lasso-mcp 1.18.1 → 1.18.2

package/.dedao-extract/analyze.mjs

@@ -0,0 +1,169 @@
+#!/usr/bin/env node
+// analyze.mjs — v2 管线共享分析器（零光栅快测 + 墨迹终检 + 文字零丢失）
+// 白盒依据：P14-探察-二轮 §3（S-A' 闭环：pdfimages 0.1s + pdftotext -bbox 0.07s/页；
+// pdftoppm 对含课程表 PDF 约 2min/份，只作出厂终检，禁入主管线）
+import { execFileSync } from "node:child_process";
+import * as fs from "node:fs";
+import * as path from "node:path";
+import * as os from "node:os";
+
+// 打印参数（extract-batch3 同款）：A4、四边 0.4in → pt 域
+export const PT_PER_IN = 72;
+export const PAGE_W_IN = 8.268, PAGE_H_IN = 11.693, MARGIN_IN = 0.4;
+export const CONTENT_TOP = MARGIN_IN * 72; // 28.8pt
+export const PAGE_H_PT = PAGE_H_IN * 72; // 841.92
+export const CONTENT_BOTTOM = PAGE_H_PT - MARGIN_IN * 72; // 813.12
+export const CONTENT_H_PT = CONTENT_BOTTOM - CONTENT_TOP; // 784.32
+export const PAGE_CSS = Math.round((PAGE_H_IN - 2 * MARGIN_IN) * 96); // 1046 css px/页
+export const PROMO_NAT = { w: 1080, h: 607 }; // P14 §1.1 尾部宣传图自然尺寸指纹
+
+function tmp(pdf) {
+  return path.join(os.tmpdir(), `dz-ana-${process.pid}-${path.basename(pdf)}`);
+}
+
+// ---- pdfimages -list 解析 ----
+export function pdfImages(pdf) {
+  const out = execFileSync("pdfimages", ["-list", pdf], { maxBuffer: 64 << 20 }).toString();
+  const rows = [];
+  for (const line of out.split("\n").slice(2)) {
+    const t = line.trim().split(/\s+/);
+    if (t.length < 14 || !/^\d+$/.test(t[0])) continue;
+    rows.push({
+      page: +t[0], num: +t[1], type: t[2], w: +t[3], h: +t[4],
+      enc: t[8], xppi: +t[t.length - 4], yppi: +t[t.length - 3],
+    });
+  }
+  return rows.filter((r) => r.type === "image"); // smask=同对象 alpha 通道，不计
+}
+
+// ---- pdftotext -bbox 解析（词级坐标） ----
+export function bboxWords(pdf) {
+  const out = execFileSync("pdftotext", ["-bbox", pdf, "-"], { maxBuffer: 128 << 20 }).toString();
+  const pages = [];
+  let page = null;
+  const re = /<page width="([\d.]+)" height="([\d.]+)">|<word xMin="([\d.-]+)" yMin="([\d.-]+)" xMax="([\d.-]+)" yMax="([\d.-]+)">([\s\S]*?)<\/word>/g;
+  let m;
+  while ((m = re.exec(out))) {
+    if (m[1]) { page = { w: +m[1], h: +m[2], words: [] }; pages.push(page); }
+    else if (page) {
+      page.words.push({ x0: +m[3], y0: +m[4], x1: +m[5], y1: +m[6], text: m[7] });
+    }
+  }
+  return pages;
+}
+
+// ---- 快测洞报告（S-A' 闭环核心） ----
+// 判定：非末页 ∧ 文本尾空 >12% ∧ 版面充满度 <0.88（文本跨距 + 图显示高 + 图间隙）
+// 课程表切片页（全页图）被充满度豁免；图片收尾页被充满度豁免（bbox 只见文字的假洞）
+// P23 修复：页脚页码（displayHeaderFooter，y≈821-827 > CONTENT_BOTTOM=813.12）必须排除——
+// 路线b 合并 PDF 带页码页脚，计入会把 textBottom 钉死在页脚 → tailPct 恒 0 → 洞检测全盲。
+// 单章管线（engine）无页脚，此过滤为无操作，行为不变。
+export function holeReport(pdf, { tailPct = 12, fullFloor = 0.88, gapPt = 24 } = {}) {
+  const imgs = pdfImages(pdf);
