Glyphra is an AI-powered CLI tool that explains Terraform plans — because reading giant diffs should not be a full-time job.
Built with Go, powered by Hugging Face, and documented via its own blog, Glyphra is your dev sidekick when infra gets noisy.
- Reads Terraform plan output
- Summarizes it using an LLM (like
flan-t5-large) - Outputs a human-readable summary
- Can generate Markdown blog posts from summaries
You can build it from source:
git clone https://github.com/danny-molnar/glyphra.git
cd glyphra
make build
./glyphra --help- Go 1.21+
- Hugging Face API token (free tier is fine)
You can store the token in .envrc or export it directly:
export HUGGINGFACE_TOKEN=your_token_heremake plan # Generates a fake Terraform plan
make run # Runs Glyphra on plan.outmake blog # Summarizes and saves to Markdown blog postmake test— run Go testsmake lint— run basic lintersmake clean— remove build artifactsmake docker-build— build Docker image
Check out the dev journey: 👉 https://danny-molnar.github.io/glyphra
Source files live in the /docs folder and use the Minimal Mistakes theme with a dark skin.
Because infrastructure shouldn't be cryptic. And AI can make tooling more helpful and more fun.
- Support for
terraform show -json - GitHub PR summarization (
glyphra pr) - Local AI model support via Ollama
- Auto-tagging and blog post scheduling
Built by @danny-molnar Infra dev turned CLI engineer with a love for 🍜, AI, and dark mode everything.
MIT License — use it, remix it, build cool tools with it. See LICENSE for details.