From cdca3b57097576665fc251a5db4b58cb37279744 Mon Sep 17 00:00:00 2001 From: myhloli Date: Wed, 27 May 2026 22:22:24 +0800 Subject: [PATCH] feat: refactor finalize_middle_json to support server-side post-processing and enhance title leveling --- mineru/backend/utils/para_block_utils.py | 9 +++++++-- 1 file changed, 7 insertions(+), 2 deletions(-) diff --git a/mineru/backend/utils/para_block_utils.py b/mineru/backend/utils/para_block_utils.py index d0d622b1..d66f2dac 100644 --- a/mineru/backend/utils/para_block_utils.py +++ b/mineru/backend/utils/para_block_utils.py @@ -23,6 +23,10 @@ TEXT_MERGE_TRANSPARENT_TYPES = { BlockType.CODE, } OCR_DET_LINES_KEY = "_ocr_det_lines" +INTERNAL_BLOCK_METADATA_KEYS = { + OCR_DET_LINES_KEY, + "line_avg_height", +} def iter_block_spans(block): @@ -376,7 +380,8 @@ def _has_mergeable_block_bbox_relation(current_block, previous_block): def _cleanup_block_internal_metadata(block): - """递归清理只供 Hybrid 内部段落合并使用的临时字段。""" - block.pop(OCR_DET_LINES_KEY, None) + """递归清理只供 finalize 内部流程使用的临时字段。""" + for metadata_key in INTERNAL_BLOCK_METADATA_KEYS: + block.pop(metadata_key, None) for sub_block in block.get("blocks", []): _cleanup_block_internal_metadata(sub_block)