+  const pages = bboxWords(pdf);
+  const n = pages.length;
+  const per = pages.map((pg, i) => {
+    const contentWords = pg.words.filter((w) => w.y1 <= CONTENT_BOTTOM + 0.75); // 页脚词排除
+    const ys = contentWords.map((w) => w.y1);
+    const textBottom = ys.length ? Math.max(...ys) : null;
+    const textTop = contentWords.length ? Math.min(...contentWords.map((w) => w.y0)) : null;
+    const imgsOnPage = imgs.filter((r) => r.page === i + 1);
+    const imgPt = Math.min(
+      CONTENT_H_PT,
+      imgsOnPage.reduce((s, r) => s + (r.yppi > 0 ? (r.h / r.yppi) * PT_PER_IN : 0), 0),
+    );
+    const textSpan = textBottom !== null ? textBottom - CONTENT_TOP : 0;
+    const used = Math.max(textSpan, 0) + imgPt + imgsOnPage.length * gapPt;
+    const fullness = +(used / CONTENT_H_PT).toFixed(3);
+    const tailPt = textBottom !== null ? CONTENT_BOTTOM - textBottom : CONTENT_H_PT;
+    const tailPctV = +(Math.max(0, tailPt) / CONTENT_H_PT * 100).toFixed(1);
+    const isLast = i === n - 1;
+    const hole = !isLast && tailPctV > tailPct && fullness < fullFloor;
+    const freePt = +Math.max(0, CONTENT_H_PT - used).toFixed(1); // 版面自由空间（保守：文本跨距+图高+图间隙的加性模型）
+    const freePct = +(freePt / CONTENT_H_PT * 100).toFixed(1);
+    return { page: i + 1, textBottom, textTop, imgCount: imgsOnPage.length, imgPt: Math.round(imgPt), fullness, tailPt: +tailPt.toFixed(1), tailPct: tailPctV, freePt, freePct, isLast, hole };
+  });
+  return { pdf, pages: n, holes: per.filter((p) => p.hole), per };
+}
+
+// ---- 宣传图断言：PDF 内不得存在 1080×607 图 ----
+export function promoInPdf(pdf) {
+  return pdfImages(pdf).filter((r) => r.w === PROMO_NAT.w && r.h === PROMO_NAT.h);
+}
+
+// ---- 文字零丢失（P14 §3 定案：cleanup 态 innerText vs pdftotext，missing 必须为空） ----
+export function textZeroLoss(pdf, innerTxt) {
+  const pdftxt = execFileSync("pdftotext", [pdf, "-"], { maxBuffer: 128 << 20 }).toString();
+  const norm = (s) => s.normalize("NFC").replace(/[\s­​‌‍]/g, "");
+  const a = norm(pdftxt), b = norm(innerTxt); // a=PDF 侧，b=DOM 侧
+  // 多重集 diff（字符级，CJK 字≈词；PDF 侧多出字符=良性：CSS 列表序号等）
+  const cnt = new Map();
+  for (const ch of a) cnt.set(ch, (cnt.get(ch) || 0) + 1);
+  const missCnt = new Map();
+  for (const ch of b) {
+    const left = cnt.get(ch) || 0;
+    if (left > 0) cnt.set(ch, left - 1);
+    else missCnt.set(ch, (missCnt.get(ch) || 0) + 1);
+  }
+  const missing = [...missCnt.entries()].map(([ch, c]) => `${ch}x${c}`);
+  // 保序检查：b 必须是 a 的子序列
+  let j = 0;
+  for (let i = 0; i < a.length && j < b.length; i++) if (a[i] === b[j]) j++;
+  return { missing, missingEmpty: missing.length === 0, inOrder: j === b.length, pdfChars: a.length, domChars: b.length };
+}
+
+// ---- 墨迹终检（慢：pdftoppm PGM，仅出厂抽检/3 章对比证据） ----
