job-pro 0.7.2 → 0.7.4

package/dist/cambricon.js

@@ -3,402 +3,30 @@
 // ============================================================
 // API DISCOVERY (probed 2026-05-16)
 //
-// www.cambricon.com (the corporate site) embeds links to Moka tenant URLs
-// in its 加入我们 / careers section. Extracted slugs:
+// www.cambricon.com embeds links to Moka tenant URLs in its 加入我们 section:
 //
 //   /campus-recruitment/cambricon/44201        ← campus + intern (main entry)
 //   /recommendation-recruitment/cambricon/42452  (referral channel, overlaps)
 //   /recommendation-recruitment/cambricon/46261  (referral channel, overlaps)
 //
 // No /social-recruitment/cambricon/<siteId> URL is published — Cambricon
-// only opens 校招 / 实习 publicly through Moka. The campus SSR HTML embeds
-// `<input id="init-data" value="{...}">` containing the full first page of
-// jobs + aesIv for subsequent AES-CBC paginated calls. Same pattern as
-// `cli/src/megvii.ts`; the heavy lifting (htmlDecode, parseInitData,
-// fetchPortalHtml two-fetch cookie dance, decryptMokaEnvelope) is
-// duplicated here for now — a shared `moka.ts` factory is worth refactoring
-// to once we have 6+ Moka tenants (currently megvii/deepseek/galaxyuniversal/
-// stepfun/moonshot/+cambricon = 6 → schedule for next pass).
-import { extractResumeSignals, scoreOverlap, checkResume } from "./tencent.js";
-import { createDecipheriv } from "node:crypto";
-export { checkResume, extractResumeSignals, scoreOverlap };
-const SOURCE = "app.mokahr.com/cambricon";
-const ORG_SLUG = "cambricon";
-const CAMPUS_SITE_ID = 44201;
-const CAMPUS_URL = `https://app.mokahr.com/campus-recruitment/${ORG_SLUG}/${CAMPUS_SITE_ID}`;
-const API_ENDPOINT = "https://app.mokahr.com/api/outer/ats-apply/website/jobs/v2";
-const DEFAULT_HEADERS = {
-    "User-Agent": "Mozilla/5.0 (Macintosh; Intel Mac OS X 10_15_7) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/124.0.0.0 Safari/537.36",
-    Accept: "text/html,application/xhtml+xml,application/xml;q=0.9,*/*;q=0.8",
-    "Accept-Language": "zh-CN,zh;q=0.9,en;q=0.8",
-};
-// ---- helpers (duplicated from megvii.ts — slated for moka.ts refactor) ----
-function htmlDecode(s) {
-    return s
-        .replace(/&quot;/g, '"')
-        .replace(/&amp;/g, "&")
-        .replace(/&lt;/g, "<")
-        .replace(/&gt;/g, ">")
-        .replace(/&#x27;/g, "'")
-        .replace(/&#39;/g, "'");
-}
-function parseInitData(html) {
-    const m = html.match(/<input[^>]*id="init-data"[^>]*value="([^"]+)"/);
-    if (!m)
-        return null;
-    try {
-        return JSON.parse(htmlDecode(m[1]));
-    }
-    catch {
-        return null;
-    }
-}
-async function fetchPortalHtml(url) {
-    let response;
-    try {
-        response = await fetch(url, { method: "GET", headers: DEFAULT_HEADERS, redirect: "manual" });
-    }
-    catch (err) {
-        return { ok: false, message: `network error: ${err instanceof Error ? err.message : err}` };
-    }
-    const cookies = [];
-    const headersAny = response.headers;
-    if (typeof headersAny.getSetCookie === "function") {
-        for (const v of headersAny.getSetCookie.call(response.headers) ?? []) {
-            const c = v.split(";")[0];
-            if (c)
-                cookies.push(c);
-        }
-    }
-    if (cookies.length === 0) {
-        const raw = response.headers.get("set-cookie");
-        if (raw)
-            cookies.push(...raw.split(/,(?=[^;]+=)/).map((c) => c.split(";")[0].trim()));
-    }
-    const cookieHeader = cookies.join("; ");
-    let r2;
-    try {
-        r2 = await fetch(url, {
-            method: "GET",
-            headers: { ...DEFAULT_HEADERS, Cookie: cookieHeader },
-            redirect: "follow",
-        });
-    }
-    catch (err) {
-        return { ok: false, message: `network error: ${err instanceof Error ? err.message : err}` };
-    }
-    if (!r2.ok)
-        return { ok: false, status: r2.status, message: `HTTP ${r2.status}` };
-    const html = await r2.text();
-    return { ok: true, html, cookieHeader, status: r2.status, message: "ok" };
-}
-function decryptMokaEnvelope(envelope, aesIv) {
-    if (!envelope.data || !envelope.necromancer)
-        return null;
-    try {
-        const key = Buffer.from(envelope.necromancer, "utf8");
