first commit

.gitignore

@@ -0,0 +1,10 @@
+data/
+downloads/
+local-records/
+test/history/
+test/artifacts/
+发布去重记录.json
+__pycache__/
+*.py[cod]
+.venv/
+.DS_Store

README.md

@@ -0,0 +1,97 @@
+# 新闻工作台
+
+将新闻下载、文章打包、浏览器导入、排版与换号联动整合到本地控制台。
+
+**当前版本：0.1。可完成下载 → 入队 → 导入 → 排版 → 等待人工发表。结果页自动核对按用户决定暂留待接入；人工提交后会记录候选结果并暂停，不会自动计数、换号或处理下一篇。** 换号调度和计数接口已完成离线验证，完整在线链路仍需补齐结果页 HTML 后验证。
+
+## 启动
+
+双击本目录的 `start.command`，或在本目录运行：
+
+```sh
+uv run --no-project --python 3.12 --with-requirements requirements.txt python app.py
+```
+
+默认打开 http://127.0.0.1:8770/ 。保留启动窗口，按 Ctrl+C 停止本地服务。修改端口使用 `--port 8772`；不自动打开浏览器使用 `--no-browser`。首次运行 uv 可能下载 Python 或依赖，后续可以复用缓存。
+
+页面必须通过本地服务打开，不要直接双击 `static/index.html`。Python 依赖仅用于下载和内容处理；浏览器脚本没有远程脚本依赖。
+
+## 浏览器安装与配对
+
+1. 在油猴安装控制台“浏览器连接”区域的两个脚本：
+   - `browser/local-bridge.user.js`：本地连接、配对与常驻终止按钮。
+   - `browser/auto-news.user.js`：导入、排版和任务执行。它由 `importer.js`、`formatter.js`、`runner.js` 构建。
+2. 停用旧的独立导入与排版油猴脚本，避免同一编辑器重复挂载。
+3. 换号插件仍由用户安装，位于项目根目录的 `dww11-account-switch-browser-extension/`，没有复制到本目录。本次在其 `content.js` 中增加了不含凭据的状态属性及终止事件监听。请在浏览器扩展管理页重新加载**这个原安装目录**，再刷新网站。旧 ZIP 未重新打包，不包含本次修改。
+4. 在控制台点击“显示本次连接码”。在论坛左下角“新闻工作台 · 本地连接”展开设置，填写端口和连接码并保存。
+5. 服务重启后连接码会改变，需要重新配对；任务和计数保留，但服务不会自动恢复运行。
+
+本机 HTTP 通信通过油猴特权请求完成，不要求论坛跨域开放本地服务。真实浏览器的脚本管理器配置、本地网络权限和网站编辑器兼容性仍需首次实机验证。配对码保存在油猴存储，不写入页面全局变量。下载 API Key 不进入文章包，不发给论坛。
+
+## 操作流程
+
+1. 填写 Key，获取频道、选择数量并下载。完整保存的文章会自动校验并进入队列；下载失败和入队失败分别显示，已保存的文章不因入队失败而删除。
+2. 也可展开“导入已有文章目录”，填写单篇目录或含多个文章子目录的根目录，点击检查。留空表示扫描本工具的 `downloads/`。扫描只处理根目录的一层文章子目录；文件须已完整落盘。
+3. 填写目标版块 `fid`、每账号最少/最多篇数和处理间隔。先核对历史发帖记录，再勾选确认并保存。版块和篇数不会替用户选择。
+4. 登录论坛，建议只保留一个执行标签。点击控制台“开始 / 继续”。执行器识别账号后领取一篇任务，进入指定新帖页、导入、排版，停止在人工审核阶段。
+5. 在论坛检查账号、版块、标题、图片位置、正文和来源，并亲手点击“发表帖子”。程序不调用发帖提交函数，不发送发帖请求，也不自动保存草稿。
+6. 当前版本会记录人工点击和提交快照，在跳转后的候选页面或等待超时后标记“发布结果待核对”并暂停。**需要补充结果页 HTML 后，才能自动核对受理结果并接续下一篇。**
+
+待补齐结果页自动核对后：审核中算网站已受理一次，但不算已确认公开；达到账号随机目标篇数且仍有待处理文章时，回首页调用插件换号，核对新 UID 与插件空闲状态后创建下一账号批次。没有待处理文章时直接结束，不为达到数量重发。
+
+## 终止、恢复与冲突
+
+- 网站左下角常驻“终止流程”，点击后立即取消本标签可取消的导入，并通知其他安装脚本的标签和换号插件；持久状态禁止继续领取任务。
+- 控制台“终止流程”还会停止下载，浏览器在下一次连接检查时停止（正常轮询约 2 秒）。网络断开时无法承诺固定延迟；网站按钮可立即停止本标签。已到达网站的图片、登录或发帖请求无法撤回。
+- 暂停/终止后必须明确点击“开始 / 继续”。下载停止等待当前网络请求/单篇保存结束，不伪装成已经撤回在途请求。
+- 非空编辑器、人工修改、账号变化、缺图、未知来源或上传失败会停在“需要处理”，不弹自动确认覆盖旧稿。
+- 导入中刷新后不会重新上传。先核对网站附件和草稿，再决定如何处理。若需要换标签，先在控制台暂停，再在目标标签点击“接管已暂停任务”。接管不会自动重新导入内容。
+- 只有人工确认网站**未受理**且已清理编辑器/草稿后，才能通过队列“人工核对 / 处理”重新排队或移出。审核中、已受理和结果不明不应重试。
+- 换号插件暂停后需要通过插件现有“继续未完成流程”恢复，不重复抽号。若账号未发生变化，调度器不会猜测换号成功。
+
+## 文件与数据
+
+```text
+app.py / store.py / articles.py   本地接口、任务队列、文章校验
+modules/downloader/              下载器副本与成功保存回调
+modules/package.py               文章包生成器副本
+browser/                         页面执行器、排版和连接脚本
+static/                          操作前端
+start.command                    启动入口
+build_browser.py                 重建合并油猴脚本
+test/                           离线测试、模拟素材、历史测试截图
+data/queue.sqlite3               运行任务、计数、受理去重记录（本地）
+downloads/                      本工具下载结果（本地）
+```
+
+- `test/history/` 保存从本目录移入的两张演示截图与演示任务 JSON。根目录其他测试文件未搬动。
+- `发布去重记录.json` 是实际受理证据，继续留在原处，并在服务启动时迁入去重库。它不代表完整历史已对账。原文件可找到时还建立内容指纹；原文件丢失时仍保留原 Markdown 摘要。
+- 原始下载文件不改写。长标题生成保守 UTF-16 长度不超过 80 的派生标题并加 `...`，原题仍展示在队列；网站真实计数规则尚需实测。
+- 排版仅处理与文章来源一致的末尾标准尾注，保留正文中的其他“来源”文字。图片顺序、标识改变会回滚排版。
+- WebP/AVIF/SVG 当前明确拦截，需用户另存支持格式并更新 Markdown；没有静默丢图或自动转码。单图须小于 20 MiB，图片 base64 总大小不超过 48 MiB。
+- 本地库含文章内容快照和账号标识，不含账号密码、Cookie 或上传凭据。运行数据和历史截图已加入忽略规则。
+
+## 验证
+
+不连接新闻 API 或真实论坛：
+
+```sh
+uv run --no-project --python 3.12 --with-requirements requirements.txt python -m unittest discover -s test -p 'test_*.py' -v
+node --test test/importer.test.cjs test/bridge.test.cjs
+python3 build_browser.py
+```
+
+浏览器回归需要 Playwright 和本机 Chrome，使用临时独立浏览器，不使用用户配置或账号：
+
+```sh
+# 终端一：临时模拟数据服务，端口 8771
+uv run --no-project --python 3.12 --with-requirements requirements.txt python test/serve_fixture.py
+# 终端二：Node 环境需能加载 playwright；CHROME_PATH 可指定浏览器可执行文件
+node test/browser.test.cjs
+```
+
+浏览器回归拦截所有真实论坛请求，仅使用模拟 HTML。截图输出到 `test/artifacts/`。本轮离线通过不代表已在用户登录态的真实网站上完成上传、换号或发布。
+
+## 下一步需要的页面资料
+
+按用户决定，等待结果页 HTML：帖子标题、首帖作者及 UID 链接、版块及 fid 链接、“审核中”提示、公开帖子与失败提示。不要提供 Cookie、密码或表单令牌。将据此完成 `runner.js` 的结果观察适配，调用已经实现的受理记录接口，联通后续计数和换号。

app.py

@@ -0,0 +1,205 @@
+#!/usr/bin/env python3
+"""Local controller: serves UI and authenticated browser bridge on loopback only."""
+import argparse
+import hmac
+import importlib.util
+import json
+from http.server import BaseHTTPRequestHandler, ThreadingHTTPServer
+from pathlib import Path
+import secrets
+import sys
+import threading
+from urllib.parse import urlsplit
+import webbrowser
+
+ROOT = Path(__file__).resolve().parent
+sys.path.insert(0, str(ROOT / 'modules/downloader'))
+spec = importlib.util.spec_from_file_location('news_downloader', ROOT / 'modules/downloader/app.py')
+downloader = importlib.util.module_from_spec(spec)
+spec.loader.exec_module(downloader)
+from store import Store
+
+
+class Controller:
+    def __init__(self, data=ROOT / 'data', downloads=ROOT / 'downloads', ledger=ROOT / '发布去重记录.json'):
+        self.store = Store(Path(data) / 'queue.sqlite3', ledger)
+        self.ui_token, self.bridge_token = secrets.token_urlsafe(32), secrets.token_urlsafe(32)
+        self.manager = downloader.Downloader(downloads, on_saved=self.saved)
+        self.scan_lock = threading.Lock()
+
+    def saved(self, article, result, keys):
+        directory = Path(result.directory)
+        marker = directory / '.article.json'
+        pending = directory / '.article.json.tmp'
+        pending.write_text(json.dumps({'fingerprints': keys, 'source_url': result.source_url}, ensure_ascii=False), encoding='utf-8')
+        pending.replace(marker)
+        self.store.enqueue(directory)
+
+    def scan(self, root):
+        if self.manager.thread and self.manager.thread.is_alive():
+            raise ValueError('下载进行中，请等待完成后扫描；新下载文章会自动入队')
+        root = Path(root).expanduser().resolve(strict=True)
+        if not root.is_dir():
+            raise ValueError('文章根目录不存在')
+        if not self.scan_lock.acquire(blocking=False):
+            raise ValueError('正在扫描，请稍候')
+        try:
+            directories = [root] if list(root.glob('*.md')) else sorted(p for p in root.iterdir() if p.is_dir() and not p.name.startswith('.'))
+            result = []
+            for directory in directories:
+                try:
+                    status = self.store.enqueue(directory)
+                except (ValueError, OSError) as error:
+                    status = str(error)
+                result.append({'directory': directory.name, 'status': status})
+            return {'results': result}
+        finally:
+            self.scan_lock.release()
+
+    def ui(self, route, body):
+        if route == 'state':
+            return {**self.store.state(), 'download': self.manager.snapshot()}
+        if route == 'pair':
+            return {'token': self.bridge_token}
+        if route == 'channels':
+            key = str(body.get('key', '')).strip()
+            if not key or len(key) > 512:
+                raise ValueError('请填写 API Key')
+            items = downloader.result_list(downloader.api_request(key, 'channel'))
+            if not items or not all(isinstance(x, str) and x for x in items):
+                raise ValueError('频道响应格式不正确')
+            return {'channels': items}
+        if route == 'download':
+            key, channel, target = str(body.get('key', '')).strip(), body.get('channel'), body.get('target')
+            if not key or len(key) > 512 or not isinstance(channel, str) or not channel.strip() or type(target) is not int or not 1 <= target <= 100000:
+                raise ValueError('请填写 Key、频道和有效下载数量')
+            self.manager.start(key, channel, target)
+            return {'ok': True}
+        if route == 'download-stop':
+            self.manager.stop.set()
+            return {'ok': True}
+        if route == 'scan':
+            return self.scan(body.get('root', '')) if body.get('root') else self.scan(self.manager.root)
+        if route == 'configure':
+            self.store.configure(body)
+            return {'ok': True}
+        if route in ('start', 'pause', 'stop'):
+            if route == 'stop':
+                self.manager.stop.set()
+            return self.store.control(route)
+        if route == 'resolve':
+            return self.store.resolve(body)
+        if route == 'history':
+            records = body.get('records')
+            if not isinstance(records, list):
+                raise ValueError('历史记录需包含 records 数组')
+            self.store.import_history(records)
+            return {'ok': True}
+        raise ValueError('未知操作')
+
+    def bridge(self, route, body):
+        methods = {'heartbeat': self.store.heartbeat, 'next': self.store.next, 'package': self.store.package,
+                   'event': self.store.event, 'accept': self.store.accept, 'switch-begin': self.store.switch_begin,
+                   'takeover': self.store.takeover, 'current': self.store.current}
+        if route == 'stop':
+            self.manager.stop.set()
+            return self.store.control('stop')
+        if route not in methods:
+            raise ValueError('未知浏览器操作')
+        return methods[route](body)
+
+
+def make_handler(controller):
+    class Handler(BaseHTTPRequestHandler):
+        def log_message(self, *args):
+            pass  # Never echo request bodies, API keys or pairing tokens.
+
+        def send(self, code, payload, mime='application/json; charset=utf-8'):
+            content = payload if isinstance(payload, bytes) else json.dumps(payload, ensure_ascii=False).encode()
+            self.send_response(code)
+            self.send_header('Content-Type', mime)
+            self.send_header('Content-Length', str(len(content)))
+            self.send_header('Cache-Control', 'no-store')
+            self.send_header('X-Content-Type-Options', 'nosniff')
+            self.send_header('Content-Security-Policy', "default-src 'self'; script-src 'self'; style-src 'self'; connect-src 'self'; frame-ancestors 'none'; base-uri 'none'")
+            self.end_headers()
+            try:
+                self.wfile.write(content)
+            except (BrokenPipeError, ConnectionResetError):
+                pass
+
+        def local(self):
+            return self.headers.get('Host') in (f'127.0.0.1:{self.server.server_port}', f'localhost:{self.server.server_port}')
+
+        def origin_ok(self):
+            origin = self.headers.get('Origin')
+            return not origin or origin in (f'http://127.0.0.1:{self.server.server_port}', f'http://localhost:{self.server.server_port}')
+
+        def do_GET(self):
+            if not self.local() or not self.origin_ok():
+                return self.send(403, {'error': '仅允许本机控制台访问'})
+            route = urlsplit(self.path).path
+            if route == '/api/session':
+                return self.send(200, {'token': controller.ui_token})
+            files = {'/': ('static/index.html', 'text/html; charset=utf-8'),
+                     '/app.js': ('static/app.js', 'text/javascript; charset=utf-8'),
+                     '/style.css': ('static/style.css', 'text/css; charset=utf-8'),
+                     '/browser/auto-news.user.js': ('browser/auto-news.user.js', 'text/javascript; charset=utf-8'),
+                     '/browser/local-bridge.user.js': ('browser/local-bridge.user.js', 'text/javascript; charset=utf-8')}
+            if route not in files:
+                return self.send(404, {'error': '不存在'})
+            name, mime = files[route]
+            file = ROOT / name
+            if not file.exists():
+                return self.send(404, {'error': '请先运行 build_browser.py'})
+            return self.send(200, file.read_bytes(), mime)
+
+        def do_POST(self):
+            if not self.local():
+                return self.send(403, {'error': 'Host 不匹配'})
+            route = urlsplit(self.path).path
+            bridge = route.startswith('/bridge/')
+            expected = controller.bridge_token if bridge else controller.ui_token
+            provided = self.headers.get('X-Session-Token', '')
+            if not hmac.compare_digest(expected, provided) or (not bridge and not self.origin_ok()):
+                return self.send(403, {'error': '连接凭据无效，请重新配对'})
+            if not bridge and not route.startswith('/api/'):
+                return self.send(404, {'error': '不存在'})
+            try:
+                length = int(self.headers.get('Content-Length', 0))
+                if not 0 < length <= 2 * 1024 * 1024:
+                    raise ValueError('请求大小无效')
+                body = json.loads(self.rfile.read(length))
+                if not isinstance(body, dict):
+                    raise ValueError('请求必须是对象')
+                result = (controller.bridge if bridge else controller.ui)(route.rsplit('/', 1)[1], body)
+                self.send(200, result)
+            except (ValueError, KeyError, OSError) as error:
+                self.send(400, {'error': str(error)[:1000]})
+            except Exception:
+                self.send(500, {'error': '内部错误；请停止当前任务并检查本地服务'})
+    return Handler
+
+
+def main():
+    parser = argparse.ArgumentParser()
+    parser.add_argument('--port', type=int, default=8770)
+    parser.add_argument('--no-browser', action='store_true')
+    args = parser.parse_args()
+    controller = Controller()
+    server = ThreadingHTTPServer(('127.0.0.1', args.port), make_handler(controller))
+    url = f'http://127.0.0.1:{args.port}/'
+    print('新闻工作台：' + url, flush=True)
+    if not args.no_browser:
+        webbrowser.open(url)
+    try:
+        server.serve_forever()
+    except KeyboardInterrupt:
+        controller.manager.stop.set()
+        controller.store.control('stop')
+    finally:
+        server.server_close()
+
+
+if __name__ == '__main__':
+    main()

articles.py

@@ -0,0 +1,116 @@
+"""Local article validation and portable packages. Never changes downloaded files."""
+import hashlib
+import json
+from pathlib import Path
+import re
+import unicodedata
+from urllib.parse import unquote
+
+from modules.package import make_package
+
+MAX_PACKAGE_BYTES = 48 * 1024 * 1024
+FOOTER = re.compile(r'(?:^|\n)来源：([^\n]+)，如有侵权，请联系删除。\s*$')
+
+
+def validate_images(body, directory):
+    """Conservative inline-image preflight; the importer remains the final parser."""
+    if re.search(r'!\[[^\]]*\]\s*\[', body):
+        raise ValueError('不支持引用式图片，请改成内联图片路径')
+    for start in re.finditer(r'!\[[^\]]*\]\(', body):
+        pos, depth, quote, angle = start.end(), 1, '', False
+        begin = pos
+        while pos < len(body):
+            ch = body[pos]
+            if ch == '\\':
+                pos += 2
+                continue
+            if quote:
+                if ch == quote:
+                    quote = ''
+            elif not angle and depth == 1 and body[pos - 1].isspace() and ch in ('"', "'"):
+                quote = ch
+            elif ch == '<':
+                angle = True
+            elif ch == '>':
+                angle = False
+            elif not angle and ch == '(':
+                depth += 1
+            elif not angle and ch == ')':
+                depth -= 1
+                if not depth:
+                    break
+            pos += 1
+        if depth:
+            raise ValueError('图片引用括号不完整')
+        target = body[begin:pos].strip()
+        target = target[1:target.index('>')] if target.startswith('<') and '>' in target else re.sub(r'\s+["\'][\s\S]*["\']\s*$', '', target)
+        target = unquote(re.sub(r'\\([()\\ ])', r'\1', target))
+        if re.match(r'^(?:[a-zA-Z][\w+.-]*:|/|\\)', target):
+            raise ValueError('图片必须使用目录内相对路径：' + target)
+        image = (directory / target).resolve()
+        if not image.is_relative_to(directory) or not image.is_file():
+            raise ValueError('缺图或图片路径超出目录：' + target)
+        if image.suffix.lower() not in ('.jpg', '.jpeg', '.png', '.gif'):
+            raise ValueError('不支持的配图格式：' + target)
+        if not 0 < image.stat().st_size < 20 * 1024 * 1024:
+            raise ValueError('配图为空或达到 20 MiB：' + target)
+
+
+def short_title(title, budget=80):
+    # Conservative UTF-16 budget: do not split astral characters / surrogate pairs.
+    if len(title.encode('utf-16-le')) // 2 <= budget:
+        return title
+    out = ''
+    for ch in title:
+        if len((out + ch).encode('utf-16-le')) // 2 > budget - 3:
+            break
+        out += ch
+    return out + '...'
+
+
+def inspect_article(directory):
+    directory = Path(directory).resolve(strict=True)
+    candidates = sorted(directory.glob('*.md'))
+    if len(candidates) != 1:
+        raise ValueError('每个文章目录必须恰好包含一个 Markdown 文件')
+    md = candidates[0]
+    if not md.resolve().is_relative_to(directory):
+        raise ValueError('Markdown 不得链接到目录外')
+    raw = md.read_text(encoding='utf-8-sig')
+    heading = re.match(r'^\s*#\s+([^\n]+)\n?', raw)
+    title = (heading[1] if heading else md.stem).strip()
+    body = raw[heading.end():].strip() if heading else raw.strip()
+    footer = FOOTER.search(body)
+    if not footer:
+        raise ValueError('文末缺少标准来源尾注，请补充来源后再扫描')
+    source = footer[1].strip()
+    body = body[:footer.start()].strip()
+    if not title or not body:
+        raise ValueError('标题或正文为空')
+    validate_images(body, directory)
+    for file in directory.rglob('*'):
+        if file.suffix.lower() in ('.webp', '.avif', '.svg'):
+            raise ValueError('发现暂不支持的图片格式：' + file.name + '；请另存为 JPG/PNG/GIF 并更新引用')
+    pack = make_package(directory)
+    if sum(len(i['base64']) for i in pack['images']) > MAX_PACKAGE_BYTES:
+        raise ValueError('文章包超过 48 MiB，请减少单篇图片体积')
+    pack['title'] = short_title(title)
+    # Preserve body and exact final footer for the formatter; the original file stays intact.
+    pack['markdown'] = '# ' + pack['title'] + '\n\n' + body + '\n\n来源：' + source + '，如有侵权，请联系删除。\n'
+    content = unicodedata.normalize('NFC', body)
+    # Hash file contents, not directory/title; renamed copies share their article identity.
+    image_hashes = []
+    for item in pack['images']:
+        import base64
+        digest = hashlib.sha256(base64.b64decode(item['base64'])).hexdigest()
+        content = content.replace(item['path'], 'image:' + digest)
+        image_hashes.append(digest)
+    identity = hashlib.sha256((content + '\n' + '\n'.join(sorted(set(image_hashes)))).encode()).hexdigest()
+    metadata = {}
+    marker = directory / '.article.json'
+    if marker.is_file():
+        metadata = json.loads(marker.read_text())
+    return dict(article_key=identity, markdown_sha256=hashlib.sha256(md.read_bytes()).hexdigest(),
+                directory=str(directory), original_title=title, title=pack['title'], source=source,
+                image_count=len(pack['images']), package=pack,
+                source_keys=metadata.get('fingerprints', []))

