How should an OpenAI Responses API client extract generated text reliably?

How should an OpenAI Responses API client extract generated text reliably?

1. Check the response status, iterate output items by type, and collect text from message content blocks
2. Always read output[0].content[0].text because the first item is guaranteed to be text
3. Serialize the entire response object and display it as the answer