-        const iv = Buffer.from(aesIv, "utf8");
-        const decipher = createDecipheriv("aes-128-cbc", key, iv);
-        const plain = Buffer.concat([
-            decipher.update(Buffer.from(envelope.data, "base64")),
-            decipher.final(),
-        ]);
-        return JSON.parse(plain.toString("utf8"));
-    }
-    catch {
-        return null;
-    }
-}
-async function fetchEncryptedPage(pageNum, pageSize, aesIv, cookieHeader) {
-    const url = `${API_ENDPOINT}?orgId=${encodeURIComponent(ORG_SLUG)}`;
-    const body = {
-        orgId: ORG_SLUG,
-        siteId: String(CAMPUS_SITE_ID),
-        pageNum,
-        pageSize,
-        needStat: true,
-    };
-    let response;
-    try {
-        response = await fetch(url, {
-            method: "POST",
-            headers: {
-                ...DEFAULT_HEADERS,
-                Accept: "application/json,*/*",
-                "Content-Type": "application/json",
-                Origin: "https://app.mokahr.com",
-                Referer: CAMPUS_URL,
-                Cookie: cookieHeader,
-            },
-            body: JSON.stringify(body),
-        });
-    }
-    catch (err) {
-        return { ok: false, message: `network error: ${err instanceof Error ? err.message : err}` };
-    }
-    if (!response.ok)
-        return { ok: false, message: `HTTP ${response.status}` };
-    let envelope;
-    try {
-        envelope = await response.json();
-    }
-    catch {
-        return { ok: false, message: "bad JSON from upstream" };
-    }
-    const decoded = decryptMokaEnvelope(envelope, aesIv);
-    if (!decoded || decoded.code !== 0 || !decoded.data) {
-        return { ok: false, message: decoded?.msg || envelope?.msg || "decrypt or upstream error" };
-    }
-    return {
-        ok: true,
-        jobs: decoded.data.jobs ?? [],
-        total: decoded.data.jobStats?.total ?? 0,
-        message: "ok",
-    };
-}
-function buildCityMap(groups) {
-    const out = {};
-    if (!groups)
-        return out;
-    for (const g of groups) {
-        if (typeof g.cityId === "number" && g.label)
-            out[g.cityId] = g.label;
-    }
-    return out;
-}
-function workCitiesFor(job, cityMap) {
-    const cities = (job.locations ?? [])
-        .map((l) => {
-        if (typeof l.cityId === "number" && cityMap[l.cityId])
-            return cityMap[l.cityId];
-        return l.country || "";
-    })
-        .filter((s) => s.length > 0);
-    const uniq = [];
-    for (const c of cities)
-        if (!uniq.includes(c))
-            uniq.push(c);
-    return uniq.join(" / ");
-}
-function commitmentFor(job) {
-    if (typeof job.commitment === "string" && job.commitment.length > 0)
-        return job.commitment;
-    if (job.hireMode === 1)
-        return "全职";
-    if (job.hireMode === 2)
-        return "实习";
-    return "";
-}
-function summarize(job, cityMap) {
-    return {
-        post_id: String(job.id),
-        title: job.title ?? "",
-        project: job.zhineng?.name ?? "",
-        recruit_label: commitmentFor(job),
-        bgs: job.department?.name ?? "",
-        work_cities: workCitiesFor(job, cityMap),
-        apply_url: `${CAMPUS_URL}#/jobs/${encodeURIComponent(job.id)}`,
-    };
-}
-function matchesKeyword(job, kw) {
-    if (!kw)
-        return true;
-    const lc = kw.toLowerCase();
-    return ((job.title ?? "").toLowerCase().includes(lc) ||
-        (job.zhineng?.name ?? "").toLowerCase().includes(lc) ||
-        (job.department?.name ?? "").toLowerCase().includes(lc));
-}
-// ---- searchPositions ----
-export async function searchPositions(opts = {}) {
-    const pageSize = opts.pageSize ?? 20;
-    const page = opts.page ?? 1;
-    const keyword = opts.keyword ?? "";
-    const portal = await fetchPortalHtml(CAMPUS_URL);
-    if (!portal.ok || !portal.html) {
-        return {
-            ok: false,
-            source: SOURCE,
-            message: portal.message,
-            query: { keyword, page, pageSize },
-            positions: [],
-            total: 0,
-        };
-    }
-    const init = parseInitData(portal.html);
-    if (!init || !init.jobs || !init.jobStats) {
-        return {
-            ok: false,
-            source: SOURCE,
-            message: "Moka init-data missing jobs/jobStats",
