Can Small Models Write and QA Client-Side A/B Test Code? I Built a Workflow to Find Out

Can we create a workflow that reliably writes JavaScript code for our client-side A/B tests? One that not only writes the code, but also automatically handles the QA for common errors and then self-heals to produce something more reliable? And can we do all of this efficiently, using small models (so sustainable) and without maxing