Sometimes infrastructure is too verbose to be meaningful.

As someone pivoting from platform engineering to software development, I wanted to build something real — something that bridges both worlds. That became Glyphra.

🤔 The Problem

Terraform plans are powerful… but also overwhelming. They’re filled with diffs, noise, and technical details — but little explanation of intent.

What does this plan actually do? Why is this change happening?

That’s the gap I wanted to close — with AI.

🤖 The Solution

I wired up my CLI tool (written in Go) to Hugging Face’s flan-t5-large model.

Now, Glyphra can do something like this:

glyphra plan --input plan.out

And produce a human-readable summary of what’s going on in a Terraform plan.

It’s not just automation — it’s interpretation.

🧠 What I Learned

  • How to integrate an LLM using Go
  • How to manage secrets securely using .envrc
  • How to test Go code and mock AI responses
  • How to blog directly from the CLI with GitHub Pages + Jekyll

🛠 What’s Next

  • Parse and explain terraform show -json output
  • Add a glyphra pr command to summarize GitHub PRs
  • Chunk large plans and explain them in parts
  • Try offline local models (Ollama? GGML?)

This was just the first step. Glyphra is going to get smarter — and so am I.

Updated: