Extract section-typed full text for up to 8 papers in one call.

Reading several papers is the normal shape of a research task, and doing it one request at a time against a per-minute limit is what pushed agents toward fetching one paper and guessing about the rest.

sections is REQUIRED when more than one paper is requested. Eight papers of all is ~108KB, which is the exact outcome this endpoint exists to prevent; a 422 teaches that in one round trip instead of silently burning the caller's context window.

Per-paper failures are ENTRIES, never a non-200 — one unextractable paper must not cost the caller the seven that worked. Each entry carries ok, and failures carry error in: invalid_id, not_extractable, extraction_failed, timeout.

Costs ONE quota unit regardless of batch size, so batching is cheaper than looping and callers have a reason to actually use it.

Rate limit: 6/minute per API key.