+export function inkReport(pdf, { first = 1, last = null, tailInkTh = 242, dpi = 36 } = {}) {
+  const dir = tmp(pdf);
+  fs.rmSync(dir, { recursive: true, force: true });
+  fs.mkdirSync(dir, { recursive: true });
+  const npages = last ?? bboxWords(pdf).length;
+  const args = ["-f", String(first), "-l", String(last ?? npages), "-r", String(dpi), "-gray", pdf, path.join(dir, "p")];
+  execFileSync("pdftoppm", args, { maxBuffer: 64 << 20, timeout: 570000 });
+  const files = fs.readdirSync(dir).filter((f) => f.endsWith(".pgm")).sort();
+  const out = [];
+  for (const f of files) {
+    const buf = fs.readFileSync(path.join(dir, f));
+    let p = 0;
+    const tok = () => {
+      while (p < buf.length && (buf[p] === 0x20 || buf[p] === 0x0a || buf[p] === 0x0d || buf[p] === 0x09)) p++;
+      if (buf[p] === 0x23) { while (p < buf.length && buf[p] !== 0x0a) p++; return tok(); }
+      const s = p;
+      while (p < buf.length && !(buf[p] === 0x20 || buf[p] === 0x0a || buf[p] === 0x0d || buf[p] === 0x09)) p++;
+      return buf.toString("latin1", s, p);
+    };
+    if (tok() !== "P5") continue;
+    const w = +tok(), h = +tok(); tok(); p++;
+    const data = buf.subarray(p, p + w * h);
+    let lastInk = -1, firstInk = -1;
+    for (let y = 0; y < h && firstInk < 0; y++)
+      for (let x = 0; x < w; x++) if (data[y * w + x] < tailInkTh) { firstInk = y; break; }
+    for (let y = h - 1; y >= 0; y--) {
+      let ink = false;
+      for (let x = 0; x < w; x++) if (data[y * w + x] < tailInkTh) { ink = true; break; }
+      if (ink) { lastInk = y; break; }
+    }
+    out.push({ page: +f.match(/(\d+)\.pgm$/)[1], tailBlankPct: +(((h - 1 - lastInk) / h) * 100).toFixed(1), headBlankPct: firstInk < 0 ? 100 : +((firstInk / h) * 100).toFixed(1) });
+  }
+  fs.rmSync(dir, { recursive: true, force: true });
+  return out;
+}
+
+// ---- Chrome 可见性守卫（P16：PID 定向，禁进程名全量） ----
+export function chromeGuard() {
+  try {
+    const pid = execFileSync("zsh", ["-c", "lsof -nP -iTCP:9226 -sTCP:LISTEN | awk 'NR>1{print $2}' | sort -u | head -1"])
+      .toString().trim();
+    if (!pid || !/^\d+$/.test(pid)) return { ok: false, err: "no-pid" };
+    const vis = execFileSync("osascript", ["-e",
+      `tell application "System Events" to get visible of (first process whose unix id is ${pid})`]).toString().trim();
+    if (vis === "true") {
+      execFileSync("osascript", ["-e",
+        `tell application "System Events" to set visible of (first process whose unix id is ${pid}) to false`]);
+      return { ok: true, pid, was: "visible", action: "rehidden" };
+    }
+    return { ok: true, pid, was: "hidden", action: "none" };
+  } catch (e) {
+    return { ok: false, err: String(e).slice(0, 120) };
+  }
+}

