dsh-plugin-lookatstudy 0.11.0 → 0.11.1

package/lib/docx-parser-BhyqPImb.mjs

@@ -0,0 +1,55 @@
+import { n as readZipText, t as readZip } from "./zip-reader-KnRrq0av.mjs";
+//#region src/vendor/docx-parser.ts
+function decodeEntities(s) {
+	return s.replace(/<[^>]+>/g, "").replace(/&lt;/g, "<").replace(/&gt;/g, ">").replace(/&quot;/g, "\"").replace(/&apos;/g, "'").replace(/&amp;/g, "&");
+}
+function parseParagraphs(documentXml) {
+	const paras = [];
+	const pRe = /<w:p\b[^>]*\/>|<w:p\b[^>]*>[\s\S]*?<\/w:p>/g;
+	let pm;
+	while ((pm = pRe.exec(documentXml)) !== null) {
+		const p = pm[0];
+		const style = p.match(/<w:pStyle w:val="([^"]+)"/i)?.[1] ?? "";
+		let text = "";
+		const tRe = /<w:t(?:\s[^>]*)?>([\s\S]*?)<\/w:t>|<w:tab\b[^>]*\/>|<w:br\b[^>]*\/>/g;
+		let tm;
+		while ((tm = tRe.exec(p)) !== null) if (tm[1] !== void 0) text += decodeEntities(tm[1]);
+		else if (/w:tab/.test(tm[0])) text += "	";
+		else text += "\n";
+		paras.push({
+			style: style.trim(),
+			text
+		});
+	}
+	return paras;
+}
+function headingLevel(style) {
+	const s = style.toLowerCase().replace(/\s+/g, "");
+	const m = s.match(/heading(\d)/) ?? s.match(/titre(\d)/) ?? s.match(/berschrift(\d)/);
+	if (m) return Math.min(6, Math.max(1, Number(m[1])));
+	if (/^(title|titre)$/.test(s)) return 1;
+	return 0;
+}
+/** .docx → markdown。Word Heading 样式保真为 # 级标题,代码样式段落围栏。 */
+function parseDocx(buf) {
+	const entries = readZip(buf);
+	const doc = readZipText(entries, "word/document.xml");
+	if (!doc) throw new Error("docx 结构异常:缺 word/document.xml");
+	const lines = [];
+	for (const para of parseParagraphs(doc)) {
+		const t = para.text.trim();
+		if (!t) {
+			lines.push("");
+			continue;
+		}
+		const lvl = headingLevel(para.style);
+		if (lvl > 0) lines.push("", "#".repeat(lvl) + " " + t, "");
+		else if (/code|sourcecode/i.test(para.style)) lines.push("```", t, "```", "");
+		else lines.push(t, "");
+	}
+	const md = lines.join("\n").replace(/\n{3,}/g, "\n\n").trim();
+	if (!md) throw new Error("docx 里没有可识别的文本");
+	return md;
+}
+//#endregion
+export { parseDocx };

package/lib/epub-parser-DvlKap-d.mjs

@@ -0,0 +1,112 @@
+import { n as htmlToMarkdown } from "./html-article-C6nhqJbX.mjs";
+import { n as readZipText, t as readZip } from "./zip-reader-KnRrq0av.mjs";
+//#region src/vendor/epub-parser.ts
+/** 提取某标签的全部出现(OPF 的属性顺序不定,先抓整标签再逐个提属性)。 */
+function tags(xml, tagName) {
+	const out = [];
+	const re = new RegExp(`<${tagName}\\b[^>]*>`, "g");
+	let m;
+	while ((m = re.exec(xml)) !== null) {
+		const raw = m[0];
+		const attrs = {};
+		const attrRe = /([\w:-]+)\s*=\s*"([^"]*)"/g;
+		let a;
+		while ((a = attrRe.exec(raw)) !== null) attrs[a[1]] = a[2];
+		out.push({
+			raw,
+			attrs
+		});
+	}
+	return out;
+}
+function firstTagText(xml, tagName) {
+	return xml.match(new RegExp(`<${tagName}[^>]*>([^<]*)</${tagName}>`, "i"))?.[1]?.trim() ?? "";
+}
+/** zip 内路径归一:posix 分隔 + href 相对 OPF 目录解析。 */
+function resolveZipPath(opfPath, href) {
+	const cleanHref = decodeURIComponent(href.split("#")[0] ?? href).replace(/\\/g, "/");
+	if (!opfPath.includes("/")) return cleanHref.replace(/^\.\//, "");
+	const parts = `${opfPath.slice(0, opfPath.lastIndexOf("/"))}/${cleanHref}`.split("/");
+	const resolved = [];
