// the find
gofireflyio/aiac
Artificial Intelligence Infrastructure-as-Code Generator.
aiac is a CLI and Go library that sends natural-language prompts to OpenAI, Amazon Bedrock, or Ollama and extracts the generated IaC/config/script from the response. You describe what you want in plain English, it returns runnable Terraform, Kubernetes manifests, Dockerfiles, etc. Aimed at DevOps engineers who want a faster first draft rather than a complete automation pipeline.
The named-backend config model in v5 is genuinely well-designed — you can point the same tool at GPT-4, a local Ollama instance, and a Bedrock staging account without changing your workflow. The interactive REPL with clipboard, file save, and follow-up prompts is more useful than a one-shot generator. The library surface is clean: `Chat()` returns a stateful object you `Send()` to, which makes it trivial to embed in your own Go tooling. Multi-provider support without a cloud middleware layer keeps it self-hostable.
Generated IaC goes straight to a file with no validation step — no `terraform validate`, no schema check, nothing to tell you the output is syntactically correct before you commit it. There is no prompt templating or reuse system, so every invocation starts from scratch; teams can't share or version their prompts. The `--list-models` output may include image/video models that will error at runtime, with no filtering or warning. Last meaningful activity was over a year ago and the v5 migration broke the env-var workflow that CI pipelines relied on, with no migration path other than switching to a config file.