JJiro Labs

AI developer tooling · Jiro Labs internal

Zero-click AI vision routing for a text-only workflow

Fixed a broken image workflow to zero extra steps with AI routing.

vision-helper — watch mode

$ npm run vision

› watching uploads/ for new images…

› screenshot-1142.png detected

› routing to vision model · claude-sonnet

› analysis saved → uploads/analysis/…md

› done. zero manual steps.

$ # the workflow never stopped

The situation

An internal AI-assisted workflow broke whenever images were involved: the underlying model was text-only, and every screenshot had to be analyzed through a separate tool by hand.

What we built

  • A lightweight AI tool that automatically detects image content in a workflow
  • Transparent routing to a vision-capable model via a single API key
  • A watcher mode: drop a file, get an analysis — no commands, no context switching
  • A transparent proxy variant that inspects requests and routes images automatically

The outcome

What used to require copying files, switching tools, and pasting results became zero extra steps — the workflow handles images invisibly.

0

Extra manual steps after routing

Auto

Image detection inside the existing workflow

1

API key to route any vision-capable model

StackNode.jsOpenRouterClaude vision

Start a project

Have a project in mind?

Let's build a working version of it — not just a proposal.

Prefer to think first? Tell us what you're trying to solve and we'll come back with a clear next step — no sales runaround.