for target frameworks like Terraform, CloudFormation, or Pulumi. The generation process handles streaming responses from the LLM backend, formats output, and presents results through an interactive CLI interface where users can refine or regenerate code.

AIAC implements an AWS Bedrock backend that integrates with AWS's managed LLM service, supporting multiple foundation models (Claude, Llama, Mistral, etc.) through a unified interface. The backend handles AWS authentication via credentials or IAM roles, manages Bedrock API calls, and abstracts model-specific differences. This enables enterprise users to leverage AWS's compliance, security, and cost management features while accessing multiple LLM providers.

AIAC implements an Ollama backend that connects to locally-running Ollama instances, enabling infrastructure code generation using open-source models (Llama 2, Mistral, etc.) without sending data to cloud providers. The backend communicates with Ollama's REST API, handles model loading and inference locally, and provides complete data privacy. This enables organizations with strict data residency or privacy requirements to generate infrastructure code entirely on-premises.

AIAC generates configuration files (Dockerfiles, Kubernetes manifests, docker-compose) and CI/CD pipeline definitions (GitHub Actions, Jenkins, GitLab CI) from English descriptions. The system uses LLM prompting to produce framework-specific configuration syntax, handling the nuances of each format (YAML indentation for Kubernetes, Dockerfile layer optimization, GitHub Actions workflow syntax). Generated configurations are returned as complete, ready-to-use files that can be immediately integrated into projects.

AIAC generates Open Policy Agent (OPA) policies from natural language descriptions of compliance and governance requirements. The system translates English specifications (e.g., 'enforce readiness probes on all Kubernetes deployments') into Rego policy language, enabling users to define infrastructure guardrails without learning OPA syntax. Generated policies can be immediately integrated into Kubernetes admission controllers or policy evaluation pipelines.

AIAC generates operational scripts (Python, Bash, SQL) and command-line utilities from English descriptions of infrastructure tasks. The system produces executable code for common operations like port scanning, database queries, log analysis, and resource enumeration. Generated scripts are returned as complete, runnable code that can be immediately executed or integrated into automation pipelines.

AIAC provides a configuration system using TOML files that allows users to define multiple named LLM backends with provider-specific settings, credentials, and default models. The configuration loader reads from ~/.config/aiac/aiac.toml (or custom path via --config flag) and instantiates the appropriate backend implementation at runtime. This enables users to manage multiple LLM provider configurations in a single file and switch between them via CLI flags without code changes.

AIAC provides a command-line interface that accepts user prompts, streams LLM responses in real-time, and presents an interactive menu for code refinement and regeneration. The CLI handles streaming from the configured backend, formats output for readability, and allows users to request modifications, regenerate with different parameters, or save generated code. The streaming architecture enables real-time feedback without waiting for complete LLM response generation.

AIAC is distributed as a Docker image that includes the compiled binary and can be deployed with pre-configured LLM backends through environment variables or mounted configuration files. The Docker image enables users to run AIAC without installing Go or managing dependencies, making it suitable for CI/CD pipelines, containerized workflows, and cloud deployments. The image supports passing configuration via environment variables or mounting TOML files at runtime.

AIAC exposes a Go library interface (libaiac) that allows developers to embed infrastructure code generation directly into Go applications. The library provides functions to initialize backends, construct prompts, and execute code generation programmatically without using the CLI. This enables custom tooling, automation frameworks, and applications to leverage AIAC's code generation capabilities through a clean Go API.

AIAC implements a dedicated OpenAI backend that communicates with OpenAI's API using the official Go client library. The backend handles authentication via API keys, supports streaming responses for real-time output, and allows users to select specific models (GPT-4, GPT-3.5-turbo, etc.) through configuration. The implementation manages request formatting, error handling, and response parsing specific to OpenAI's API contract.