+	for (const p of parts) if (p === "..") resolved.pop();
+	else if (p !== "." && p !== "") resolved.push(p);
+	return resolved.join("/");
+}
+/** EPUB3 nav.xhtml 或 EPUB2 toc.ncx → href(去 fragment)→ 章节标题。 */
+function parseTocLabels(tocXml, isNcx) {
+	const labels = /* @__PURE__ */ new Map();
+	if (isNcx) {
+		const blockRe = /<navPoint\b[\s\S]*?<\/navPoint>/g;
+		let m;
+		while ((m = blockRe.exec(tocXml)) !== null) {
+			const label = firstTagText(m[0], "text");
+			const src = m[0].match(/<content[^>]+src="([^"]+)"/i)?.[1];
+			if (label && src) labels.set(decodeURIComponent(src.split("#")[0].replace(/\\/g, "/")), label);
+		}
+	} else {
+		const aRe = /<a\b[^>]*href="([^"]+)"[^>]*>([\s\S]*?)<\/a>/gi;
+		let m;
+		while ((m = aRe.exec(tocXml)) !== null) {
+			const label = (m[2] ?? "").replace(/<[^>]+>/g, "").trim();
+			if (label) labels.set(decodeURIComponent((m[1] ?? "").split("#")[0].replace(/\\/g, "/")), label);
+		}
+	}
+	return labels;
+}
+function sanitizeFileName(title) {
+	return (title || "chapter").replace(/[\\/:*?"<>|#\s]+/g, "-").replace(/^-+|-+$/g, "").slice(0, 50) || "chapter";
+}
+function parseEpub(buf) {
+	const entries = readZip(buf);
+	const read = (p) => readZipText(entries, p);
+	const opfPath = read("META-INF/container.xml").match(/full-path="([^"]+)"/i)?.[1];
+	if (!opfPath) throw new Error("epub 结构异常:找不到 container.xml 里的 OPF 路径");
+	const opf = read(opfPath);
+	if (!opf) throw new Error(`epub 结构异常:OPF 文件缺失(${opfPath})`);
+	const bookTitle = firstTagText(opf, "dc:title") || "未命名电子书";
+	const manifest = /* @__PURE__ */ new Map();
+	for (const t of tags(opf, "item")) manifest.set(t.attrs["id"] ?? "", {
+		href: t.attrs["href"] ?? "",
+		mediaType: (t.attrs["media-type"] ?? "").toLowerCase(),
+		properties: t.attrs["properties"] ?? ""
+	});
+	const spineIds = tags(opf, "itemref").map((t) => t.attrs["idref"] ?? "").filter(Boolean);
+	const spineTocId = opf.match(/<spine\b[^>]*\btoc="([^"]+)"/i)?.[1];
+	let tocLabels = /* @__PURE__ */ new Map();
+	const navItem = [...manifest.values()].find((it) => it.properties.split(/\s+/).includes("nav"));
+	if (navItem?.href) tocLabels = parseTocLabels(read(resolveZipPath(opfPath, navItem.href)), false);
+	if (tocLabels.size === 0 && spineTocId && manifest.has(spineTocId)) tocLabels = parseTocLabels(read(resolveZipPath(opfPath, manifest.get(spineTocId).href)), true);
+	const opfDirKey = (href) => decodeURIComponent(href.split("#")[0] ?? href).replace(/\\/g, "/");
+	const chapters = [];
+	let n = 0;
+	for (const id of spineIds) {
+		const item = manifest.get(id);
+		if (!item?.href) continue;
+		if (!(item.mediaType === "application/xhtml+xml" || item.mediaType === "text/html" || /\.(xhtml|html|htm)$/i.test(item.href))) continue;
+		if (item.properties.split(/\s+/).includes("nav")) continue;
+		const xhtml = read(resolveZipPath(opfPath, item.href));
+		if (!xhtml) continue;
+		const md = htmlToMarkdown(xhtml, { stripImages: true });
+		if (!md || md.replace(/[#\s>*-]/g, "").length < 8) continue;
+		n++;
+		const body = md.startsWith("# ") && md.includes("\n") ? md.slice(md.indexOf("\n") + 1).trim() : md;
+		const firstHeading = md.startsWith("# ") ? md.split("\n")[0].slice(2).trim() : "";
+		const title = tocLabels.get(opfDirKey(item.href)) || firstHeading || `第 ${n} 章`;