browser/auto-news.user.js

@@ -0,0 +1,850 @@
+// ==UserScript==
+// @name         新闻工作台 · 导入排版与流程
+// @namespace    local.dww11.auto-news
+// @version      0.1.0
+// @match        https://www.dww11.com/*
+// @match        https://dww11.com/*
+// @match        https://house.dww11.com/*
+// @grant        none
+// @run-at       document-idle
+// @noframes
+// ==/UserScript==
+
+
+(function (root, factory) {
+    'use strict';
+    const core = factory();
+    if (typeof module === 'object' && module.exports) module.exports = core;
+    else core.mount(root);
+})(typeof window === 'undefined' ? null : window, function () {
+    'use strict';
+    const VERSION = '1.0.0';
+    const MAX_IMAGE_SIZE = 20 * 1024 * 1024;
+    const IMAGE_EXT = /\.(?:jpe?g|png|gif)$/i;
+
+    function normalizePath(path) {
+        const parts = [];
+        const value = String(path).replace(/\\/g, '/').normalize('NFC');
+        if (/^(?:\/|[a-z]:|[a-z]+:)/i.test(value)) throw new Error('请使用目录内的相对路径：' + path);
+        for (const part of value.split('/')) {
+            if (!part || part === '.') continue;
+            if (part === '..') {
+                if (!parts.length) throw new Error('图片路径超出所选目录：' + path);
+                parts.pop();
+            } else parts.push(part);
+        }
+        return parts.join('/');
+    }
+
+    function imagePath(mdPath, target) {
+        if (/^(?:[a-z][\w+.-]*:|\/\/|\/)/i.test(target)) throw new Error('图片必须在所选目录内，请先下载：' + target);
+        let decoded;
+        try { decoded = decodeURIComponent(target); } catch { throw new Error('图片路径编码无效：' + target); }
+        return normalizePath(mdPath.slice(0, mdPath.lastIndexOf('/') + 1) + decoded);
+    }
+
+    // A small, intentionally bounded Markdown reader. Raw HTML is never executed.
+    // Links/images support balanced parentheses, angle destinations and optional titles.
+    function readLink(text, start) {
+        const image = text.startsWith('![', start);
+        if (!image && text[start] !== '[') return null;
+        const left = start + (image ? 2 : 1);
+        let end = left;
+        for (; end < text.length; end++) {
+            if (text[end] === '\\') { end++; continue; }
+            if (text[end] === ']') break;
+        }
+        if (text.slice(end, end + 2) !== '](') return null;
+        let pos = end + 2, depth = 1, quoted = '', angle = false;
+        for (; pos < text.length; pos++) {
+            const ch = text[pos];
+            if (ch === '\\') { pos++; continue; }
+            if (quoted) { if (ch === quoted) quoted = ''; continue; }
+            if (!angle && depth === 1 && /\s/.test(text[pos - 1]) && (ch === '"' || ch === "'")) { quoted = ch; continue; }
+            if (ch === '<') { angle = true; continue; }
+            if (ch === '>') { angle = false; continue; }
+            if (!angle && ch === '(') depth++;
+            if (!angle && ch === ')' && --depth === 0) break;
+        }
+        if (depth) return null;
+        let target = text.slice(end + 2, pos).trim();
+        const destination = target.match(/^<([^>]+)>\s*(?:["'][\s\S]*["'])?$/);
+        target = destination ? destination[1] : target.replace(/\s+["'][\s\S]*["']\s*$/, '');
+        target = target.replace(/\\([()\\ ])/g, '$1');
+        return { image, label: text.slice(left, end), target, end: pos + 1 };
+    }
+
+    function inline(text, onImage) {
+        let out = '';
+        for (let pos = 0; pos < text.length;) {
+            if (text[pos] === '\\' && pos + 1 < text.length) { out += text[pos + 1]; pos += 2; continue; }
+            const link = readLink(text, pos);
+            if (link) {
+                if (link.image) out += onImage(link.target, link.label);
+                else if (/^(https?:\/\/|mailto:)/i.test(link.target) && !/[\[\]\r\n"']/.test(link.target)) {
+                    out += '[url=' + link.target + ']' + inline(link.label, onImage) + '[/url]';
+                } else out += link.label + ' (' + link.target + ')';
+                pos = link.end;
+                continue;
+            }
+            if (text[pos] === '`') {
+                const end = text.indexOf('`', pos + 1);
+                if (end > pos + 1) { out += text.slice(pos + 1, end); pos = end + 1; continue; }
+            }
+            let matched = false;
+            for (const [mark, tag] of [['**', 'b'], ['__', 'b'], ['~~', 's'], ['*', 'i'], ['_', 'i']]) {
+                if (!text.startsWith(mark, pos) || (mark === '_' && /\w/.test(text[pos - 1] || ''))) continue;
+                const end = text.indexOf(mark, pos + mark.length);
+                if (end <= pos + mark.length) continue;
+                out += '[' + tag + ']' + inline(text.slice(pos + mark.length, end), onImage) + '[/' + tag + ']';
+                pos = end + mark.length; matched = true; break;
+            }
+            if (!matched) out += text[pos++];
+        }
+        return out;
+    }
+
+    function markdownToBBCode(markdown, onImage) {
+        const lines = markdown.replace(/\r\n?/g, '\n').split('\n');
+        const output = [];
+        let list = '', quote = false, code = false, fence = '';
+        const closeList = () => { if (list) output.push('[/list]'); list = ''; };
+        for (let index = 0; index < lines.length; index++) {
+            const line = lines[index];
+            const codeMark = line.match(/^\s*(`{3,}|~{3,})/);
+            if (codeMark && (!code || codeMark[1][0] === fence)) {
+                closeList();
+                output.push(code ? '[/code]' : '[code]');
+                code = !code; fence = codeMark[1][0]; continue;
+            }
+            if (code) { output.push(line.replace(/\[\/code\]/gi, '［/code］')); continue; }
+            const quoted = /^\s*>\s?/.test(line);
+            if (quoted !== quote) { closeList(); output.push(quoted ? '[quote]' : '[/quote]'); quote = quoted; }
+            const content = quoted ? line.replace(/^\s*>\s?/, '') : line;
+            const item = content.match(/^\s*(?:([-+*])|\d+[.)])\s+(.+)$/);
+            if (item) {
+                const type = item[1] ? '[list]' : '[list=1]';
+                if (list !== type) { closeList(); output.push(type); list = type; }
+                output.push('[*]' + inline(item[2], onImage)); continue;
+            }
+            closeList();
+            const heading = content.match(/^\s{0,3}(#{1,6})\s+(.+?)\s*#*$/);
+            if (heading) output.push('[b]' + inline(heading[2], onImage) + '[/b]');
+            else if (/^\s*(?:---+|\*\*\*+|___+)\s*$/.test(content)) output.push('[hr]');
+            else output.push(inline(content, onImage));
+        }
+        closeList();
+        if (quote) output.push('[/quote]');
+        if (code) output.push('[/code]');
+        return output.join('\n').trim();
+    }
+
+    function compile(markdown, mdPath, fileMap, titleOverride) {
+        let body = String(markdown).replace(/^\uFEFF/, '').replace(/\r\n?/g, '\n');
+        const firstHeading = body.match(/^\s*#\s+([^\n]+)\n?/);
+        const title = String(titleOverride || firstHeading?.[1] || mdPath.split('/').pop().replace(/\.md$/i, '')).trim();
+        if (firstHeading) body = body.slice(firstHeading[0].length).trim();
+        if (!title || Array.from(title).length > 80) throw new Error('标题为空或超过 80 字，请修改 Markdown 标题后重试。');
+        if (!body.trim()) throw new Error('正文为空。');
+        const warnings = [];
+        if (/<\/?[a-z][^>]*>/i.test(body)) warnings.push('存在 HTML，将作为文字保留；请检查排版。');
+        if (/^\s*\|.*\|\s*$/m.test(body)) warnings.push('表格暂按原文保留，需手动检查。');
+        if (/!\[[^\]]*\]\s*\[[^\]]*\]/.test(body) || /^\s*\[[^\]]+\]:/m.test(body)) {
+            throw new Error('暂不支持引用式 Markdown 图片/链接，请改成 ![](图片路径) 或 [文字](网址)。');
+        }
+        const images = new Map();
+        // Validate every image before any network request; repeated references upload once.
+        markdownToBBCode(body, (target) => {
+            const path = imagePath(mdPath, target);
+            const file = fileMap.get(path);
+            if (!file) throw new Error('找不到配图：' + path);
+            if (!IMAGE_EXT.test(path)) throw new Error('网站仅支持 JPG、JPEG、PNG、GIF：' + path);
+            if (!file.size || file.size >= MAX_IMAGE_SIZE) throw new Error('配图为空或达到 20 MB：' + path);
+            images.set(path, file);
+            return '[图片：' + path + ']';
+        });
+        return { title, body, mdPath, images, warnings,
+            render(aids) {
+                return markdownToBBCode(body, target => {
+                    const path = imagePath(mdPath, target);
+                    const aid = aids.get(path);
+                    if (!/^[1-9]\d*$/.test(String(aid))) throw new Error('配图尚未上传：' + path);
+                    return '[attachimg]' + aid + '[/attachimg]';
+                });
+            }
+        };
+    }
+
+    async function prepareFiles(files, markdownPath) {
+        const fileMap = new Map();
+        for (const file of Array.from(files)) {
+            const path = normalizePath(file.webkitRelativePath || file.name);
+            if (fileMap.has(path)) throw new Error('文件路径重复：' + path);
+            fileMap.set(path, file);
+        }
+        const candidates = Array.from(fileMap.keys()).filter(p => /\.md$/i.test(p)).sort();
+        if (!candidates.length) throw new Error('所选目录中没有 .md 文件。');
+        if (!markdownPath && candidates.length > 1) return { candidates, files: Array.from(files) };
+        const mdPath = markdownPath || candidates[0];
+        const md = fileMap.get(mdPath);
+        if (!md) throw new Error('指定的 Markdown 文件不存在。');
+        return compile(await md.text(), mdPath, fileMap);
+    }
+
+    async function preparePackage(data) {
+        if (!data || data.schemaVersion !== 1 || typeof data.markdown !== 'string' || !Array.isArray(data.images)) {
+            throw new Error('文章包格式错误，需要 schemaVersion: 1、markdown 和 images。');
+        }
+        const map = new Map();
+        for (const item of data.images) {
+            const path = normalizePath(item.path);
+            if (map.has(path)) throw new Error('文章包图片路径重复：' + path);
+            let blob = item.blob;
+            if (!blob && typeof item.base64 === 'string') {
+                const binary = atob(item.base64);
+                blob = new Blob([Uint8Array.from(binary, ch => ch.charCodeAt(0))], { type: item.mime || 'application/octet-stream' });
+            }
+            if (!blob || typeof blob.arrayBuffer !== 'function') throw new Error('文章包缺少图片数据：' + path);
+            map.set(path, blob);
+        }
+        return compile(data.markdown, normalizePath(data.markdownPath || 'article.md'), map, data.title);
+    }
+
+    function parseAid(text) {
+        const value = String(text).trim();
+        // The site's actual WebUploader adapter uses a positive integer response.
+        if (!/^(?:[1-9]\d*|"[1-9]\d*")$/.test(value)) throw new Error('服务器未返回有效附件编号（可能登录过期、超出权限或上传失败）。');
+        return value.replace(/"/g, '');
+    }
+
+    function mount(win) {
+        const doc = win.document;
+        const params = new URLSearchParams(win.location.search);
+        if (params.get('mod') !== 'post' || params.get('action') !== 'newthread' || !doc.querySelector('#subject') || !doc.querySelector('#e_textarea')) return;
+        if (win.DWW11Importer) return;
+        let article = null, busy = false, candidates = null, lastResult = null, savedBackup = null, stageRevision = 0;
+        const cache = new Map(), urls = [], logs = [];
+        const host = doc.createElement('div');
+        host.id = 'dww11-importer';
+        host.style.cssText = 'position:fixed;right:18px;bottom:18px;z-index:2147483000';
+        const shadow = host.attachShadow({ mode: 'open' });
+        shadow.innerHTML = `<style>
+            :host{all:initial}*{box-sizing:border-box}details{font:14px/1.6 system-ui,sans-serif;color:#203044;background:#fff;border:1px solid #d4dee8;border-radius:12px;box-shadow:0 6px 28px #0003;width:360px;max-width:92vw}summary{cursor:pointer;padding:12px 16px;font-weight:650}section{padding:0 16px 16px;max-height:70vh;overflow:auto}button,select{font:inherit;padding:7px 10px;border-radius:6px;border:1px solid #b8c9da;background:#f4f8fc;color:#203044;cursor:pointer;margin:4px 4px 4px 0}button.primary{background:#1767b3;color:white;border-color:#1767b3}button:disabled{opacity:.5;cursor:default}p{margin:8px 0}small{color:#617183}pre{font:12px/1.6 system-ui;white-space:pre-wrap;overflow-wrap:anywhere;max-height:170px;overflow:auto;background:#f4f6f9;padding:8px}img{max-width:90px;max-height:80px;margin:4px;object-fit:contain}.error{color:#b32424}.success{color:#18713e}select{max-width:100%}[hidden]{display:none!important}
+            </style><details open><summary>本地文章导入 · 1.0</summary><section>
+            <small>选择整个文章文件夹，检查后上传配图并填写。发布由你操作。网站自带的自动保存仍可能运行。</small>
+            <p><button id="choose" type="button">选择文章目录</button><button id="package" type="button">导入文章包</button></p>
+            <input id="folder" type="file" webkitdirectory multiple hidden><input id="json" type="file" accept=".json,application/json" hidden>
+            <select id="md" hidden aria-label="选择 Markdown"></select>
+            <p id="status" role="status" aria-live="polite">尚未选择目录</p><div id="preview" hidden><strong id="title"></strong><pre id="body"></pre><div id="images"></div><p id="warnings"></p></div>
+            <button id="import" class="primary" type="button" disabled>上传配图并填写</button>
+            <button id="diagnostics" type="button">导出诊断</button><pre id="log" hidden></pre>
+            </section></details>`;
+        doc.body.appendChild(host);
+        const $ = id => shadow.getElementById(id);
+        const summary = () => ({ version: VERSION, busy, ready: !!article, title: article?.title || null,
+            markdownPath: article?.mdPath || null, imageCount: article?.images.size || 0,
+            warnings: article?.warnings || [], result: lastResult });
+        function notify(phase, message, extra = {}) {
+            $('status').textContent = message;
+            $('status').className = phase === 'error' ? 'error' : phase === 'complete' ? 'success' : '';
+            logs.push({ time: new Date().toISOString(), phase, message, ...extra });
+            $('log').hidden = false;
+            $('log').textContent = logs.slice(-8).map(l => l.message).join('\n');
+            win.dispatchEvent(new CustomEvent('dww11-importer:progress', { detail: { phase, message, ...extra } }));
+        }
+        function releasePreviews() { while (urls.length) URL.revokeObjectURL(urls.pop()); }
+        function clearStage() {
+            article = null; lastResult = null; candidates = null; releasePreviews();
+            $('preview').hidden = true; $('md').hidden = true; $('import').disabled = true;
+        }
+        function showArticle(prepared) {
+            article = prepared; lastResult = null; releasePreviews();
+            $('preview').hidden = false;
+            $('title').textContent = article.title;
+            $('body').textContent = article.body;
+            $('warnings').textContent = article.warnings.join('；');
+            $('images').replaceChildren();
+            for (const [path, blob] of article.images) {
+                const img = doc.createElement('img');
+                img.src = URL.createObjectURL(blob); urls.push(img.src); img.alt = path; img.title = path;
+                $('images').appendChild(img);
+            }
+            $('import').disabled = false;
+            notify('ready', `已检查：${article.images.size} 张图片。点击“上传配图并填写”开始导入。`);
+            return summary();
+        }
+        async function stageFiles(files, options = {}) {
+            if (busy) throw new Error('正在导入，请稍候。');
+            clearStage();
+            const revision = ++stageRevision;
+            const list = Array.from(files);
+            const prepared = await prepareFiles(list, options.markdownPath);
+            if (revision !== stageRevision) return { cancelled: true };
+            if (prepared.candidates) {
+                candidates = prepared.files;
+                $('md').replaceChildren(new Option('请选择目录中的 Markdown 文件', ''));
+                for (const path of prepared.candidates) $('md').appendChild(new Option(path, path));
+                $('md').hidden = false;
+                notify('choose-markdown', '目录中有多个 Markdown 文件，请选择一篇。');
+                return { candidates: prepared.candidates };
+            }
+            return showArticle(prepared);
+        }
+        async function stagePackage(data) {
+            if (busy) throw new Error('正在导入，请稍候。');
+            clearStage();
+            const revision = ++stageRevision;
+            const prepared = await preparePackage(data);
+            if (revision !== stageRevision) return { cancelled: true };
+            return showArticle(prepared);
+        }
+        const snapshot = () => ({ title: doc.querySelector('#subject').value,
+            mode: Number(win.wysiwyg), text: win.getEditorContents() });
+        function assertEditor() {
+            for (const name of ['getEditorContents', 'switchEditor']) {
+                if (typeof win[name] !== 'function') throw new Error('未识别到网站编辑器：' + name + '。请刷新页面后重试。');
+            }
+            if (!win.imgUpload?.settings || typeof win.updateImageList !== 'function') throw new Error('网站图片上传组件尚未就绪，请刷新后重试。');
+            if (Number(win.allowbbcode) === 0 || Number(win.allowimgcode) === 0) throw new Error('当前版块或账号不支持正文格式/图片。');
+        }
+        async function uploadImage(blob, path, settings, signal) {
+            signal?.throwIfAborted();
+            const url = new URL(settings.upload_url, win.location.href);
+            if (url.origin !== win.location.origin || url.pathname !== '/misc.php' || url.searchParams.get('operation') !== 'upload') {
+                throw new Error('上传地址与预期不符，已停止。');
+            }
+            const hash = Array.from(new Uint8Array(await win.crypto.subtle.digest('SHA-256', await blob.arrayBuffer())), n => n.toString(16).padStart(2, '0')).join('');
+            const key = url.href + ':' + settings.post_params.uid + ':' + hash;
+            if (cache.has(key)) { notify('upload', '复用已上传图片：' + path); return cache.get(key); }
+            const data = new FormData();
+            for (const [key, value] of Object.entries(settings.post_params)) data.append(key, String(value));
+            data.set('type', 'image');
+            data.set('filetype', blob.type || (/\.png$/i.test(path) ? 'image/png' : /\.gif$/i.test(path) ? 'image/gif' : 'image/jpeg'));
+            data.append(settings.file_post_name || 'Filedata', blob, path.split('/').pop());
+            const control = new AbortController(), timer = setTimeout(() => control.abort(), 90000);
+            try {
+                signal?.throwIfAborted();
+                signal?.addEventListener('abort', () => control.abort(), { once: true });
+                const response = await win.fetch(url.href, { method: 'POST', body: data, credentials: 'same-origin', redirect: 'error', signal: control.signal });
+                if (!response.ok) throw new Error('图片上传请求失败：HTTP ' + response.status);
+                const aid = parseAid(await response.text());
+                cache.set(key, aid);
+                return aid;
+            } catch (error) {
+                if (error.name === 'AbortError') throw new Error('上传超时；该图片可能已到达服务器，请检查网站图片列表后再重试。');
+                throw error;
+            } finally { clearTimeout(timer); }
+        }
+        async function registerImages(aids, signal) {
+            signal?.throwIfAborted();
+            if (!aids.length) return;
+            const ids = new Set(aids);
+            doc.querySelectorAll('#imgattachlist [id^="image_td_"],#imgattachlist [id^="image_"]').forEach(el => {
+                const id = el.id.match(/^(?:image_td_|image_)(\d+)$/)?.[1]; if (id) ids.add(id);
+            });
+            const ready = () => aids.every(id => doc.getElementById('image_' + id));
+            if (ready()) return;
+            // Use the site's own list loader, which registers attachment form fields too.
+            win.updateImageList(1, Array.from(ids).join('|'));
+            await new Promise((resolve, reject) => {
+                const timer = setTimeout(() => { observer.disconnect(); reject(new Error('图片已上传，但网站图片列表加载超时。重试会复用本页已上传图片。')); }, 25000);
+                signal?.addEventListener('abort', () => { clearTimeout(timer); observer.disconnect(); reject(new Error('流程已终止')); }, { once: true });
+                const check = () => { if (ready()) { clearTimeout(timer); observer.disconnect(); resolve(); } };
+                const observer = new MutationObserver(check);
+                observer.observe(doc.body, { childList: true, subtree: true }); check();
+            });
+        }
+        function setEditor(title, bbcode) {
+            win.switchEditor(0);
+            if (Number(win.wysiwyg) !== 0) throw new Error('无法切换编辑器为纯文本模式。');
+            const textarea = doc.querySelector('#e_textarea');
+            textarea.value = bbcode;
+            doc.querySelector('#subject').value = title;
+            for (const el of [textarea, doc.querySelector('#subject')]) {
+                el.dispatchEvent(new Event('input', { bubbles: true })); el.dispatchEvent(new Event('change', { bubbles: true }));
+            }
+            // Let Discuz generate the WYSIWYG content with native attachment metadata.
+            win.switchEditor(1);
+            const switcher = doc.querySelector('#e_switchercheck');
+            if (switcher) switcher.checked = Number(win.wysiwyg) === 0;
+        }
+        async function importPrepared(options = {}) {
+            const { signal, automatic = false } = options;
+            signal?.throwIfAborted();
+            if (busy) throw new Error('正在导入，请勿重复点击。');
+            if (!article) throw new Error('请先选择文章目录或文章包。');
+            assertEditor();
+            for (const id of ['bbcodeoff', 'htmlon']) {
+                if (doc.getElementById(id)?.checked) throw new Error('请先关闭“禁用 Discuz! 代码”和“HTML 代码”，再导入。');
+            }
+            const before = snapshot();
+            if ((before.title.trim() || before.text.replace(/<[^>]*>/g, '').trim() || /<img\b/i.test(before.text)) &&
+                (automatic ? (() => { throw new Error('编辑器已有内容，请人工处理旧稿后重试'); })() : !win.confirm('当前编辑器已有内容。导入将替换标题和正文，是否继续？'))) return { cancelled: true };
+            const selected = article;
+            busy = true;
+            const controls = ['choose', 'package', 'import', 'md'].map($);
+            controls.forEach(el => { el.disabled = true; });
+            // Prevent accidental submission of a partly imported article.
+            const blockSubmit = event => { event.preventDefault(); event.stopImmediatePropagation(); notify('busy', '正在导入，请等待完成后再操作发布。'); };
+            const form = doc.querySelector('#postform');
+            form.addEventListener('submit', blockSubmit, true);
+            const submitButtons = Array.from(form.querySelectorAll('button[type="submit"],button[name],input[type="submit"]'));
+            const previousDisabled = submitButtons.map(el => el.disabled);
+            submitButtons.forEach(el => { el.disabled = true; });
+            let committed = false;
+            try {
+                const aids = new Map();
+                const settings = win.imgUpload.settings;
+                const max = Math.min(MAX_IMAGE_SIZE, Number(settings.file_size_limit) * 1024 || MAX_IMAGE_SIZE);
+                for (const [path, blob] of selected.images) if (blob.size >= max) throw new Error('配图超过当前网站限制：' + path);
+                let index = 0;
+                for (const [path, blob] of selected.images) {
+                    notify('upload', `上传 ${++index}/${selected.images.size}：${path}`, { completed: index - 1, total: selected.images.size });
+                    aids.set(path, await uploadImage(blob, path, settings, signal));
+                }
+                notify('register', '正在加载网站图片列表…');
+                await registerImages(Array.from(new Set(aids.values())), signal);
+                signal?.throwIfAborted();
+                const now = snapshot();
+                if ((now.title !== before.title || now.text !== before.text) && (automatic ? (() => { throw new Error('上传期间编辑器被修改，已保留当前内容'); })() : !win.confirm('上传期间你修改了编辑器。是否用导入文章替换最新内容？'))) {
+                    notify('cancelled', '已保留当前正文；图片已上传，可稍后再次导入。'); return { cancelled: true };
+                }
+                const backup = snapshot();
+                const bbcode = selected.render(aids);
+                // Preserve a local, downloadable snapshot before changing title or body.
+                savedBackup = { title: backup.title, content: backup.text, mode: backup.mode, createdAt: new Date().toISOString() };
+                try {
+                    for (const id of ['bbcodeoff', 'htmlon']) {
+                        const el = doc.getElementById(id);
+                        if (el && id === 'bbcodeoff' && el.checked) throw new Error('请先取消“禁用 Discuz! 代码”，再重试。');
+                        if (el && id === 'htmlon' && el.checked) throw new Error('请先关闭“HTML 代码”，再重试。');
+                    }
+                    signal?.throwIfAborted();
+                    setEditor(selected.title, bbcode);
+                    const raw = win.getEditorContents();
+                    if (Number(win.wysiwyg) === 1) {
+                        for (const id of new Set(aids.values())) if (!raw.includes('attachimg_' + id)) throw new Error('富文本图片转换未通过检查。');
+                    } else if (raw !== bbcode) throw new Error('正文填写校验失败。');
+                    if (doc.querySelector('#subject').value !== selected.title) throw new Error('标题填写校验失败。');
+                    committed = true;
+                } catch (error) {
+                    // Restore the exact editor representation saved immediately before filling.
+                    win.switchEditor(backup.mode);
+                    if (backup.mode) win.editdoc.body.innerHTML = backup.text;
+                    else doc.querySelector('#e_textarea').value = backup.text;
+                    doc.querySelector('#subject').value = backup.title;
+                    const switcher = doc.querySelector('#e_switchercheck'); if (switcher) switcher.checked = !backup.mode;
+                    throw error;
+                }
+                lastResult = { title: selected.title, images: Object.fromEntries(aids), completedAt: new Date().toISOString() };
+                notify('complete', `已填写标题、正文和 ${aids.size} 张配图。请检查排版，随后手动发表。`);
+                return lastResult;
+            } catch (error) {
+                notify('error', error.message + (committed ? '' : ' 标题和正文未替换；已上传的图片会保留。'));
+                throw error;
+            } finally {
+                busy = false; form.removeEventListener('submit', blockSubmit, true);
+                submitButtons.forEach((el, index) => { el.disabled = previousDisabled[index]; });
+                controls.forEach(el => { el.disabled = false; });
+                $('import').disabled = !article;
+            }
+        }
+        function downloadJSON(value, filename) {
+            const url = URL.createObjectURL(new Blob([JSON.stringify(value, null, 2)], { type: 'application/json' }));
+            const link = doc.createElement('a'); link.href = url; link.download = filename; link.click();
+            setTimeout(() => URL.revokeObjectURL(url), 1000);
+        }
+        // Public v1 interface. Staging is read-only; importPrepared is the explicit upload/fill step.
+        // No postMessage listener, localhost service, publishing endpoint or saved credentials.
+        const api = Object.freeze({ version: VERSION, schemaVersion: 1, stageFiles, stagePackage,
+            getState: summary, openPicker: () => { if (!busy) $('folder').click(); }, importPrepared,
+            getBackup: () => savedBackup });
+        win.DWW11Importer = api;
+        $('choose').onclick = () => { $('folder').value = ''; $('folder').click(); };
+        $('package').onclick = () => { $('json').value = ''; $('json').click(); };
+        const handleError = error => notify('error', error.message || String(error));
+        $('folder').onchange = () => stageFiles($('folder').files).catch(handleError);
+        $('json').onchange = async () => {
+            try { const file = $('json').files[0]; if (file) await stagePackage(JSON.parse(await file.text())); }
+            catch (error) { handleError(error); }
+        };
+        $('md').onchange = () => {
+            const files = candidates, path = $('md').value;
+            if (files && path) stageFiles(files, { markdownPath: path }).catch(handleError);
+        };
+        $('import').onclick = () => importPrepared().catch(handleError);
+        $('diagnostics').onclick = () => downloadJSON({ ...summary(), logs }, 'dww11-import-diagnostics.json');
+        win.addEventListener('pagehide', releasePreviews, { once: true });
+    }
+    return { VERSION, normalizePath, imagePath, readLink, markdownToBBCode, compile, prepareFiles, preparePackage, parseAid, mount };
+});
+
+// Extracted from dww11-wechat-oneclick.user.js v1.0.9; package source is authoritative.
+(() => {
+'use strict';
+const cleanText = value => String(value || '').replace(/\s+/g, ' ').trim();
+    function flattenParagraphs(body) {
+        const selector = 'p,div,section,article,figure,figcaption,h1,h2,h3,h4,h5,h6';
+        const marker = () => {
+            const br = body.ownerDocument.createElement('br');
+            br.dataset.dww11Boundary = '1'; return br;
+        };
+        for (const block of Array.from(body.querySelectorAll(selector)).reverse()) {
+            if (block.closest('table,ul,ol,blockquote')) continue;
+            block.replaceWith(marker(), ...block.childNodes, marker());
+        }
+        let previous = false, pending = [];
+        function finish() {
+            const actual = pending.filter(br => !br.hasAttribute('data-dww11-boundary'));
+            let retained = false;
+            for (const br of pending) {
+                if (!br.hasAttribute('data-dww11-boundary')) continue;
+                if (previous && !actual.length && !retained) { br.removeAttribute('data-dww11-boundary'); retained = true; }
+                else br.remove();
+            }
+            pending = [];
+        }
+        const walker = body.ownerDocument.createTreeWalker(body, NodeFilter.SHOW_ELEMENT | NodeFilter.SHOW_TEXT);
+        const nodes = []; while (walker.nextNode()) nodes.push(walker.currentNode);
+        for (const node of nodes) {
+            if (node.nodeType === 1 && node.tagName === 'BR') pending.push(node);
+            else if ((node.nodeType === 3 && node.textContent.trim()) || (node.nodeType === 1 && ['IMG', 'VIDEO', 'HR', 'IFRAME'].includes(node.tagName))) {
+                finish(); previous = true;
+            }
+        }
+        // 末尾不保留由容器边界制造的额外换行。
+        for (const br of pending) if (br.hasAttribute('data-dww11-boundary')) br.remove();
+    }
+    function spaceImages(body) {
+        let previous = null, gapNodes = [], blankLines = 0;
+        const changes = [];
+        function content(kind, node) {
+            if (previous && previous.kind !== kind) {
+                let anchor = node;
+                while (anchor.parentElement && anchor.parentElement !== body &&
+                       !['TD', 'TH', 'LI', 'BLOCKQUOTE'].includes(anchor.parentElement.tagName) && !anchor.parentElement.contains(previous.node)) anchor = anchor.parentElement;
+                changes.push({ anchor, nodes: gapNodes.slice(), count: Math.max(2, blankLines) });
+            }
+            previous = { kind, node }; gapNodes = []; blankLines = 0;
+        }
+        function visit(node) {
+            if (node.nodeType === 3) {
+                if (node.textContent.trim()) content('text', node); else gapNodes.push(node);
+                return;
+            }
+            if (node.nodeType !== 1 || ['SCRIPT', 'STYLE', 'NOSCRIPT'].includes(node.tagName)) return;
+            if (node.tagName === 'BR') { gapNodes.push(node); blankLines++; return; }
+            if (node.tagName === 'IMG') { content('image', node); return; }
+            for (const child of node.childNodes) visit(child);
+        }
+        for (const child of body.childNodes) visit(child);
+        for (const { anchor, nodes, count } of changes) {
+            for (const node of nodes) node.remove();
+            const fragment = body.ownerDocument.createDocumentFragment();
+            for (let i = 0; i < count; i++) fragment.append(body.ownerDocument.createElement('br'));
+            anchor.parentNode.insertBefore(fragment, anchor);
+        }
+    }
+    function formatBody(body, author, sourceText = null) {
+        const doc = body.ownerDocument;
+        body.querySelectorAll('[data-dww11-wx-source="1"]').forEach(el => el.remove());
+        // Discuz 会把内联 color 值直接转为 [color=…]。
+        // !important 不能出现在该值中；颜色和字号只用原生 FONT 属性。
+        for (const el of [body, ...body.querySelectorAll('*')]) {
+            if (['SCRIPT', 'STYLE'].includes(el.tagName)) continue;
+            for (const property of ['font', 'font-size', 'color', '-webkit-text-fill-color']) {
+                el.style.removeProperty(property);
+            }
+            // 其他排版属性也不保留 CSS 优先级，避免它泄漏进论坛标签。
+            for (const property of Array.from(el.style)) {
+                if (el.style.getPropertyPriority(property)) {
+                    el.style.setProperty(property, el.style.getPropertyValue(property));
+                }
+            }
+            if (!el.style.cssText) el.removeAttribute('style');
+        }
+        // 去掉采集器或旧版脚本生成的嵌套 FONT，保留其中的内容和图片。
+        for (const font of Array.from(body.querySelectorAll('font'))) {
+            font.replaceWith(...font.childNodes);
+        }
+        flattenParagraphs(body);
+        spaceImages(body);
+        // 所有正文内容和换行放在同一 FONT 中。不能使用只有 BR 的空 FONT：
+        // 论坛转换器可能清除空格式标签，导致空行退回默认1号或直接丢失。
+        const articleFont = doc.createElement('font');
+        articleFont.size = '4'; articleFont.color = '#000000';
+        articleFont.dataset.dww11WxBody = '1';
+        articleFont.append(...body.childNodes);
+        body.append(articleFont);
+        // A 自身的浏览器默认蓝色会覆盖外层继承色，链接文字另设黑色。
+        for (const link of articleFont.querySelectorAll('a')) {
+            const linkFont = doc.createElement('font');
+            linkFont.size = '4'; linkFont.color = '#000000';
+            linkFont.append(...link.childNodes); link.append(linkFont);
+        }
+        // 只用于编辑器图片适配；帖子展示仍交给网站原有图片/附件规则。
+        for (const img of articleFont.querySelectorAll('img')) {
+            img.style.maxWidth = '100%'; img.style.height = 'auto';
+        }
+        const source = doc.createElement('p');
+        source.dataset.dww11WxSource = '1';
+        const font = doc.createElement('font');
+        font.size = '1'; font.color = '#c0c0c0';
+        const lines = (sourceText ?? `来源：${author}，如有侵权，请联系删除。`).split('\n');
+        lines.forEach((line, index) => {
+            if (index) font.append(doc.createElement('br'));
+            font.append(doc.createTextNode(line));
+        });
+        source.append(font); body.append(source);
+    }
+    function existingSource(body) {
+        // 按段落和 BR 读取末尾逻辑行，兼容嵌套 FONT、纯文本及末尾空段落。
+        const lines = [];
+        let parts = [];
+        function flush() {
+            const text = cleanText(parts.map(part => part.node.textContent.slice(part.start, part.end)).join(''));
+            if (text) lines.push({ text, parts });
+            parts = [];
+        }
+        function visit(node) {
+            if (node.nodeType === 3) {
+                // 保留文本中的显式换行，并记录精确位置，避免删除同容器中的正文。
+                const value = node.textContent;
+                let start = 0;
+                for (let i = 0; i <= value.length; i++) {
+                    if (i !== value.length && value[i] !== '\n') continue;
+                    // HTML 缩进换行通常只是空白；只有后续是来源或版权行才分行。
+                    if (i < value.length && !/^\s*(?:(?:来源|文章来源|本文来源|图文来源|文字来源|作者|作者信息|来源作者|转自|转载自|出处)\s*[:：]|如有侵权|若有侵权|侵权请|转载请|版权归)/.test(value.slice(i + 1))) continue;
+                    if (i > start) parts.push({ node, start, end: i });
+                    if (i < value.length) flush();
+                    start = i + 1;
+                }
+                return;
+            }
+            if (node.nodeType !== 1 || ['SCRIPT', 'STYLE', 'NOSCRIPT'].includes(node.tagName)) return;
+            if (node.tagName === 'BR') { flush(); return; }
+            if (['IMG', 'VIDEO', 'AUDIO', 'IFRAME', 'HR'].includes(node.tagName)) {
+                flush(); lines.push({ text: '\ufffc', parts: [] }); return;
+            }
+            const block = /^(P|DIV|SECTION|ARTICLE|FIGURE|FIGCAPTION|H[1-6]|LI|TR|BLOCKQUOTE|PRE)$/.test(node.tagName);
+            if (block) flush();
+            for (const child of node.childNodes) visit(child);
+            if (block) flush();
+        }
+        for (const child of body.childNodes) visit(child);
+        flush();
+        let first = -1;
+        for (let i = lines.length - 1; i >= 0; i--) {
+            if (/^(?:来源|文章来源|本文来源|图文来源|文字来源|作者|作者信息|来源作者|转自|转载自|出处)\s*[:：]\s*\S/.test(lines[i].text)) {
+                first = i;
+            } else if (!/^(?:如有侵权|若有侵权|侵权请|转载请|版权归)/.test(lines[i].text)) {
+                break;
+            }
+        }
+        if (first < 0) return null;
+        const start = lines[first].parts[0];
+        const range = body.ownerDocument.createRange();
+        range.setStart(start.node, start.start);
+        range.setEnd(body, body.childNodes.length);
+        return {
+            text: lines.slice(first).map(line => line.text).join('\n'),
+            remove: () => range.deleteContents()
+        };
+    }
+
+window.DWW11Formatter = Object.freeze({
+  format(expectedSource) {
+    if (Number(window.wysiwyg) !== 1) throw new Error('请切换到可视化编辑器后排版');
+    const body = document.querySelector('#e_iframe')?.contentDocument?.body;
+    if (!body) throw new Error('编辑器未就绪');
+    const footer = existingSource(body);
+    const expected = `来源：${expectedSource}，如有侵权，请联系删除。`;
+    if (!footer || footer.text !== expected) throw new Error('文末来源与文章包不一致，请人工检查');
+    const images = [...body.querySelectorAll('img')].map(x => x.id + ':' + x.getAttribute('src'));
+    const backup = body.innerHTML;
+    try {
+      footer.remove();
+      formatBody(body, expectedSource, expected);
+      const after = [...body.querySelectorAll('img')].map(x => x.id + ':' + x.getAttribute('src'));
+      if (JSON.stringify(images) !== JSON.stringify(after)) throw new Error('排版后图片标识或顺序改变');
+      if (body.querySelectorAll('[data-dww11-wx-source="1"]').length !== 1) throw new Error('来源尾注数量异常');
+      body.dispatchEvent(new Event('input', {bubbles:true}));
+      return {images: after, source: expected};
+    } catch (error) { body.innerHTML = backup; throw error; }
+  }
+});
+})();
+
+/* Page-side orchestrator. This file never clicks or submits the publication form. */
+(() => {
+  'use strict';
+  if (window.DWW11NewsRunner) return;
+  window.DWW11NewsRunner = true;
+  const SLOT='auto-news-current';
+  const owner=sessionStorage.getItem('auto-news-owner') || crypto.randomUUID();
+  sessionStorage.setItem('auto-news-owner',owner);
+  let run=null, working=false, halted=true, aborter=null, active=null, tickBusy=false;
+  const pending=new Map();
+  function rpc(route,body={}) {
+    return new Promise((resolve,reject)=>{
+      const id=crypto.randomUUID();
+      const timer=setTimeout(()=>{pending.delete(id);reject(new Error('本地连接无响应'));},65000);
+      pending.set(id,{resolve,reject,timer});
+      document.dispatchEvent(new CustomEvent('dww11:local-request',{detail:JSON.stringify({id,route,body:{owner,...body}})}));
+    });
+  }
+  document.addEventListener('dww11:local-response',event=>{
+    let data;try{data=JSON.parse(event.detail);}catch{return;}
+    const waiter=pending.get(data.id);if(!waiter)return;
+    clearTimeout(waiter.timer);pending.delete(data.id);
+    data.error?waiter.reject(new Error(data.error)):waiter.resolve(data.value);
+  });
+  const status=message=>document.dispatchEvent(new CustomEvent('dww11:connection-status',{detail:message}));
+  document.addEventListener('dww11:auto-stop',()=>{halted=true;aborter?.abort(new Error('流程已终止'));});
+  function account() {
+    const element=document.querySelector('#mumucms_username');
+    if(!element)return null;
+    const name=[...element.childNodes].filter(n=>n.nodeType===Node.TEXT_NODE).map(n=>n.textContent).join('').trim();
+    const links=[element,...element.querySelectorAll('a')].filter(e=>e.href);
+    let uid=null;
+    for(const link of links) {
+      const href=link.href;
+      uid=new URL(href,location.href).searchParams.get('uid') || href.match(/space-uid-(\d+)/)?.[1];
+      if(uid)break;
+    }
+    return uid&&name?{uid,name}:null;
+  }
+  function extension() {
+    try{return JSON.parse(document.querySelector('#dww11-extension-panel')?.getAttribute('data-auto-news-state') || '{}');}catch{return {};}
+  }
+  function assertLive(epoch) {
+    if(halted||!run?.running||run.epoch!==epoch)throw new Error('流程已暂停或终止');
+    if(active?.account?.uid!==account()?.uid)throw new Error('当前账号与任务不一致');
+  }
+  function editorSnapshot() {
+    const body=document.querySelector('#e_iframe')?.contentDocument?.body;
+    if(!body)throw new Error('找不到可视化编辑器');
+    return {title:document.querySelector('#subject')?.value || '',
+      images:[...body.querySelectorAll('img')].map(x=>({id:x.id,src:x.getAttribute('src')})),
+      // Snapshot only final submitted content; never includes hidden tokens or the full form.
+      html:body.innerHTML.slice(0,500000), text:body.innerText.slice(0,100000)};
+  }
+  function saveLocal(value) {sessionStorage.setItem(SLOT,JSON.stringify(value));}
+  function loadLocal(){try{return JSON.parse(sessionStorage.getItem(SLOT)||'null');}catch{return null;}}
+  async function event(state,extra={}) {return rpc('event',{id:active.id,state,epoch:run.epoch,...extra});}
+  async function fail(error) {
+    status(error.message);
+    if(active)try{await event('needs_attention',{error:error.message});}catch{}
+    halted=true;aborter?.abort();document.dispatchEvent(new CustomEvent('dww11:auto-stop'));
+  }
+  function editorPage(job) {
+    const u=new URL(location.href);
+    return u.searchParams.get('mod')==='post'&&u.searchParams.get('action')==='newthread'&&u.searchParams.get('fid')===String(job.fid);
+  }
+  async function prepare(job) {
+    working=true;active=job;aborter=new AbortController();const epoch=run.epoch;
+    try {
+      assertLive(epoch);
+      if(!editorPage(job)) {
+        saveLocal({id:job.id});
+        location.assign(`https://www.dww11.com/forum.php?mod=post&action=newthread&fid=${job.fid}`);return;
+      }
+      if(!window.DWW11Importer||!window.DWW11Formatter)throw new Error('导入或排版模块尚未就绪，请检查安装');
+      status('正在检查文章包…');
+      const pack=await rpc('package',{id:job.id});assertLive(epoch);
+      const prepared=await window.DWW11Importer.stagePackage(pack);assertLive(epoch);
+      if(prepared.warnings?.length)throw new Error('文章需要人工处理：'+prepared.warnings.join('；'));
+      await event('importing');assertLive(epoch);status('上传图片并填写文章…');
+      const result=await window.DWW11Importer.importPrepared({automatic:true,signal:aborter.signal});assertLive(epoch);
+      if(result?.cancelled)throw new Error('导入已取消');
+      await event('formatting');assertLive(epoch);
+      window.DWW11Formatter.format(job.source);assertLive(epoch);
+      const snapshot=editorSnapshot();
+      if(snapshot.title!==job.title)throw new Error('标题校验失败');
+      const aids=new Set(Object.values(result.images));
+      for(const aid of aids) if(!snapshot.images.some(x=>x.id==='attachimg_'+aid))throw new Error('配图附件缺失：'+aid);
+      await event('awaiting_user_review',{snapshot});assertLive(epoch);
+      active={...job,state:'awaiting_user_review'};saveLocal({id:job.id});
+      status('已完成排版，请核对内容并亲手点击“发表帖子”。');
+    }catch(error){await fail(error);}finally{working=false;}
+  }
+  // Record only trusted user activation. Never replay a click, call submit(), or send a post request.
+  document.addEventListener('click',event=>{
+    if(!event.isTrusted||!event.target.closest?.('#postsubmit')||active?.state!=='awaiting_user_review')return;
+    try {
+      const snapshot=editorSnapshot();
+      const local={id:active.id,snapshot,clickedAt:Date.now(),observed:false};saveLocal(local);
+      rpc('event',{id:active.id,state:'user_submitting',epoch:run?.epoch,snapshot})
+        .then(()=>{const latest=loadLocal();if(latest?.clickedAt===local.clickedAt)saveLocal({...local,observed:true});})
+        .catch(error=>status('提交记录待恢复：'+error.message));
+      active.state='user_submitting';status('已观察到人工点击，等待结果页核对。');
+    }catch(error){status('提交快照记录失败：'+error.message);}
+  },true);
+  async function observeResult(job) {
+    const local=loadLocal();
+    if(local?.id===job.id&&local.snapshot&&job.state==='awaiting_user_review') {
+      await rpc('event',{id:job.id,state:'user_submitting',epoch:run.epoch,snapshot:local.snapshot});
+      job.state='user_submitting';
+    }
+    if(job.state==='user_submitting') {
+      // Result selectors intentionally await user-provided HTML; URL alone is not proof.
+      if(!editorPage(job) || (local?.clickedAt && Date.now()-local.clickedAt>45000)) {
+        await rpc('event',{id:job.id,state:'submit_unknown',error:'等待结果页 HTML 适配；候选页面：'+location.href});
+        await rpc('stop');
+        status('已记录候选结果页；尚不能自动核对。流程已暂停，请补充结果页 HTML。');
+      }
+    }else if(job.state==='submit_unknown')status('发布结果待核对；不会重发、计数或换号。');
+  }
+  async function switchAccount() {
+    if(working||halted)return;
+    if(location.origin!=='https://www.dww11.com'||location.pathname!=='/') {
+      location.assign('https://www.dww11.com/');return;
+    }
+    const ext=extension(),button=document.querySelector('#dww11-extension-next');
+    if(!ext.ready||!ext.idle||!button||button.disabled) {status('请安装并重新加载带联动接口的换号插件');return;}
+    await rpc('switch-begin');
+    if(halted||!run?.running)return;
+    button.click();status('正在换号；等待插件完成并核对新账号');
+  }
+  async function tick() {
+    if(tickBusy)return;tickBusy=true;
+    try {
+      run=await rpc('heartbeat',{account:account(),url:location.href,extension:extension()});
+      halted=!run.running;
+      if(!run.running) {aborter?.abort();document.dispatchEvent(new CustomEvent('dww11:auto-stop'));}
+      if(working)return;
+      // Passive result recording is permitted after Stop, but no new business actions.
+      if(run.active) {
+        const local=loadLocal();
+        if(local?.id===run.active&&local.snapshot&&!run.running)status('流程已暂停；提交结果仍需核对');
+      }
+      if(!run.running) {
+        if(run.active && run.owner===owner) {const saved=await rpc('current');if(saved.job) {active=saved.job;await observeResult(saved.job);}}
+        status(run.reason || '流程已暂停');return;
+      }
+      const next=await rpc('next',{account:account(),extension:extension()});run=next.run;
+      if(next.job)active=next.job;
+      if(next.action==='prepare'||(next.action==='resume'&&next.job.state==='preparing'))void prepare(next.job);
+      else if(next.action==='resume') {
+        await observeResult(next.job);
+        if(next.job.state==='awaiting_user_review') {
+          if(!editorPage(next.job)||!document.querySelector('#subject')?.value) await fail(new Error('待审稿页面已丢失，请人工核对旧稿，不自动重新导入'));
+          else status('等待你亲手发表；程序不再修改编辑器');
+        }
+        else if(['importing','formatting'].includes(next.job.state))await fail(new Error('导入阶段发生中断，请检查编辑器和已上传图片后处理任务'));
+        else if(next.job.state==='needs_attention')status(next.job.error);
+      }else if(next.action==='switch')await switchAccount();
+      else if(next.action==='switch_wait')status('等待换号结果；若插件已暂停，请人工处理后继续');
+      else if(next.action==='login')status('未能读取当前账号 UID，请登录并确认账号元素');
+      else if(next.action==='other_tab')status('另一个标签持有任务；如需接管请先暂停');
+      else if(next.action==='interval')status('等待设定的处理间隔');
+      else status(run.reason || '等待任务');
+    }catch(error){status(error.message);halted=true;aborter?.abort();document.dispatchEvent(new CustomEvent('dww11:auto-stop'));}
+    finally{tickBusy=false;}
+  }
+  const takeover=document.createElement('button');takeover.type='button';takeover.textContent='接管已暂停任务';
+  takeover.style.cssText='position:fixed;left:16px;top:16px;z-index:2147483646;padding:8px;background:#fff;border:1px solid #aab;border-radius:6px';
+  takeover.onclick=async()=>{try{await rpc('takeover');status('已接管；核对当前页面后在工作台继续');}catch(error){status(error.message);}};
+  document.body.append(takeover);
+  setInterval(tick,2000);tick();
+})();

