jXSrSSKJrJrJrJr SSKJr SSKJ r SSK J r SSK J r /SQrg) u  Table-only extractors. `TableExtractor` is the seam for engines whose job is *table extraction*, not document parsing — explicitly decoupled from `Parser` (see core/README.md). They answer different questions and produce different shapes: - `Parser` -> `ParseResult` (whole document, structured; the `DoclingDocument` is on `.document`) - `TableExtractor` -> `list[ExtractedTable]` (tables only, Pydantic models) The two are composed in `quber.core.fusion` (`fuse_artifacts`, `DocumentFusion`), which grafts the extracted tables into a unified `DoclingDocument`. The active engine is `SetOfMarkExtractor` (vision locates each table and its region; Camelot fills it in-region; grounded correction cleans structure). `CamelotCorrespondenceExtractor` and `CamelotLLMTableExtractor` are DEPRECATED — not for new use. Both stay reachable through `quber table --engine`, and `quber validate` always runs `CamelotLLMTableExtractor` (with mock classifier and unifier) for its raw Camelot counts, so removing it breaks that command. )ExtractedTableExtractionRecord FilledCellTableExtractor)CamelotCorrespondenceExtractor)CamelotLLMTableExtractor)MockTableExtractor)SetOfMarkExtractor)r rrrr rrrN)__doc__quber.core.extractors.baserrrr,quber.core.extractors.camelot.correspondencer!quber.core.extractors.camelot.llmrquber.core.extractors.mockr !quber.core.extractors.set_of_markr __all__rs&0dcWF9@ r