+		chapters.push({
+			path: `chapters/${String(n).padStart(2, "0")}-${sanitizeFileName(title)}.md`,
+			title,
+			markdown: `# ${title}\n\n${body}`
+		});
+	}
+	if (chapters.length === 0) throw new Error("epub 里没有可识别的章节文本");
+	return {
+		title: bookTitle,
+		chapters
+	};
+}
+/** 文件夹导入路径用:整本书压平成一个 markdown(全部 H1 降为 H2,给结构设计当 anchor 拆章)。 */
+function parseEpubFlat(buf) {
+	return parseEpub(buf).chapters.map((c) => c.markdown.replace(/^# /gm, "## ")).join("\n\n");
+}
+//#endregion
+export { parseEpubFlat };

package/lib/html-article-C6nhqJbX.mjs

@@ -0,0 +1,426 @@
+//#region src/vendor/html-article.ts
+const ENTITIES = {
+	amp: "&",
+	lt: "<",
+	gt: ">",
+	quot: "\"",
+	apos: "'",
+	nbsp: "\xA0",
+	hellip: "…",
+	mdash: "—",
+	ndash: "–",
+	rsquo: "'",
+	lsquo: "'",
+	rdquo: "\"",
+	ldquo: "\"",
+	middot: "·",
+	bull: "•",
+	copy: "©",
+	times: "×",
+	divide: "÷",
+	deg: "°"
+};
+function decodeEntities(s) {
+	return s.replace(/&(#x?[0-9a-fA-F]+|[a-zA-Z]+);/g, (_m, ent) => {
+		if (ent.startsWith("#x") || ent.startsWith("#X")) return String.fromCodePoint(parseInt(ent.slice(2), 16));
+		if (ent.startsWith("#")) return String.fromCodePoint(parseInt(ent.slice(1), 10));
+		return ENTITIES[ent.toLowerCase()] ?? _m;
+	});
+}
+const VOID_TAGS = /* @__PURE__ */ new Set([
+	"br",
+	"hr",
+	"img",
+	"input",
+	"meta",
+	"link",
+	"area",
+	"base",
+	"col",
+	"embed",
+	"source",
+	"track",
+	"wbr"
+]);
+/** Stack-parse well-formed-enough HTML/XHTML into a forest of nodes. */
+function parseHtml(html) {
+	const roots = [];
+	const stack = [];
+	const push = (n) => {
+		(stack.length ? stack[stack.length - 1].children : roots).push(n);
+	};
+	const tokenRe = /<!--[\s\S]*?-->|<!\[CDATA\[([\s\S]*?)\]\]>|<\/([a-zA-Z][-\w:]*)\s*>|<([a-zA-Z][-\w:]*)((?:"[^"]*"|'[^']*'|[^>"'])*?)(\/?)>|([^<]+)/g;
+	let m;
+	while ((m = tokenRe.exec(html)) !== null) if (m[2] !== void 0) {
+		const close = m[2].toLowerCase();
+		for (let i = stack.length - 1; i >= 0; i--) if (stack[i].tag === close) {
+			stack.length = i;
+			break;
+		}
+	} else if (m[3] !== void 0) {
+		const tag = m[3].toLowerCase();
+		const attrs = {};
+		const attrRe = /([-\w:]+)\s*=\s*("([^"]*)"|'([^']*)')/g;
+		let a;
+		while ((a = attrRe.exec(m[4] ?? "")) !== null) attrs[a[1].toLowerCase()] = decodeEntities(a[3] ?? a[4] ?? "");
+		const node = {
+			tag,
+			attrs,
+			children: [],
+			text: ""
+		};
+		push(node);
+		if (!VOID_TAGS.has(tag) && m[5] !== "/") stack.push(node);
+	} else if (m[1] !== void 0) push({
+		tag: "",
+		attrs: {},
+		children: [],
+		text: m[1]
+	});
+	else if (m[6] !== void 0) {
+		const text = decodeEntities(m[6]);
+		if (text.trim()) push({
+			tag: "",
+			attrs: {},
+			children: [],
+			text
+		});
+	}
+	return roots;
+}
+const BLOCK_TAGS = /* @__PURE__ */ new Set([
+	"p",
+	"div",
+	"section",
+	"article",
+	"header",
+	"footer",
+	"main",
+	"aside",
+	"nav",
+	"figure",
+	"figcaption",
+	"h1",
+	"h2",
+	"h3",
+	"h4",
+	"h5",
+	"h6",
+	"ul",
+	"ol",
+	"li",
+	"blockquote",
+	"pre",
+	"table",
+	"tr",
+	"thead",
+	"tbody",
+	"hr",
+	"br",
+	"address",
+	"details",
+	"summary"
+]);
+function nodeText(node) {
+	if (node.tag === "") return node.text;
+	return node.children.map(nodeText).join("");
+}