package/.dedao-extract/dbg-fixscreen.mjs

@@ -0,0 +1,43 @@
+#!/usr/bin/env node
+// dbg-fixscreen.mjs — 捕获 JS_FIX_SCREEN 的原始 evaluate 返回（定位 dropped=undefined 根因）
+import { Client } from "@modelcontextprotocol/sdk/client/index.js";
+import { StdioClientTransport } from "@modelcontextprotocol/sdk/client/stdio.js";
+import * as path from "node:path";
+import * as fs from "node:fs";
+
+const ROOT = "/Users/wangdong/Documents/Project/cc-control-all/lasso";
+const OUT = path.join(ROOT, ".dedao-extract");
+const TARGET = "https://www.dedao.cn/course/article?id=6EBOqDNZ27YlVdg8oKm4bQ1odMAkgz";
+
+// 从 extract-v2.mjs 源码原样抽取 JS_FIX_SCREEN（保证测的就是产线代码）
+const src = fs.readFileSync(path.join(OUT, "extract-v2.mjs"), "utf8");
+const m = src.match(/const JS_FIX_SCREEN = `([\s\S]*?)`;/);
+if (!m) { console.error("extract failed"); process.exit(1); }
+const PAGE_CSS = 1046;
+const JS_FIX_SCREEN = m[1].replace("${PAGE_CSS}", String(PAGE_CSS));
+
+const client = new Client({ name: "dedao-dbg-fs", version: "1.0.0" });
+const transport = new StdioClientTransport({
+  command: "/bin/zsh",
+  args: ["-c", `exec node dist/index.js 2>>${JSON.stringify(path.join(OUT, "server-dbg-stderr.log"))}`],
+  cwd: ROOT, env: { ...process.env, LASSO_CDP_PORT: "9226" },
+});
+await client.connect(transport);
+async function browse(action, options = {}, timeoutMs = 180000) {
+  const res = await client.callTool({ name: "browse_logged_in", arguments: { url: TARGET, action, options } }, undefined, { timeout: timeoutMs });
+  return JSON.parse(res.content[0].text);
+}
+
+// navigate 先行（P6：新 server 会话需先绑定 page）
+const nav = await browse("navigate", {});
+console.log("[nav]", JSON.stringify(nav).slice(0, 120));
+await new Promise((r) => setTimeout(r, 2500));
+const t = await browse("evaluate", { js: `return JSON.stringify({title:(document.querySelector('div.article-body-wrap .article-title')||{textContent:''}).textContent.trim().slice(0,30),figs:document.querySelectorAll('div.article-body-wrap figure').length});` });
+console.log("[state]", t.data?.preview ?? JSON.stringify(t).slice(0, 200));
+
+const r = await browse("evaluate", { js: JS_FIX_SCREEN }, undefined, { timeout: 120000 });
+console.log("[outcome]", r.outcome, "| [error]", r.error ?? null);
+console.log("[preview]", (r.data?.preview ?? "").slice(0, 1500));
+await transport.close();
+await new Promise((r) => setTimeout(r, 300));
+process.exit(0);

package/.dedao-extract/dbg-text.mjs

@@ -0,0 +1,25 @@
+#!/usr/bin/env node
+// debug text fetch — 看 evaluate 原始返回
+import { Client } from "@modelcontextprotocol/sdk/client/index.js";
+import { StdioClientTransport } from "@modelcontextprotocol/sdk/client/stdio.js";
+const ROOT = "/Users/wangdong/Documents/Project/cc-control-all/lasso";
+const client = new Client({ name: "dbg", version: "1" });
+const transport = new StdioClientTransport({
+  command: "/bin/zsh",
+  args: ["-c", `exec node dist/index.js 2>>/tmp/dbg-stderr.log`],
+  cwd: ROOT,
+  env: { ...process.env, LASSO_CDP_PORT: "9226" },
+});
+await client.connect(transport);
+const call = (js) =>
+  client
+    .callTool(
+      { name: "browse_logged_in", arguments: { url: "https://www.dedao.cn/course/article?id=1", action: "evaluate", options: { js } } },
+      undefined,
+      { timeout: 60000 },
+    )
+    .then((r) => r.content[0].text);
+console.log("A:", await call(`return JSON.stringify((document.body.innerText).slice(0,60));`));
+console.log("B:", await call(`return (document.body.innerText||'').length;`));
+await transport.close();
+setTimeout(() => process.exit(0), 300);

