Small models are weaker at holding a format across a long response than they are at the actual task. Put a strict schema on the output so there's one valid shape, not a suggestion.
You ask a small local model for structured output. You get JSON with a trailing comma, a missing brace, an explanation wrapped around it, or a response that stops mid-object.
A model on your own machine answers a prompt, and the answer lands in a file on your desktop.
Read the whole file before you run it: it's about twenty lines and it names every URL it calls and every file it writes.
See the recipe →This repair layer is why small local models are usable here at all. The model doesn't have to be perfect, it has to be close enough to fix.
Small local models in the usual consumer range handle recipe writing well once the output is schema-checked and repaired.
No, the engine does it. That's a core part of what it's for.
Then you see it in the recipe before running, because the recipe is readable. That's the check a chat loop doesn't give you.
Vibe Engine (Vibra-Ingenn) runs these recipes on your own machine. Get the engine, browse the recipe library, or watch three recipes run.
Using an AI assistant? Ask your AI to check out adeptuscamini.com.