feat: export core Hermes skills
This commit is contained in:
@@ -0,0 +1,117 @@
|
||||
## 批注注入（comments.xml 机制）
|
||||
|
||||
ContractEditor 库**不支持批注**。批注需要直接操作 docx 的 OOXML 批注三件套。验证可用（2026-06-17 培训合同实战）。
|
||||
|
||||
### 批注的三个组成部分
|
||||
1. **word/comments.xml**：批注内容本体（每条 `<w:comment>`，含 id/author/date/initials）。
|
||||
2. **word/document.xml**：在被批注的段落里插入锚点：
|
||||
- `<w:commentRangeStart w:id="X"/>` —— 放在段落第一个 run/ins 之前
|
||||
- `<w:commentRangeEnd w:id="X"/>` —— 放在段落末尾
|
||||
- 一个含 `<w:commentReference w:id="X"/>` 的 run —— 放在 rangeEnd 之后
|
||||
3. **关系声明**：
|
||||
- `[Content_Types].xml` 加 Override（comments.xml 的 ContentType）
|
||||
- `word/_rels/document.xml.rels` 加 Relationship（指向 comments.xml）
|
||||
|
||||
### 可复用代码（在 ContractEditor 修订并 save 之后，对成品 docx 注入批注）
|
||||
|
||||
```python
|
||||
import zipfile, io
|
||||
from lxml import etree
|
||||
|
||||
W = 'http://schemas.openxmlformats.org/wordprocessingml/2006/main'
|
||||
Wq = '{' + W + '}'
|
||||
|
||||
# 批注清单：anchor 是"接受修订后能精确匹配（唯一）"的段落片段
|
||||
comments = [
|
||||
{"id": "201", "anchor": "甲方扣除相应服务费后", "text": "建议……"},
|
||||
{"id": "202", "anchor": "向甲方住所地人民法院提起诉讼", "text": "建议……"},
|
||||
]
|
||||
DATE = "2026-06-17T10:00:00Z" # ISO格式
|
||||
|
||||
def build_comments_xml(comments):
|
||||
parts = ['<?xml version="1.0" encoding="UTF-8" standalone="yes"?>']
|
||||
parts.append(f'<w:comments xmlns:w="{W}" xmlns:r="http://schemas.openxmlformats.org/officeDocument/2006/relationships">')
|
||||
for c in comments:
|
||||
parts.append(f'<w:comment w:id="{c["id"]}" w:author="WB" w:date="{DATE}" w:initials="WB">')
|
||||
# 批注文字字号建议比正文小（sz=18=9pt），字体跟随文档（宋体+hint=eastAsia）
|
||||
parts.append('<w:p><w:r><w:rPr><w:rFonts w:ascii="宋体" w:hAnsi="宋体" w:eastAsia="宋体" w:cs="宋体" w:hint="eastAsia"/><w:sz w:val="18"/><w:szCs w:val="18"/></w:rPr>')
|
||||
parts.append(f'<w:t xml:space="preserve">{c["text"]}</w:t></w:r></w:p>')
|
||||
parts.append('</w:comment>')
|
||||
parts.append('</w:comments>')
|
||||
return ''.join(parts)
|
||||
|
||||
def inject_comments(in_path, out_path, comments):
|
||||
comments_xml = build_comments_xml(comments)
|
||||
with open(in_path, 'rb') as f:
|
||||
data = f.read()
|
||||
buf_in, buf_out = io.BytesIO(data), io.BytesIO()
|
||||
inserted = {c["id"]: False for c in comments}
|
||||
with zipfile.ZipFile(buf_in, 'r') as zin, zipfile.ZipFile(buf_out, 'w', zipfile.ZIP_DEFLATED) as zout:
|
||||
for item in zin.infolist():
|
||||
raw = zin.read(item.filename)
|
||||
if item.filename == 'word/document.xml':
|
||||
tree = etree.fromstring(raw)
|
||||
body = tree.find(f'{Wq}body')
|
||||
for para in body.findall(f'.//{Wq}p'):
|
||||
# 接受修订后文本（跳过 w:del 内的 t）做锚点匹配
|
||||
ptext = ''
|
||||
for t in para.findall(f'.//{Wq}t'):
|
||||
if not any(a.tag == f'{Wq}del' for a in t.iterancestors()):
|
||||
ptext += (t.text or '')
|
||||
for c in comments:
