dsh-zju-lecture-note 2.2.4 → 2.2.5
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
- package/CHANGELOG.md +61 -1
- package/README.md +97 -150
- package/SECURITY.md +136 -44
- package/docs/technical.md +152 -44
- package/lib/client.js +14 -13
- package/lib/index.js +53 -19
- package/lib/tools.js +28 -3
- package/lib/tools.schema.json +28 -3
- package/mcp_server.py +285 -34
- package/package.json +6 -2
- package/skills/scholia-notes/references/prelude.typ +13 -0
- package/skills/scholia-notes.manifest.json +3 -3
- package/skills/zju-lecture-summarizer/SKILL.md +45 -12
- package/skills/zju-lecture-summarizer-help/SKILL.md +95 -0
- package/skills/zju-lecture-summarizer-help.manifest.json +14 -0
- package/skills/zju-lecture-summarizer-login/SKILL.md +40 -0
- package/skills/zju-lecture-summarizer-login.manifest.json +14 -0
- package/skills/zju-lecture-summarizer.manifest.json +4 -4
- package/zju_lecture/__init__.py +16 -0
- package/zju_lecture/browser_login.py +923 -0
- package/zju_lecture/cli.py +98 -3
- package/zju_lecture/client.py +91 -24
- package/zju_lecture/handout.py +51 -17
- package/zju_lecture/handout_prelude.typ +12 -0
- package/zju_lecture/note_lint.py +12 -0
- package/zju_lecture/prelude.typ +13 -0
- package/zju_lecture/study_notes.py +126 -10
|
@@ -291,6 +291,32 @@ _PSEUDO_HEADING_RE = re.compile(
|
|
|
291
291
|
r"^\*\*[^*\n]{1,40}\*\*(?:\s*`?\[\d{1,2}:\d{2}(?::\d{2})?\]`?)?\s*$"
|
|
292
292
|
)
|
|
293
293
|
|
|
294
|
+
#: 行内粗体小标题:`**矛盾从哪里来** \`[00:05]\`–\`[00:06]\`:…`。
|
|
295
|
+
#: 它和整行粗体的伪标题是同一个意思(下一个话题的抬头),只是后面直接接了正文。
|
|
296
|
+
#: 例题四段式的标签(`**题目**:` 等)必须排除,否则 `[例]` 会被拆散。
|
|
297
|
+
_BOLD_LEAD_RE = re.compile(
|
|
298
|
+
r"^\*\*([^*\n]{1,40})\*\*\s*"
|
|
299
|
+
r"(?:`?\[\d{1,2}:\d{2}(?::\d{2})?\]`?\s*"
|
|
300
|
+
r"(?:[–—~-]\s*`?\[\d{1,2}:\d{2}(?::\d{2})?\]`?)?\s*)?"
|
|
301
|
+
r"[::]"
|
|
302
|
+
)
|
|
303
|
+
_REVIEW_HEAD_LABELS = frozenset({"题目", "切入点", "关键步骤", "易错点", "易错"})
|
|
304
|
+
|
|
305
|
+
#: 短引导词 + 时间码 + 冒号:`推导要点 \`[00:43]\`–\`[00:45]\`:…`。
|
|
306
|
+
#: 与粗体小标题同一个意思,只是作者没加粗;框在这一行收尾。证明块例外 ——
|
|
307
|
+
#: 证明的 qed 必须落在整段推导的末尾,不能因为一个「第一步 [00:12]:」就截断。
|
|
308
|
+
_PLAIN_LEAD_RE = re.compile(
|
|
309
|
+
r"^([\u4e00-\u9fff][^\s::,。;、()()]{0,11})\s*"
|
|
310
|
+
r"`?\[\d{1,2}:\d{2}(?::\d{2})?\]`?\s*"
|
|
311
|
+
r"(?:[–—~-]\s*`?\[\d{1,2}:\d{2}(?::\d{2})?\]`?\s*)?"