browser/formatter-original.user.js

@@ -0,0 +1,397 @@
+// ==UserScript==
+// @name         广谈大湾网 · 微信文章一键导入并排版
+// @namespace    local.dww11.wechat-oneclick
+// @version      1.0.9
+// @description  使用网站现有微信采集器，修正标题，统一正文4号黑色，追加1号银色来源；不点击保存或发布。
+// @match        https://www.dww11.com/forum.php*
+// @match        https://dww11.com/forum.php*
+// @grant        GM_xmlhttpRequest
+// @grant        unsafeWindow
+// @connect      mp.weixin.qq.com
+// @run-at       document-idle
+// @noframes
+// ==/UserScript==
+
+(() => {
+    'use strict';
+    const ID = 'dww11-wx-oneclick';
+    const FORMAT_ID = 'dww11-format-oneclick';
+    let busy = false;
+    function setBusy(value) {
+        busy = value;
+        for (const id of [ID, FORMAT_ID]) {
+            const button = document.getElementById(id);
+            if (button) button.disabled = value;
+        }
+    }
+    const sleep = ms => new Promise(resolve => setTimeout(resolve, ms));
+    const visible = el => el && el.getClientRects().length && getComputedStyle(el).display !== 'none';
+    async function waitFor(fn, timeout = 10000) {
+        const deadline = Date.now() + timeout;
+        while (Date.now() < deadline) {
+            const result = fn();
+            if (result) return result;
+            await sleep(100);
+        }
+        throw new Error('等待编辑器或采集窗口超时，请关闭遮挡窗口后重试。');
+    }
+    function normalizeURL(raw) {
+        const url = new URL(raw.trim().replace(/[\u200b-\u200d\ufeff]/g, ''));
+        if (url.protocol !== 'https:' || url.hostname !== 'mp.weixin.qq.com' || url.port || url.username || url.password || !/^\/(s(?:\/|$)|s$)/.test(url.pathname)) {
+            throw new Error('请输入 https://mp.weixin.qq.com/s/… 或 /s?… 微信文章地址。');
+        }
+        url.hash = '';
+        return url.href;
+    }
+    function cleanText(value) {
+        return String(value || '').replace(/\s+/g, ' ').trim();
+    }
+    function metadata(html) {
+        const doc = new DOMParser().parseFromString(html, 'text/html');
+        const text = selector => cleanText(doc.querySelector(selector)?.textContent);
+        const meta = selector => cleanText(doc.querySelector(selector)?.getAttribute('content'));
+        // 与下载组件一致，先确认是真正的文章，避免把验证页标题当成文章。
+        if (!doc.querySelector('#js_content')) throw new Error('未找到微信文章正文，可能需要验证、登录，或文章已失效。');
+        const title = text('#activity-name .js_title_inner') || text('#activity-name') || meta('meta[property="og:title"]') || meta('meta[name="twitter:title"]') || text('h1') || text('title');
+        // 不读取 rich_media_meta_text 容器：它可能包含重复的隐藏作者，或实际上是日期/IP。
+        // 来源使用蓝色链接 #js_name 的公众号名称，不使用灰色署名作者。
+        const author = text('#js_name');
+        if (!title || !author) throw new Error('微信未返回完整标题和公众号名称，可能需要验证或登录。');
+        return { title, author };
+    }
+    function fetchMetadata(url) {
+        return new Promise((resolve, reject) => {
+            GM_xmlhttpRequest({
+                method: 'GET', url, timeout: 25000, anonymous: true,
+                onload(response) {
+                    try {
+                        if (response.status !== 200) throw new Error('微信返回状态 ' + response.status);
+                        if (response.finalUrl && new URL(response.finalUrl).hostname !== 'mp.weixin.qq.com') throw new Error('微信地址发生了非预期跳转。');
+                        resolve(metadata(response.responseText));
+                    } catch (error) { reject(error); }
+                },
+                onerror: () => reject(new Error('无法读取微信文章。')),
+                ontimeout: () => reject(new Error('读取微信文章超时。'))
+            });
+        });
+    }
+    async function getMetadata(url) {
+        try { return await fetchMetadata(url); }
+        catch (error) {
+            const title = prompt(error.message + '\n请从源文章复制完整标题（取消则停止导入）：');
+            if (title === null) return null;
+            const author = prompt('请填写源文章蓝色链接处的公众号名称（取消则停止导入）：');
+            if (author === null) return null;
+            if (!cleanText(title) || !cleanText(author)) throw new Error('标题和作者不能为空。');
+            return { title: cleanText(title), author: cleanText(author) };
+        }
+    }
+    function editorBody() {
+        return document.querySelector('#e_iframe')?.contentDocument?.body;
+    }
+    async function ensureVisualEditor() {
+        const checkbox = document.querySelector('#e_switchercheck');
+        if (checkbox?.checked) checkbox.click();
+        await waitFor(() => visible(document.querySelector('#e_iframe')) && editorBody());
+    }
+    // 只监听本次现有采集器的响应，不重写其请求或图片附件处理。
+    function watchImport(jq, url) {
+        let timer, resolveResult, rejectResult;
+        const result = new Promise((resolve, reject) => { resolveResult = resolve; rejectResult = reject; });
+        const events = jq(document);
+        const cleanup = () => { clearTimeout(timer); events.off('.dww11wx'); };
+        const finish = error => { cleanup(); error ? rejectResult(error) : resolveResult(); };
+        events.on('ajaxComplete.dww11wx', (_event, xhr, settings) => {
+            let requestURL, sentURL;
+            try {
+                requestURL = new URL(settings.url, location.href);
+                sentURL = typeof settings.data === 'string' ? new URLSearchParams(settings.data).get('url') : settings.data?.url;
+            } catch { return; }
+            if (requestURL.origin !== location.origin || requestURL.searchParams.get('id') !== 'wxcj:cj' || sentURL !== url) return;
+            const parts = String(xhr.responseText || '').split('----wxcj----');
+            if (xhr.status < 200 || xhr.status >= 300 || parts.length < 4 || !parts[2]?.trim()) {
+                finish(new Error('网站采集失败或返回内容不完整。未执行排版，请检查采集提示。'));
+                return;
+            }
+            finish();
+        });
+        timer = setTimeout(() => finish(new Error('网站采集超过90秒，未执行排版。采集器仍可能在后台返回，请勿连续重复导入。')), 90000);
+        return { result, cleanup };
+    }
+    // 将微信 P/SECTION 的隐式换行改成真实 BR，避免论坛忽略 SECTION 边界。
+    function flattenParagraphs(body) {
+        const selector = 'p,div,section,article,figure,figcaption,h1,h2,h3,h4,h5,h6';
+        const marker = () => {
+            const br = body.ownerDocument.createElement('br');
+            br.dataset.dww11Boundary = '1'; return br;
+        };
+        for (const block of Array.from(body.querySelectorAll(selector)).reverse()) {
+            if (block.closest('table,ul,ol,blockquote')) continue;
+            block.replaceWith(marker(), ...block.childNodes, marker());
+        }
+        let previous = false, pending = [];
+        function finish() {
+            const actual = pending.filter(br => !br.hasAttribute('data-dww11-boundary'));
+            let retained = false;
+            for (const br of pending) {
+                if (!br.hasAttribute('data-dww11-boundary')) continue;
+                if (previous && !actual.length && !retained) { br.removeAttribute('data-dww11-boundary'); retained = true; }
+                else br.remove();
+            }
+            pending = [];
+        }
+        const walker = body.ownerDocument.createTreeWalker(body, NodeFilter.SHOW_ELEMENT | NodeFilter.SHOW_TEXT);
+        const nodes = []; while (walker.nextNode()) nodes.push(walker.currentNode);
+        for (const node of nodes) {
+            if (node.nodeType === 1 && node.tagName === 'BR') pending.push(node);
+            else if ((node.nodeType === 3 && node.textContent.trim()) || (node.nodeType === 1 && ['IMG', 'VIDEO', 'HR', 'IFRAME'].includes(node.tagName))) {
+                finish(); previous = true;
+            }
+        }
+        // 末尾不保留由容器边界制造的额外换行。
+        for (const br of pending) if (br.hasAttribute('data-dww11-boundary')) br.remove();
+    }
+    function spaceImages(body) {
+        let previous = null, gapNodes = [], blankLines = 0;
+        const changes = [];
+        function content(kind, node) {
+            if (previous && previous.kind !== kind) {
+                let anchor = node;
+                while (anchor.parentElement && anchor.parentElement !== body &&
+                       !['TD', 'TH', 'LI', 'BLOCKQUOTE'].includes(anchor.parentElement.tagName) && !anchor.parentElement.contains(previous.node)) anchor = anchor.parentElement;
+                changes.push({ anchor, nodes: gapNodes.slice(), count: Math.max(2, blankLines) });
+            }
+            previous = { kind, node }; gapNodes = []; blankLines = 0;
+        }
+        function visit(node) {
+            if (node.nodeType === 3) {
+                if (node.textContent.trim()) content('text', node); else gapNodes.push(node);
+                return;
+            }
+            if (node.nodeType !== 1 || ['SCRIPT', 'STYLE', 'NOSCRIPT'].includes(node.tagName)) return;
+            if (node.tagName === 'BR') { gapNodes.push(node); blankLines++; return; }
+            if (node.tagName === 'IMG') { content('image', node); return; }
+            for (const child of node.childNodes) visit(child);
+        }
+        for (const child of body.childNodes) visit(child);
+        for (const { anchor, nodes, count } of changes) {
+            for (const node of nodes) node.remove();
+            const fragment = body.ownerDocument.createDocumentFragment();
+            for (let i = 0; i < count; i++) fragment.append(body.ownerDocument.createElement('br'));
+            anchor.parentNode.insertBefore(fragment, anchor);
+        }
+    }
+    function formatBody(body, author, sourceText = null) {
+        const doc = body.ownerDocument;
+        body.querySelectorAll('[data-dww11-wx-source="1"]').forEach(el => el.remove());
+        // Discuz 会把内联 color 值直接转为 [color=…]。
+        // !important 不能出现在该值中；颜色和字号只用原生 FONT 属性。
+        for (const el of [body, ...body.querySelectorAll('*')]) {
+            if (['SCRIPT', 'STYLE'].includes(el.tagName)) continue;
+            for (const property of ['font', 'font-size', 'color', '-webkit-text-fill-color']) {
+                el.style.removeProperty(property);
+            }
+            // 其他排版属性也不保留 CSS 优先级，避免它泄漏进论坛标签。
+            for (const property of Array.from(el.style)) {
+                if (el.style.getPropertyPriority(property)) {
+                    el.style.setProperty(property, el.style.getPropertyValue(property));
+                }
+            }
+            if (!el.style.cssText) el.removeAttribute('style');
+        }
+        // 去掉采集器或旧版脚本生成的嵌套 FONT，保留其中的内容和图片。
+        for (const font of Array.from(body.querySelectorAll('font'))) {
+            font.replaceWith(...font.childNodes);
+        }
+        flattenParagraphs(body);
+        spaceImages(body);
+        // 所有正文内容和换行放在同一 FONT 中。不能使用只有 BR 的空 FONT：
+        // 论坛转换器可能清除空格式标签，导致空行退回默认1号或直接丢失。
+        const articleFont = doc.createElement('font');
+        articleFont.size = '4'; articleFont.color = '#000000';
+        articleFont.dataset.dww11WxBody = '1';
+        articleFont.append(...body.childNodes);
+        body.append(articleFont);
+        // A 自身的浏览器默认蓝色会覆盖外层继承色，链接文字另设黑色。
+        for (const link of articleFont.querySelectorAll('a')) {
+            const linkFont = doc.createElement('font');
+            linkFont.size = '4'; linkFont.color = '#000000';
+            linkFont.append(...link.childNodes); link.append(linkFont);
+        }
+        // 只用于编辑器图片适配；帖子展示仍交给网站原有图片/附件规则。
+        for (const img of articleFont.querySelectorAll('img')) {
+            img.style.maxWidth = '100%'; img.style.height = 'auto';
+        }
+        const source = doc.createElement('p');
+        source.dataset.dww11WxSource = '1';
+        const font = doc.createElement('font');
+        font.size = '1'; font.color = '#c0c0c0';
+        const lines = (sourceText ?? `来源：${author}，如有侵权，请联系删除。`).split('\n');
+        lines.forEach((line, index) => {
+            if (index) font.append(doc.createElement('br'));
+            font.append(doc.createTextNode(line));
+        });
+        source.append(font); body.append(source);
+    }
+    function existingSource(body) {
+        // 按段落和 BR 读取末尾逻辑行，兼容嵌套 FONT、纯文本及末尾空段落。
+        const lines = [];
+        let parts = [];
+        function flush() {
+            const text = cleanText(parts.map(part => part.node.textContent.slice(part.start, part.end)).join(''));
+            if (text) lines.push({ text, parts });
+            parts = [];
+        }
+        function visit(node) {
+            if (node.nodeType === 3) {
+                // 保留文本中的显式换行，并记录精确位置，避免删除同容器中的正文。
+                const value = node.textContent;
+                let start = 0;
+                for (let i = 0; i <= value.length; i++) {
+                    if (i !== value.length && value[i] !== '\n') continue;
+                    // HTML 缩进换行通常只是空白；只有后续是来源或版权行才分行。
+                    if (i < value.length && !/^\s*(?:(?:来源|文章来源|本文来源|图文来源|文字来源|作者|作者信息|来源作者|转自|转载自|出处)\s*[:：]|如有侵权|若有侵权|侵权请|转载请|版权归)/.test(value.slice(i + 1))) continue;
+                    if (i > start) parts.push({ node, start, end: i });
+                    if (i < value.length) flush();
+                    start = i + 1;
+                }
+                return;
+            }
+            if (node.nodeType !== 1 || ['SCRIPT', 'STYLE', 'NOSCRIPT'].includes(node.tagName)) return;
+            if (node.tagName === 'BR') { flush(); return; }
+            if (['IMG', 'VIDEO', 'AUDIO', 'IFRAME', 'HR'].includes(node.tagName)) {
+                flush(); lines.push({ text: '\ufffc', parts: [] }); return;
+            }
+            const block = /^(P|DIV|SECTION|ARTICLE|FIGURE|FIGCAPTION|H[1-6]|LI|TR|BLOCKQUOTE|PRE)$/.test(node.tagName);
+            if (block) flush();
+            for (const child of node.childNodes) visit(child);
+            if (block) flush();
+        }
+        for (const child of body.childNodes) visit(child);
+        flush();
+        let first = -1;
+        for (let i = lines.length - 1; i >= 0; i--) {
+            if (/^(?:来源|文章来源|本文来源|图文来源|文字来源|作者|作者信息|来源作者|转自|转载自|出处)\s*[:：]\s*\S/.test(lines[i].text)) {
+                first = i;
+            } else if (!/^(?:如有侵权|若有侵权|侵权请|转载请|版权归)/.test(lines[i].text)) {
+                break;
+            }
+        }
+        if (first < 0) return null;
+        const start = lines[first].parts[0];
+        const range = body.ownerDocument.createRange();
+        range.setStart(start.node, start.start);
+        range.setEnd(body, body.childNodes.length);
+        return {
+            text: lines.slice(first).map(line => line.text).join('\n'),
+            remove: () => range.deleteContents()
+        };
+    }
+    async function runFormat(button) {
+        if (busy) return;
+        setBusy(true);
+        try {
+            await ensureVisualEditor();
+            const body = editorBody();
+            const source = existingSource(body);
+            const hasText = cleanText(body.textContent).replace(source ? cleanText(source.text) : '', '').trim();
+            if (!hasText && !body.querySelector('img,video,audio,iframe,hr,table')) {
+                throw new Error('正文为空，请先输入或粘贴文章。');
+            }
+            const raw = source ? '' : prompt('未在文末检测到来源／作者信息，请补充来源公众号名称（取消则停止排版）：');
+            if (raw === null) return;
+            const author = cleanText(raw);
+            if (!source && !author) throw new Error('来源公众号名称不能为空。');
+            button.textContent = '正在排版…';
+            // 已有来源保留原文，仅调整字号颜色；没有时才生成来源行。
+            source?.remove();
+            formatBody(body, author, source?.text ?? null);
+            alert(`排版完成。\n正文：4号黑色；来源：1号银色。\n${source ? source.text : `来源：${author}`}\n请人工检查后自行决定是否保存或发布。`);
+        } catch (error) { alert('排版未完成：' + error.message); }
+        finally {
+            setBusy(false);
+            button.textContent = '一键排版';
+        }
+    }
+    async function run(button) {
+        if (busy) return;
+        const raw = prompt('粘贴微信公众号文章 URL：');
+        if (raw === null || !raw.trim()) return;
+        setBusy(true);
+        let watcher;
+        try {
+            const url = normalizeURL(raw);
+            const jq = unsafeWindow.myJQ;
+            if (typeof jq !== 'function' || !jq.fn?.on) throw new Error('没有找到网站微信采集器的 myJQ，请刷新页面后重试。');
+            const plugin = document.querySelector('#e_controls .wxcj');
+            if (!plugin) throw new Error('当前页面没有微信采集组件。');
+            button.textContent = '读取标题和作者…';
+            const info = await getMetadata(url);
+            if (!info) return;
+            const subject = document.querySelector('#subject');
+            if (!subject) throw new Error('没有找到标题框。');
+            if (subject.maxLength > 0 && info.title.length > subject.maxLength) throw new Error('完整标题超过标题框长度限制，请先处理标题长度。');
+            await ensureVisualEditor();
+            button.textContent = '采集文章中…';
+            plugin.click();
+            const input = await waitFor(() => Array.from(document.querySelectorAll('.wxjc-dialog input')).find(visible));
+            const submit = input.closest('.wxjc-dialog').querySelector('.wxjc-subimt');
+            if (!submit) throw new Error('没有找到采集按钮。');
+            input.value = url;
+            input.dispatchEvent(new Event('input', { bubbles: true }));
+            input.dispatchEvent(new Event('change', { bubbles: true }));
+            watcher = watchImport(jq, url);
+            // 先登记失败处理，避免点击失败时留下未处理的 Promise。
+            watcher.result.catch(() => {});
+            submit.click();
+            await watcher.result;
+            button.textContent = '正在排版…';
+            const body = await waitFor(() => editorBody()?.innerHTML.trim() && editorBody());
+            formatBody(body, info.author);
+            subject.value = info.title;
+            subject.dispatchEvent(new Event('input', { bubbles: true }));
+            subject.dispatchEvent(new Event('change', { bubbles: true }));
+            subject.dispatchEvent(new KeyboardEvent('keyup', { bubbles: true }));
+            // 正文保持在可视化编辑器，由网站正常提交流程转成 BBCode。
+            // 不把 HTML 写入纯文本 message，也不调用任何保存／发布入口。
+            alert(`导入并排版完成。\n标题：${info.title}\n作者：${info.author}\n正文：4号黑色；来源：1号银色。\n请人工检查后自行决定是否保存或发布。`);
+        } catch (error) { alert('导入未完成：' + error.message); }
+        finally {
+            watcher?.cleanup();
+            setBusy(false);
+            button.textContent = '一键导入微信公众号并排版';
+        }
+    }
+    function mount() {
+        const toolbar = document.querySelector('#e_controls');
+        if (!toolbar || !document.querySelector('#subject')) return;
+        const style = 'position:relative;z-index:2;padding:0 12px;height:32px;line-height:30px;border:1px solid #159447;border-radius:4px;background:#159447;color:white;font-size:13px;cursor:pointer;';
+        // 右浮动的同一组内使用 flex，保证排版按钮始终位于导入按钮左侧。
+        let group = document.getElementById('dww11-wx-actions');
+        if (!group) {
+            group = document.createElement('span');
+            group.id = 'dww11-wx-actions';
+            group.style.cssText = 'float:right;display:inline-flex;align-items:center;gap:8px;margin:6px 8px;';
+            toolbar.append(group);
+        }
+        if (!document.getElementById(FORMAT_ID)) {
+            const button = document.createElement('button');
+            button.id = FORMAT_ID; button.type = 'button';
+            button.textContent = '一键排版'; button.style.cssText = style;
+            button.disabled = busy;
+            button.addEventListener('click', () => runFormat(button));
+            group.prepend(button);
+        }
+        if (toolbar.querySelector('.wxcj') && !document.getElementById(ID)) {
+            const button = document.createElement('button');
+            button.id = ID; button.type = 'button';
+            button.textContent = '一键导入微信公众号并排版'; button.style.cssText = style;
+            button.disabled = busy;
+            button.addEventListener('click', () => run(button));
+            group.append(button);
+        }
+    }
+    mount();
+    const observer = new MutationObserver(mount);
+    observer.observe(document.documentElement, { childList: true, subtree: true });
+})();