+function findDescendant(node, tag, cls) {
+	for (const c of node.children) {
+		if (c.tag === tag && (!cls || (c.attrs["class"] ?? "").split(/\s+/).includes(cls))) return c;
+		const hit = findDescendant(c, tag, cls);
+		if (hit) return hit;
+	}
+	return null;
+}
+function inlineRuns(nodes, stripImages) {
+	let out = "";
+	for (const n of nodes) switch (n.tag) {
+		case "":
+			out += n.text.replace(/\s+/g, " ");
+			break;
+		case "br":
+			out += "\n";
+			break;
+		case "strong":
+		case "b": {
+			const inner = inlineRuns(n.children, stripImages).trim();
+			if (inner) out += `**${inner}**`;
+			break;
+		}
+		case "em":
+		case "i": {
+			const inner = inlineRuns(n.children, stripImages).trim();
+			if (inner) out += `*${inner}*`;
+			break;
+		}
+		case "code":
+			out += "`" + nodeText(n).trim() + "`";
+			break;
+		case "a": {
+			const inner = inlineRuns(n.children, stripImages).trim();
+			const href = n.attrs["href"] ?? "";
+			out += inner && href && !href.startsWith("#") ? `[${inner}](${href})` : inner;
+			break;
+		}
+		case "img":
+		case "picture":
+		case "svg":
+		case "figure":
+			if (!stripImages && n.tag === "img") {
+				const alt = n.attrs["alt"] ?? "";
+				const src = n.attrs["src"] ?? "";
+				if (src) out += `![${alt}](${src})`;
+			}
+			break;
+		case "script":
+			if ((n.attrs["type"] ?? "").includes("math/tex")) {
+				const tex = nodeText(n).trim();
+				if (tex) out += tex.includes("\n") ? `$$${tex}$$` : `$${tex}$`;
+			}
+			break;
+		case "style":
+		case "head": break;
+		case "span":
+			if ((n.attrs["class"] ?? "").split(/\s+/).includes("katex")) {
+				const ann = findDescendant(n, "annotation");
+				const tex = ann ? nodeText(ann).trim() : "";
+				if (tex) out += tex.includes("\n") ? `$$${tex}$$` : `$${tex}$`;
+				break;
+			}
+			out += inlineRuns(n.children, stripImages);
+			break;
+		default: out += inlineRuns(n.children, stripImages);
+	}
+	return out.replace(/[ \t]+\n/g, "\n");
+}
+function emitBlocks(nodes, stripImages, lines, listDepth = 0, quotePrefix = "") {
+	const indent = "  ".repeat(listDepth);
+	for (const n of nodes) {
+		const prefix = quotePrefix;
+		switch (n.tag) {
+			case "": {
+				const t = n.text.trim();
+				if (t) lines.push(prefix + t);
+				break;
+			}
+			case "h1":
+			case "h2":
+			case "h3":
+			case "h4":
+			case "h5":
+			case "h6": {
+				const level = Number(n.tag[1]);
+				const t = inlineRuns(n.children, stripImages).trim();
+				if (t) lines.push(prefix + "#".repeat(level) + " " + t);
+				break;
+			}
+			case "p":
+			case "div":
+			case "section":
+			case "article":
+			case "main":
+			case "figcaption":
+			case "summary":
+			case "details":
+			case "address": {
+				if (n.children.some((c) => BLOCK_TAGS.has(c.tag) && c.tag !== "br")) {
+					emitBlocks(n.children, stripImages, lines, listDepth, prefix);
+					break;
+				}
+				const t = inlineRuns(n.children, stripImages).trim();
+				if (t) lines.push(prefix + indent + t);
+				break;
+			}
+			case "ul":
+			case "ol": {
+				let idx = 1;
+				for (const c of n.children) if (c.tag === "li") {
+					const marker = n.tag === "ol" ? `${idx++}. ` : "- ";
+					if (!c.children.some((cc) => BLOCK_TAGS.has(cc.tag) && cc.tag !== "br")) {
