02 · Public app · Personal lab · 2025

SQL Enhancer

A T-SQL bench with two live SQL Server targets.

The public app talks to Azure SQL and to SQL Server on Kubernetes at the same time. You write a statement, ask for an estimated plan, and switch the target. The interface is in Portuguese.

Open SQL Enhancer

SQL Enhancer hub with the object explorer, a SELECT against dbo.Clientes, and an empty result pane. — The live hub. Left: object explorer for the Azure SQL simulator and the k3s instance. Center: the statement, execute, estimated plan, and the target switch. The k3s instance is sometimes idle and reconnecting. That state is visible on purpose.

How SQL Enhancer is wired

What you can do in it

Pick a part of the application. The note says what it is for, what you should see, and one thing worth trying.

What it does

Two targets

Keeps Azure SQL and SQL Server on K3s in one session, with a status for each.

On the screen: A header chip when both engines answer, plus Alive on Azure SQL and a separate state for the local k3s pod.
Try this: Switch the statement target between Azure and k3s before you execute. If k3s says it is idle or reconnecting, that is the lab telling you the pod is not warm.

The running bench

This frame is the public application, not a picture of it. The interface is in Portuguese. Execute is Executar. The Kubernetes target is sometimes idle.

Running application

Open in a new tab

What the application is

SQL Enhancer is a Python and FastAPI service behind a React interface. Most of my paid database work has been SQL Server, from Kyndryl’s corporate estates to Azure, and this is where I keep that practice sharp: write T-SQL, look at the shape of a query, and also do the administration that sits around it.

GitHub Actions runs the API tests and builds the frontend. The interface is deployed on Vercel. The Kubernetes instance stays in the lab, which is why it is not always warm.

What you can do in it

The page below embeds the running application. The shell is in Portuguese: Executar runs the statement, Plano estimado asks for a plan, and the target buttons choose Azure or k3s. Results, I/O metrics, the .sqlplan and messages are separate tabs, so a clean grid is not the only answer the engine gave you.

What this does not prove

The tutor is a practice aid. I do not claim a model replaces an execution plan, and I do not attach a client metric to this app. The source repository is not public yet.