browser/formatter.js

@@ -0,0 +1,195 @@
+// Extracted from dww11-wechat-oneclick.user.js v1.0.9; package source is authoritative.
+(() => {
+'use strict';
+const cleanText = value => String(value || '').replace(/\s+/g, ' ').trim();
+    function flattenParagraphs(body) {
+        const selector = 'p,div,section,article,figure,figcaption,h1,h2,h3,h4,h5,h6';
+        const marker = () => {
+            const br = body.ownerDocument.createElement('br');
+            br.dataset.dww11Boundary = '1'; return br;
+        };
+        for (const block of Array.from(body.querySelectorAll(selector)).reverse()) {
+            if (block.closest('table,ul,ol,blockquote')) continue;
+            block.replaceWith(marker(), ...block.childNodes, marker());
+        }
+        let previous = false, pending = [];
+        function finish() {
+            const actual = pending.filter(br => !br.hasAttribute('data-dww11-boundary'));
+            let retained = false;
+            for (const br of pending) {
+                if (!br.hasAttribute('data-dww11-boundary')) continue;
+                if (previous && !actual.length && !retained) { br.removeAttribute('data-dww11-boundary'); retained = true; }
+                else br.remove();
+            }
+            pending = [];
+        }
+        const walker = body.ownerDocument.createTreeWalker(body, NodeFilter.SHOW_ELEMENT | NodeFilter.SHOW_TEXT);
+        const nodes = []; while (walker.nextNode()) nodes.push(walker.currentNode);
+        for (const node of nodes) {
+            if (node.nodeType === 1 && node.tagName === 'BR') pending.push(node);
+            else if ((node.nodeType === 3 && node.textContent.trim()) || (node.nodeType === 1 && ['IMG', 'VIDEO', 'HR', 'IFRAME'].includes(node.tagName))) {
+                finish(); previous = true;
+            }
+        }
+        // 末尾不保留由容器边界制造的额外换行。
+        for (const br of pending) if (br.hasAttribute('data-dww11-boundary')) br.remove();
+    }
+    function spaceImages(body) {
+        let previous = null, gapNodes = [], blankLines = 0;
+        const changes = [];
+        function content(kind, node) {
+            if (previous && previous.kind !== kind) {
+                let anchor = node;
+                while (anchor.parentElement && anchor.parentElement !== body &&
+                       !['TD', 'TH', 'LI', 'BLOCKQUOTE'].includes(anchor.parentElement.tagName) && !anchor.parentElement.contains(previous.node)) anchor = anchor.parentElement;
+                changes.push({ anchor, nodes: gapNodes.slice(), count: Math.max(2, blankLines) });
+            }
+            previous = { kind, node }; gapNodes = []; blankLines = 0;
+        }
+        function visit(node) {
+            if (node.nodeType === 3) {
+                if (node.textContent.trim()) content('text', node); else gapNodes.push(node);
+                return;
+            }
+            if (node.nodeType !== 1 || ['SCRIPT', 'STYLE', 'NOSCRIPT'].includes(node.tagName)) return;
+            if (node.tagName === 'BR') { gapNodes.push(node); blankLines++; return; }
+            if (node.tagName === 'IMG') { content('image', node); return; }
+            for (const child of node.childNodes) visit(child);
+        }
+        for (const child of body.childNodes) visit(child);
+        for (const { anchor, nodes, count } of changes) {
+            for (const node of nodes) node.remove();
+            const fragment = body.ownerDocument.createDocumentFragment();
+            for (let i = 0; i < count; i++) fragment.append(body.ownerDocument.createElement('br'));
+            anchor.parentNode.insertBefore(fragment, anchor);
+        }
+    }
+    function formatBody(body, author, sourceText = null) {
+        const doc = body.ownerDocument;
+        body.querySelectorAll('[data-dww11-wx-source="1"]').forEach(el => el.remove());
+        // Discuz 会把内联 color 值直接转为 [color=…]。
+        // !important 不能出现在该值中；颜色和字号只用原生 FONT 属性。
+        for (const el of [body, ...body.querySelectorAll('*')]) {
+            if (['SCRIPT', 'STYLE'].includes(el.tagName)) continue;
+            for (const property of ['font', 'font-size', 'color', '-webkit-text-fill-color']) {
+                el.style.removeProperty(property);
+            }
+            // 其他排版属性也不保留 CSS 优先级，避免它泄漏进论坛标签。
+            for (const property of Array.from(el.style)) {
+                if (el.style.getPropertyPriority(property)) {
+                    el.style.setProperty(property, el.style.getPropertyValue(property));
+                }
+            }
+            if (!el.style.cssText) el.removeAttribute('style');
+        }
+        // 去掉采集器或旧版脚本生成的嵌套 FONT，保留其中的内容和图片。
+        for (const font of Array.from(body.querySelectorAll('font'))) {
+            font.replaceWith(...font.childNodes);
+        }
+        flattenParagraphs(body);
+        spaceImages(body);
+        // 所有正文内容和换行放在同一 FONT 中。不能使用只有 BR 的空 FONT：
+        // 论坛转换器可能清除空格式标签，导致空行退回默认1号或直接丢失。
+        const articleFont = doc.createElement('font');
+        articleFont.size = '4'; articleFont.color = '#000000';
+        articleFont.dataset.dww11WxBody = '1';
+        articleFont.append(...body.childNodes);
+        body.append(articleFont);
+        // A 自身的浏览器默认蓝色会覆盖外层继承色，链接文字另设黑色。
+        for (const link of articleFont.querySelectorAll('a')) {
+            const linkFont = doc.createElement('font');
+            linkFont.size = '4'; linkFont.color = '#000000';
+            linkFont.append(...link.childNodes); link.append(linkFont);
+        }
+        // 只用于编辑器图片适配；帖子展示仍交给网站原有图片/附件规则。
+        for (const img of articleFont.querySelectorAll('img')) {
+            img.style.maxWidth = '100%'; img.style.height = 'auto';
+        }
+        const source = doc.createElement('p');
+        source.dataset.dww11WxSource = '1';
+        const font = doc.createElement('font');
+        font.size = '1'; font.color = '#c0c0c0';
+        const lines = (sourceText ?? `来源：${author}，如有侵权，请联系删除。`).split('\n');
+        lines.forEach((line, index) => {
+            if (index) font.append(doc.createElement('br'));
+            font.append(doc.createTextNode(line));
+        });
+        source.append(font); body.append(source);
+    }
+    function existingSource(body) {
+        // 按段落和 BR 读取末尾逻辑行，兼容嵌套 FONT、纯文本及末尾空段落。
+        const lines = [];
+        let parts = [];
+        function flush() {
+            const text = cleanText(parts.map(part => part.node.textContent.slice(part.start, part.end)).join(''));
+            if (text) lines.push({ text, parts });
+            parts = [];
+        }
+        function visit(node) {
+            if (node.nodeType === 3) {
+                // 保留文本中的显式换行，并记录精确位置，避免删除同容器中的正文。
+                const value = node.textContent;
+                let start = 0;
+                for (let i = 0; i <= value.length; i++) {
+                    if (i !== value.length && value[i] !== '\n') continue;
+                    // HTML 缩进换行通常只是空白；只有后续是来源或版权行才分行。
+                    if (i < value.length && !/^\s*(?:(?:来源|文章来源|本文来源|图文来源|文字来源|作者|作者信息|来源作者|转自|转载自|出处)\s*[:：]|如有侵权|若有侵权|侵权请|转载请|版权归)/.test(value.slice(i + 1))) continue;
+                    if (i > start) parts.push({ node, start, end: i });
+                    if (i < value.length) flush();
+                    start = i + 1;
+                }
+                return;
+            }
+            if (node.nodeType !== 1 || ['SCRIPT', 'STYLE', 'NOSCRIPT'].includes(node.tagName)) return;
+            if (node.tagName === 'BR') { flush(); return; }
+            if (['IMG', 'VIDEO', 'AUDIO', 'IFRAME', 'HR'].includes(node.tagName)) {
+                flush(); lines.push({ text: '\ufffc', parts: [] }); return;
+            }
+            const block = /^(P|DIV|SECTION|ARTICLE|FIGURE|FIGCAPTION|H[1-6]|LI|TR|BLOCKQUOTE|PRE)$/.test(node.tagName);
+            if (block) flush();
+            for (const child of node.childNodes) visit(child);
+            if (block) flush();
+        }
+        for (const child of body.childNodes) visit(child);
+        flush();
+        let first = -1;
+        for (let i = lines.length - 1; i >= 0; i--) {
+            if (/^(?:来源|文章来源|本文来源|图文来源|文字来源|作者|作者信息|来源作者|转自|转载自|出处)\s*[:：]\s*\S/.test(lines[i].text)) {
+                first = i;
+            } else if (!/^(?:如有侵权|若有侵权|侵权请|转载请|版权归)/.test(lines[i].text)) {
+                break;
+            }
+        }
+        if (first < 0) return null;
+        const start = lines[first].parts[0];
+        const range = body.ownerDocument.createRange();
+        range.setStart(start.node, start.start);
+        range.setEnd(body, body.childNodes.length);
+        return {
+            text: lines.slice(first).map(line => line.text).join('\n'),
+            remove: () => range.deleteContents()
+        };
+    }
+
+window.DWW11Formatter = Object.freeze({
+  format(expectedSource) {
+    if (Number(window.wysiwyg) !== 1) throw new Error('请切换到可视化编辑器后排版');
+    const body = document.querySelector('#e_iframe')?.contentDocument?.body;
+    if (!body) throw new Error('编辑器未就绪');
+    const footer = existingSource(body);
+    const expected = `来源：${expectedSource}，如有侵权，请联系删除。`;
+    if (!footer || footer.text !== expected) throw new Error('文末来源与文章包不一致，请人工检查');
+    const images = [...body.querySelectorAll('img')].map(x => x.id + ':' + x.getAttribute('src'));
+    const backup = body.innerHTML;
+    try {
+      footer.remove();
+      formatBody(body, expectedSource, expected);
+      const after = [...body.querySelectorAll('img')].map(x => x.id + ':' + x.getAttribute('src'));
+      if (JSON.stringify(images) !== JSON.stringify(after)) throw new Error('排版后图片标识或顺序改变');
+      if (body.querySelectorAll('[data-dww11-wx-source="1"]').length !== 1) throw new Error('来源尾注数量异常');
+      body.dispatchEvent(new Event('input', {bubbles:true}));
+      return {images: after, source: expected};
+    } catch (error) { body.innerHTML = backup; throw error; }
+  }
+});
+})();

browser/importer.js

@@ -0,0 +1,480 @@
+// ==UserScript==
+// @name         广谈大湾网 · 本地文章目录导入
+// @namespace    local.dww11.article-importer
+// @version      1.0.0
+// @description  选择整个文章目录，导入 Markdown 标题、正文与本地配图；保留人工发布与程序接口。
+// @match        https://www.dww11.com/forum.php*
+// @match        https://dww11.com/forum.php*
+// @run-at       document-idle
+// @grant        none
+// @noframes
+// @license      MIT
+// ==/UserScript==
+
+(function (root, factory) {
+    'use strict';
+    const core = factory();
+    if (typeof module === 'object' && module.exports) module.exports = core;
+    else core.mount(root);
+})(typeof window === 'undefined' ? null : window, function () {
+    'use strict';
+    const VERSION = '1.0.0';
+    const MAX_IMAGE_SIZE = 20 * 1024 * 1024;
+    const IMAGE_EXT = /\.(?:jpe?g|png|gif)$/i;
+
+    function normalizePath(path) {
+        const parts = [];
+        const value = String(path).replace(/\\/g, '/').normalize('NFC');
+        if (/^(?:\/|[a-z]:|[a-z]+:)/i.test(value)) throw new Error('请使用目录内的相对路径：' + path);
+        for (const part of value.split('/')) {
+            if (!part || part === '.') continue;
+            if (part === '..') {
+                if (!parts.length) throw new Error('图片路径超出所选目录：' + path);
+                parts.pop();
+            } else parts.push(part);
+        }
+        return parts.join('/');
+    }
+
+    function imagePath(mdPath, target) {
+        if (/^(?:[a-z][\w+.-]*:|\/\/|\/)/i.test(target)) throw new Error('图片必须在所选目录内，请先下载：' + target);
+        let decoded;
+        try { decoded = decodeURIComponent(target); } catch { throw new Error('图片路径编码无效：' + target); }
+        return normalizePath(mdPath.slice(0, mdPath.lastIndexOf('/') + 1) + decoded);
+    }
+
+    // A small, intentionally bounded Markdown reader. Raw HTML is never executed.
+    // Links/images support balanced parentheses, angle destinations and optional titles.
+    function readLink(text, start) {
+        const image = text.startsWith('![', start);
+        if (!image && text[start] !== '[') return null;
+        const left = start + (image ? 2 : 1);
+        let end = left;
+        for (; end < text.length; end++) {
+            if (text[end] === '\\') { end++; continue; }
+            if (text[end] === ']') break;
+        }
+        if (text.slice(end, end + 2) !== '](') return null;
+        let pos = end + 2, depth = 1, quoted = '', angle = false;
+        for (; pos < text.length; pos++) {
+            const ch = text[pos];
+            if (ch === '\\') { pos++; continue; }
+            if (quoted) { if (ch === quoted) quoted = ''; continue; }
+            if (!angle && depth === 1 && /\s/.test(text[pos - 1]) && (ch === '"' || ch === "'")) { quoted = ch; continue; }
+            if (ch === '<') { angle = true; continue; }
+            if (ch === '>') { angle = false; continue; }
+            if (!angle && ch === '(') depth++;
+            if (!angle && ch === ')' && --depth === 0) break;
+        }
+        if (depth) return null;
+        let target = text.slice(end + 2, pos).trim();
+        const destination = target.match(/^<([^>]+)>\s*(?:["'][\s\S]*["'])?$/);
+        target = destination ? destination[1] : target.replace(/\s+["'][\s\S]*["']\s*$/, '');
+        target = target.replace(/\\([()\\ ])/g, '$1');
+        return { image, label: text.slice(left, end), target, end: pos + 1 };
+    }
+
+    function inline(text, onImage) {
+        let out = '';
+        for (let pos = 0; pos < text.length;) {
+            if (text[pos] === '\\' && pos + 1 < text.length) { out += text[pos + 1]; pos += 2; continue; }
+            const link = readLink(text, pos);
+            if (link) {
+                if (link.image) out += onImage(link.target, link.label);
+                else if (/^(https?:\/\/|mailto:)/i.test(link.target) && !/[\[\]\r\n"']/.test(link.target)) {
+                    out += '[url=' + link.target + ']' + inline(link.label, onImage) + '[/url]';
+                } else out += link.label + ' (' + link.target + ')';
+                pos = link.end;
+                continue;
+            }
+            if (text[pos] === '`') {
+                const end = text.indexOf('`', pos + 1);
+                if (end > pos + 1) { out += text.slice(pos + 1, end); pos = end + 1; continue; }
+            }
+            let matched = false;
+            for (const [mark, tag] of [['**', 'b'], ['__', 'b'], ['~~', 's'], ['*', 'i'], ['_', 'i']]) {
+                if (!text.startsWith(mark, pos) || (mark === '_' && /\w/.test(text[pos - 1] || ''))) continue;
+                const end = text.indexOf(mark, pos + mark.length);
+                if (end <= pos + mark.length) continue;
+                out += '[' + tag + ']' + inline(text.slice(pos + mark.length, end), onImage) + '[/' + tag + ']';
+                pos = end + mark.length; matched = true; break;
+            }
+            if (!matched) out += text[pos++];
+        }
+        return out;
+    }
+
+    function markdownToBBCode(markdown, onImage) {
+        const lines = markdown.replace(/\r\n?/g, '\n').split('\n');
+        const output = [];
+        let list = '', quote = false, code = false, fence = '';
+        const closeList = () => { if (list) output.push('[/list]'); list = ''; };
+        for (let index = 0; index < lines.length; index++) {
+            const line = lines[index];
+            const codeMark = line.match(/^\s*(`{3,}|~{3,})/);
+            if (codeMark && (!code || codeMark[1][0] === fence)) {
+                closeList();
+                output.push(code ? '[/code]' : '[code]');
+                code = !code; fence = codeMark[1][0]; continue;
+            }
+            if (code) { output.push(line.replace(/\[\/code\]/gi, '［/code］')); continue; }
+            const quoted = /^\s*>\s?/.test(line);
+            if (quoted !== quote) { closeList(); output.push(quoted ? '[quote]' : '[/quote]'); quote = quoted; }
+            const content = quoted ? line.replace(/^\s*>\s?/, '') : line;
+            const item = content.match(/^\s*(?:([-+*])|\d+[.)])\s+(.+)$/);
+            if (item) {
+                const type = item[1] ? '[list]' : '[list=1]';
+                if (list !== type) { closeList(); output.push(type); list = type; }
+                output.push('[*]' + inline(item[2], onImage)); continue;
+            }
+            closeList();
+            const heading = content.match(/^\s{0,3}(#{1,6})\s+(.+?)\s*#*$/);
+            if (heading) output.push('[b]' + inline(heading[2], onImage) + '[/b]');
+            else if (/^\s*(?:---+|\*\*\*+|___+)\s*$/.test(content)) output.push('[hr]');
+            else output.push(inline(content, onImage));
+        }
+        closeList();
+        if (quote) output.push('[/quote]');
+        if (code) output.push('[/code]');
+        return output.join('\n').trim();
+    }
+
+    function compile(markdown, mdPath, fileMap, titleOverride) {
+        let body = String(markdown).replace(/^\uFEFF/, '').replace(/\r\n?/g, '\n');
+        const firstHeading = body.match(/^\s*#\s+([^\n]+)\n?/);
+        const title = String(titleOverride || firstHeading?.[1] || mdPath.split('/').pop().replace(/\.md$/i, '')).trim();
+        if (firstHeading) body = body.slice(firstHeading[0].length).trim();
+        if (!title || Array.from(title).length > 80) throw new Error('标题为空或超过 80 字，请修改 Markdown 标题后重试。');
+        if (!body.trim()) throw new Error('正文为空。');
+        const warnings = [];
+        if (/<\/?[a-z][^>]*>/i.test(body)) warnings.push('存在 HTML，将作为文字保留；请检查排版。');
+        if (/^\s*\|.*\|\s*$/m.test(body)) warnings.push('表格暂按原文保留，需手动检查。');
+        if (/!\[[^\]]*\]\s*\[[^\]]*\]/.test(body) || /^\s*\[[^\]]+\]:/m.test(body)) {
+            throw new Error('暂不支持引用式 Markdown 图片/链接，请改成 ![](图片路径) 或 [文字](网址)。');
+        }
+        const images = new Map();
+        // Validate every image before any network request; repeated references upload once.
+        markdownToBBCode(body, (target) => {
+            const path = imagePath(mdPath, target);
+            const file = fileMap.get(path);
+            if (!file) throw new Error('找不到配图：' + path);
+            if (!IMAGE_EXT.test(path)) throw new Error('网站仅支持 JPG、JPEG、PNG、GIF：' + path);
+            if (!file.size || file.size >= MAX_IMAGE_SIZE) throw new Error('配图为空或达到 20 MB：' + path);
+            images.set(path, file);
+            return '[图片：' + path + ']';
+        });
+        return { title, body, mdPath, images, warnings,
+            render(aids) {
+                return markdownToBBCode(body, target => {
+                    const path = imagePath(mdPath, target);
+                    const aid = aids.get(path);
+                    if (!/^[1-9]\d*$/.test(String(aid))) throw new Error('配图尚未上传：' + path);
+                    return '[attachimg]' + aid + '[/attachimg]';
+                });
+            }
+        };
+    }
+
+    async function prepareFiles(files, markdownPath) {
+        const fileMap = new Map();
+        for (const file of Array.from(files)) {
+            const path = normalizePath(file.webkitRelativePath || file.name);
+            if (fileMap.has(path)) throw new Error('文件路径重复：' + path);
+            fileMap.set(path, file);
+        }
+        const candidates = Array.from(fileMap.keys()).filter(p => /\.md$/i.test(p)).sort();
+        if (!candidates.length) throw new Error('所选目录中没有 .md 文件。');
+        if (!markdownPath && candidates.length > 1) return { candidates, files: Array.from(files) };
+        const mdPath = markdownPath || candidates[0];
+        const md = fileMap.get(mdPath);
+        if (!md) throw new Error('指定的 Markdown 文件不存在。');
+        return compile(await md.text(), mdPath, fileMap);
+    }
+
+    async function preparePackage(data) {
+        if (!data || data.schemaVersion !== 1 || typeof data.markdown !== 'string' || !Array.isArray(data.images)) {
+            throw new Error('文章包格式错误，需要 schemaVersion: 1、markdown 和 images。');
+        }
+        const map = new Map();
+        for (const item of data.images) {
+            const path = normalizePath(item.path);
+            if (map.has(path)) throw new Error('文章包图片路径重复：' + path);
+            let blob = item.blob;
+            if (!blob && typeof item.base64 === 'string') {
+                const binary = atob(item.base64);
+                blob = new Blob([Uint8Array.from(binary, ch => ch.charCodeAt(0))], { type: item.mime || 'application/octet-stream' });
+            }
+            if (!blob || typeof blob.arrayBuffer !== 'function') throw new Error('文章包缺少图片数据：' + path);
+            map.set(path, blob);
+        }
+        return compile(data.markdown, normalizePath(data.markdownPath || 'article.md'), map, data.title);
+    }
+
+    function parseAid(text) {
+        const value = String(text).trim();
+        // The site's actual WebUploader adapter uses a positive integer response.
+        if (!/^(?:[1-9]\d*|"[1-9]\d*")$/.test(value)) throw new Error('服务器未返回有效附件编号（可能登录过期、超出权限或上传失败）。');
+        return value.replace(/"/g, '');
+    }
+
+    function mount(win) {
+        const doc = win.document;
+        const params = new URLSearchParams(win.location.search);
+        if (params.get('mod') !== 'post' || params.get('action') !== 'newthread' || !doc.querySelector('#subject') || !doc.querySelector('#e_textarea')) return;
+        if (win.DWW11Importer) return;
+        let article = null, busy = false, candidates = null, lastResult = null, savedBackup = null, stageRevision = 0;
+        const cache = new Map(), urls = [], logs = [];
+        const host = doc.createElement('div');
+        host.id = 'dww11-importer';
+        host.style.cssText = 'position:fixed;right:18px;bottom:18px;z-index:2147483000';
+        const shadow = host.attachShadow({ mode: 'open' });
+        shadow.innerHTML = `<style>
+            :host{all:initial}*{box-sizing:border-box}details{font:14px/1.6 system-ui,sans-serif;color:#203044;background:#fff;border:1px solid #d4dee8;border-radius:12px;box-shadow:0 6px 28px #0003;width:360px;max-width:92vw}summary{cursor:pointer;padding:12px 16px;font-weight:650}section{padding:0 16px 16px;max-height:70vh;overflow:auto}button,select{font:inherit;padding:7px 10px;border-radius:6px;border:1px solid #b8c9da;background:#f4f8fc;color:#203044;cursor:pointer;margin:4px 4px 4px 0}button.primary{background:#1767b3;color:white;border-color:#1767b3}button:disabled{opacity:.5;cursor:default}p{margin:8px 0}small{color:#617183}pre{font:12px/1.6 system-ui;white-space:pre-wrap;overflow-wrap:anywhere;max-height:170px;overflow:auto;background:#f4f6f9;padding:8px}img{max-width:90px;max-height:80px;margin:4px;object-fit:contain}.error{color:#b32424}.success{color:#18713e}select{max-width:100%}[hidden]{display:none!important}
+            </style><details open><summary>本地文章导入 · 1.0</summary><section>
+            <small>选择整个文章文件夹，检查后上传配图并填写。发布由你操作。网站自带的自动保存仍可能运行。</small>
+            <p><button id="choose" type="button">选择文章目录</button><button id="package" type="button">导入文章包</button></p>
+            <input id="folder" type="file" webkitdirectory multiple hidden><input id="json" type="file" accept=".json,application/json" hidden>
+            <select id="md" hidden aria-label="选择 Markdown"></select>
+            <p id="status" role="status" aria-live="polite">尚未选择目录</p><div id="preview" hidden><strong id="title"></strong><pre id="body"></pre><div id="images"></div><p id="warnings"></p></div>
+            <button id="import" class="primary" type="button" disabled>上传配图并填写</button>
+            <button id="diagnostics" type="button">导出诊断</button><pre id="log" hidden></pre>
+            </section></details>`;
+        doc.body.appendChild(host);
+        const $ = id => shadow.getElementById(id);
+        const summary = () => ({ version: VERSION, busy, ready: !!article, title: article?.title || null,
+            markdownPath: article?.mdPath || null, imageCount: article?.images.size || 0,
+            warnings: article?.warnings || [], result: lastResult });
+        function notify(phase, message, extra = {}) {
+            $('status').textContent = message;
+            $('status').className = phase === 'error' ? 'error' : phase === 'complete' ? 'success' : '';
+            logs.push({ time: new Date().toISOString(), phase, message, ...extra });
+            $('log').hidden = false;
+            $('log').textContent = logs.slice(-8).map(l => l.message).join('\n');
+            win.dispatchEvent(new CustomEvent('dww11-importer:progress', { detail: { phase, message, ...extra } }));
+        }
+        function releasePreviews() { while (urls.length) URL.revokeObjectURL(urls.pop()); }
+        function clearStage() {
+            article = null; lastResult = null; candidates = null; releasePreviews();
+            $('preview').hidden = true; $('md').hidden = true; $('import').disabled = true;
+        }
+        function showArticle(prepared) {
+            article = prepared; lastResult = null; releasePreviews();
+            $('preview').hidden = false;
+            $('title').textContent = article.title;
+            $('body').textContent = article.body;
+            $('warnings').textContent = article.warnings.join('；');
+            $('images').replaceChildren();
+            for (const [path, blob] of article.images) {
+                const img = doc.createElement('img');
+                img.src = URL.createObjectURL(blob); urls.push(img.src); img.alt = path; img.title = path;
+                $('images').appendChild(img);
+            }
+            $('import').disabled = false;
+            notify('ready', `已检查：${article.images.size} 张图片。点击“上传配图并填写”开始导入。`);
+            return summary();
+        }
+        async function stageFiles(files, options = {}) {
+            if (busy) throw new Error('正在导入，请稍候。');
+            clearStage();
+            const revision = ++stageRevision;
+            const list = Array.from(files);
+            const prepared = await prepareFiles(list, options.markdownPath);
+            if (revision !== stageRevision) return { cancelled: true };
+            if (prepared.candidates) {
+                candidates = prepared.files;
+                $('md').replaceChildren(new Option('请选择目录中的 Markdown 文件', ''));
+                for (const path of prepared.candidates) $('md').appendChild(new Option(path, path));
+                $('md').hidden = false;
+                notify('choose-markdown', '目录中有多个 Markdown 文件，请选择一篇。');
+                return { candidates: prepared.candidates };
+            }
+            return showArticle(prepared);
+        }
+        async function stagePackage(data) {
+            if (busy) throw new Error('正在导入，请稍候。');
+            clearStage();
+            const revision = ++stageRevision;
+            const prepared = await preparePackage(data);
+            if (revision !== stageRevision) return { cancelled: true };
+            return showArticle(prepared);
+        }
+        const snapshot = () => ({ title: doc.querySelector('#subject').value,
+            mode: Number(win.wysiwyg), text: win.getEditorContents() });
+        function assertEditor() {
+            for (const name of ['getEditorContents', 'switchEditor']) {
+                if (typeof win[name] !== 'function') throw new Error('未识别到网站编辑器：' + name + '。请刷新页面后重试。');
+            }
+            if (!win.imgUpload?.settings || typeof win.updateImageList !== 'function') throw new Error('网站图片上传组件尚未就绪，请刷新后重试。');
+            if (Number(win.allowbbcode) === 0 || Number(win.allowimgcode) === 0) throw new Error('当前版块或账号不支持正文格式/图片。');
+        }
+        async function uploadImage(blob, path, settings, signal) {
+            signal?.throwIfAborted();
+            const url = new URL(settings.upload_url, win.location.href);
+            if (url.origin !== win.location.origin || url.pathname !== '/misc.php' || url.searchParams.get('operation') !== 'upload') {
+                throw new Error('上传地址与预期不符，已停止。');
+            }
+            const hash = Array.from(new Uint8Array(await win.crypto.subtle.digest('SHA-256', await blob.arrayBuffer())), n => n.toString(16).padStart(2, '0')).join('');
+            const key = url.href + ':' + settings.post_params.uid + ':' + hash;
+            if (cache.has(key)) { notify('upload', '复用已上传图片：' + path); return cache.get(key); }
+            const data = new FormData();
+            for (const [key, value] of Object.entries(settings.post_params)) data.append(key, String(value));
+            data.set('type', 'image');
+            data.set('filetype', blob.type || (/\.png$/i.test(path) ? 'image/png' : /\.gif$/i.test(path) ? 'image/gif' : 'image/jpeg'));
+            data.append(settings.file_post_name || 'Filedata', blob, path.split('/').pop());
+            const control = new AbortController(), timer = setTimeout(() => control.abort(), 90000);
+            try {
+                signal?.throwIfAborted();
+                signal?.addEventListener('abort', () => control.abort(), { once: true });
+                const response = await win.fetch(url.href, { method: 'POST', body: data, credentials: 'same-origin', redirect: 'error', signal: control.signal });
+                if (!response.ok) throw new Error('图片上传请求失败：HTTP ' + response.status);
+                const aid = parseAid(await response.text());
+                cache.set(key, aid);
+                return aid;
+            } catch (error) {
+                if (error.name === 'AbortError') throw new Error('上传超时；该图片可能已到达服务器，请检查网站图片列表后再重试。');
+                throw error;
+            } finally { clearTimeout(timer); }
+        }
+        async function registerImages(aids, signal) {
+            signal?.throwIfAborted();
+            if (!aids.length) return;
+            const ids = new Set(aids);
+            doc.querySelectorAll('#imgattachlist [id^="image_td_"],#imgattachlist [id^="image_"]').forEach(el => {
+                const id = el.id.match(/^(?:image_td_|image_)(\d+)$/)?.[1]; if (id) ids.add(id);
+            });
+            const ready = () => aids.every(id => doc.getElementById('image_' + id));
+            if (ready()) return;
+            // Use the site's own list loader, which registers attachment form fields too.
+            win.updateImageList(1, Array.from(ids).join('|'));
+            await new Promise((resolve, reject) => {
+                const timer = setTimeout(() => { observer.disconnect(); reject(new Error('图片已上传，但网站图片列表加载超时。重试会复用本页已上传图片。')); }, 25000);
+                signal?.addEventListener('abort', () => { clearTimeout(timer); observer.disconnect(); reject(new Error('流程已终止')); }, { once: true });
+                const check = () => { if (ready()) { clearTimeout(timer); observer.disconnect(); resolve(); } };
+                const observer = new MutationObserver(check);
+                observer.observe(doc.body, { childList: true, subtree: true }); check();
+            });
+        }
+        function setEditor(title, bbcode) {
+            win.switchEditor(0);
+            if (Number(win.wysiwyg) !== 0) throw new Error('无法切换编辑器为纯文本模式。');
+            const textarea = doc.querySelector('#e_textarea');
+            textarea.value = bbcode;
+            doc.querySelector('#subject').value = title;
+            for (const el of [textarea, doc.querySelector('#subject')]) {
+                el.dispatchEvent(new Event('input', { bubbles: true })); el.dispatchEvent(new Event('change', { bubbles: true }));
+            }
+            // Let Discuz generate the WYSIWYG content with native attachment metadata.
+            win.switchEditor(1);
+            const switcher = doc.querySelector('#e_switchercheck');
+            if (switcher) switcher.checked = Number(win.wysiwyg) === 0;
+        }
+        async function importPrepared(options = {}) {
+            const { signal, automatic = false } = options;
+            signal?.throwIfAborted();
+            if (busy) throw new Error('正在导入，请勿重复点击。');
+            if (!article) throw new Error('请先选择文章目录或文章包。');
+            assertEditor();
+            for (const id of ['bbcodeoff', 'htmlon']) {
+                if (doc.getElementById(id)?.checked) throw new Error('请先关闭“禁用 Discuz! 代码”和“HTML 代码”，再导入。');
+            }
+            const before = snapshot();
+            if ((before.title.trim() || before.text.replace(/<[^>]*>/g, '').trim() || /<img\b/i.test(before.text)) &&
+                (automatic ? (() => { throw new Error('编辑器已有内容，请人工处理旧稿后重试'); })() : !win.confirm('当前编辑器已有内容。导入将替换标题和正文，是否继续？'))) return { cancelled: true };
+            const selected = article;
+            busy = true;
+            const controls = ['choose', 'package', 'import', 'md'].map($);
+            controls.forEach(el => { el.disabled = true; });
+            // Prevent accidental submission of a partly imported article.
+            const blockSubmit = event => { event.preventDefault(); event.stopImmediatePropagation(); notify('busy', '正在导入，请等待完成后再操作发布。'); };
+            const form = doc.querySelector('#postform');
+            form.addEventListener('submit', blockSubmit, true);
+            const submitButtons = Array.from(form.querySelectorAll('button[type="submit"],button[name],input[type="submit"]'));
+            const previousDisabled = submitButtons.map(el => el.disabled);
+            submitButtons.forEach(el => { el.disabled = true; });
+            let committed = false;
+            try {
+                const aids = new Map();
+                const settings = win.imgUpload.settings;
+                const max = Math.min(MAX_IMAGE_SIZE, Number(settings.file_size_limit) * 1024 || MAX_IMAGE_SIZE);
+                for (const [path, blob] of selected.images) if (blob.size >= max) throw new Error('配图超过当前网站限制：' + path);
+                let index = 0;
+                for (const [path, blob] of selected.images) {
+                    notify('upload', `上传 ${++index}/${selected.images.size}：${path}`, { completed: index - 1, total: selected.images.size });
+                    aids.set(path, await uploadImage(blob, path, settings, signal));
+                }
+                notify('register', '正在加载网站图片列表…');
+                await registerImages(Array.from(new Set(aids.values())), signal);
+                signal?.throwIfAborted();
+                const now = snapshot();
+                if ((now.title !== before.title || now.text !== before.text) && (automatic ? (() => { throw new Error('上传期间编辑器被修改，已保留当前内容'); })() : !win.confirm('上传期间你修改了编辑器。是否用导入文章替换最新内容？'))) {
+                    notify('cancelled', '已保留当前正文；图片已上传，可稍后再次导入。'); return { cancelled: true };
+                }
+                const backup = snapshot();
+                const bbcode = selected.render(aids);
+                // Preserve a local, downloadable snapshot before changing title or body.
+                savedBackup = { title: backup.title, content: backup.text, mode: backup.mode, createdAt: new Date().toISOString() };
+                try {
+                    for (const id of ['bbcodeoff', 'htmlon']) {
+                        const el = doc.getElementById(id);
+                        if (el && id === 'bbcodeoff' && el.checked) throw new Error('请先取消“禁用 Discuz! 代码”，再重试。');
+                        if (el && id === 'htmlon' && el.checked) throw new Error('请先关闭“HTML 代码”，再重试。');
+                    }
+                    signal?.throwIfAborted();
+                    setEditor(selected.title, bbcode);
+                    const raw = win.getEditorContents();
+                    if (Number(win.wysiwyg) === 1) {
+                        for (const id of new Set(aids.values())) if (!raw.includes('attachimg_' + id)) throw new Error('富文本图片转换未通过检查。');
+                    } else if (raw !== bbcode) throw new Error('正文填写校验失败。');
+                    if (doc.querySelector('#subject').value !== selected.title) throw new Error('标题填写校验失败。');
+                    committed = true;
+                } catch (error) {
+                    // Restore the exact editor representation saved immediately before filling.
+                    win.switchEditor(backup.mode);
+                    if (backup.mode) win.editdoc.body.innerHTML = backup.text;
+                    else doc.querySelector('#e_textarea').value = backup.text;
+                    doc.querySelector('#subject').value = backup.title;
+                    const switcher = doc.querySelector('#e_switchercheck'); if (switcher) switcher.checked = !backup.mode;
+                    throw error;
+                }
+                lastResult = { title: selected.title, images: Object.fromEntries(aids), completedAt: new Date().toISOString() };
+                notify('complete', `已填写标题、正文和 ${aids.size} 张配图。请检查排版，随后手动发表。`);
+                return lastResult;
+            } catch (error) {
+                notify('error', error.message + (committed ? '' : ' 标题和正文未替换；已上传的图片会保留。'));
+                throw error;
+            } finally {
+                busy = false; form.removeEventListener('submit', blockSubmit, true);
+                submitButtons.forEach((el, index) => { el.disabled = previousDisabled[index]; });
+                controls.forEach(el => { el.disabled = false; });
+                $('import').disabled = !article;
+            }
+        }
+        function downloadJSON(value, filename) {
+            const url = URL.createObjectURL(new Blob([JSON.stringify(value, null, 2)], { type: 'application/json' }));
+            const link = doc.createElement('a'); link.href = url; link.download = filename; link.click();
+            setTimeout(() => URL.revokeObjectURL(url), 1000);
+        }
+        // Public v1 interface. Staging is read-only; importPrepared is the explicit upload/fill step.
+        // No postMessage listener, localhost service, publishing endpoint or saved credentials.
+        const api = Object.freeze({ version: VERSION, schemaVersion: 1, stageFiles, stagePackage,
+            getState: summary, openPicker: () => { if (!busy) $('folder').click(); }, importPrepared,
+            getBackup: () => savedBackup });
+        win.DWW11Importer = api;
+        $('choose').onclick = () => { $('folder').value = ''; $('folder').click(); };
+        $('package').onclick = () => { $('json').value = ''; $('json').click(); };
+        const handleError = error => notify('error', error.message || String(error));
+        $('folder').onchange = () => stageFiles($('folder').files).catch(handleError);
+        $('json').onchange = async () => {
+            try { const file = $('json').files[0]; if (file) await stagePackage(JSON.parse(await file.text())); }
+            catch (error) { handleError(error); }
+        };
+        $('md').onchange = () => {
+            const files = candidates, path = $('md').value;
+            if (files && path) stageFiles(files, { markdownPath: path }).catch(handleError);
+        };
+        $('import').onclick = () => importPrepared().catch(handleError);
+        $('diagnostics').onclick = () => downloadJSON({ ...summary(), logs }, 'dww11-import-diagnostics.json');
+        win.addEventListener('pagehide', releasePreviews, { once: true });
+    }
+    return { VERSION, normalizePath, imagePath, readLink, markdownToBBCode, compile, prepareFiles, preparePackage, parseAid, mount };
+});

