The #1 reason LLM automations break in production: you trusted the model to return clean data. As...The #1 reason LLM automations break in production: you trusted the model to return clean data. As...
The network for creativity
Join 1.25M professional creatives like you
Connect with clients, get discovered, and run your business 100% commission-free
Creatives on Contra have earned over $150M and we are just getting started
Back to feedPost
fernando's avatar
fernando aporta franco
pro
• Aug 22
The #1 reason LLM automations break in production: you trusted the model to return clean data.
Ask an LLM for JSON and most of the time you get JSON — then one call comes back with a markdown fence, a missing field, or a chatty "Sure! Here's your data:" and your whole pipeline falls over at 3am.
What actually works: → Define a strict schema and validate every response against it → On a mismatch, don't crash — retry with the error fed back to the model → Add a graceful fallback for the call that still won't comply
Do this and "AI in production" stops being a coin flip — typed, reliable output instead of vibes.
I open-sourced a small framework for exactly this: github.com/ferinazuma/structllm
Claude
Back to feed
The network for creativity
Join 1.25M professional creatives like you
Connect with clients, get discovered, and run your business 100% commission-free
Creatives on Contra have earned over $150M and we are just getting started