package/.dedao-extract/dbg-text2.mjs

@@ -0,0 +1,20 @@
+import { Client } from "@modelcontextprotocol/sdk/client/index.js";
+import { StdioClientTransport } from "@modelcontextprotocol/sdk/client/stdio.js";
+const ROOT = "/Users/wangdong/Documents/Project/cc-control-all/lasso";
+const client = new Client({ name: "dbg2", version: "1" });
+const transport = new StdioClientTransport({
+  command: "/bin/zsh",
+  args: ["-c", `exec node dist/index.js 2>>/tmp/dbg2-stderr.log`],
+  cwd: ROOT,
+  env: { ...process.env, LASSO_CDP_PORT: "9226" },
+});
+await client.connect(transport);
+const call = (action, options) =>
+  client.callTool({ name: "browse_logged_in", arguments: { url: "https://www.dedao.cn/course/article?id=6EBOqDNZ27YlVdg8oKm4bQ1odMAkgz", action, options } }, undefined, { timeout: 90000 }).then((r) => r.content[0].text);
+console.log("nav:", (await call("navigate", {})).slice(0, 120));
+await new Promise(r => setTimeout(r, 4000));
+console.log("A body.innerText len:", await call("evaluate", { js: `return (document.body.innerText||'').length + ' | head=' + (document.body.innerText||'').slice(0,30);` }));
+console.log("B wrap innerText len:", await call("evaluate", { js: `return (document.querySelector('div.article-body-wrap').innerText||'').length;` }));
+console.log("C textContent len:", await call("evaluate", { js: `return (document.querySelector('div.article-body-wrap').textContent||'').length;` }));
+await transport.close();
+setTimeout(() => process.exit(0), 300);

package/.dedao-extract/dbg-text3.mjs

@@ -0,0 +1,18 @@
+import { Client } from "@modelcontextprotocol/sdk/client/index.js";
+import { StdioClientTransport } from "@modelcontextprotocol/sdk/client/stdio.js";
+const ROOT = "/Users/wangdong/Documents/Project/cc-control-all/lasso";
+const client = new Client({ name: "dbg3", version: "1" });
+const transport = new StdioClientTransport({
+  command: "/bin/zsh",
+  args: ["-c", `exec node dist/index.js 2>>/tmp/dbg3-stderr.log`],
+  cwd: ROOT,
+  env: { ...process.env, LASSO_CDP_PORT: "9226" },
+});
+await client.connect(transport);
+const call = (action, options) =>
+  client.callTool({ name: "browse_logged_in", arguments: { url: "https://www.dedao.cn/course/article?id=6EBOqDNZ27YlVdg8oKm4bQ1odMAkgz", action, options } }, undefined, { timeout: 90000 }).then((r) => r.content[0].text);
+console.log("nav FULL:", await call("navigate", {}));
+await new Promise(r => setTimeout(r, 5000));
+console.log("A:", await call("evaluate", { js: `return (document.body.innerText||'').length;` }));
+await transport.close();
+setTimeout(() => process.exit(0), 300);