browser/local-bridge.user.js

@@ -0,0 +1,73 @@
+// ==UserScript==
+// @name         新闻工作台 · 本地连接
+// @namespace    local.dww11.auto-news
+// @version      0.1.0
+// @match        https://www.dww11.com/*
+// @match        https://dww11.com/*
+// @match        https://house.dww11.com/*
+// @grant        GM_xmlhttpRequest
+// @grant        GM_getValue
+// @grant        GM_setValue
+// @grant        GM_addValueChangeListener
+// @connect      127.0.0.1
+// @run-at       document-start
+// @noframes
+// ==/UserScript==
+(() => {
+  'use strict';
+  const allowed = new Set(['heartbeat','next','package','event','accept','switch-begin','takeover','stop','current']);
+  let token = GM_getValue('pairToken',''), port = GM_getValue('pairPort',8770);
+  let stopRevision = 0;
+  const signalStop = () => { stopRevision++; document.dispatchEvent(new CustomEvent('dww11:auto-stop')); };
+  GM_addValueChangeListener('halted', (_key, _old, value) => { if (value) signalStop(); });
+  function request(route, body) {
+    return new Promise((resolve, reject) => {
+      if (!token) return reject(new Error('请在左下角填入本地工作台的连接码'));
+      GM_xmlhttpRequest({method:'POST',url:`http://127.0.0.1:${port}/bridge/${route}`,
+        headers:{'Content-Type':'application/json','X-Session-Token':token},data:JSON.stringify(body),timeout:60000,
+        onload(response) { try { const value=JSON.parse(response.responseText); if(response.status!==200) throw new Error(value.error || '本地连接失败'); resolve(value); } catch(error) { reject(error); } },
+        onerror:()=>reject(new Error('本地服务未连接，请检查服务与脚本权限')),
+        ontimeout:()=>reject(new Error('本地服务超时，任务已暂停'))});
+    });
+  }
+  document.addEventListener('dww11:local-request', async event => {
+    let message;
+    try {message=JSON.parse(event.detail);} catch {return;}
+    if (!allowed.has(message.route) || typeof message.id !== 'string') return;
+    try {
+      if (message.route==='stop') { GM_setValue('halted',true); signalStop(); }
+      const revision = stopRevision;
+      let value=await request(message.route,message.body || {});
+      if (message.route==='heartbeat') {
+        if (revision !== stopRevision) value = {...value, running:false, reason:'终止后忽略旧连接响应'};
+        GM_setValue('halted',!value.running);
+        if (!value.running) signalStop();
+      }
+      document.dispatchEvent(new CustomEvent('dww11:local-response',{detail:JSON.stringify({id:message.id,value})}));
+    } catch(error) {
+      if (message.route==='heartbeat') {GM_setValue('halted',true); signalStop();}
+      document.dispatchEvent(new CustomEvent('dww11:local-response',{detail:JSON.stringify({id:message.id,error:error.message})}));
+    }
+  });
+  function mount() {
+    if (!document.body || document.getElementById('auto-news-connection')) return;
+    const host=document.createElement('div');host.id='auto-news-connection';
+    host.style.cssText='position:fixed;left:16px;bottom:16px;z-index:2147483647';
+    const shadow=host.attachShadow({mode:'closed'});
+    shadow.innerHTML='<style>section{font:13px/1.6 system-ui;background:white;color:#24364a;border:1px solid #ccd7e0;border-radius:10px;padding:12px;width:250px;box-shadow:0 3px 20px #0002}button,input{font:inherit;box-sizing:border-box;padding:7px;margin:4px 0;width:100%}button{cursor:pointer}#halt{background:#ad302d;color:white;border:0;border-radius:6px}small{display:block}</style><section><strong>新闻工作台 · 本地连接</strong><button id="halt">终止流程</button><details><summary>连接设置</summary><label>本地端口<input id="port" type="number" min="1" max="65535"></label><label>连接码<input id="token" type="password" autocomplete="off"></label><button id="save">保存连接设置</button></details><small id="status">请启动本地工作台并配对</small></section>';
+    const $=id=>shadow.getElementById(id);$('port').value=port;
+    $('save').onclick=()=>{
+      const value=Number($('port').value);if(!Number.isInteger(value)||value<1||value>65535){$('status').textContent='端口无效';return;}
+      if($('token').value.trim()) token=$('token').value.trim();port=value;
+      GM_setValue('pairToken',token);GM_setValue('pairPort',port);$('token').value='';$('status').textContent='已保存；服务重启后需要更新连接码';
+    };
+    $('halt').onclick=async()=>{
+      GM_setValue('halted',true);signalStop();$('status').textContent='已终止；在途请求仍可能完成';
+      try {await request('stop',{});} catch(error) {$('status').textContent+='。'+error.message;}
+    };
+    document.addEventListener('dww11:connection-status',event=>{$('status').textContent=String(event.detail).slice(0,250);});
+    document.body.append(host);
+    if(GM_getValue('halted',true)) signalStop();
+  }
+  if(document.readyState==='loading') document.addEventListener('DOMContentLoaded',mount,{once:true}); else mount();
+})();

browser/runner.js

@@ -0,0 +1,173 @@
+/* Page-side orchestrator. This file never clicks or submits the publication form. */
+(() => {
+  'use strict';
+  if (window.DWW11NewsRunner) return;
+  window.DWW11NewsRunner = true;
+  const SLOT='auto-news-current';
+  const owner=sessionStorage.getItem('auto-news-owner') || crypto.randomUUID();
+  sessionStorage.setItem('auto-news-owner',owner);
+  let run=null, working=false, halted=true, aborter=null, active=null, tickBusy=false;
+  const pending=new Map();
+  function rpc(route,body={}) {
+    return new Promise((resolve,reject)=>{
+      const id=crypto.randomUUID();
+      const timer=setTimeout(()=>{pending.delete(id);reject(new Error('本地连接无响应'));},65000);
+      pending.set(id,{resolve,reject,timer});
+      document.dispatchEvent(new CustomEvent('dww11:local-request',{detail:JSON.stringify({id,route,body:{owner,...body}})}));
+    });
+  }
+  document.addEventListener('dww11:local-response',event=>{
+    let data;try{data=JSON.parse(event.detail);}catch{return;}
+    const waiter=pending.get(data.id);if(!waiter)return;
+    clearTimeout(waiter.timer);pending.delete(data.id);
+    data.error?waiter.reject(new Error(data.error)):waiter.resolve(data.value);
+  });
+  const status=message=>document.dispatchEvent(new CustomEvent('dww11:connection-status',{detail:message}));
+  document.addEventListener('dww11:auto-stop',()=>{halted=true;aborter?.abort(new Error('流程已终止'));});
+  function account() {
+    const element=document.querySelector('#mumucms_username');
+    if(!element)return null;
+    const name=[...element.childNodes].filter(n=>n.nodeType===Node.TEXT_NODE).map(n=>n.textContent).join('').trim();
+    const links=[element,...element.querySelectorAll('a')].filter(e=>e.href);
+    let uid=null;
+    for(const link of links) {
+      const href=link.href;
+      uid=new URL(href,location.href).searchParams.get('uid') || href.match(/space-uid-(\d+)/)?.[1];
+      if(uid)break;
+    }
+    return uid&&name?{uid,name}:null;
+  }
+  function extension() {
+    try{return JSON.parse(document.querySelector('#dww11-extension-panel')?.getAttribute('data-auto-news-state') || '{}');}catch{return {};}
+  }
+  function assertLive(epoch) {
+    if(halted||!run?.running||run.epoch!==epoch)throw new Error('流程已暂停或终止');
+    if(active?.account?.uid!==account()?.uid)throw new Error('当前账号与任务不一致');
+  }
+  function editorSnapshot() {
+    const body=document.querySelector('#e_iframe')?.contentDocument?.body;
+    if(!body)throw new Error('找不到可视化编辑器');
+    return {title:document.querySelector('#subject')?.value || '',
+      images:[...body.querySelectorAll('img')].map(x=>({id:x.id,src:x.getAttribute('src')})),
+      // Snapshot only final submitted content; never includes hidden tokens or the full form.
+      html:body.innerHTML.slice(0,500000), text:body.innerText.slice(0,100000)};
+  }
+  function saveLocal(value) {sessionStorage.setItem(SLOT,JSON.stringify(value));}
+  function loadLocal(){try{return JSON.parse(sessionStorage.getItem(SLOT)||'null');}catch{return null;}}
+  async function event(state,extra={}) {return rpc('event',{id:active.id,state,epoch:run.epoch,...extra});}
+  async function fail(error) {
+    status(error.message);
+    if(active)try{await event('needs_attention',{error:error.message});}catch{}
+    halted=true;aborter?.abort();document.dispatchEvent(new CustomEvent('dww11:auto-stop'));
+  }
+  function editorPage(job) {
+    const u=new URL(location.href);
+    return u.searchParams.get('mod')==='post'&&u.searchParams.get('action')==='newthread'&&u.searchParams.get('fid')===String(job.fid);
+  }
+  async function prepare(job) {
+    working=true;active=job;aborter=new AbortController();const epoch=run.epoch;
+    try {
+      assertLive(epoch);
+      if(!editorPage(job)) {
+        saveLocal({id:job.id});
+        location.assign(`https://www.dww11.com/forum.php?mod=post&action=newthread&fid=${job.fid}`);return;
+      }
+      if(!window.DWW11Importer||!window.DWW11Formatter)throw new Error('导入或排版模块尚未就绪，请检查安装');
+      status('正在检查文章包…');
+      const pack=await rpc('package',{id:job.id});assertLive(epoch);
+      const prepared=await window.DWW11Importer.stagePackage(pack);assertLive(epoch);
+      if(prepared.warnings?.length)throw new Error('文章需要人工处理：'+prepared.warnings.join('；'));
+      await event('importing');assertLive(epoch);status('上传图片并填写文章…');
+      const result=await window.DWW11Importer.importPrepared({automatic:true,signal:aborter.signal});assertLive(epoch);
+      if(result?.cancelled)throw new Error('导入已取消');
+      await event('formatting');assertLive(epoch);
+      window.DWW11Formatter.format(job.source);assertLive(epoch);
+      const snapshot=editorSnapshot();
+      if(snapshot.title!==job.title)throw new Error('标题校验失败');
+      const aids=new Set(Object.values(result.images));
+      for(const aid of aids) if(!snapshot.images.some(x=>x.id==='attachimg_'+aid))throw new Error('配图附件缺失：'+aid);
+      await event('awaiting_user_review',{snapshot});assertLive(epoch);
+      active={...job,state:'awaiting_user_review'};saveLocal({id:job.id});
+      status('已完成排版，请核对内容并亲手点击“发表帖子”。');
+    }catch(error){await fail(error);}finally{working=false;}
+  }
+  // Record only trusted user activation. Never replay a click, call submit(), or send a post request.
+  document.addEventListener('click',event=>{
+    if(!event.isTrusted||!event.target.closest?.('#postsubmit')||active?.state!=='awaiting_user_review')return;
+    try {
+      const snapshot=editorSnapshot();
+      const local={id:active.id,snapshot,clickedAt:Date.now(),observed:false};saveLocal(local);
+      rpc('event',{id:active.id,state:'user_submitting',epoch:run?.epoch,snapshot})
+        .then(()=>{const latest=loadLocal();if(latest?.clickedAt===local.clickedAt)saveLocal({...local,observed:true});})
+        .catch(error=>status('提交记录待恢复：'+error.message));
+      active.state='user_submitting';status('已观察到人工点击，等待结果页核对。');
+    }catch(error){status('提交快照记录失败：'+error.message);}
+  },true);
+  async function observeResult(job) {
+    const local=loadLocal();
+    if(local?.id===job.id&&local.snapshot&&job.state==='awaiting_user_review') {
+      await rpc('event',{id:job.id,state:'user_submitting',epoch:run.epoch,snapshot:local.snapshot});
+      job.state='user_submitting';
+    }
+    if(job.state==='user_submitting') {
+      // Result selectors intentionally await user-provided HTML; URL alone is not proof.
+      if(!editorPage(job) || (local?.clickedAt && Date.now()-local.clickedAt>45000)) {
+        await rpc('event',{id:job.id,state:'submit_unknown',error:'等待结果页 HTML 适配；候选页面：'+location.href});
+        await rpc('stop');
+        status('已记录候选结果页；尚不能自动核对。流程已暂停，请补充结果页 HTML。');
+      }
+    }else if(job.state==='submit_unknown')status('发布结果待核对；不会重发、计数或换号。');
+  }
+  async function switchAccount() {
+    if(working||halted)return;
+    if(location.origin!=='https://www.dww11.com'||location.pathname!=='/') {
+      location.assign('https://www.dww11.com/');return;
+    }
+    const ext=extension(),button=document.querySelector('#dww11-extension-next');
+    if(!ext.ready||!ext.idle||!button||button.disabled) {status('请安装并重新加载带联动接口的换号插件');return;}
+    await rpc('switch-begin');
+    if(halted||!run?.running)return;
+    button.click();status('正在换号；等待插件完成并核对新账号');
+  }
+  async function tick() {
+    if(tickBusy)return;tickBusy=true;
+    try {
+      run=await rpc('heartbeat',{account:account(),url:location.href,extension:extension()});
+      halted=!run.running;
+      if(!run.running) {aborter?.abort();document.dispatchEvent(new CustomEvent('dww11:auto-stop'));}
+      if(working)return;
+      // Passive result recording is permitted after Stop, but no new business actions.
+      if(run.active) {
+        const local=loadLocal();
+        if(local?.id===run.active&&local.snapshot&&!run.running)status('流程已暂停；提交结果仍需核对');
+      }
+      if(!run.running) {
+        if(run.active && run.owner===owner) {const saved=await rpc('current');if(saved.job) {active=saved.job;await observeResult(saved.job);}}
+        status(run.reason || '流程已暂停');return;
+      }
+      const next=await rpc('next',{account:account(),extension:extension()});run=next.run;
+      if(next.job)active=next.job;
+      if(next.action==='prepare'||(next.action==='resume'&&next.job.state==='preparing'))void prepare(next.job);
+      else if(next.action==='resume') {
+        await observeResult(next.job);
+        if(next.job.state==='awaiting_user_review') {
+          if(!editorPage(next.job)||!document.querySelector('#subject')?.value) await fail(new Error('待审稿页面已丢失，请人工核对旧稿，不自动重新导入'));
+          else status('等待你亲手发表；程序不再修改编辑器');
+        }
+        else if(['importing','formatting'].includes(next.job.state))await fail(new Error('导入阶段发生中断，请检查编辑器和已上传图片后处理任务'));
+        else if(next.job.state==='needs_attention')status(next.job.error);
+      }else if(next.action==='switch')await switchAccount();
+      else if(next.action==='switch_wait')status('等待换号结果；若插件已暂停，请人工处理后继续');
+      else if(next.action==='login')status('未能读取当前账号 UID，请登录并确认账号元素');
+      else if(next.action==='other_tab')status('另一个标签持有任务；如需接管请先暂停');
+      else if(next.action==='interval')status('等待设定的处理间隔');
+      else status(run.reason || '等待任务');
+    }catch(error){status(error.message);halted=true;aborter?.abort();document.dispatchEvent(new CustomEvent('dww11:auto-stop'));}
+    finally{tickBusy=false;}
+  }
+  const takeover=document.createElement('button');takeover.type='button';takeover.textContent='接管已暂停任务';
+  takeover.style.cssText='position:fixed;left:16px;top:16px;z-index:2147483646;padding:8px;background:#fff;border:1px solid #aab;border-radius:6px';
+  takeover.onclick=async()=>{try{await rpc('takeover');status('已接管；核对当前页面后在工作台继续');}catch(error){status(error.message);}};
+  document.body.append(takeover);
+  setInterval(tick,2000);tick();
+})();

