“이 80-page financial report의 모든 table을 separate CSV files로 extract해 줘.”
MoClaw가 하는 일
- 1 PDF를 index합니다. pages 12~76에서 23개의 tables를 detect합니다.
- 2 각 table에 대해 headers를 infer하고 Camelot extraction을 실행합니다(low-quality scans는 vision OCR로 fallback).
- 3 23개의 CSV files를 section heading 기준 이름으로 ~/reports/에 작성합니다.
- 4 각 file link와 5-row preview가 포함된 manifest를 chat에 post합니다.
다음처럼 답합니다. '23 tables extracted to ~/reports/. Largest is page 47 (Quarterly revenue by region, 8 cols × 47 rows). Two tables on pages 22 and 58 had layout issues — I used vision OCR instead of Camelot for those, double-check before using.'