package/.dedao-extract/dbg-text4.mjs

@@ -0,0 +1,27 @@
+import { Client } from "@modelcontextprotocol/sdk/client/index.js";
+import { StdioClientTransport } from "@modelcontextprotocol/sdk/client/stdio.js";
+const ROOT = "/Users/wangdong/Documents/Project/cc-control-all/lasso";
+const client = new Client({ name: "dbg4", version: "1" });
+const transport = new StdioClientTransport({
+  command: "/bin/zsh",
+  args: ["-c", `exec node dist/index.js 2>>/tmp/dbg4-stderr.log`],
+  cwd: ROOT,
+  env: { ...process.env, LASSO_CDP_PORT: "9226" },
+});
+await client.connect(transport);
+const call = (action, options) =>
+  client.callTool({ name: "browse_logged_in", arguments: { url: "https://www.dedao.cn/course/article?id=6EBOqDNZ27YlVdg8oKm4bQ1odMAkgz", action, options } }, undefined, { timeout: 90000 }).then((r) => r.content[0].text);
+console.log("nav:", (await call("navigate", {})).slice(0, 80));
+for (let i = 0; i < 10; i++) {
+  await new Promise(r => setTimeout(r, 1200));
+  const j = JSON.parse(await call("evaluate", { js: `return JSON.stringify({rs:document.readyState,t:!!document.querySelector('.article-title'),len:(document.body.innerText||'').length});` }));
+  const p = j?.data?.preview;
+  console.log(i, p);
+  if (p && p.includes('"t":true')) break;
+}
+// 精确复刻 evText 第一片
+const raw = await call("evaluate", { js: `return JSON.stringify((document.body.innerText).slice(0,1800));` });
+const rj = JSON.parse(raw);
+console.log("outcome:", rj.outcome, "| preview len:", (rj.data?.preview ?? "").length, "| preview head:", JSON.stringify((rj.data?.preview ?? "").slice(0, 80)));
+await transport.close();
+setTimeout(() => process.exit(0), 300);

package/.dedao-extract/dbg-text5.mjs

@@ -0,0 +1,31 @@
+import { Client } from "@modelcontextprotocol/sdk/client/index.js";
+import { StdioClientTransport } from "@modelcontextprotocol/sdk/client/stdio.js";
+import * as fs from "node:fs";
+const ROOT = "/Users/wangdong/Documents/Project/cc-control-all/lasso";
+const client = new Client({ name: "dbg5", version: "1" });
+const transport = new StdioClientTransport({
+  command: "/bin/zsh",
+  args: ["-c", `exec node dist/index.js 2>>/tmp/dbg5-stderr.log`],
+  cwd: ROOT,
+  env: { ...process.env, LASSO_CDP_PORT: "9226" },
+});
+await client.connect(transport);
+const call = (action, options) =>
+  client.callTool({ name: "browse_logged_in", arguments: { url: "https://www.dedao.cn/course/article?id=6EBOqDNZ27YlVdg8oKm4bQ1odMAkgz", action, options } }, undefined, { timeout: 90000 }).then((r) => r.content[0].text);
+console.log("nav ok:", JSON.parse(await call("navigate", {})).outcome);
+await new Promise(r => setTimeout(r, 4000));
+// b3 JS_CLEANUP 原样（从 scout7 文件抽取，保证一字不差）
+const src = fs.readFileSync(".dedao-extract/scout7-round2.mjs", "utf8");
+const m = src.match(/const JS_CLEANUP = `return \(\(\)=>\{[\s\S]*?\}\)\(\)`;/);
+await call("evaluate", { js: m[0].replace("const JS_CLEANUP = ", "").replace(/;$/, "") });
+const raw = await call("evaluate", { js: `return JSON.stringify((document.body.innerText).slice(0,1800));` });
+const rj = JSON.parse(raw);
+const p = rj?.data?.preview ?? "";
+console.log("outcome:", rj.outcome, "| preview len:", p.length);
+console.log("preview head:", JSON.stringify(p.slice(0, 120)));
+console.log("preview tail:", JSON.stringify(p.slice(-120)));
+try { const s = JSON.parse(p); console.log("parse OK, string len:", s.length, "| head:", JSON.stringify(s.slice(0, 40))); } catch (e) { console.log("parse FAIL:", e.message); }
+const raw2 = await call("evaluate", { js: `return JSON.stringify((document.body.innerText||'').length);` });
+console.log("innerText len direct:", JSON.parse(raw2)?.data?.preview);
+await transport.close();
+setTimeout(() => process.exit(0), 300);