-            query: { keyword, page, pageSize },
-            positions: [],
-            total: 0,
-        };
-    }
-    const cityMap = buildCityMap(init.jobsGroupedByLocation);
-    let jobs = init.jobs;
-    const total = init.jobStats.total ?? jobs.length;
-    if (page > 1 && init.aesIv && portal.cookieHeader) {
-        const more = await fetchEncryptedPage(page, pageSize, init.aesIv, portal.cookieHeader);
-        if (!more.ok || !more.jobs) {
-            return {
-                ok: false,
-                source: SOURCE,
-                message: `pagination failed: ${more.message}`,
-                query: { keyword, page, pageSize },
-                positions: [],
-                total,
-            };
-        }
-        jobs = more.jobs;
-    }
-    const filtered = jobs.filter((j) => matchesKeyword(j, keyword));
-    const sliced = filtered.slice(0, pageSize);
-    const positions = sliced.map((j) => summarize(j, cityMap));
-    return {
-        ok: true,
-        source: SOURCE,
-        query: { keyword, page, pageSize },
-        page,
-        page_size: pageSize,
-        total,
-        positions,
-    };
-}
-// ---- fetchAllPositions ----
-export async function fetchAllPositions(opts = {}) {
-    const pageSize = opts.pageSize ?? 20;
-    const maxPages = Math.max(1, opts.maxPages ?? 50);
-    const keyword = opts.keyword ?? "";
-    const portal = await fetchPortalHtml(CAMPUS_URL);
-    if (!portal.ok || !portal.html) {
-        return {
-            ok: false,
-            source: SOURCE,
-            message: portal.message,
-            total: 0,
-            fetched: 0,
-            positions: [],
-        };
-    }
-    const init = parseInitData(portal.html);
-    if (!init || !init.jobs || !init.jobStats || !init.aesIv) {
-        return {
-            ok: false,
-            source: SOURCE,
-            message: "Moka init-data missing required fields",
-            total: 0,
-            fetched: 0,
-            positions: [],
-        };
-    }
-    const cityMap = buildCityMap(init.jobsGroupedByLocation);
-    const total = init.jobStats.total ?? 0;
-    const collected = [...init.jobs];
-    let page = 2;
-    while (collected.length < total && page <= maxPages) {
-        const more = await fetchEncryptedPage(page, pageSize, init.aesIv, portal.cookieHeader ?? "");
-        if (!more.ok || !more.jobs || more.jobs.length === 0)
-            break;
-        collected.push(...more.jobs);
-        page += 1;
-    }
-    const filtered = collected.filter((j) => matchesKeyword(j, keyword));
-    return {
-        ok: true,
-        source: SOURCE,
-        total,
-        fetched: filtered.length,
-        positions: filtered.map((j) => summarize(j, cityMap)),
-    };
-}
-// ---- fetchPositionDetail ----
-export async function fetchPositionDetail(postId) {
-    return {
-        ok: false,
-        source: SOURCE,
-        message: "Moka detail endpoint requires the same encrypted-session flow; not implemented. " +
-            "Use the apply_url deeplink for the full JD.",
-        post_id: postId,
-        apply_url: `${CAMPUS_URL}#/jobs/${encodeURIComponent(postId)}`,
-    };
-}
-// ---- fetchDictionaries ----
-export async function fetchDictionaries() {
-    const portal = await fetchPortalHtml(CAMPUS_URL);
-    if (!portal.ok || !portal.html) {
-        return { ok: false, source: SOURCE, message: portal.message };
-    }
-    const init = parseInitData(portal.html);
-    if (!init) {
-        return { ok: false, source: SOURCE, message: "Moka init-data missing" };
-    }
-    return {
-        ok: true,
-        source: SOURCE,
-        locations: init.jobsGroupedByLocation ?? [],
-        moka_org: { slug: ORG_SLUG, id: CAMPUS_SITE_ID, url: CAMPUS_URL },
-    };
-}
-// ---- notices (no public endpoint) ----
-const NOTICES_STUB_MSG = "Cambricon (寒武纪): no public notices endpoint on Moka tenant";
-export async function listNotices() {
-    return {
-        ok: false,
-        source: SOURCE,
-        message: NOTICES_STUB_MSG,
-        notices: [],
-    };
-}
-export async function getNotice(noticeId) {
-    return {
-        ok: false,
-        source: SOURCE,
-        message: NOTICES_STUB_MSG,