|
|
312
|
+
r"[::]"
|
|
313
|
+
)
|
|
314
|
+
|
|
315
|
+
#: 未知的中文块标记:`[推导] …`。只认中文标签 —— 行首的 `[1]` 引用、公式里
|
|
316
|
+
#: 的 `[S]` 都不是标记。跨段收集在这里收尾,渲染时按加粗小标题排,
|
|
317
|
+
#: 不再把方括号原样印进 PDF。
|
|
318
|
+
_UNKNOWN_MARKER_RE = re.compile(r"^\[([\u4e00-\u9fff][^\]\n]{0,7})\]\s*(.*)$")
|
|
319
|
+
|
|
294
320
|
#: 斜杠分式:`\partial D/\partial t`、`\mathrm{d}i/\mathrm{d}t`、
|
|
295
321
|
#: `\sqrt{\mu_0/\varepsilon_0}`。这些写法在行内会与上下标挤在一起,
|
|
296
322
|
#: 渲染层统一改成 `\frac{…}{…}`;只处理**明确的微分/根式**,不碰一般 `a/b`。
|
|
@@ -1393,21 +1419,60 @@ def _is_marker_line(text: str) -> bool:
|
|
|
1393
1419
|
)
|
|
1394
1420
|
|
|
1395
1421
|
|
|
1422
|
+
def block_boundary(text: str, proof: bool = False) -> bool:
|
|
1423
|
+
"""跨段装置是否应该在这一行**收尾**(两条 PDF 链路共用)。
|
|
1424
|
+
|
|
1425
|
+
框与证明本来就该跨段(v2.2.4:公式与段落不能掉出框外),但「跨段」不能
|
|
1426
|
+
等于「吞到下一个标题」:一份讲义里 `[定理]` 后面紧跟的普通段落往往是
|
|
1427
|
+
下一段讲解,`[定义]` 后面常跟着推导与结论。旧规则只认标题、已知标记与
|
|
1428
|
+
整行粗体,实地讲义里出现过一个定义框吞掉推导、提示块甚至列表的情况。
|
|
1429
|
+
|
|
1430
|
+
这里补四条**结构性**边界,都是「作者显然在开一个新块」的信号:
|
|
1431
|
+
|
|
1432
|
+
* 引用 / 提示块(``>``)—— 独立块,不属于上一个装置;
|
|
1433
|
+
* 以 ``⚠`` 开头的提醒段 —— 同样是独立提示,不应该被 qed 或方框包住;
|
|
1434
|
+
* 粗体小标题 + 冒号(可带时间码)—— 例题四段式标签例外;
|
|
1435
|
+
* 短引导词 + 时间码 + 冒号 —— 与上一条同义,只是没加粗;
|
|
1436
|
+
``proof=True`` 时跳过,避免把证明从中间截断(qed 会停在半路);
|
|
1437
|
+
* 未知的中文标记(``[推导]``)—— 作者想开新块,只是名字不在表里。
|
|
1438
|
+
|
|
1439
|
+
普通段落、块级公式、列表、表格仍然留在装置里,证明与例题的多段正文不受影响。
|
|
1440
|
+
"""
|
|
1441
|
+
stripped = text.strip()
|
|
1442
|
+
if not stripped:
|
|
1443
|
+
return False
|
|
1444
|
+
if re.match(r"^#{1,6}\s+", stripped):
|
|
1445
|
+
return True
|
|
1446
|
+
if _is_marker_line(stripped) or _PSEUDO_HEADING_RE.match(stripped):
|
|
1447
|
+
return True
|
|
1448
|
+
if stripped.startswith(">") or stripped.startswith("\u26a0"):
|
|
1449
|
+
return True
|
|
1450
|
+
bold = _BOLD_LEAD_RE.match(stripped)
|
|
1451
|
+
if bold and bold.group(1).strip() not in _REVIEW_HEAD_LABELS:
|
|
1452
|
+
return True
|
|
1453
|
+
if not proof:
|
|
1454
|
+
lead = _PLAIN_LEAD_RE.match(stripped)
|
|
1455
|
+
if lead and lead.group(1).strip() not in _REVIEW_HEAD_LABELS:
|
|
1456
|
+
return True
|
|
1457
|
+
unknown = _UNKNOWN_MARKER_RE.match(stripped)
|
|
1458
|
+
if unknown and unknown.group(1) not in MARKERS:
|
