VADA – Veeam Architecture Diagram Assistant

So this one’s been in my head for a while. Something I kept running into, whether it’s tenders, customer conversations, or just internal planning, is that someone always needs a diagram. Everyone has their tool of choice: Draw.io, Excalidraw, Visio, and none of them know anything about Veeam. You’re always starting from scratch, hunting for the right icons, trying to remember which components connect to what.

So I thought: what if you could just describe your architecture in plain English, and have something that actually understands Veeam build the diagram for you?

That’s VADA, the Veeam Architecture Diagram Assistant.

What It Is

VADA is an AI-powered diagramming tool built specifically for Veeam architectures. Think Excalidraw, but it knows what a hardened repository is, it knows how a gateway server connects to Veeam Data Cloud Vault, and it knows what lines should be management vs. backup. All of that knowledge comes from a full RAG of the Veeam Help Center documentation running in the background, powered by Claude.

The left pane is a natural language chat. You describe what you want: “single site with a VBR, a hardened repository, and a backup copy job to Veeam Data Cloud Vault” and it goes off, checks it against the documentation, works out what components are required and how they connect, and builds the diagram. It’s not perfect yet, this is very much v1, but it gets you there fast.

How It Works in Practice

In the demo I run through a few scenarios. Starting from a prompt, VADA builds out a primary site with a VBR server, gateway server, backup proxy, hardened repository, and the Veeam Data Cloud Vault tenancy, all with the correct connection lines drawn. Hover over any component and you get a description of what it is and what it does.

From there you can keep chatting to it. Ask it to add VMware to the primary site and it’ll drop in a vSphere cluster with the correct backup connections to the hardened repository. Ask it to add a secondary data center and you can start manually dropping in components too, add a VBR, add a hardened repo, and it’ll prompt you with the right validation connections and apply them automatically.

It also has an architecture analysis panel. This is where it gets interesting: a second AI layer that looks at your completed diagram, checks it against Veeam best practices, and starts making recommendations. In the demo it flags that I’ve connected two VBRs to the design but haven’t added Veeam ONE for centralized monitoring, which is exactly the kind of thing you’d miss when you’re moving fast.

The Other Bits

A few other things I’ve baked in:

  • Rollback: same idea as ChatGPT’s conversation history. If the AI does something you didn’t want, you can roll it back to the previous state before it made changes.
  • BOM generation: it builds out a full bill of materials for each site automatically as you add components.
  • Export options: SVG, PNG, and Draw.io. The Draw.io export is important because not everyone wants to live in VADA, and some people need that extra level of control. The idea was never to replace Excalidraw or Draw.io, it was to get you 80% of the way there in a few minutes, then let you finish it in whatever tool you prefer.
  • Templates: if you don’t know where to start, there are pre-built templates. The multi-site one in the demo shows three data centers (A, B, C) all replicating to Vault with correctly named proxies per site.
  • Save and library: save your design, download the JSON, share it with someone else, or come back to it later.
  • Dark mode: obviously.

Where It’s At

This is v1 and it shows in places. Diagrams can jump around a bit when the AI is making changes, components don’t always land exactly where you want them, and some of the layout cleanup is still manual. That’s fine, the snapping and drag behaviour works well once things settle, and the recompute layout option helps realign things.

The feedback button is genuinely the most important feature right now. If a line connects somewhere it shouldn’t, or a component gets placed in a way that doesn’t make sense, I want to know about it. All of that goes into a log bundle I can review and use to tighten up the model.

If you’ve got thoughts or want to see something added, reach out, keen to hear what people think.