Phase 1 · Building with LLMs · Prompting
Structured output
Get reliable JSON back that your code can use.
Open this lesson Create a free account
Why it matters
Most LLM features that do real work end with code reading the model's answer: a lead gets added to the CRM, a review gets a tag, a product gets a category. That only works if every reply arrives in the exact shape your code expects. Structured output is how you get that guarantee. Validating what comes back is what keeps a well-formed but wrong answer from flowing into your systems.
What you'll be able to do
- Explain why asking for JSON in the prompt isn't enough when code depends on the reply
- Describe the data you want back as a JSON schema
- Request schema-constrained output with `output_config.format` and parse the reply
- Validate parsed data against business rules, and handle replies you can't use
Before this lesson
Next in this phase: Prompt chaining