|
1459
|
+
return True
|
|
1460
|
+
return False
|
|
1461
|
+
|
|
1462
|
+
|
|
1396
1463
|
def _collect_block_lines(
|
|
1397
1464
|
lines: Sequence[str],
|
|
1398
1465
|
cursor: int,
|
|
1466
|
+
name: str = "",
|
|
1399
1467
|
) -> Tuple[List[str], int]:
|
|
1400
|
-
"""
|
|
1468
|
+
"""收集标记行之后、到下一个标题/标记/新话题为止的底稿行(空行保留)。"""
|
|
1401
1469
|
body: List[str] = []
|
|
1402
1470
|
total = len(lines)
|
|
1471
|
+
proof = name == "证明"
|
|
1403
1472
|
while cursor < total:
|
|
1404
1473
|
raw = lines[cursor]
|
|
1405
1474
|
stripped = raw.strip()
|
|
1406
|
-
if stripped and (
|
|
1407
|
-
re.match(r"^#{1,6}\s+", stripped)
|
|
1408
|
-
or _is_marker_line(stripped)
|
|
1409
|
-
or _PSEUDO_HEADING_RE.match(stripped)
|
|
1410
|
-
):
|
|
1475
|
+
if stripped and block_boundary(stripped, proof=proof):
|
|
1411
1476
|
break
|
|
1412
1477
|
body.append(raw)
|
|
1413
1478
|
cursor += 1
|
|
@@ -1869,7 +1934,7 @@ def render_typst(
|
|
|
1869
1934
|
continue
|
|
1870
1935
|
if _marker_absorbs_blocks(name):
|
|
1871
1936
|
# 框与证明跨段:同一装置的公式、段落、列表都留在块内。
|
|
1872
|
-
extra, cursor = _collect_block_lines(input_lines, cursor)
|
|
1937
|
+
extra, cursor = _collect_block_lines(input_lines, cursor, name)
|
|
1873
1938
|
fragment, label, body_paragraphs = _render_block_marker(
|
|
1874
1939
|
name, body, extra, known_labels, warnings
|
|
1875
1940
|
)
|
|
@@ -1885,20 +1950,71 @@ def render_typst(
|
|
|
1885
1950
|
knots.append({"call": call, "section": section})
|
|
1886
1951
|
continue
|
|
1887
1952
|
|
|
1953
|
+
# 未知的中文标记(`[推导] …`):作者想开一个新块,但名字不在 MARKERS 表里。
|
|
1954
|
+
# 推导按旧规范的 Proof 排(斜体 Proof. + 块末 qed,跨段、不套框);
|
|
1955
|
+
# 其余未知标记按「加粗小标题 + 正文」兜底。note_lint 仍会提示改成已支持的标记。
|
|
1956
|
+
unknown = _UNKNOWN_MARKER_RE.match(stripped)
|
|
1957
|
+
if unknown:
|
|
1958
|
+
if unknown.group(1) == "推导":
|
|
1959
|
+
extra, cursor = _collect_block_lines(input_lines, cursor, "证明")
|
|
1960
|
+
fragment, label, body_paragraphs = _render_block_marker(
|
|
1961
|
+
"证明", unknown.group(2), extra, known_labels, warnings
|
|
1962
|
+
)
|
|
1963
|
+
lines_out.append(fragment)
|
|
1964
|
+
paragraphs += body_paragraphs
|
|
1965
|
+
if label:
|
|
1966
|
+
labels.append(label)
|
|
1967
|
+
_count("proof")
|
|
1968
|
+
continue
|
|
1969
|
