問題文
An extraction pipeline must always produce structured output for each document, never conversational text. Several extraction schemas exist and the document type is not known in advance. Which tool_choice setting fits?
選択肢
- No tool_choice setting, with a system prompt instruction to always return JSON.
- Forced selection with {"type": "tool", "name": "extract_invoice"} for every document.
- tool_choice: "auto", which lets the model decide whether to call a tool at all.
- tool_choice: "any", which guarantees the model calls a tool but lets it choose which schema applies.