aiac

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 architecture uses Go's interface-based polymorphism to achieve interchangeability without conditional logic scattered throughout the codebase.

AIAC uses a TOML configuration file (located at ~/.config/aiac/aiac.toml by default) to define multiple named backends, each with provider-specific settings, API keys, and default models. The configuration system supports environment variable substitution and custom config paths via CLI flags, enabling both local development workflows and containerized/CI deployments. The configuration loader parses the TOML structure into Go structs that are validated and used to instantiate the appropriate backend at runtime.

AIAC integrates with OpenAI's API by implementing the Backend interface for OpenAI models (GPT-3.5, GPT-4, etc.). The backend handles authentication via API keys, request formatting, streaming response handling, and error management. Users can select specific OpenAI models via configuration, enabling cost/performance tradeoffs. The implementation uses OpenAI's official Go client library for API communication.

AIAC integrates with AWS Bedrock by implementing the Backend interface for Bedrock's managed LLM service. The backend handles AWS authentication via IAM credentials, request formatting for Bedrock's API, and response parsing. Users can access multiple LLM providers (Anthropic Claude, Cohere, etc.) through Bedrock's unified API. This enables organizations with existing AWS infrastructure to leverage Bedrock without managing separate API accounts.

AIAC integrates with Ollama, an open-source tool for running LLMs locally. The Ollama backend implementation communicates with a local Ollama instance via HTTP API, enabling code generation without sending prompts to external services. Users can run open-source models (Llama 2, Mistral, etc.) locally, providing complete data privacy and no API costs. This backend is ideal for organizations with strict data governance requirements or offline environments.

AIAC accepts natural language prompts describing infrastructure requirements and generates production-ready IaC code by sending the prompt to an LLM backend with provider-specific context. The system uses prompt engineering to guide the LLM toward generating valid Terraform, CloudFormation, Pulumi, or other IaC syntax. The generated code is returned as plain text that users can validate, modify, and commit to version control. This capability bridges the gap between human intent and machine-readable infrastructure definitions.

AIAC generates configuration files (Dockerfiles, Kubernetes manifests, GitHub Actions workflows, Jenkins pipelines) and CI/CD pipeline definitions from natural language descriptions. The LLM uses provider-specific knowledge to generate syntactically correct YAML, JSON, or Dockerfile content. This capability extends beyond infrastructure code to cover the operational and deployment layers, enabling users to define entire deployment pipelines through conversational prompts.

AIAC generates Open Policy Agent (OPA) Rego policies and other policy-as-code artifacts from natural language descriptions of compliance or security requirements. The LLM understands OPA syntax and generates policies that can be evaluated against infrastructure definitions, Kubernetes resources, or other policy-evaluable objects. This enables users to express security policies in plain English and automatically generate the corresponding Rego code.

AIAC generates utility scripts (Python, Bash, Go) and database queries (SQL) from natural language descriptions of desired functionality. The LLM generates syntactically correct, executable code that performs specific operational tasks. This capability extends AIAC beyond infrastructure and configuration to cover operational automation and data querying, enabling users to generate one-off scripts without writing code manually.

AIAC provides an interactive mode where users can view generated code, request modifications, and regenerate code based on feedback without restarting the CLI. The system maintains the conversation context and allows users to iteratively refine generated code through natural language instructions. This interactive loop enables rapid prototyping and refinement of infrastructure and configuration code.

AIAC provides a command-line interface that accepts natural language prompts and streams generated code to stdout with optional formatting. The CLI handles argument parsing, backend selection, and output formatting. Generated code is displayed in real-time as the LLM produces it, and users can pipe output to files or other tools. The CLI supports flags for backend selection, config path, and output formatting options.

AIAC can be deployed as a Docker container with pre-configured LLM backends, enabling users to run AIAC without installing Go or managing local configuration. The container image includes the AIAC binary and can be configured via environment variables or mounted config files. This enables easy integration into containerized workflows and cloud-native environments.

AIAC exposes a Go library (libaiac) that allows developers to integrate code generation capabilities directly into Go applications. The library provides functions to initialize backends, send prompts, and retrieve generated code programmatically. This enables developers to build custom tools, IDE extensions, or applications that leverage AIAC's code generation without using the CLI.