Schemas, Pydantic, and validation — making the model return real data (step 9/9) · structured output

Checkpoint

One last thing before we move on. Same surface as a write step — but the lesson doesn't complete until this passes.

Final drill. Write safe_extract(raw) that parses a JSON string the model returned and validates a Ticket shape:

email must be a non-empty str
severity must be an int between 1 and 5 inclusive

If any check fails, return the dict {"ok": False, "error": "<reason>"} where <reason> is one of "bad json", "missing email", "bad severity". On success, return {"ok": True, "data": <parsed dict>}.

Then the script will run it on three inputs and print the result of each. Expected output:

ok=True email=maya@promptdojo.dev
ok=False error=missing email
ok=False error=bad severity

Checkpoint

One last thing before we move on. Same surface as a write step — but the lesson doesn't complete until this passes.

Final drill. Write safe_extract(raw) that parses a JSON string the model returned and validates a Ticket shape:

email must be a non-empty str
severity must be an int between 1 and 5 inclusive

Then the script will run it on three inputs and print the result of each. Expected output:

ok=True email=maya@promptdojo.dev
ok=False error=missing email
ok=False error=bad severity

full-screen editor opens — close anytime to keep reading.

Schemas, Pydantic, and validation — making the model return real data — step 9 of 9