-        notice_id: noticeId,
-    };
-}
-export async function findNoticesByQuestion(question, _opts = {}) {
-    return {
-        ok: false,
-        source: SOURCE,
-        question,
-        message: NOTICES_STUB_MSG,
-        matches: [],
-    };
-}
-// ---- matchResume ----
-export async function matchResume(text, opts = {}) {
-    const { terms, cities } = extractResumeSignals(text ?? "");
-    const candidates = Math.max(20, opts.candidates ?? 100);
-    const search = await fetchAllPositions({
-        pageSize: 20,
-        maxPages: Math.ceil(candidates / 15),
-    });
-    if (!search.ok) {
-        return {
-            ok: false,
-            source: SOURCE,
-            extracted_terms: terms,
-            city_preferences: cities,
-            matches: [],
-            message: search.message,
-        };
-    }
-    const topN = Math.max(1, opts.topN ?? 10);
-    const scored = search.positions
-        .map((p) => ({
-        p,
-        score: scoreOverlap(`${p.title} ${p.project} ${p.bgs}`, terms, cities).score,
-    }))
-        .sort((a, b) => b.score - a.score)
-        .slice(0, topN)
-        .map((x) => x.p);
-    return {
-        ok: true,
-        source: SOURCE,
-        extracted_terms: terms,
-        city_preferences: cities,
-        matches: scored,
-    };
-}
+// only opens 校招 / 实习 publicly through Moka. Same factory as
+// `cli/src/moka.ts` (used by megvii / geely / etc.).
+import { createAdapter } from "./moka.js";
+const adapter = createAdapter({
+    orgSlug: "cambricon",
+    label: "Cambricon",
+    channels: [
+        { siteId: 44201, kind: "campus-recruitment", recruitType: "campus" },
+    ],
+    defaultRecruitType: "campus",
+});
+export const searchPositions = adapter.searchPositions;
+export const fetchAllPositions = adapter.fetchAllPositions;
+export const fetchPositionDetail = adapter.fetchPositionDetail;
+export const fetchDictionaries = adapter.fetchDictionaries;
+export const listNotices = adapter.listNotices;
+export const getNotice = adapter.getNotice;
+export const findNoticesByQuestion = adapter.findNoticesByQuestion;
+export const matchResume = adapter.matchResume;
+export const checkResume = adapter.checkResume;

package/dist/cdp.js

@@ -0,0 +1,191 @@
+// Headless-browser helper for adapters whose upstream is gated by anti-bot
+// signatures that the CLI can't reproduce from raw HTTP.
+//
+// Usage pattern:
+//   1. `await getBrowser()` returns a process-singleton puppeteer-core Browser
+//      attached to the user's system Chrome.
+//   2. Call `viaBrowser(url, async page => …)` to navigate and run a fn in
+//      the page context, then receive the return value.
+//
+// Why puppeteer-core (not puppeteer): we attach to the user's existing
+// Chrome installation; no 100MB Chromium download. Trade-off: we need a
+// working Chrome executable path.
+//
+// Failure modes:
+//   * puppeteer-core not installed → ENOENT on dynamic import → caller
+//     receives `{ ok:false, reason:"puppeteer-not-installed", message: … }`
+//     and renders it as the canonical ok:false stub.
+//   * No Chrome found at any well-known path → same error shape with
+//     `reason:"chrome-not-found"`.
+//   * Browser launch failed (sandbox, profile lock, …) → `reason:"launch-failed"`.
+import { existsSync } from "node:fs";
+const CHROME_PATHS = [
+    // macOS
+    "/Applications/Google Chrome.app/Contents/MacOS/Google Chrome",
+    "/Applications/Google Chrome Canary.app/Contents/MacOS/Google Chrome Canary",
+    "/Applications/Chromium.app/Contents/MacOS/Chromium",
+    // Linux
+    "/usr/bin/google-chrome",
+    "/usr/bin/google-chrome-stable",
+    "/usr/bin/chromium",
+    "/usr/bin/chromium-browser",
+    // Windows (when running under WSL / Git Bash)
+    "/c/Program Files/Google/Chrome/Application/chrome.exe",
+];
+const USER_AGENT = "Mozilla/5.0 (Macintosh; Intel Mac OS X 10_15_7) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/138.0.0.0 Safari/537.36";
+// ---------- singleton browser ----------
+let _browser = null;
+let _browserError = null;
+let _launching = null;
+async function loadPuppeteer() {
+    try {
+        // Dynamic import; if puppeteer-core was tree-shaken or uninstalled,
+        // this rejects with ERR_MODULE_NOT_FOUND.
