.jpg?width=1920&height=1920&fit=bounds&quality=80&format=jpg&auto=webp#)
.jpg?width=1920&height=1920&fit=bounds&quality=80&format=jpg&auto=webp#){kind=tracking}
/https://tf-cmsv2-smithsonianmag-media.s3.amazonaws.com/filer_public/ec/fc/ecfc9eb9-6049-45ee-8e36-4b616aa8dbba/pxl_20250113_122303511_web.jpg?#)
How to Validate AI API Responses with Zod and AJV (and Avoid Breaking Your Frontend)
Why Validate API Responses from AI? Integrating AI APIs (like OpenAI, Gemini, or Claude) comes with an annoying risk—responses can be inconsistent, missing fields, or contain unexpected data that breaks your frontend. For example: Dev: "Return JSON with name, age, and email fields." AI: Returns a JSON... with nickname instead of name and no email.
Why Validate API Responses from AI?
Integrating AI APIs (like OpenAI, Gemini, or Claude) comes with an annoying risk—responses can be inconsistent, missing fields, or contain unexpected data that breaks your frontend.
For example:
Dev: "Return JSON with name, age, and email fields."
AI: Returns a JSON... with nickname instead of name and no email.