|
||||
if not inserted[c["id"]] and c["anchor"] in ptext:
|
||||
cid = c["id"]
|
||||
# 第一个挂靠点：段落第一个 r 或 ins
|
||||
first_child = None
|
||||
for child in para:
|
||||
if child.tag in (f'{Wq}r', f'{Wq}ins'):
|
||||
first_child = child; break
|
||||
if first_child is None:
|
||||
continue
|
||||
crs = etree.Element(f'{Wq}commentRangeStart'); crs.set(f'{Wq}id', cid)
|
||||
first_child.addprevious(crs)
|
||||
cre = etree.Element(f'{Wq}commentRangeEnd'); cre.set(f'{Wq}id', cid)
|
||||
para.append(cre)
|
||||
rr = etree.SubElement(para, f'{Wq}r')
|
||||
rrp = etree.SubElement(rr, f'{Wq}rPr')
|
||||
rs = etree.SubElement(rrp, f'{Wq}rStyle'); rs.set(f'{Wq}val', 'CommentReference')
|
||||
cref = etree.SubElement(rr, f'{Wq}commentReference'); cref.set(f'{Wq}id', cid)
|
||||
inserted[cid] = True
|
||||
raw = etree.tostring(tree, xml_declaration=True, encoding='UTF-8', standalone=True)
|
||||
elif item.filename == '[Content_Types].xml':
|
||||
ct = etree.fromstring(raw)
|
||||
NS = 'http://schemas.openxmlformats.org/package/2006/content-types'
|
||||
ov = etree.SubElement(ct, f'{{{NS}}}Override')
|
||||
ov.set('PartName', '/word/comments.xml')
|
||||
ov.set('ContentType', 'application/vnd.openxmlformats-officedocument.wordprocessingml.comments+xml')
|
||||
raw = etree.tostring(ct, xml_declaration=True, encoding='UTF-8', standalone=True)
|
||||
elif item.filename == 'word/_rels/document.xml.rels':
|
||||
rt = etree.fromstring(raw)
|
||||
RNS = 'http://schemas.openxmlformats.org/package/2006/relationships'
|
||||
r = etree.SubElement(rt, f'{{{RNS}}}Relationship')
|
||||
r.set('Id', 'rIdComments1')
|
||||
r.set('Type', 'http://schemas.openxmlformats.org/officeDocument/2006/relationships/comments')
|
||||
r.set('Target', 'comments.xml')
|
||||
raw = etree.tostring(rt, xml_declaration=True, encoding='UTF-8', standalone=True)
|
||||
zout.writestr(item, raw)
|
||||
zout.writestr('word/comments.xml', comments_xml.encode('utf-8'))
|
||||
with open(out_path, 'wb') as f:
|
||||
f.write(buf_out.getvalue())
|
||||
return inserted # 检查是否全部 True
|
||||
```
|
||||
|
||||
### 锚点选择要点
|
||||
- anchor 必须是**接受修订后文本里唯一**的片段（先用脚本验证 hits==1，多处命中会挂错段）。
|
||||
- 避免选被修订（w:del/w:ins）切割的文字做anchor——优先选未被改动的稳定片段。
|
||||
- 若 anchor 落在被修订段，匹配用"接受修订后文本"（跳过 w:del），与上面代码一致。
|
||||
|
||||
### 验证（终审必做）
|
||||
```python
|
||||
# commentRangeStart/End/Reference 与 comments.xml 的 id 必须全部对应
|
||||
crs = set(e.get(Wq+'id') for e in doc_root.iter(Wq+'commentRangeStart'))
|
||||
cre = set(e.get(Wq+'id') for e in doc_root.iter(Wq+'commentRangeEnd'))
|
||||
cref = set(e.get(Wq+'id') for e in doc_root.iter(Wq+'commentReference'))
|
||||
com = set(c.get(Wq+'id') for c in com_root.iter(Wq+'comment'))
|
||||
assert crs == cre == cref == com
|
||||
```
|
||||
末了用 python-docx `Document(out)` 能打开（XML合法）+ OnlyOffice 渲染确认批注气泡显示。
|
||||
Reference in New Issue
Block a user