build_browser.py

@@ -0,0 +1,23 @@
+"""Build the page-context userscript from reviewed local components; no external assets."""
+from pathlib import Path
+ROOT = Path(__file__).resolve().parent
+header = '''// ==UserScript==
+// @name         新闻工作台 · 导入排版与流程
+// @namespace    local.dww11.auto-news
+// @version      0.1.0
+// @match        https://www.dww11.com/*
+// @match        https://dww11.com/*
+// @match        https://house.dww11.com/*
+// @grant        none
+// @run-at       document-idle
+// @noframes
+// ==/UserScript==
+'''
+parts = []
+for name in ('importer.js', 'formatter.js', 'runner.js'):
+    source = (ROOT / 'browser' / name).read_text()
+    if '// ==/UserScript==' in source:
+        source = source.split('// ==/UserScript==', 1)[1]
+    parts.append(source)
+(ROOT / 'browser/auto-news.user.js').write_text(header + '\n'.join(parts))
+print('已生成 browser/auto-news.user.js')

modules/downloader/README.md

@@ -0,0 +1,57 @@
+# 探数新闻下载器
+
+在本机浏览器中填写 API Key、获取频道、选择数量，下载新闻正文和图片。所有项目文件与下载结果都位于本目录，不修改微信下载器。
+
+## 启动
+
+安装 `uv` 后，macOS 可以双击 `start.command`。也可以在本目录运行：
+
+```sh
+uv run --no-project --python 3.12 --with-requirements requirements.txt python app.py
+```
+
+启动后自动打开 http://127.0.0.1:8765 。保留启动窗口，退出时按 Ctrl+C。端口被占用时加 `--port 8766`；不自动打开浏览器时加 `--no-browser`。
+
+1. 填写探数 API Key，点击「获取新闻频道」。频道来自真实接口，不使用写死的列表。
+2. 选择频道、输入目标数量，点击「开始下载」。
+3. 页面显示成功保存的数量、百分比、重复数、失败数和最近 300 条日志。日志固定高度，可滚动；向上查看时不会强制滚回底部。
+4. 下载结果保存在本目录的 `downloads` 文件夹。点击「停止下载」后，等待当前请求或文章保存结束。
+5. 「调试终端」默认折叠，支持频道查询以及列表查询的 `type`、`num`、`start` 参数。测试请求只显示响应，不保存文章；同样消耗 API 调用额度。输出最多保留 100000 字符。
+
+密钥仅在页面和服务运行内存中使用，不保存到配置、历史数据库或浏览器存储。服务仅监听本机地址。刷新页面可查看当前任务进度，开始新任务需重新填写密钥。单个服务同时只运行一个下载任务，请勿同时启动多个实例操作同一下载目录。
+
+## 保存格式
+
+沿用 `wechat_official_acc_article_downloader` 的格式和图片处理规则：
+
+```text
+downloads/
+├── .history.sqlite3
+└── 新闻标题/
+    ├── 新闻标题.md
+    ├── image001.jpg
+    └── image002.png
+```
+
+Markdown 开头是一级标题，空行后直接接由接口 `content` 转换的正文，不额外添加接口 `time` 对应的发布时间或分隔线。末尾为 `来源：{src}，如有侵权，请联系删除。`，来源为空时使用「未知作者」。不在开头添加原文链接或作者。正文中的图片下载到同一目录、使用相对引用、按真实文件类型命名，同一图片地址只下载一次。接口的 `pic` 封面不单独插入正文。
+
+标题中的不合法字符会替换；不同文章碰巧同名时添加 ` (2)` 等后缀，目录和 Markdown 文件保持同名。任何正文图片下载失败，该篇不计入完成数，并清理半成品，再继续其他新闻；下次任务可以重试。
+
+## 去重与分页
+
+- 下载图片前，检查原文 URL / PC URL（移除常见跟踪参数和片段）及规范化正文指纹。任一匹配即跳过，标题相同但正文和网址不同不会单凭标题去重。
+- SQLite 保存成功文章的指纹，跨任务、跨频道和服务重启后继续去重。被删除的文章目录不会阻止重新下载。不要删除 `.history.sqlite3`，否则会失去对应历史记录。
+- 每次任务从最新位置 `start=0` 开始，按返回数量推进，每页最多 40 条；重复和失败不计入成功数量，并继续后面的新闻。
+- 接口限制 `start ≤ 400`，一次任务最多覆盖约 440 条，具体还受返回数量和实时更新影响。达到分页边界或收到空列表时停止，未达目标就显示「未达到目标」，不会伪装成 100%。
+- 列表接口不自动重试，防止隐蔽消耗计费额度。API 业务错误、网络错误或响应格式错误会中断任务，已成功保存的文章保留。图片请求沿用原下载器的重试策略。
+- 新闻列表会实时变化，无法保证服务器在不同请求中不返回同一新闻；本工具在保存前去重，尽量避免重复图片下载与文件保存。
+
+接口依据：[探数官方新闻 API 文档](https://www.tanshuapi.com/market/detail-85)。
+
+## 验证
+
+```sh
+uv run --no-project --python 3.12 --with-requirements requirements.txt python -m unittest discover -s tests -v
+```
+
+测试使用模拟新闻响应与临时目录，不消耗真实 API 额度。覆盖格式与图片引用、图片失败清理、历史去重、分页边界、失败重试、取消、日志上限和本地 HTTP 接口。真实账号权限、余额、返回内容与远程图片可达性，需要在页面填入有效 Key 后验证。

modules/downloader/app.py

@@ -0,0 +1,299 @@
+#!/usr/bin/env python3
+"""仅在本机运行的探数新闻下载器。"""
+from contextlib import contextmanager
+from collections import deque
+from datetime import datetime
+from hashlib import sha256
+from http.server import BaseHTTPRequestHandler, ThreadingHTTPServer
+from pathlib import Path
+from urllib.parse import urlsplit, urlunsplit, parse_qsl, urlencode
+import argparse
+import json
+import secrets
+import sqlite3
+import threading
+import time
+import unicodedata
+import webbrowser
+
+import requests
+from bs4 import BeautifulSoup
+from article_writer import save_article
+
+ROOT = Path(__file__).resolve().parent
+API = 'https://api.tanshuapi.com/api/toutiao/v1/'
+
+
+def api_request(key, endpoint, params=None):
+    if endpoint not in ('channel', 'index'):
+        raise ValueError('未知接口')
+    try:
+        # 不自动重试计费接口，避免一次操作产生不可见的重复调用。
+        response = requests.get(API + endpoint, params={'key': key, **(params or {})}, timeout=(10, 40))
+        response.raise_for_status()
+        payload = response.json()
+    except requests.RequestException:
+        raise ValueError('接口网络请求失败或超时，请在调试终端检查后重试') from None
+    except ValueError:
+        raise ValueError('接口未返回有效 JSON') from None
+    if not isinstance(payload, dict):
+        raise ValueError('接口响应格式错误')
+    return payload
+
+
+def result_list(payload):
+    if payload.get('code') != 1:
+        raise ValueError(f"探数接口错误 {payload.get('code')}：{payload.get('msg', '未知错误')}")
+    data = payload.get('data')
+    if not isinstance(data, dict) or not isinstance(data.get('list'), list):
+        raise ValueError('接口响应缺少 data.list')
+    return data['list']
+
+
+def fingerprints(article):
+    values = []
+    for field in ('url', 'weburl'):
+        value = str(article.get(field) or '').strip()
+        if value:
+            u = urlsplit(value)
+            query = [(k, v) for k, v in parse_qsl(u.query) if not k.lower().startswith('utm_') and k.lower() not in ('spm', 'from', 'pos', 'vt')]
+            values.append('url:' + urlunsplit(('', u.netloc.lower(), u.path, urlencode(sorted(query)), '')))
+    body = BeautifulSoup(str(article.get('content') or ''), 'html.parser').get_text(' ', strip=True)
+    body = ''.join(unicodedata.normalize('NFKC', body).split())
+    if body:
+        values.append('body:' + body)
+    return [sha256(value.encode()).hexdigest() for value in values]
+
+
+class Downloader:
+    def __init__(self, root=ROOT / 'downloads', requester=api_request, writer=save_article, on_saved=None):
+        self.root = Path(root).resolve()
+        self.root.mkdir(parents=True, exist_ok=True)
+        self.requester, self.writer = requester, writer
+        self.on_saved = on_saved
+        self.lock = threading.RLock()
+        self.stop = threading.Event()
+        self.thread = None
+        self.logs = deque(maxlen=300)
+        self.seq = 0
+        self.state = dict(status='idle', saved=0, target=0, skipped=0, failed=0, scanned=0, requests=0, channel='', message='等待开始')
+        with self.db() as db:
+            db.execute('CREATE TABLE IF NOT EXISTS seen (fingerprint TEXT PRIMARY KEY, directory TEXT NOT NULL)')
+
+    @contextmanager
+    def db(self):
+        connection = sqlite3.connect(self.root / '.history.sqlite3')
+        try:
+            with connection:
+                yield connection
+        finally:
+            connection.close()
+
+    def log(self, message):
+        with self.lock:
+            self.seq += 1
+            self.logs.append(dict(id=self.seq, time=datetime.now().strftime('%H:%M:%S'), message=message))
+
+    def update(self, **values):
+        with self.lock:
+            self.state.update(values)
+
+    def snapshot(self):
+        with self.lock:
+            return {**self.state, 'logs': list(self.logs), 'output': str(self.root)}
+
+    def start(self, key, channel, target):
+        with self.lock:
+            if self.thread and self.thread.is_alive():
+                raise ValueError('已有任务正在运行')
+            self.stop.clear()
+            self.logs.clear()
+            self.state = dict(status='running', saved=0, target=target, skipped=0, failed=0, scanned=0, requests=0, channel=channel, message='正在下载')
+            self.thread = threading.Thread(target=self.run, args=(key, channel, target), daemon=True)
+            self.thread.start()
+
+    def run(self, key, channel, target):
+        saved = skipped = failed = scanned = calls = 0
+        seen_in_run = set()
+        self.log(f'开始下载频道「{channel}」，目标 {target} 条；历史重复新闻将跳过')
+        try:
+            with self.db() as db:
+                offset = 0
+                while offset <= 400 and saved < target and not self.stop.is_set():
+                    count = min(40, target - saved)
+                    calls += 1
+                    self.update(requests=calls)
+                    self.log(f'请求第 {calls} 页：start={offset}，num={count}')
+                    items = result_list(self.requester(key, 'index', {'type': channel, 'num': count, 'start': offset}))
+                    if not items:
+                        self.log('接口返回空列表，已停止')
+                        break
+                    for article in items:
+                        if saved >= target or self.stop.is_set():
+                            break
+                        scanned += 1
+                        try:
+                            if not isinstance(article, dict):
+                                raise ValueError('新闻条目格式错误')
+                            keys = fingerprints(article)
+                            existing = any(k in seen_in_run for k in keys)
+                            for k in keys:
+                                row = db.execute('SELECT directory FROM seen WHERE fingerprint=?', (k,)).fetchone()
+                                if row and Path(row[0]).is_dir():
+                                    existing = True
+                            seen_in_run.update(keys)
+                            if existing:
+                                skipped += 1
+                                self.log('跳过重复新闻：' + str(article.get('title', '无标题'))[:160])
+                                continue
+                            result = self.writer(article, self.root)
+                            try:
+                                db.executemany('INSERT OR REPLACE INTO seen VALUES (?, ?)', [(k, result.directory) for k in keys])
+                                db.commit()
+                            except Exception:
+                                # 保存目录尚未登记成功，回滚本次产物，避免下次重复保存。
+                                import shutil
+                                shutil.rmtree(result.directory, ignore_errors=True)
+                                raise
+                            saved += 1
+                            if self.on_saved:
+                                try:
+                                    self.on_saved(article, result, keys)
+                                except Exception as error:
+                                    self.log('文章已保存，入队需处理：' + str(error).replace(key, '[已隐藏]')[:300])
+                            self.log(f'下载新闻 {saved}/{target} 条：{result.title}')
+                        except Exception as exc:
+                            failed += 1
+                            self.log(f'保存失败：{str(exc).replace(key, "[已隐藏]")[:400]}')
+                        finally:
+                            self.update(saved=saved, skipped=skipped, failed=failed, scanned=scanned)
+                    # 避免越过 start=400；已消费的条目决定下一起始位置。
+                    offset += len(items)
+                    if offset > 400:
+                        self.log('已到达接口允许的分页范围')
+                    if saved < target and offset <= 400:
+                        self.stop.wait(0.3)
+            status = 'completed' if saved >= target else 'cancelled' if self.stop.is_set() else 'partial'
+            message = f'已保存 {saved}/{target} 条，跳过重复 {skipped} 条，失败 {failed} 条'
+            if status == 'partial':
+                message += '；当前可获取新闻不足，未达到目标数量'
+            if status == 'cancelled':
+                message += '；任务已停止'
+            self.update(status=status, message=message)
+            self.log(message)
+        except Exception as exc:
+            message = str(exc).replace(key, '[已隐藏]')[:500]
+            self.update(status='error', message=message)
+            self.log('任务中断：' + message)
+
+
+def make_handler(manager, token):
+    class Handler(BaseHTTPRequestHandler):
+        def log_message(self, *args):
+            pass
+
+        def send(self, status, payload, mime='application/json; charset=utf-8'):
+            data = json.dumps(payload, ensure_ascii=False).encode() if isinstance(payload, dict) else payload
+            self.send_response(status)
+            self.send_header('Content-Type', mime)
+            self.send_header('Content-Length', str(len(data)))
+            self.send_header('Cache-Control', 'no-store')
+            self.send_header('X-Content-Type-Options', 'nosniff')
+            self.end_headers()
+            self.wfile.write(data)
+
+        def local(self):
+            return self.headers.get('Host') in (f'127.0.0.1:{self.server.server_port}', f'localhost:{self.server.server_port}')
+
+        def do_GET(self):
+            if not self.local():
+                return self.send(403, {'error': '仅允许本机访问'})
+            path = urlsplit(self.path).path
+            if path == '/api/state':
+                return self.send(200, manager.snapshot())
+            if path == '/api/session':
+                return self.send(200, {'token': token})
+            files = {'/': ('index.html', 'text/html; charset=utf-8'), '/app.js': ('app.js', 'text/javascript; charset=utf-8'), '/style.css': ('style.css', 'text/css; charset=utf-8')}
+            if path not in files:
+                return self.send(404, {'error': '未找到页面'})
+            name, mime = files[path]
+            self.send(200, (ROOT / 'static' / name).read_bytes(), mime)
+
+        def do_POST(self):
+            if not self.local() or self.headers.get('X-Session-Token') != token:
+                return self.send(403, {'error': '页面会话失效，请刷新页面'})
+            key = ''
+            try:
+                length = int(self.headers.get('Content-Length', 0))
+                if not 0 < length <= 16384:
+                    raise ValueError('请求大小无效')
+                body = json.loads(self.rfile.read(length))
+                if not isinstance(body, dict):
+                    raise ValueError('请求必须是 JSON 对象')
+                if self.path == '/api/stop':
+                    manager.stop.set()
+                    return self.send(200, {'ok': True})
+                key = str(body.get('key') or '').strip()
+                if not key or len(key) > 512:
+                    raise ValueError('请填写有效 API Key')
+                if self.path == '/api/channels':
+                    channels = result_list(manager.requester(key, 'channel'))
+                    if not channels or not all(isinstance(c, str) and c for c in channels):
+                        raise ValueError('频道列表为空或格式错误')
+                    return self.send(200, {'channels': channels})
+                if self.path == '/api/start':
+                    target = body.get('target')
+                    channel = body.get('channel')
+                    if type(target) is not int or not 1 <= target <= 100000:
+                        raise ValueError('下载数量须为 1 至 100000 的整数')
+                    if not isinstance(channel, str) or not channel.strip() or len(channel) > 100:
+                        raise ValueError('请选择新闻频道')
+                    manager.start(key, channel, target)
+                    return self.send(200, {'ok': True})
+                if self.path == '/api/debug':
+                    endpoint = body.get('endpoint')
+                    params = {}
+                    if endpoint == 'index':
+                        for field, low, high in [('num', 1, 40), ('start', 0, 400)]:
+                            value = body.get(field)
+                            if type(value) is not int or not low <= value <= high:
+                                raise ValueError(f'{field} 必须在 {low}–{high} 之间')
+                            params[field] = value
+                        params['type'] = str(body.get('channel') or '')
+                    before = time.monotonic()
+                    payload = manager.requester(key, endpoint, params)
+                    # 服务端响应同样隐藏 key，且只作为文本展示，不渲染上游 HTML。
+                    clean = json.loads(json.dumps(payload, ensure_ascii=False).replace(key, '[已隐藏]'))
+                    return self.send(200, {'elapsed': round(time.monotonic() - before, 2), 'request': API + str(endpoint) + '?' + urlencode({'key': '[已隐藏]', **params}), 'response': clean})
+                self.send(404, {'error': '未知操作'})
+            except Exception as exc:
+                message = str(exc)
+                if key:
+                    message = message.replace(key, '[已隐藏]')
+                self.send(400, {'error': message[:500]})
+    return Handler
+
+
+def main():
+    parser = argparse.ArgumentParser(description=__doc__)
+    parser.add_argument('--port', type=int, default=8765)
+    parser.add_argument('--no-browser', action='store_true')
+    args = parser.parse_args()
+    manager = Downloader()
+    server = ThreadingHTTPServer(('127.0.0.1', args.port), make_handler(manager, secrets.token_urlsafe(32)))
+    url = f'http://127.0.0.1:{server.server_port}'
+    print(f'新闻下载器：{url}\n保存目录：{manager.root}\n按 Ctrl+C 退出', flush=True)
+    if not args.no_browser:
+        webbrowser.open(url)
+    try:
+        server.serve_forever()
+    except KeyboardInterrupt:
+        manager.stop.set()
+        if manager.thread:
+            manager.thread.join(timeout=45)
+    finally:
+        server.server_close()
+
+
+if __name__ == '__main__':
+    main()

modules/downloader/article_writer.py

@@ -0,0 +1,212 @@
+#!/usr/bin/env python3
+"""将探数新闻正文保存为与微信下载器相同的 Markdown 目录格式。"""
+
+from __future__ import annotations
+
+import base64
+from dataclasses import asdict, dataclass
+from pathlib import Path
+import re
+import shutil
+import tempfile
+from urllib.parse import unquote, urljoin, urlsplit
+
+import requests
+from bs4 import BeautifulSoup
+from markdownify import markdownify
+from requests.adapters import HTTPAdapter
+from urllib3.util.retry import Retry
+
+
+class ArticleDownloadError(RuntimeError):
+    """页面不是文章、图片下载失败等可向调用方报告的错误。"""
+
+
+@dataclass
+class ArticleResult:
+    title: str
+    source_url: str
+    directory: str
+    markdown: str
+    images: list[str]
+    warnings: list[str]
+
+    def to_dict(self) -> dict:
+        return asdict(self)
+
+
+def safe_name(title: str) -> str:
+    """跨平台目录/文件名；目录名与其中 Markdown 的文件名保持一致。"""
+    name = re.sub(r'[<>:"/\\|?*\x00-\x1f\x7f]', '_', title)
+    name = re.sub(r'\s+', ' ', name).strip(' .')
+    # 同时限制 UTF-8 字节数，给 .md 和重名后缀保留空间。
+    name = name.encode('utf-8')[:180].decode('utf-8', errors='ignore').rstrip(' .')
+    if not name:
+        name = '未命名文章'
+    if re.fullmatch(r'(?i:CON|PRN|AUX|NUL|COM[1-9]|LPT[1-9])(?:\..*)?', name):
+        name = '_' + name
+    return name
+
+
+def _session() -> requests.Session:
+    session = requests.Session()
+    session.headers.update({
+        'User-Agent': 'Mozilla/5.0 (Macintosh; Intel Mac OS X 10_15_7) '
+                      'AppleWebKit/537.36 (KHTML, like Gecko) Chrome/130.0.0.0 Safari/537.36',
+        'Accept-Language': 'zh-CN,zh;q=0.9,en;q=0.8',
+    })
+    retry = Retry(total=3, backoff_factor=0.6,
+                  status_forcelist=[429, 500, 502, 503, 504], allowed_methods=['GET'])
+    session.mount('https://', HTTPAdapter(max_retries=retry))
+    session.mount('http://', HTTPAdapter(max_retries=retry))
+    return session
+
+
+def _get(session: requests.Session, url: str, timeout: float, limit: int,
+         referer: str | None = None) -> tuple[bytes, str, str]:
+    if urlsplit(url).scheme not in ('http', 'https'):
+        raise ArticleDownloadError(f'不支持的下载地址：{url[:200]}')
+    headers = {'Referer': referer} if referer else {}
+    with session.get(url, timeout=(min(timeout, 15), timeout),
+                     headers=headers, stream=True) as response:
+        response.raise_for_status()
+        chunks, size = [], 0
+        for chunk in response.iter_content(65536):
+            size += len(chunk)
+            if size > limit:
+                raise ArticleDownloadError(f'文件超过大小限制（{limit // 1024 // 1024} MB）：{url[:200]}')
+            chunks.append(chunk)
+        return b''.join(chunks), response.headers.get('Content-Type', ''), response.url
+
+
+def _image_ext(data: bytes, mime: str) -> str:
+    if data.startswith(b'\xff\xd8\xff'):
+        return '.jpg'
+    if data.startswith(b'\x89PNG\r\n\x1a\n'):
+        return '.png'
+    if data.startswith((b'GIF87a', b'GIF89a')):
+        return '.gif'
+    if data[:4] == b'RIFF' and data[8:12] == b'WEBP':
+        return '.webp'
+    if len(data) >= 12 and data[4:8] == b'ftyp' and data[8:12] in (b'avif', b'avis'):
+        return '.avif'
+    if b'<svg' in data[:1024].lower() and mime.split(';')[0] == 'image/svg+xml':
+        return '.svg'
+    raise ArticleDownloadError(f'返回内容不是支持的图片（Content-Type: {mime or "未知"}）')
+
+
+def _image_source(img) -> str:
+    for attr in ('data-src', 'data-original', 'data-actualsrc', 'data-lazy-src', 'src'):
+        value = img.get(attr)
+        if value and str(value).strip():
+            return str(value).strip()
+    srcset = img.get('data-srcset') or img.get('srcset')
+    if srcset:
+        return str(srcset).split(',')[-1].strip().split()[0]
+    return ''
+
+
+def save_article(article: dict, output_root: str | Path = '.', *, timeout: float = 30,
+                 allow_missing_images: bool = False) -> ArticleResult:
+    """从 API 正文保存文章；图片失败时清理半成品，不覆盖已有同名目录。"""
+    url = str(article.get('weburl') or article.get('url') or '')
+    title = str(article.get('title') or '').strip()
+    raw_content = str(article.get('content') or '').strip()
+    if not title or not raw_content:
+        raise ArticleDownloadError('新闻标题或正文为空')
+    if timeout <= 0:
+        raise ValueError('timeout 必须大于 0')
+    with _session() as session:
+        final_url = url
+        soup = BeautifulSoup('<article>' + raw_content + '</article>', 'html.parser')
+        content = soup.article
+        for tag in content.select('script, style, noscript, iframe, form, button'):
+            tag.decompose()
+        # 部分微信排版把插图放在行内 CSS 背景中，也转为普通图片参与下载。
+        for tag in list(content.find_all(style=True)):
+            for match in re.finditer(r'url\(\s*[\'\"]?(.*?)[\'\"]?\s*\)', str(tag['style'])):
+                source = match.group(1).strip()
+                if source and not source.startswith('#'):
+                    background = soup.new_tag('img', src=source, alt='')
+                    tag.insert(0, background)
+        if not content.get_text(strip=True) and not content.find('img'):
+            raise ArticleDownloadError('文章正文为空，可能需要浏览器执行 JavaScript')
+        for link in content.find_all('a', href=True):
+            href = str(link['href']).strip()
+            if href and not href.startswith('#'):
+                absolute = urljoin(final_url, href)
+                if urlsplit(absolute).scheme in ('http', 'https', 'mailto', 'tel'):
+                    link['href'] = absolute
+                else:
+                    del link['href']
+
+        root = Path(output_root).expanduser().resolve()
+        root.mkdir(parents=True, exist_ok=True)
+        stage = Path(tempfile.mkdtemp(prefix='.article-', dir=root))
+        images, warnings, downloaded = [], [], {}
+        try:
+            for img in list(content.find_all('img')):
+                source = _image_source(img)
+                if not source:
+                    img.replace_with(soup.new_string('[图片缺少下载地址]'))
+                    if not allow_missing_images:
+                        raise ArticleDownloadError('正文中存在没有下载地址的图片')
+                    warnings.append('正文中存在没有下载地址的图片')
+                    continue
+                image_url = urljoin(final_url, source)
+                try:
+                    if image_url not in downloaded:
+                        if source.startswith('data:image/'):
+                            header, payload = source.split(',', 1)
+                            data = base64.b64decode(payload, validate=True) if ';base64' in header else unquote(payload).encode()
+                            mime = header[5:].split(';')[0]
+                            if len(data) > 50 * 1024 * 1024:
+                                raise ArticleDownloadError('内嵌图片超过 50 MB')
+                        else:
+                            data, mime, _ = _get(session, image_url, timeout, 50 * 1024 * 1024, final_url)
+                        ext = _image_ext(data, mime)
+                        filename = f'image{len(images) + 1:03d}{ext}'
+                        (stage / filename).write_bytes(data)
+                        images.append(filename)
+                        downloaded[image_url] = filename
+                    # 清除懒加载/响应式属性，使转换器只引用本地文件。
+                    alt = str(img.get('alt') or '')
+                    img.attrs = {'src': downloaded[image_url], 'alt': alt}
+                except (requests.RequestException, ArticleDownloadError, ValueError) as exc:
+                    if not allow_missing_images:
+                        raise ArticleDownloadError(f'图片下载失败：{image_url[:200]}；{exc}') from exc
+                    warnings.append(f'图片下载失败：{image_url[:200]}；{exc}')
+                    img.replace_with(soup.new_string(f'[图片下载失败：{img.get("alt") or "未命名图片"}]'))
+
+            # 新闻图片后常带空的图注标题，避免生成单独的 ## 和多余空行。
+            for heading in content.select('h1, h2, h3, h4, h5, h6'):
+                if not heading.get_text(strip=True) and not heading.find('img'):
+                    heading.decompose()
+            body = markdownify(str(content), heading_style='ATX', bullets='-',
+                               strip=['picture', 'source']).strip()
+            # 微信富文本中各段常没有标准 p 标签；markdownify 仍会保留文字和图片顺序。
+            author = str(article.get('src') or '未知作者').strip() or '未知作者'
+            text = f'# {title}\n\n'
+            text += body + f'\n\n来源：{author}，如有侵权，请联系删除。\n'
+            basename = safe_name(title)
+            # 原子认领新目录，支持多个进程并发保存同名文章。
+            index = 1
+            while True:
+                name = basename if index == 1 else f'{basename} ({index})'
+                destination = root / name
+                try:
+                    destination.mkdir()
+                    break
+                except FileExistsError:
+                    index += 1
+            try:
+                (stage / f'{name}.md').write_text(text, encoding='utf-8')
+                for item in stage.iterdir():
+                    shutil.move(str(item), destination / item.name)
+            except BaseException:
+                shutil.rmtree(destination, ignore_errors=True)
+                raise
+            return ArticleResult(title, url, str(destination), str(destination / f'{name}.md'),
+                                 [str(destination / x) for x in images], warnings)
+        finally:
+            shutil.rmtree(stage, ignore_errors=True)