+						const t = inlineRuns(c.children, stripImages).trim();
+						if (t) lines.push(prefix + indent + marker + t);
+					} else {
+						const nested = [];
+						emitBlocks(c.children, stripImages, nested, listDepth + 1, prefix);
+						if (nested.length) lines.push(prefix + indent + marker + nested[0].trimStart());
+						lines.push(...nested.slice(1));
+					}
+				} else emitBlocks([c], stripImages, lines, listDepth, prefix);
+				lines.push("");
+				break;
+			}
+			case "blockquote": {
+				const inner = [];
+				emitBlocks(n.children, stripImages, inner, 0, "");
+				for (const l of inner) if (l.trim()) lines.push(prefix + "> " + l);
+				lines.push("");
+				break;
+			}
+			case "pre": {
+				const code = n.children.map((c) => nodeText(c)).join("").replace(/\n+$/, "");
+				if (code.trim()) {
+					lines.push(prefix + "```", code.split("\n").map((l) => prefix + l).join("\n"), prefix + "```");
+					lines.push("");
+				}
+				break;
+			}
+			case "table":
+				for (const c of n.children) if (c.tag === "tr") {
+					const cells = c.children.filter((cc) => cc.tag === "td" || cc.tag === "th").map((cc) => inlineRuns(cc.children, stripImages).trim().replace(/\|/g, "\\|"));
+					if (cells.length) lines.push(prefix + "| " + cells.join(" | ") + " |");
+				} else emitBlocks([c], stripImages, lines, listDepth, prefix);
+				lines.push("");
+				break;
+			case "hr":
+				lines.push(prefix + "---");
+				lines.push("");
+				break;
+			case "br":
+				lines.push("");
+				break;
+			case "script":
+				if ((n.attrs["type"] ?? "").includes("math/tex")) {
+					const tex = nodeText(n).trim();
+					if (tex) lines.push(prefix + (tex.includes("\n") ? `$$${tex}$$` : `$${tex}$`));
+				}
+				break;
+			case "span": {
+				if ((n.attrs["class"] ?? "").split(/\s+/).includes("katex")) {
+					const ann = findDescendant(n, "annotation");
+					const tex = ann ? nodeText(ann).trim() : "";
+					if (tex) lines.push(prefix + (tex.includes("\n") ? `$$${tex}$$` : `$${tex}$`));
+					break;
+				}
+				const t = inlineRuns([n], stripImages).trim();
+				if (t) lines.push(prefix + t);
+				break;
+			}
+			case "img":
+			case "picture":
+			case "svg":
+			case "figure":
+				if (!stripImages && n.tag === "img") {
+					const src = n.attrs["src"] ?? "";
+					if (src) lines.push(prefix + `![${n.attrs["alt"] ?? ""}](${src})`);
+				}
+				break;
+			case "style":
+			case "head":
+			case "nav":
+			case "button":
+			case "form":
+			case "noscript": break;
+			default: emitBlocks(n.children, stripImages, lines, listDepth, prefix);
+		}
+	}
+}
+/** 任意 HTML/XHTML → markdown（epub 章节用；数学源回收在 script 剥除之前，同上游）。 */
+function htmlToMarkdown(html, opts = {}) {
+	const stripImages = opts.stripImages ?? false;
+	let roots = parseHtml(html);
+	if (roots.length === 0) roots = parseHtml(`<html><body>${html}</body></html>`);
+	const lines = [];
+	emitBlocks(roots, stripImages, lines);
+	return lines.join("\n").replace(/\n{3,}/g, "\n\n").trim();
+}
+const STRIP_FOR_ARTICLE = /* @__PURE__ */ new Set([
+	"script",
+	"style",
+	"nav",
+	"header",
+	"footer",
+	"aside",
+	"form",
+	"noscript",
+	"button",
+	"iframe",
+	"svg"
+]);
+function textDensity(nodes) {
+	let chars = 0, tags = 0;
+	const walk = (ns) => {
+		for (const n of ns) {
+			if (n.tag === "") {
