A Meridian Line pipeline extracts vessel name, laycan start, laycan end and demurrage rate from charter fixtures and writes the result to a Delta table. The model returns correct values but wraps them in prose that varies from call to call, breaking the parser. Which prompt change most reliably produces machine-parseable output?
Select an answer to reveal the explanation.
Short Explanation
You’re right to focus on the output’s shape: asking for concision still leaves the format open. An exact JSON schema, a JSON-only instruction, and one filled example give your parser a clear target.
Full Explanation
Output-format enforcement is one of the named Application Development objectives, and the reliable recipe has three parts: state the exact structure, including key names and value types; explicitly forbid anything outside it; and show one filled example so the model has a concrete target. Asking the model to be concise constrains length, not structure, so it can still vary the output format between calls. Lowering the temperature to zero is a useful extraction setting, but it controls sampling variance, not format: a consistent response can still have the wrong shape and break the parser. A Markdown table is more structured than prose, but it is a rendering rather than a data contract; column alignment, escaped pipes, or header variation can break naive parsers, and nested values have nowhere to go. The exam-relevant caveat is that prompt-level schema specification is the associate-level answer, while a serving-side structured-output or function-calling constraint offers a stronger guarantee when available. In practice, validate the parsed result in code before writing it to Delta: a prompt is a request, not an enforcement mechanism.