modules/downloader/requirements.txt

@@ -0,0 +1,3 @@
+requests>=2.32,<3
+beautifulsoup4>=4.12,<5
+markdownify>=1.1,<2

modules/downloader/static/app.js

@@ -0,0 +1,79 @@
+'use strict';
+const $ = id => document.getElementById(id);
+let token = '', running = false, starting = false, loadingChannels = false, loadedKey = '', lastLog = -1;
+const labels = {idle:'等待开始',running:'下载中',completed:'已完成',partial:'未达到目标',cancelled:'已停止',error:'任务中断'};
+const key = () => $('key').value.trim();
+function controls() {
+  const busy = running || starting;
+  $('start').disabled = busy || !loadedKey || loadedKey !== key() || !$('channel').value;
+  $('stop').disabled = !running;
+  $('channels').disabled = busy || loadingChannels;
+  for (const id of ['key','channel','target']) $(id).disabled = busy || (id === 'channel' && !loadedKey);
+}
+async function post(path, body) {
+  const response = await fetch('/api/' + path, {method:'POST',headers:{'Content-Type':'application/json','X-Session-Token':token},body:JSON.stringify(body)});
+  const data = await response.json();
+  if (!response.ok) throw new Error(data.error || '请求失败');
+  return data;
+}
+$('key').addEventListener('input', () => {loadedKey=''; $('channel').replaceChildren(new Option('请重新获取频道','')); controls();});
+$('channels').addEventListener('click', async () => {
+  loadingChannels=true; controls(); $('notice').textContent='正在获取频道…';
+  const submittedKey = key();
+  try {
+    const data = await post('channels',{key:submittedKey});
+    if (key() !== submittedKey) return;
+    $('channel').replaceChildren(...data.channels.map(c => new Option(c,c)));
+    loadedKey=submittedKey; $('notice').textContent=`已获取 ${data.channels.length} 个频道。`;
+    $('debug-channel').value=$('channel').value;
+  } catch(error) {$('notice').textContent=error.message;}
+  finally {loadingChannels=false; controls();}
+});
+$('channel').addEventListener('change', () => {$('debug-channel').value=$('channel').value; controls();});
+$('download-form').addEventListener('submit', async event => {
+  event.preventDefault(); starting=true; controls(); $('notice').textContent='';
+  try {await post('start',{key:key(),channel:$('channel').value,target:Number($('target').value)}); await refresh();}
+  catch(error) {$('notice').textContent=error.message;}
+  finally {starting=false; controls();}
+});
+$('stop').addEventListener('click', async () => {
+  $('stop').disabled=true;
+  try {await post('stop',{}); $('notice').textContent='正在停止，等待当前请求或文章保存结束…';}
+  catch(error) {$('notice').textContent=error.message;}
+});
+$('debug-form').addEventListener('submit', async event => {
+  event.preventDefault(); $('debug-submit').disabled=true; $('debug-output').textContent='正在请求…';
+  try {
+    const data=await post('debug',{key:key(),endpoint:$('endpoint').value,channel:$('debug-channel').value,num:Number($('debug-num').value),start:Number($('debug-start').value)});
+    const output=JSON.stringify(data,null,2);
+    $('debug-output').textContent=output.slice(0,100000)+(output.length>100000?'\n…输出超过 100000 字符，已截断。':'');
+  } catch(error) {$('debug-output').textContent=error.message;}
+  finally {$('debug-submit').disabled=false;}
+});
+async function refresh() {
+  const response=await fetch('/api/state');
+  if (!response.ok) throw new Error('无法读取下载状态');
+  const data=await response.json();
+  running=data.status==='running'; controls();
+  const percent=data.target ? Math.floor(data.saved/data.target*100) : 0;
+  $('status').textContent=labels[data.status] || data.status;
+  $('progress-text').textContent=`下载新闻 ${data.saved}/${data.target} 条`;
+  $('percent').textContent=`已完成 ${percent}%`; $('progress').value=percent;
+  for (const id of ['saved','skipped','failed','requests','message','output']) $(id).textContent=data[id];
+  const latest=data.logs.length ? data.logs[data.logs.length-1].id : 0;
+  if (latest!==lastLog) {
+    const logs=$('logs'); const bottom=logs.scrollHeight-logs.scrollTop-logs.clientHeight<40;
+    logs.replaceChildren(...data.logs.map(log => {const row=document.createElement('div'); const stamp=document.createElement('time'); stamp.textContent=log.time; row.append(stamp,document.createTextNode(log.message)); return row;}));
+    if (bottom) logs.scrollTop=logs.scrollHeight;
+    lastLog=latest;
+  }
+}
+async function poll() {
+  try {await refresh();}
+  catch(error) {$('notice').textContent='与本地服务连接中断，请检查启动窗口。';}
+  setTimeout(poll,700);
+}
+(async () => {
+  try {const response=await fetch('/api/session'); token=(await response.json()).token; await poll();}
+  catch(error) {$('notice').textContent='初始化失败，请刷新页面。';}
+})();

modules/downloader/static/index.html

@@ -0,0 +1,14 @@
+<!doctype html>
+<html lang="zh-CN">
+<head><meta charset="UTF-8"><meta name="viewport" content="width=device-width, initial-scale=1"><title>探数 · 新闻下载器</title><link rel="stylesheet" href="/style.css"><script src="/app.js" defer></script></head>
+<body><main>
+<header><div class="brand">TANSHU / NEWS ARCHIVE</div><h1>把新闻，保存到本地。</h1><p>选择频道，收集新闻。正文与图片一起保存，自动跳过已下载的内容。</p></header>
+<section class="card"><div class="section-heading"><h2>下载设置</h2><span class="badge">本地运行</span></div>
+<form id="download-form"><label for="key">API Key</label><div class="key-row"><input id="key" type="password" placeholder="填写探数 API Key" autocomplete="off" required maxlength="512"><button id="channels" type="button" class="secondary">获取新闻频道</button></div><p class="hint">密钥仅用于本次页面会话，不写入文件或浏览器存储。</p>
+<div class="fields"><div><label for="channel">新闻频道</label><select id="channel" required disabled><option value="">先填写密钥并获取频道</option></select></div><div><label for="target">下载数量</label><input id="target" type="number" min="1" max="100000" step="1" value="10" required></div></div>
+<p class="hint">接口每次最多 40 条，起始位置最多 400。遇到重复会继续向后查找；可用新闻不足时会说明原因。</p><div class="actions"><button id="start" disabled>开始下载</button><button id="stop" type="button" class="secondary" disabled>停止下载</button></div><p id="notice" role="alert"></p></form></section>
+<section class="card"><div class="section-heading"><h2>下载进度</h2><span id="status" class="badge">等待开始</span></div><div class="progress-heading"><strong id="progress-text">下载新闻 0/0 条</strong><span id="percent">已完成 0%</span></div><progress id="progress" max="100" value="0" aria-label="下载进度"></progress><div class="stats"><span>已保存 <b id="saved">0</b></span><span>重复跳过 <b id="skipped">0</b></span><span>保存失败 <b id="failed">0</b></span><span>接口请求 <b id="requests">0</b></span></div><p id="message">等待开始</p><p class="hint path">保存位置：<span id="output">加载中…</span></p></section>
+<section class="card"><div class="section-heading"><h2>运行输出</h2><span class="hint">保留最近 300 条</span></div><div id="logs" class="console" role="log" aria-label="运行输出"><div class="empty">运行情况将在这里显示。</div></div></section>
+<details class="card"><summary>调试终端 <span>手动请求接口</span></summary><p class="hint">使用上方 API Key。请求会消耗接口调用额度，但不会保存新闻。</p><form id="debug-form"><div class="fields"><div><label for="endpoint">请求接口</label><select id="endpoint"><option value="channel">新闻频道查询 /channel</option><option value="index">新闻列表查询 /index</option></select></div><div><label for="debug-channel">type · 频道</label><input id="debug-channel" value="头条"></div><div><label for="debug-num">num · 数量（1–40）</label><input id="debug-num" type="number" min="1" max="40" value="1" required></div><div><label for="debug-start">start · 起始位置（0–400）</label><input id="debug-start" type="number" min="0" max="400" value="0" required></div></div><button id="debug-submit" class="secondary">发送测试请求</button></form><pre id="debug-output" class="console">尚未发送请求。</pre></details>
+<footer>Markdown 正文 · 本地图片 · 历史去重</footer>
+</main></body></html>

modules/downloader/static/style.css

@@ -0,0 +1 @@
+:root{font-family:Inter,-apple-system,BlinkMacSystemFont,"Segoe UI","PingFang SC",sans-serif;color:#20312e;background:#f3f6f3;font-size:15px}*{box-sizing:border-box}body{margin:0}main{max-width:960px;margin:0 auto;padding:48px 24px 24px}header{margin-bottom:32px}.brand{font-size:12px;letter-spacing:2px;font-weight:700;color:#437b69}h1{font-size:36px;letter-spacing:-1px;margin:14px 0}header p{color:#687a74;line-height:1.8}h2{font-size:17px;margin:0}.card{padding:26px 28px;background:#fff;border:1px solid #dfe7e1;border-radius:16px;margin-bottom:20px;box-shadow:0 4px 16px #243c2904}.section-heading{display:flex;align-items:center;justify-content:space-between;gap:10px;margin-bottom:24px}.badge{background:#edf4ef;color:#4c7564;padding:6px 10px;border-radius:20px;font-size:12px}.key-row{display:flex;gap:12px}.key-row input{flex:1;min-width:0}label{display:block;font-weight:600;margin-bottom:9px;font-size:13px}input,select,button{font:inherit;border-radius:8px}input,select{background:#fbfcfb;border:1px solid #d7e0d9;width:100%;padding:12px;color:inherit}input:focus,select:focus,button:focus-visible,summary:focus-visible{outline:3px solid #8dcab5;outline-offset:2px}button{cursor:pointer;border:1px solid #286850;background:#286850;color:#fff;padding:12px 22px;font-weight:600;white-space:nowrap}.secondary{background:#fff;color:#345e4f;border-color:#cedbd3}button:disabled{opacity:.45;cursor:not-allowed}.fields{display:grid;grid-template-columns:1fr 1fr;gap:20px;margin:24px 0 16px}.hint{font-size:12px;color:#79867f;line-height:1.8;margin:10px 0}.actions{display:flex;gap:12px;margin-top:22px}#notice{color:#ac4c32;font-size:13px;white-space:pre-wrap}#notice:empty{display:none}.progress-heading{display:flex;justify-content:space-between;gap:16px;font-size:14px}#percent{color:#3f7b60}progress{display:block;width:100%;height:12px;border:0;border-radius:10px;overflow:hidden;margin:18px 0 20px;background:#edf2ef}progress::-webkit-progress-bar{background:#edf2ef}progress::-webkit-progress-value{background:#3c8b68;transition:width .25s}progress::-moz-progress-bar{background:#3c8b68}.stats{display:flex;gap:28px;flex-wrap:wrap;color:#7a8880;font-size:13px}.stats b{color:#304e40;margin-left:8px;font-size:17px}#message{font-size:13px;color:#52685b;line-height:1.8}.path{overflow-wrap:anywhere}.console{height:260px;overflow:auto;background:#15251f;color:#c5dacf;border-radius:10px;padding:18px;font:12px/1.9 ui-monospace,SFMono-Regular,Consolas,monospace;white-space:pre-wrap;overflow-wrap:anywhere}.console time{color:#7eac94;margin-right:14px}.empty{color:#7e9889}summary{cursor:pointer;font-weight:600}summary span{font-size:12px;font-weight:400;color:#79867f;margin-left:12px}#debug-output{margin-bottom:0;height:300px}footer{text-align:center;color:#89958e;font-size:12px;margin:32px 0}@media(max-width:600px){main{padding:26px 14px}h1{font-size:28px}.card{padding:22px 18px}.key-row{flex-direction:column}.fields{grid-template-columns:1fr}.stats{gap:14px}.progress-heading{font-size:12px}}

modules/package.py

@@ -0,0 +1,63 @@
+#!/usr/bin/env python3
+"""Build a portable v1 article package for the DWW11 userscript (standard library only)."""
+from __future__ import annotations
+
+import argparse
+import base64
+import json
+import mimetypes
+from pathlib import Path
+
+
+def make_package(directory: str | Path, markdown: str | None = None) -> dict:
+    """Return JSON-serializable data; no network access or publishing side effects."""
+    root = Path(directory).expanduser().resolve(strict=True)
+    if not root.is_dir():
+        raise ValueError("文章目录不存在")
+    choices = sorted(path for path in root.rglob("*") if path.is_file() and path.suffix.lower() == ".md")
+    if markdown:
+        selected = (root / markdown).resolve(strict=True)
+    elif len(choices) == 1:
+        selected = choices[0].resolve()
+    else:
+        raise ValueError("目录必须包含一个 Markdown 文件，或通过 --markdown 指定目录内相对路径")
+    if not selected.is_relative_to(root) or selected.suffix.lower() != ".md" or not selected.is_file():
+        raise ValueError("Markdown 必须位于所选文章目录内")
+    images = []
+    # Package local supported images. The importer uploads only those referenced by Markdown.
+    for path in sorted(root.rglob("*")):
+        if not path.is_file() or path.suffix.lower() not in {".jpg", ".jpeg", ".png", ".gif"}:
+            continue
+        if not path.resolve().is_relative_to(root):
+            raise ValueError(f"图片符号链接指向目录外：{path.name}")
+        if not 0 < path.stat().st_size < 20 * 1024 * 1024:
+            raise ValueError(f"图片为空或达到 20 MB：{path.name}")
+        images.append({"path": path.relative_to(root).as_posix(),
+                       "mime": mimetypes.guess_type(path.name)[0] or "application/octet-stream",
+                       "base64": base64.b64encode(path.read_bytes()).decode("ascii")})
+    return {"schemaVersion": 1, "markdownPath": selected.relative_to(root).as_posix(),
+            "markdown": selected.read_text(encoding="utf-8-sig"), "images": images}
+
+
+def main() -> None:
+    parser = argparse.ArgumentParser(description="把文章目录打包为油猴脚本可导入的 JSON；不上传或发布")
+    parser.add_argument("directory", help="包含 Markdown 和图片的目录")
+    parser.add_argument("--markdown", help="多个 Markdown 时指定目录内的相对路径")
+    parser.add_argument("-o", "--output", help="输出文件路径；未指定时输出到标准输出")
+    args = parser.parse_args()
+    try:
+        content = json.dumps(make_package(args.directory, args.markdown), ensure_ascii=False, indent=2)
+        if args.output:
+            output = Path(args.output).expanduser()
+            # Never silently overwrite an article or an existing package.
+            with output.open("x", encoding="utf-8") as handle:
+                handle.write(content + "\n")
+            print(json.dumps({"ok": True, "package": str(output.resolve())}, ensure_ascii=False))
+        else:
+            print(content)
+    except (OSError, ValueError) as error:
+        parser.exit(1, f"打包失败：{error}\n")
+
+
+if __name__ == "__main__":
+    main()

requirements.txt

@@ -0,0 +1,3 @@
+requests>=2.32,<3
+beautifulsoup4>=4.12,<5
+markdownify>=1.1,<2

start.command

@@ -0,0 +1,8 @@
+#!/bin/zsh
+cd "${0:A:h}"
+if ! command -v uv >/dev/null 2>&1; then
+  print '未找到 uv，请安装 uv 后重试。'
+  read '?按回车关闭…'
+  exit 1
+fi
+uv run --no-project --python 3.12 --with-requirements requirements.txt python app.py

static/app.js

@@ -0,0 +1,51 @@
+'use strict';
+const $=id=>document.getElementById(id);
+let token='',state=null,loadedKey='',refreshing=false;
+const labels={queued:'待处理',preparing:'准备中',importing:'上传与导入',formatting:'排版中',awaiting_user_review:'等待人工审核',user_submitting:'已观察到人工提交',submit_unknown:'发布结果待核对',accepted_pending_moderation:'已受理 · 审核中',published:'已确认公开',needs_attention:'需要处理',cancelled:'已移出'};
+function notice(message){$('notice').hidden=!message;$('notice').textContent=message;}
+async function api(route,body={}) {
+ const response=await fetch('/api/'+route,{method:'POST',headers:{'Content-Type':'application/json','X-Session-Token':token},body:JSON.stringify(body)});
+ const data=await response.json();if(!response.ok)throw new Error(data.error||'请求失败');return data;
+}
+function action(id,fn){$(id).addEventListener('click',async()=>{const button=$(id);button.disabled=true;try{await fn();await refresh();}catch(error){notice(error.message);}finally{button.disabled=false;}});}
+function text(tag,value,className){const e=document.createElement(tag);e.textContent=value;if(className)e.className=className;return e;}
+function render(data){
+ state=data;const {jobs,run,download}=data;
+ $('total').textContent=jobs.length;$('review').textContent=jobs.filter(j=>j.state==='awaiting_user_review').length;
+ $('accepted').textContent=jobs.filter(j=>['published','accepted_pending_moderation'].includes(j.state)).length;
+ $('batch').textContent=run.batch?`${run.batch.count} / ${run.batch.target}`:'—';$('account').textContent=run.batch?run.batch.account.name:'尚未绑定账号';
+ $('run-state').textContent=run.running?'流程运行中':'流程已暂停';$('run-reason').textContent=run.reason;
+ $('run-light').classList.toggle('gray',!run.running);$('empty').hidden=jobs.length>0;
+ $('jobs').replaceChildren(...jobs.map(job=>{
+   const tr=document.createElement('tr'),title=document.createElement('td');title.append(text('div',job.title),text('small',`${job.image_count} 张配图 · ${job.source}`));if(job.original_title!==job.title)title.append(text('small','已缩短标题，原题：'+job.original_title));
+   const status=document.createElement('td');status.append(text('span',labels[job.state]||job.state,'badge'));if(job.error)status.append(text('small',job.error));
+   const account=text('td',job.account?`${job.account.name} / ${job.fid}`:'待领取');const result=document.createElement('td');
+   if(job.post_url){const link=text('a','查看帖子');link.href=job.post_url;link.target='_blank';link.rel='noopener';result.append(link);}
+   if(!['published','accepted_pending_moderation'].includes(job.state)){
+     const retry=text('button','人工核对 / 处理');retry.className='secondary';retry.onclick=async()=>{
+       if(state.run.running){notice('请先暂停流程，再处理任务');return;}
+       if(!confirm('只有确认网站未受理这篇文章，且已清理编辑器/草稿后才能继续。\n已受理、审核中或结果不明的文章请取消。\n你是否已经确认未受理？'))return;
+       const again=confirm('重新加入队列？\n确定：重新排队。取消：移出队列。');
+       try{await api('resolve',{id:job.id,confirmed_not_accepted:true,retry:again});await refresh();}catch(error){notice(error.message);}
+     };result.append(retry);
+   }
+   tr.append(title,status,account,result);return tr;
+ }));
+ const dlLabels={idle:'等待开始',running:'下载中',completed:'下载完成',partial:'未达到目标',cancelled:'已停止',error:'下载出错'};
+ $('download-state').textContent=dlLabels[download.status]||download.status;$('download-progress').textContent=`已保存 ${download.saved}/${download.target} 篇 · 跳过 ${download.skipped} · 失败 ${download.failed}`;
+ $('progress').value=download.target?Math.floor(download.saved/download.target*100):0;
+ $('download').disabled=download.status==='running';$('logs').textContent=download.logs.map(x=>`${x.time} ${x.message}`).join('\n')||'暂无日志';
+ $('browser-state').textContent=data.clients.length?`${data.clients.length} 个标签在线`:'尚未连接';
+ $('clients').textContent=data.clients.map(c=>`${c.account?.name||'未识别账号'} · 换号插件${c.extension?.ready?'已就绪':'未连接或需更新'}`).join('\n')||'没有在线执行标签';
+}
+async function refresh(){if(refreshing||!token)return;refreshing=true;try{render(await api('state'));}catch(error){notice(error.message);$('run-state').textContent='本地服务连接中断';}finally{refreshing=false;}}
+action('pair',async()=>{const value=await api('pair');$('pair-code').hidden=false;$('pair-code').type='text';$('pair-code').value=value.token;$('pair-code').select();});
+action('channels',async()=>{const value=await api('channels',{key:$('key').value.trim()});loadedKey=$('key').value.trim();$('channel').replaceChildren(...value.channels.map(c=>new Option(c,c)));});
+$('key').oninput=()=>{loadedKey='';$('channel').replaceChildren(new Option('请重新获取频道',''));};
+$('download-form').onsubmit=async event=>{event.preventDefault();try{if(!loadedKey||loadedKey!==$('key').value.trim())throw new Error('请先获取频道');await api('download',{key:loadedKey,channel:$('channel').value,target:Number($('target').value)});notice('下载已开始，完整保存的文章会自动入队');await refresh();}catch(error){notice(error.message);}};
+action('download-stop',()=>api('download-stop'));
+action('scan',async()=>{notice('正在检查文章目录…');const value=await api('scan',{root:$('root').value.trim()});$('scan-result').textContent=value.results.map(x=>`${x.directory}：${x.status}`).join('\n')||'目录中未找到文章';notice('目录检查完成');});
+$('config-form').onsubmit=async event=>{event.preventDefault();try{await api('configure',{fid:Number($('fid').value),minimum:Number($('minimum').value),maximum:Number($('maximum').value),interval:Number($('interval').value),history_checked:$('history-checked').checked});notice('处理规则已保存');await refresh();}catch(error){notice(error.message);}};
+for(const name of ['start','pause','stop'])action(name,async()=>{await api(name);notice(name==='stop'?'流程已终止；浏览器将在下一次连接检查时停止后续操作。已发出的请求可能仍已完成。':'');});
+$('history-file').onchange=async()=>{try{const file=$('history-file').files[0];if(!file)return;if(file.size>2*1024*1024)throw new Error('记录文件过大');const value=JSON.parse(await file.text());await api('history',{records:value.records});notice('历史记录已导入；仅使用标记为 block_repost 的记录');await refresh();}catch(error){notice(error.message);}};
+(async()=>{try{const response=await fetch('/api/session');const session=await response.json();if(!response.ok)throw new Error(session.error);token=session.token;$('port').value=location.port;await refresh();if(state?.run.config?.fid){for(const id of ['fid','minimum','maximum','interval'])$(id).value=state.run.config[id];$('history-checked').checked=state.run.config.history_checked;}}catch(error){notice(error.message);}setInterval(refresh,2500);})();

static/index.html

@@ -0,0 +1,26 @@
+<!doctype html>
+<html lang="zh-CN">
+<head><meta charset="utf-8"><meta name="viewport" content="width=device-width, initial-scale=1"><title>新闻工作台</title><link rel="stylesheet" href="/style.css"></head>
+<body>
+<aside><div class="brand"><span class="mark">闻</span><div>新闻工作台<small>本地内容工作流</small></div></div><nav><a href="#overview" class="selected">工作概览</a><a href="#source">下载与文章</a><a href="#settings">处理规则</a><a href="#queue">文章队列</a><a href="#connection">浏览器连接</a></nav><div class="aside-note"><span class="dot"></span> 数据保存在本机<p>文章自动准备与排版<br>发表由你亲手完成</p></div></aside>
+<main>
+<header id="overview"><div><div class="eyebrow">CONTENT WORKSPACE</div><h1>把内容准备好，把审核留给你。</h1><p>下载、导入、排版与账号切换，在一个工作台管理。</p></div><button id="stop" class="danger">终止流程</button></header>
+<div id="notice" role="status" aria-live="polite" class="notice" hidden></div>
+<section class="flow"><div><span>01</span>下载新闻</div><i>→</i><div><span>02</span>导入与排版</div><i>→</i><div class="highlight"><span>03</span>人工审核与发表</div><i>→</i><div><span>04</span>结果核对与换号</div></section>
+<div class="pending-note"><b>当前接入进度</b> 下载、导入、排版与换号接口已整合；结果页自动核对等待页面 HTML。人工发表后会暂停，不自动计数或继续下一篇。</div>
+<section class="metrics"><article><small>队列文章</small><strong id="total">0</strong><span>完整文件检查后入队</span></article><article><small>等待你审核</small><strong id="review">0</strong><span>请在网站检查并亲手发表</span></article><article><small>明确受理</small><strong id="accepted">0</strong><span>审核中与已公开分别记录</span></article><article><small>当前账号批次</small><strong id="batch">—</strong><span id="account">尚未绑定账号</span></article></section>
+<div class="workspace-grid">
+<section id="source" class="card"><div class="section-head"><div><span class="eyebrow">SOURCE</span><h2>下载与文章</h2></div><span id="download-state" class="badge">等待开始</span></div>
+<form id="download-form"><label>探数 API Key<input id="key" type="password" autocomplete="off" placeholder="仅在本次运行中使用"></label><div class="row"><label class="grow">新闻频道<select id="channel" required><option value="">先获取频道</option></select></label><button id="channels" type="button" class="secondary">获取频道</button></div><div class="row"><label class="grow">下载篇数<input id="target" type="number" min="1" max="100000" value="5" required></label><button id="download" type="submit">开始下载</button><button id="download-stop" type="button" class="secondary">停止下载</button></div></form>
+<div class="progress"><progress id="progress" max="100" value="0"></progress><small id="download-progress">等待下载</small></div>
+<details><summary>导入已有文章目录</summary><p class="muted">填写包含文章子文件夹的完整路径。留空时扫描本工具的 downloads 目录。请在文件完整保存后扫描。</p><label>文章根目录<input id="root" placeholder="/完整路径/新闻目录"></label><button id="scan" class="secondary">检查并加入队列</button><pre id="scan-result"></pre></details>
+</section>
+<section id="settings" class="card"><div class="section-head"><div><span class="eyebrow">WORKFLOW</span><h2>处理规则</h2></div><span class="badge">人工发表</span></div>
+<form id="config-form"><label>目标版块 ID（fid）<input id="fid" type="number" min="1" placeholder="从新帖页地址中的 fid 获取" required></label><div class="row"><label class="grow">每账号最少篇数<input id="minimum" type="number" min="1" max="1000" placeholder="最小值" required></label><label class="grow">每账号最多篇数<input id="maximum" type="number" min="1" max="1000" placeholder="最大值" required></label></div><label>两篇之间间隔（秒）<input id="interval" type="number" min="0" max="86400" placeholder="请输入间隔" required></label><label class="check"><input id="history-checked" type="checkbox" required><span>我已核对待处理文章与历史发布记录，避免重复发表。</span></label><button type="submit" class="secondary">保存处理规则</button></form><p class="muted">每个账号只抽取一次目标篇数；只有明确受理的文章才计数。历史演示去重记录会自动读取。</p>
+</section></div>
+<section class="runbar"><div><span id="run-light" class="dot gray"></span><b id="run-state">正在连接本地服务…</b><p id="run-reason">等待状态</p></div><div class="actions"><button id="start">开始 / 继续</button><button id="pause" class="secondary">暂停</button></div></section>
+<section id="queue" class="card"><div class="section-head"><div><span class="eyebrow">ARTICLE QUEUE</span><h2>文章队列</h2></div><label class="file-button">导入历史去重记录<input id="history-file" type="file" accept=".json,application/json" hidden></label></div><div class="table-wrap"><table><thead><tr><th>文章</th><th>状态</th><th>账号 / 版块</th><th>结果与操作</th></tr></thead><tbody id="jobs"></tbody></table><div id="empty" class="empty"><span>＋</span><b>从第一篇文章开始</b><p>下载新闻，或检查已有文章目录。准备好的文章会出现在这里。</p></div></div></section>
+<section id="connection" class="card"><div class="section-head"><div><span class="eyebrow">CONNECTION</span><h2>连接你的浏览器</h2></div><span id="browser-state" class="badge">尚未连接</span></div><div class="connection-grid"><div><ol><li>在油猴中安装以下两个脚本，并停用原导入、排版脚本。</li><li>重新加载原目录中的换号插件，然后刷新网站。</li><li>在网站左下角展开连接设置，填入本地端口和连接码。</li></ol><div class="actions"><a class="button secondary" href="/browser/local-bridge.user.js">安装本地连接脚本</a><a class="button secondary" href="/browser/auto-news.user.js">安装导入排版脚本</a></div></div><div><label>本地端口<input id="port" readonly></label><button id="pair" class="secondary">显示本次连接码</button><input id="pair-code" type="password" readonly hidden aria-label="本次连接码"><small class="muted">服务重启后需要更新连接码。仅复制到你安装的本地连接脚本。</small><p id="clients" class="muted">没有在线执行标签</p></div></div></section>
+<details class="card"><summary>下载日志</summary><pre id="logs">暂无日志</pre></details><footer>新闻工作台 · 本机运行 · v0.1 &nbsp; / &nbsp; 终止后不会自动恢复，已发出的请求可能仍已完成。</footer>
+</main><script src="/app.js"></script>
+</body></html>

