parsing nested api responses — without crashing on a missing key

Claude wrote a parser that pulls the assistant text out of a fake LLM response. The chain uses strict [] access at every level. If the response was filtered (no choices), the parser crashes with KeyError.

Replace the strict chain on line 6 with the safe pattern: default choices to [], return the string "(blocked)" if the list is empty, otherwise pull message.content out with chained .gets.

The starter has a content-filtered response (no choices key). Expected output:

(blocked)