+
lead = "**" + unknown.group(1) + "**"
|
|
1970
|
+
if unknown.group(2):
|
|
1971
|
+
lead += " " + unknown.group(2)
|
|
1972
|
+
lines_out.append(BODY_INDENT + escape_typst(lead, known_labels))
|
|
1973
|
+
paragraph_raw[len(lines_out) - 1] = lead
|
|
1974
|
+
paragraphs += 1
|
|
1975
|
+
_count("paragraph")
|
|
1976
|
+
continue
|
|
1977
|
+
|
|
1888
1978
|
# Markdown 引用块。Typst **没有** `>` 引用语法(0.15 实测:`>` 会被
|
|
1889
1979
|
# 当成字面量印出来,整段抬头变成「> 来源:… > 上课时间:…」),
|
|
1890
1980
|
# 所以走 prelude 里的 `#quoted` 内容块;字段之间用行间断行。
|
|
1981
|
+
# `> [!IMPORTANT]` 这类 GitHub alert 先摘掉标记行:重点走 `#keypoint`
|
|
1982
|
+
# (规范蓝色的文字段落,不是方框),其余 alert 退化成带标签的灰色引用。
|
|
1891
1983
|
if stripped.startswith(">"):
|
|
1892
1984
|
quote_lines = [raw_line]
|
|
1893
1985
|
while cursor < len(input_lines) and input_lines[cursor].strip().startswith(">"):
|
|
1894
1986
|
quote_lines.append(input_lines[cursor])
|
|
1895
1987
|
cursor += 1
|
|
1896
|
-
|
|
1988
|
+
kind = ""
|
|
1989
|
+
cleaned: List[str] = []
|
|
1990
|
+
for line in quote_lines:
|
|
1991
|
+
text_line = line.strip()
|
|
1992
|
+
if text_line.startswith(">"):
|
|
1993
|
+
text_line = text_line[1:].lstrip()
|
|
1994
|
+
alert = re.match(
|
|
1995
|
+
r"^\[!(IMPORTANT|WARNING|CAUTION|NOTE|TIP)\]$", text_line, re.I
|
|
1996
|
+
)
|
|
1997
|
+
if alert and not cleaned:
|
|
1998
|
+
kind = alert.group(1).upper()
|
|
1999
|
+
continue
|
|
2000
|
+
cleaned.append(line)
|
|
2001
|
+
items = split_quote_items(cleaned)
|
|
1897
2002
|
body = " \\\n".join(
|
|
1898
2003
|
escape_typst(join_cjk_lines(item), known_labels) for item in items
|
|
1899
2004
|
)
|
|
1900
|
-
|
|
1901
|
-
|
|
2005
|
+
if kind == "IMPORTANT":
|
|
2006
|
+
lines_out.append("#keypoint[*重点* " + body + "]")
|
|
2007
|
+
_count("keypoint")
|
|
2008
|
+
elif kind:
|
|
2009
|
+
label = {
|
|
2010
|
+
"WARNING": "注意", "CAUTION": "注意",
|
|
2011
|
+
"NOTE": "说明", "TIP": "提示",
|
|
2012
|
+
}.get(kind, "提示")
|
|
2013
|
+
lines_out.append("#quoted[*" + label + "* " + body + "]")
|
|
2014
|
+
_count("quote")
|
|
2015
|
+
else:
|
|
2016
|
+
lines_out.append("#quoted[\n" + body + "\n]")
|
|
2017
|
+
_count("quote")
|
|
1902
2018
|
continue
|
|
1903
2019
|
|
|
1904
2020
|
# Markdown 水平分隔线 → Typst 的细分隔线。不处理的话 `---` 会被
|