+				chars += n.text.trim().length;
+				continue;
+			}
+			if (STRIP_FOR_ARTICLE.has(n.tag) || n.tag === "div" && (n.attrs["id"] ?? n.attrs["class"] ?? "").match(/comment|sidebar|related|share|footer|nav/i)) continue;
+			tags++;
+			walk(n.children);
+		}
+	};
+	walk(nodes);
+	return {
+		chars,
+		tags
+	};
+}
+/** Pick the densest content root: <article> > <main> > <body> > whole forest. */
+function pickContentRoot(roots) {
+	const findFirst = (tag) => {
+		const walk = (ns) => {
+			for (const n of ns) {
+				if (n.tag === tag) return n;
+				const hit = walk(n.children);
+				if (hit) return hit;
+			}
+			return null;
+		};
+		return walk(roots);
+	};
+	return [findFirst("article") ?? findFirst("main") ?? findFirst("body") ?? {
+		tag: "",
+		attrs: {},
+		children: roots,
+		text: ""
+	}];
+}
+/** 从完整 HTML 抽取文章正文并转 markdown。非文章页返回 null（诚实失败，同上游契约）。 */
+function extractArticle(html, baseUrl = "") {
+	let roots = parseHtml(html);
+	if (roots.length === 0) return null;
+	const titleNode = findDescendant({
+		tag: "root",
+		attrs: {},
+		children: roots,
+		text: ""
+	}, "title");
+	const h1 = findDescendant({
+		tag: "root",
+		attrs: {},
+		children: roots,
+		text: ""
+	}, "h1");
+	const title = (titleNode ? nodeText(titleNode) : h1 ? inlineRuns(h1.children, false).trim() : "").trim() || "无标题文章";
+	const contentRoots = pickContentRoot(roots);
+	if (baseUrl) {
+		const walk = (ns) => {
+			for (const n of ns) {
+				if (n.tag === "img") {
+					const raw = n.attrs["src"] ?? "";
+					if (raw && !raw.startsWith("data:") && !/^https?:/i.test(raw)) try {
+						n.attrs["src"] = new URL(raw, baseUrl).toString();
+					} catch {}
+				}
+				walk(n.children);
+			}
+		};
+		walk(contentRoots);
+	}
+	const { chars } = textDensity(contentRoots);
+	if (chars < 120) return null;
+	const body = htmlToMarkdownOf(contentRoots, false);
+	if (!body || body.replace(/[#\s>*|-]/g, "").length < 40) return null;
+	return {
+		title,
+		markdown: body.split("\n")[0]?.replace(/^#\s+/, "").trim() === title && body.startsWith("# ") ? body : `# ${title}\n\n${body}`
+	};
+}
+function htmlToMarkdownOf(roots, stripImages) {
+	const lines = [];
+	emitBlocks(roots, stripImages, lines);
+	return lines.join("\n").replace(/\n{3,}/g, "\n\n").trim();
+}
+//#endregion
+export { htmlToMarkdown as n, extractArticle as t };

package/lib/index.d.mts

@@ -32,10 +32,11 @@ declare const Config: z<Config>;
 declare const name = "lookatstudy-plugin";
 declare const inject: string[];
 /**
- * Register the activation-gated study surface: the 20 `study_*` tools (kept
- * unregistered while dormant), the tutor persona (stable core + soul), and
- * the dynamic learner-snapshot context — every prompt text renders empty
- * while inactive, and empty sections are dropped at assembly.
+ * Register the activation-gated study surface: the 25 `study_*` tools (kept
+ * unregistered while dormant), the tutor persona (stable core + soul), the
+ * dynamic learner-snapshot context, and the `/study` command — every prompt
+ * text renders empty while inactive, and empty sections are dropped at
+ * assembly.
  * @param ctx - plugin context carrying the tool registry and system prompt.
  * @param config - validated plugin configuration.
  */