Module 03 Summary

What this module established

System carries durable rules, user carries the request and any data you supply, and assistant messages are input on the next call - including anything wrong the model said.

Carry forward

• Read the whole response object. finish_reason catches truncation and filtering, usage gives cost per success, model explains old runs.
• A truncated response is valid text that stops mid-thought and raises nothing. Size the limit from the longest legitimate answer.
• Time to first chunk is what users feel. A change that halves total time while doubling it feels slower.

Before moving on

Move on when finish_reason is checked on every call and your log line explains a call without containing its content.

Mark complete