Structured output

By default an LLM node hands back free text. Often you want structured data instead — a category, a score, a yes/no — so the rest of the workflow can branch on it, pass it along, or store it. Structured output turns the model’s reply into typed fields you can wire like any other value.

Define the shape

Add fields to the LLM node’s Output Schema in the config panel — each with a name and a data type. The node asks the model to return JSON matching that shape, and Wayflow parses and validates the reply before it leaves the node. (The LLM page shows the Output Schema in the editor.)

Each field becomes a port

A schema with a category (text) and an urgency (number) turns the model’s answer into two typed output ports:

{ "category": "billing", "urgency": 3 }

Now category and urgency are separate, typed values — wire category into a Conditional to branch the run, pass urgency to the next step, collect either in an Output field. With a single field, the port simply carries that value directly — no wrapper object to unpack.

Works with any provider

Some model backends enforce a JSON schema natively; others can’t. Wayflow handles both — it describes your schema to the model and validates the result — so structured output behaves the same whichever provider you connect. See Providers & models.

When the model doesn’t comply

If the reply isn’t valid JSON, is missing a field, or returns the wrong shape, the run fails with a clear error rather than passing malformed data downstream — so you find out immediately instead of debugging a silent bad value three nodes later.

Where next

• LLM — where the Output Schema lives
• Conditional — branch on a structured field
• On a server — read the result from a headless run