github-featured / unstructured-io--unstructured

发布

0 发布 249 Git 标签

RSS 订阅源
  • 0.22.26 199f25555e
    比较

    feat: Track the table extraction method (#4346)

    frostbyte_neo 发布于 2026-04-28 21:26:08 +00:00

    Note

    Low Risk
    Adds a new optional metadata field and wires it through PDF
    partitioning; behavior is additive and low risk aside from potential
    downstream consumers expecting a fixed metadata schema.

    Overview
    Adds table-provenance metadata for extracted tables.
    ElementMetadata now includes an optional table_extraction_method
    (e.g., grid, tatr, vlm) and includes it in metadata consolidation.

    During PDF partitioning, the value is propagated from each
    LayoutElement onto the resulting element metadata, enabling downstream
    consumers to identify which table-extraction algorithm produced a given
    table. Version is bumped to 0.22.26 and the changelog is updated
    accordingly.

    Reviewed by Cursor Bugbot for commit
    fe6135af306bc3fb6fbe28db3aa5dc477c4f2ff0. Bugbot is set up for automated
    code reviews on this repo. Configure
    here.

    下载附件