+        const mod = (await import("puppeteer-core"));
+        return { ok: true, mod: mod.default };
+    }
+    catch (err) {
+        return {
+            ok: false,
+            error: {
+                reason: "puppeteer-not-installed",
+                message: "`puppeteer-core` is not installed. Install it locally with " +
+                    "`npm i puppeteer-core` (or `pnpm add puppeteer-core`). " +
+                    `Original error: ${err instanceof Error ? err.message : String(err)}`,
+            },
+        };
+    }
+}
+function findChrome() {
+    if (process.env.JOB_PRO_CHROME && existsSync(process.env.JOB_PRO_CHROME)) {
+        return process.env.JOB_PRO_CHROME;
+    }
+    for (const p of CHROME_PATHS) {
+        if (existsSync(p))
+            return p;
+    }
+    return null;
+}
+async function launchOnce() {
+    const pp = await loadPuppeteer();
+    if (!pp.ok)
+        return pp.error;
+    const chrome = findChrome();
+    if (!chrome) {
+        return {
+            reason: "chrome-not-found",
+            message: "No Chrome/Chromium executable found. Tried: " +
+                CHROME_PATHS.join(", ") +
+                ". Set $JOB_PRO_CHROME=/path/to/chrome to override.",
+        };
+    }
+    try {
+        const browser = await pp.mod.launch({
+            executablePath: chrome,
+            headless: true,
+            args: [
+                "--no-sandbox",
+                "--disable-blink-features=AutomationControlled",
+                "--disable-features=IsolateOrigins,site-per-process",
+            ],
+        });
+        return browser;
+    }
+    catch (err) {
+        return {
+            reason: "launch-failed",
+            message: `Chrome failed to launch: ${err instanceof Error ? err.message : String(err)}`,
+        };
+    }
+}
+/** Get a process-singleton headless browser. Subsequent calls reuse it. */
+export async function getBrowser() {
+    if (_browser)
+        return { ok: true, browser: _browser };
+    if (_browserError)
+        return { ok: false, error: _browserError };
+    if (!_launching) {
+        _launching = launchOnce();
+    }
+    const result = await _launching;
+    _launching = null;
+    if ("reason" in result) {
+        _browserError = result;
+        return { ok: false, error: result };
+    }
+    _browser = result;
+    return { ok: true, browser: result };
+}
+/** Close the singleton browser (call before process exit). */
+export async function closeBrowser() {
+    if (_browser) {
+        try {
+            await _browser.close();
+        }
+        catch {
+            /* ignore */
+        }
+        _browser = null;
+    }
+}
+// On Node exit, best-effort close the browser to avoid zombie processes.
+let _exitHookInstalled = false;
+function ensureExitHook() {
+    if (_exitHookInstalled)
+        return;
+    _exitHookInstalled = true;
+    const cleanup = () => {
+        if (_browser) {
+            try {
+                // synchronous best-effort kill; puppeteer launches Chrome as a child
+                // process tracked by the Browser object, so close() handles SIGTERM.
+                void _browser.close().catch(() => undefined);
+            }
+            catch {
+                /* ignore */
+            }
+        }
+    };
+    process.on("exit", cleanup);
+    process.on("SIGINT", () => {
+        cleanup();
+        process.exit(130);
+    });
+    process.on("SIGTERM", () => {
+        cleanup();
+        process.exit(143);
+    });
+}
+/** Open a page, run fn against it, and close the page. The singleton browser stays open. */
+export async function withPage(fn) {
+    ensureExitHook();
+    const b = await getBrowser();
+    if (!b.ok)
+        return b;
+    let page = null;
+    try {
+        page = await b.browser.newPage();
+        await page.setUserAgent(USER_AGENT);
+        const value = await fn(page);
+        return { ok: true, value };
+    }
+    catch (err) {
+        return {
+            ok: false,
+            error: {
+                reason: "launch-failed",
+                message: `page operation failed: ${err instanceof Error ? err.message : String(err)}`,
+            },
+        };
+    }
+    finally {
+        if (page) {
+            try {
+                await page.close();
+            }
+            catch {
+                /* ignore */
+            }
+        }
+    }
+}