gua-docx-pipeline

Orchestration entry point that turns one or several Word documents (.docx) into finished gua LaTeX documents in a single run: extract body images, stage them into the working directory fig/, read the docx text directly from OOXML, recover ALL formulas (Word-native OMML / MathType OLE / image-encoded) via a strict source cascade — same-name PDF first, else ask the user for one manual Save-As, else render the docx-embedded formula metafiles to PNG and transcribe visually — then build a verified figure-to-problem mapping, hand off to the gua template skill for typesetting, xelatex compile and clean up, and finally run a source-level / render-level / PDF-level acceptance checklist. Triggers include "把这份 word 整理成学案", "word 转 gua 学案", "docx 一条龙整理", "提取图片并整理成学案", "把 word 的图片和正文一起排版", "word2tex 全流程", or whenever the user supplies a .docx together with a request to produce a gua 学案/试卷/周测卷/练习卷 and does not want to run the atomic skills by hand. Use it INSTEAD of manually invoking docx-image-extractor and gua-learningplan-organizer in sequence. It is a thin SOP layer: it owns only the seams (figure staging, source-of-truth choice, formula recovery cascade, figure-to-problem mapping, per-type figure alignment) and delegates all template conventions to the atomic skills. STRUCTURE — READ THIS FILE FIRST: it is the main SOP and holds only the decisions, the command lines and the pre-delivery acceptance checklist. The deep technique catalogs (formula-recovery cascade, the figure-to-problem mapping toolbox 5a/5a′/5a-补1-16, typesetting policy, rework procedure) and their empirical case notes live in references/*.md and are read ON DEMAND — every step in the main SOP states exactly when to open which reference, so do NOT read the references up front.

存档时间线

版本索引永久保留；内容副本只保留最近 2 版，更早版本仅留索引与哈希（存档时间线的证据链不会因此断裂）。

下载存档内容副本

相关资产