static/style.css

@@ -0,0 +1 @@
+:root{font-family:Inter,-apple-system,BlinkMacSystemFont,"PingFang SC","Microsoft YaHei",sans-serif;color:#203b38;background:#f3f6f4;font-synthesis:none;font-size:14px}*{box-sizing:border-box}body{margin:0}aside{position:fixed;inset:0 auto 0 0;width:220px;background:#113e37;color:#f4fbf7;padding:32px 22px;display:flex;flex-direction:column}.brand{display:flex;align-items:center;gap:12px;font-size:20px;font-weight:650;letter-spacing:.5px}.mark{display:grid;place-items:center;background:#d8efb9;color:#17483e;width:42px;height:42px;border-radius:13px}.brand small{display:block;font-size:11px;font-weight:400;color:#b3cdc5;margin-top:5px}nav{display:grid;gap:9px;margin-top:54px}nav a{color:#bad1c9;text-decoration:none;padding:12px 16px;border-radius:8px}nav a:hover,nav .selected{background:#28534a;color:#fff}.aside-note{margin-top:auto;color:#bad1c9;font-size:12px;line-height:1.9}.dot{display:inline-block;width:8px;height:8px;border-radius:50%;background:#a3db80;margin-right:7px}.dot.gray{background:#a1ada7}main{max-width:1660px;margin-left:220px;padding:42px 40px}header{display:flex;align-items:center;justify-content:space-between;gap:20px}.eyebrow{font-size:10px;letter-spacing:2px;color:#6b8980;font-weight:700}h1{font-size:28px;letter-spacing:-.6px;margin:12px 0}header p{color:#73857e;margin:0;font-size:13px}button,.button,.file-button{font:inherit;font-weight:550;border:1px solid #1c6554;border-radius:7px;background:#1c6554;color:white;padding:10px 15px;cursor:pointer;white-space:nowrap;text-decoration:none;display:inline-block}button:hover,.button:hover{filter:brightness(.94)}button:disabled{opacity:.45;cursor:wait}.secondary,.file-button{background:white;color:#36574d;border-color:#d4dfd7}.danger{background:#fff8f6;border-color:#e7c5bb;color:#a34d37}.flow{display:flex;align-items:center;justify-content:space-between;gap:12px;margin-top:34px;background:white;border:1px solid #e2e9e2;padding:22px 25px;border-radius:12px;font-size:12px}.flow span{background:#edf3ef;color:#5f7d70;padding:5px 7px;border-radius:5px;font-size:10px;margin-right:8px}.flow i{color:#b1c0b6;font-style:normal}.flow .highlight{font-weight:650;color:#2e6947}.flow .highlight span{background:#e3efcc;color:#476635}.pending-note{background:#f8f2df;border:1px solid #eae0bd;color:#7b6938;padding:12px 16px;border-radius:8px;margin:16px 0;font-size:12px;line-height:1.8}.pending-note b{margin-right:10px}.metrics{display:grid;grid-template-columns:repeat(4,1fr);gap:16px;margin:22px 0}.metrics article{padding:20px 23px;background:white;border:1px solid #e0e8e0;border-radius:10px}.metrics small{font-size:12px;color:#6b7f73}.metrics strong{display:block;font-size:30px;font-weight:550;margin:12px 0 8px}.metrics span{font-size:11px;color:#8a988f}.workspace-grid{display:grid;grid-template-columns:1.1fr 1fr;gap:20px}.card{background:white;border:1px solid #e0e8e0;border-radius:12px;padding:24px;margin-bottom:20px;min-width:0}.section-head{display:flex;align-items:center;justify-content:space-between;gap:12px;margin-bottom:20px}h2{font-size:17px;margin:6px 0 0;font-weight:600}.badge{font-size:11px;border-radius:20px;background:#f0f5ef;color:#6c826d;padding:5px 10px}.row{display:flex;gap:12px;align-items:flex-end}.grow{flex:1;min-width:0}label{display:block;font-size:12px;color:#61756b;margin-bottom:14px}input,select{display:block;width:100%;font:inherit;color:#304a3f;padding:10px 12px;margin-top:7px;border:1px solid #dbe4dd;border-radius:6px;background:#fcfdfb;outline:none}input:focus,select:focus{border-color:#659981;box-shadow:0 0 0 3px #6599811a}input::placeholder{color:#a4afa8}.row button{margin-bottom:14px}.check{display:flex;align-items:flex-start;gap:8px;line-height:1.7}.check input{width:15px;height:15px;margin:2px 0 0;accent-color:#276b54}.muted{font-size:12px;line-height:1.8;color:#8b988e}.progress{margin:12px 0 20px}progress{appearance:none;width:100%;height:5px;border:0;border-radius:4px;display:block;background:#edf2ec}progress::-webkit-progress-bar{background:#edf2ec}progress::-webkit-progress-value{background:#76a77f;border-radius:4px}.progress small{display:block;color:#82917f;margin-top:8px;font-size:11px}summary{cursor:pointer;color:#617d6d;font-size:12px}details label{margin-top:12px}.runbar{display:flex;align-items:center;justify-content:space-between;border-radius:10px;background:#e5eee5;padding:19px 23px;margin-bottom:22px;gap:16px}.runbar b{font-weight:550}.runbar p{margin:7px 0 0 15px;font-size:12px;color:#758b76}.actions{display:flex;gap:10px;flex-wrap:wrap}table{border-collapse:collapse;width:100%;text-align:left;font-size:12px}th{color:#8a998f;font-weight:500;background:#f7f9f5;padding:12px}td{padding:16px 12px;border-bottom:1px solid #eef1eb;vertical-align:top;line-height:1.8}td:first-child{max-width:350px}td small{display:block;color:#91a093;font-size:11px}td button{font-size:11px;padding:4px 8px}td a{color:#1c6554}.table-wrap{overflow-x:auto}.empty{padding:44px 16px;text-align:center;color:#81927f}.empty span{display:block;foLine truncated

store.py

@@ -0,0 +1,294 @@
+"""Persistent serial queue. Browser submissions are always initiated by the user."""
+import json
+from contextlib import contextmanager
+from pathlib import Path
+import secrets
+import sqlite3
+import threading
+import time
+from urllib.parse import urlsplit
+
+from articles import inspect_article
+
+ACCEPTED = {'published', 'accepted_pending_moderation'}
+SUBMITTED = {'user_submitting', 'submit_unknown'}
+
+
+class Store:
+    def __init__(self, path, ledger=None):
+        self.path = str(path)
+        Path(path).parent.mkdir(parents=True, exist_ok=True)
+        self.lock = threading.RLock()
+        with self.db() as db:
+            db.executescript('''
+                CREATE TABLE IF NOT EXISTS jobs(id TEXT PRIMARY KEY, article_key TEXT UNIQUE, value TEXT NOT NULL);
+                CREATE TABLE IF NOT EXISTS ledger(key TEXT PRIMARY KEY, value TEXT NOT NULL);
+                CREATE TABLE IF NOT EXISTS settings(id INTEGER PRIMARY KEY CHECK(id=1), value TEXT NOT NULL);
+                CREATE TABLE IF NOT EXISTS posts(id TEXT PRIMARY KEY, job TEXT NOT NULL);
+            ''')
+            row = db.execute('SELECT value FROM settings WHERE id=1').fetchone()
+            self.run = json.loads(row[0]) if row else dict(running=False, epoch=0, config={}, batch=None, active=None, owner=None, switch=None, reason='请配置任务后开始')
+            self.run.update(running=False, epoch=self.run['epoch'] + 1, reason='服务已启动，请明确点击开始/继续；未决任务不会重发')
+            self._persist(db)
+        self.clients = {}
+        if ledger and Path(ledger).is_file():
+            self.import_history(json.loads(Path(ledger).read_text()).get('records', []))
+
+    @contextmanager
+    def db(self):
+        connection = sqlite3.connect(self.path, timeout=20)
+        try:
+            with connection:
+                yield connection
+        finally:
+            connection.close()
+
+    def _persist(self, db):
+        db.execute('INSERT OR REPLACE INTO settings VALUES(1,?)', (json.dumps(self.run, ensure_ascii=False),))
+
+    def _job(self, db, id):
+        row = db.execute('SELECT value FROM jobs WHERE id=?', (id,)).fetchone()
+        if not row:
+            raise ValueError('任务不存在')
+        return json.loads(row[0])
+
+    def _save(self, db, job):
+        job['updated_at'] = time.time()
+        db.execute('INSERT OR REPLACE INTO jobs VALUES(?,?,?)', (job['id'], job['article_key'], json.dumps(job, ensure_ascii=False)))
+
+    def import_history(self, records):
+        with self.lock, self.db() as db:
+            for record in records:
+                digest = record.get('markdown_sha256')
+                if digest and record.get('block_repost'):
+                    db.execute('INSERT OR IGNORE INTO ledger VALUES(?,?)', ('md:' + digest, json.dumps(record, ensure_ascii=False)))
+                    # Add content identity where the original historical file is still present.
+                    try:
+                        article = inspect_article(Path(record['markdown_path']).parent)
+                        db.execute('INSERT OR IGNORE INTO ledger VALUES(?,?)', ('article:' + article['article_key'], json.dumps(record, ensure_ascii=False)))
+                    except (OSError, ValueError, KeyError):
+                        pass
+
+    def enqueue(self, directory):
+        article = inspect_article(directory)
+        article.pop('package')
+        keys = ['article:' + article['article_key'], 'md:' + article['markdown_sha256']] + ['source:' + str(k) for k in article['source_keys']]
+        with self.lock, self.db() as db:
+            if any(db.execute('SELECT 1 FROM ledger WHERE key=?', (k,)).fetchone() for k in keys):
+                return '已受理，跳过'
+            if db.execute('SELECT 1 FROM jobs WHERE article_key=?', (article['article_key'],)).fetchone():
+                return '已在队列，跳过'
+            # Source URL identities prevent different renderings of one downloaded article re-entering.
+            for row in db.execute('SELECT value FROM jobs'):
+                if set(article['source_keys']) & set(json.loads(row[0]).get('source_keys', [])):
+                    return '相同来源已在队列，跳过'
+            article.update(id=secrets.token_hex(12), state='queued', created_at=time.time(), error='', post_url='')
+            self._save(db, article)
+            return '已入队'
+
+    def state(self):
+        with self.lock, self.db() as db:
+            jobs = [json.loads(r[0]) for r in db.execute('SELECT value FROM jobs ORDER BY rowid')]
+            return dict(run=json.loads(json.dumps(self.run)), jobs=jobs,
+                        clients=[v for v in self.clients.values() if time.time() - v['seen'] < 20],
+                        history_count=db.execute('SELECT count(*) FROM ledger').fetchone()[0])
+
+    def configure(self, config):
+        fid, minimum, maximum = config.get('fid'), config.get('minimum'), config.get('maximum')
+        interval = config.get('interval')
+        if any(type(x) is not int for x in (fid, minimum, maximum, interval)) or not (fid > 0 and 1 <= minimum <= maximum <= 1000 and 0 <= interval <= 86400):
+            raise ValueError('请填写有效版块 ID、篇数范围（1～1000）和间隔秒数')
+        if config.get('history_checked') is not True:
+            raise ValueError('请先核对待处理文章与历史发布记录，并勾选确认')
+        with self.lock, self.db() as db:
+            if self.run['active'] or self.run['switch'] or self.run['running']:
+                raise ValueError('当前流程未结束，不能更换配置')
+            self.run.update(config={k: config[k] for k in ('fid', 'minimum', 'maximum', 'interval', 'history_checked')}, batch=None)
+            self._persist(db)
+
+    def control(self, action):
+        with self.lock, self.db() as db:
+            if action == 'start':
+                if not self.run['config']:
+                    raise ValueError('请先保存配置')
+                self.run.update(running=True, reason='等待浏览器连接')
+            else:
+                self.run.update(running=False, reason='已终止' if action == 'stop' else '已暂停')
+            self.run['epoch'] += 1
+            self._persist(db)
+            return dict(self.run)
+
+    def _batch(self, account):
+        cfg = self.run['config']
+        return dict(account=account, target=cfg['minimum'] + secrets.randbelow(cfg['maximum'] - cfg['minimum'] + 1), count=0)
+
+    def _check_owner(self, owner):
+        if self.run['owner'] != owner:
+            raise ValueError('任务由另一个浏览器标签持有')
+
+    def heartbeat(self, data):
+        owner = str(data.get('owner', ''))[:100]
+        if not owner:
+            raise ValueError('缺少执行标签标识')
+        with self.lock:
+            self.clients[owner] = dict(owner=owner, account=data.get('account'), url=str(data.get('url', ''))[:1000],
+                                       extension=data.get('extension'), seen=time.time())
+            return self.state()['run']
+
+    def next(self, data):
+        owner, account = data.get('owner'), data.get('account')
+        with self.lock, self.db() as db:
+            if not self.run['running']:
+                return dict(action='idle', run=self.run)
+            if not account or not str(account.get('uid', '')).isdigit() or not account.get('name'):
+                return dict(action='login', run=self.run)
+            if self.run['owner'] not in (None, owner):
+                return dict(action='other_tab', run=self.run)
+            self.run['owner'] = owner
+            if self.run['switch']:
+                switch = self.run['switch']
+                if account['uid'] != switch['from'] and data.get('extension', {}).get('idle') is True:
+                    self.run.update(batch=self._batch(account), switch=None)
+                    self._persist(db)
+                else:
+                    return dict(action='switch' if switch['state'] == 'pending' else 'switch_wait', run=self.run)
+            if self.run['batch'] is None:
+                self.run['batch'] = self._batch(account)
+            if self.run['batch']['account']['uid'] != account['uid']:
+                self.run.update(running=False, reason='检测到账号变化，请先核对当前任务')
+                self._persist(db)
+                return dict(action='idle', run=self.run)
+            if self.run['active']:
+                job = self._job(db, self.run['active'])
+                return dict(action='resume', job=job, run=self.run)
+            rows = [json.loads(r[0]) for r in db.execute('SELECT value FROM jobs ORDER BY rowid')]
+            job = next((j for j in rows if j['state'] == 'queued'), None)
+            if not job:
+                self.run.update(running=False, reason='当前队列已处理完；新文章入队后请点击继续')
+                self._persist(db)
+                return dict(action='idle', run=self.run)
+            if time.time() < self.run.get('not_before', 0):
+                return dict(action='interval', run=self.run)
+            if self.run['batch']['count'] >= self.run['batch']['target']:
+                self.run['switch'] = dict(state='pending', **{'from': account['uid']}, created_at=time.time())
+                self._persist(db)
+                return dict(action='switch', run=self.run)
+            job.update(state='preparing', account=account, fid=self.run['config']['fid'], error='')
+            self.run['active'] = job['id']
+            self._save(db, job)
+            self._persist(db)
+            return dict(action='prepare', job=job, run=self.run)
+
+    def current(self, data):
+        with self.lock, self.db() as db:
+            self._check_owner(data['owner'])
+            return {'job': self._job(db, self.run['active']) if self.run['active'] else None, 'run': self.run}
+
+    def switch_begin(self, data):
+        with self.lock, self.db() as db:
+            self._check_owner(data['owner'])
+            if not self.run['running'] or not self.run['switch'] or self.run['switch']['state'] != 'pending':
+                raise ValueError('换号状态已改变，禁止重复点击')
+            self.run['switch']['state'] = 'clicked'
+            self._persist(db)
+            return {'ok': True}
+
+    def package(self, data):
+        with self.lock, self.db() as db:
+            self._check_owner(data['owner'])
+            if not self.run['running'] or self.run['active'] != data['id']:
+                raise ValueError('流程已停止或任务不匹配')
+            job = self._job(db, data['id'])
+            if job['state'] != 'preparing':
+                raise ValueError('当前任务不允许重新导入，请先核对已上传内容')
+        article = inspect_article(job['directory'])
+        if article['markdown_sha256'] != job['markdown_sha256'] or article['article_key'] != job['article_key']:
+            raise ValueError('入队后原文章已改变，请移出并重新扫描')
+        return article['package']
+
+    def event(self, data):
+        with self.lock, self.db() as db:
+            self._check_owner(data['owner'])
+            job = self._job(db, data['id'])
+            if self.run['active'] != job['id']:
+                raise ValueError('不是当前任务')
+            target = data['state']
+            transitions = {'preparing': {'importing', 'needs_attention'}, 'importing': {'formatting', 'needs_attention'},
+                           'formatting': {'awaiting_user_review', 'needs_attention'},
+                           'awaiting_user_review': {'user_submitting', 'needs_attention'},
+                           'user_submitting': {'submit_unknown', 'needs_attention'}, 'submit_unknown': {'needs_attention'}}
+            if target not in transitions.get(job['state'], set()):
+                raise ValueError('任务状态不允许此操作：' + job['state'] + ' → ' + target)
+            if target in ('importing', 'formatting', 'awaiting_user_review') and (not self.run['running'] or data.get('epoch') != self.run['epoch']):
+                raise ValueError('流程已暂停/终止，过期回调不能继续')
+            job.update(state=target, error=str(data.get('error', ''))[:1000])
+            if target == 'user_submitting':
+                job.update(submitted_at=time.time(), snapshot=data.get('snapshot', {}), attempt=secrets.token_hex(8))
+            if target == 'awaiting_user_review':
+                job['prepared'] = data.get('snapshot', {})
+            if target == 'needs_attention':
+                self.run.update(running=False, reason=job['error'] or '当前任务需要处理')
+                self._persist(db)
+            self._save(db, job)
+            return {'ok': True}
+
+    def accept(self, data):
+        with self.lock, self.db() as db:
+            self._check_owner(data['owner'])
+            job = self._job(db, data['id'])
+            if job['state'] in ACCEPTED:
+                return {'ok': True, 'duplicate': True}
+            if self.run['active'] != job['id'] or job['state'] not in SUBMITTED:
+                raise ValueError('只有已观察到人工提交的任务才能登记结果')
+            url = urlsplit(str(data.get('post_url', '')))
+            import re
+            match = re.fullmatch(r'/thread/(\d+)\.html', url.path)
+            if url.scheme != 'https' or url.netloc not in ('www.dww11.com', 'dww11.com') or not match:
+                raise ValueError('未识别到帖子详情页')
+            if data.get('author_uid') != job['account']['uid'] or data.get('fid') != job['fid']:
+                raise ValueError('结果页作者或版块与任务不符，请人工核对')
+            if data.get('title') != job.get('snapshot', {}).get('title'):
+                raise ValueError('结果页标题与人工提交快照不一致')
+            if data.get('state') not in ACCEPTED:
+                raise ValueError('需要明确的公开或审核中状态')
+            if not data.get('evidence'):
+                raise ValueError('缺少页面结果证据')
+            post_id = match[1]
+            existing = db.execute('SELECT job FROM posts WHERE id=?', (post_id,)).fetchone()
+            if existing and existing[0] != job['id']:
+                raise ValueError('帖子已绑定其他任务，不能重复计数')
+            db.execute('INSERT OR IGNORE INTO posts VALUES(?,?)', (post_id, job['id']))
+            job.update(state=data['state'], post_id=post_id, post_url=url.geturl(), result=data['evidence'])
+            keys = ['article:' + job['article_key'], 'md:' + job['markdown_sha256']] + ['source:' + str(k) for k in job['source_keys']]
+            for key in keys:
+                db.execute('INSERT OR REPLACE INTO ledger VALUES(?,?)', (key, json.dumps(job, ensure_ascii=False)))
+            self.run['batch']['count'] += 1
+            self.run.update(active=None, not_before=time.time() + self.run['config']['interval'])
+            self._save(db, job)
+            self._persist(db)
+            return {'ok': True}
+
+    def resolve(self, data):
+        """Explicit local UI resolution, never an automatic retry of an unknown submission."""
+        with self.lock, self.db() as db:
+            job = self._job(db, data['id'])
+            if job['state'] in ACCEPTED:
+                raise ValueError('已受理文章不能重置')
+            if self.run['running']:
+                raise ValueError('请先暂停流程')
+            if data.get('confirmed_not_accepted') is not True:
+                raise ValueError('必须先人工核对网站未受理，并清理编辑器/草稿')
+            job.update(state='queued' if data.get('retry') else 'cancelled', error='', snapshot={})
+            if self.run['active'] == job['id']:
+                self.run.update(active=None, owner=None)
+            self._save(db, job)
+            self._persist(db)
+            return {'ok': True}
+
+    def takeover(self, data):
+        with self.lock, self.db() as db:
+            if self.run['running']:
+                raise ValueError('请先在控制台暂停流程，再接管')
+            self.run['owner'] = data['owner']
+            self._persist(db)
+            return {'ok': True}

test/README.md

@@ -0,0 +1,12 @@
+# 测试与归档
+
+- `history/`：从 auto_news_upload 根目录移入的历史截图和演示任务 JSON，保留原内容。不是自动测试，不能用作本次测试结论。
+- `fixtures/article/`：自造 Markdown 和三张 1×1 PNG，其中两张内容相同，用于验证上传去重及插图顺序。
+- `test_queue.py`：持久队列、去重、重启、终止、未知提交和按篇数换号。
+- `test_controller.py`：本地接口权限、模拟下载到队列、静态页面路由。
+- `test_download_core.py`：复制的下载器原有离线测试，针对合并项目中的下载器副本运行。
+- `importer.test.cjs`：导入器回归与自动模式的停止/冲突保护。
+- `browser.test.cjs`：本地页面、移动布局与模拟编辑器；从不请求真实论坛。
+- `artifacts/`：本轮页面截图，属于生成产物。
+
+原始受理去重记录继续保存在 `../发布去重记录.json`，服务启动时自动导入；没有作为测试残留移走或删除。

test/bridge.test.cjs

@@ -0,0 +1,13 @@
+'use strict';
+const {test}=require('node:test');const assert=require('node:assert/strict');const fs=require('node:fs');const path=require('node:path');const vm=require('node:vm');
+test('终止后较早的 heartbeat 响应不能恢复流程',async()=>{
+ const handlers=new Map(),requests=[],responses=[],values={pairToken:'test-token',pairPort:8770};
+ class Event{constructor(type,options={}){this.type=type;this.detail=options.detail;}}
+ const doc={readyState:'loading',addEventListener(type,fn){handlers.set(type,fn);},dispatchEvent(event){if(event.type==='dww11:local-response')responses.push(JSON.parse(event.detail));else handlers.get(event.type)?.(event);}};
+ vm.runInNewContext(fs.readFileSync(path.join(__dirname,'../browser/local-bridge.user.js'),'utf8'),{document:doc,CustomEvent:Event,GM_getValue:(k,d)=>values[k]??d,GM_setValue:(k,v)=>values[k]=v,GM_addValueChangeListener:()=>{},GM_xmlhttpRequest:r=>requests.push(r)});
+ const call=(id,route)=>handlers.get('dww11:local-request')(new Event('',{detail:JSON.stringify({id,route})}));
+ const heartbeat=call('old','heartbeat');const stop=call('stop','stop');
+ requests[0].onload({status:200,responseText:JSON.stringify({running:true,epoch:1})});await heartbeat;
+ assert.equal(responses.find(r=>r.id==='old').value.running,false);assert.equal(values.halted,true);
+ requests[1].onload({status:200,responseText:JSON.stringify({running:false,epoch:2})});await stop;
+});

test/browser.test.cjs

@@ -0,0 +1,91 @@
+'use strict';
+const {chromium}=require('playwright');
+const fs=require('node:fs');
+const path=require('node:path');
+const assert=require('node:assert/strict');
+const ROOT=path.resolve(__dirname,'..');
+const output=path.join(__dirname,'artifacts');fs.mkdirSync(output,{recursive:true});
+(async()=>{
+ const browser=await chromium.launch({headless:true,executablePath:process.env.CHROME_PATH||'/Applications/Google Chrome.app/Contents/MacOS/Google Chrome'});
+ try{
+  const page=await browser.newPage({viewport:{width:1440,height:1100}}),errors=[];
+  page.on('pageerror',error=>errors.push(error.message));
+  await page.route('**/*',route=>new URL(route.request().url()).hostname==='127.0.0.1'?route.continue():route.abort());
+  await page.goto('http://127.0.0.1:8771/');
+  await page.getByText('原创导入测试稿',{exact:true}).waitFor();
+  await page.screenshot({path:path.join(output,'desktop.png'),fullPage:true});
+  await page.locator('#fid').fill('125');await page.locator('#minimum').fill('1');await page.locator('#maximum').fill('2');await page.locator('#interval').fill('10');await page.locator('#history-checked').check();
+  await page.getByRole('button',{name:'保存处理规则'}).click();await page.getByText('处理规则已保存',{exact:true}).waitFor();
+  await page.locator('#start').click();await page.getByText('流程运行中',{exact:true}).waitFor();
+  await page.locator('#stop').click();await page.getByText('流程已暂停',{exact:true}).waitFor();
+  await page.setViewportSize({width:390,height:844});await page.screenshot({path:path.join(output,'mobile.png'),fullPage:true});
+  assert.equal(await page.evaluate(()=>document.documentElement.scrollWidth>innerWidth),false,'mobile overflow');
+  assert.deepEqual(errors,[]);console.log('PASS local UI: configuration, start/stop, desktop/mobile layout');
+  await page.close();
+  const mock=await browser.newPage();
+  await mock.route('**/*',route=>route.fulfill({contentType:'text/html; charset=utf-8',body:'<!doctype html><a id="mumucms_username" href="/home.php?mod=space&uid=123">测试账号</a><input id="subject"><iframe id="e_iframe"></iframe><button id="postsubmit">发表帖子</button>'}));
+  await mock.goto('https://www.dww11.com/forum.php?mod=post&action=newthread&fid=125');
+  await mock.evaluate(()=>{
+   window.wysiwyg=1;window.events=[];window.clicks=0;window.formats=0;
+   window.job={id:'fixture',state:'preparing',title:'测试标题',source:'测试来源',fid:125,account:{uid:'123',name:'测试账号'}};
+   window.flow={running:true,epoch:1,active:'fixture',config:{fid:125}};
+   document.querySelector('#postsubmit').addEventListener('click',()=>window.clicks++);
+   window.DWW11Importer={stagePackage:async()=>({warnings:[]}),importPrepared:async()=>{
+    document.querySelector('#subject').value='测试标题';
+    document.querySelector('#e_iframe').contentDocument.body.innerHTML='<p>正文第一段</p><img id="attachimg_1" src="data:image/gif;base64,R0lGODlhAQABAIAAAAAAAP///yH5BAEAAAAALAAAAAABAAEAAAIBRAA7"><p>正文第二段</p><p>来源：测试来源，如有侵权，请联系删除。</p>';
+    return {images:{'a.gif':'1'}};
+   }};
+   document.addEventListener('dww11:local-request',event=>{
+    const {id,route,body}=JSON.parse(event.detail);let value={};
+    window.events.push({route,body});
+    if(route==='heartbeat'){window.flow.owner=body.owner;value=window.flow;}
+    if(route==='next')value={action:window.job.state==='preparing'?'prepare':'resume',job:window.job,run:window.flow};
+    if(route==='current')value={job:window.job,run:window.flow};
+    if(route==='event'){window.job.state=body.state;value={ok:true};}
+    if(route==='stop'){window.flow.running=false;value=window.flow;}
+    document.dispatchEvent(new CustomEvent('dww11:local-response',{detail:JSON.stringify({id,value})}));
+   });
+  });
+  await mock.addScriptTag({path:path.join(ROOT,'browser/formatter.js')});
+  await mock.addScriptTag({path:path.join(ROOT,'browser/runner.js')});
+  await mock.waitForFunction(()=>window.job.state==='awaiting_user_review');
+  await mock.waitForTimeout(2400);
+  assert.equal(await mock.evaluate(()=>window.clicks),0,'runner must never publish');
+  assert.equal(await mock.evaluate(()=>document.querySelector('#e_iframe').contentDocument.body.querySelectorAll('[data-dww11-wx-source]').length),1);
+  const before=await mock.evaluate(()=>document.querySelector('#e_iframe').contentDocument.body.innerHTML);
+  await mock.evaluate(()=>window.DWW11Formatter.format('测试来源'));
+  const after=await mock.evaluate(()=>document.querySelector('#e_iframe').contentDocument.body.innerHTML);
+  assert.equal(after,before,'repeated formatting should be idempotent');
+  await mock.locator('#postsubmit').click();
+  await mock.waitForFunction(()=>window.job.state==='user_submitting');
+  await mock.evaluate(()=>history.pushState({},'', '/thread/999.html'));
+  await mock.waitForFunction(()=>window.job.state==='submit_unknown');
+  assert.equal(await mock.evaluate(()=>window.flow.running),false);
+  assert.equal(await mock.evaluate(()=>window.events.some(e=>e.route==='accept')),false,'unknown page must not count accepted');
+  assert.equal(await mock.evaluate(()=>window.clicks),1);
+  console.log('PASS browser fixture: import/format, source uniqueness, repeated formatting, manual-only publish, unknown result stops');
+  await mock.close();
+  const ext=await browser.newPage();
+  await ext.route('**/*',route=>route.fulfill({contentType:'text/html; charset=utf-8',body:'<!doctype html><div class="topwp"><a href="#">立即登录</a></div>'}));
+  await ext.goto('https://www.dww11.com/');
+  await ext.evaluate(()=>{
+    window.routes=[];window.chrome.runtime={sendMessage:async message=>{
+      window.routes.push(message.route);
+      if(message.route==='/hello')return {ok:true,data:{total:1,active:1,round:1,remaining:1,flow:null}};
+      if(message.route==='/start')return new Promise(resolve=>window.finishStart=()=>resolve({ok:true,data:{flow:{id:'test-run',phase:'logout',paused:false}}}));
+      if(message.route==='/suspend')return {ok:true,data:{}};
+      throw new Error('Unexpected extension operation: '+message.route);
+    }};
+  });
+  await ext.addScriptTag({path:path.join(ROOT,'../dww11-account-switch-browser-extension/content.js')});
+  await ext.waitForFunction(()=>JSON.parse(document.querySelector('#dww11-extension-panel').dataset.autoNewsState).idle);
+  await ext.locator('#dww11-extension-next').click();
+  await ext.waitForFunction(()=>!!window.finishStart);
+  await ext.evaluate(()=>{document.dispatchEvent(new CustomEvent('dww11:auto-stop'));window.finishStart();});
+  await ext.waitForFunction(()=>window.routes.includes('/suspend'));
+  assert.equal(await ext.evaluate(()=>JSON.parse(document.querySelector('#dww11-extension-panel').dataset.autoNewsState).paused),true);
+  assert.deepEqual(await ext.evaluate(()=>window.routes),['/hello','/start','/suspend']);
+  console.log('PASS extension bridge: stop during pending start prevents login and persists pause');
+  await ext.close();
+ }finally{await browser.close();}
+})().catch(error=>{console.error(error);process.exitCode=1;});

test/fixtures/article/原创导入测试稿.md

@@ -0,0 +1,15 @@
+# 原创导入测试稿
+
+这是项目自有的离线测试稿，用于验证正文、段落和图片顺序。
+
+![](image001.png)
+
+第二段保留原文标点，以及普通文字。
+
+![](image002.png)
+
+最后一段说明图片上传不等于发表文章。
+
+![](image003.png)
+
+来源：项目测试素材，如有侵权，请联系删除。

test/importer.test.cjs

@@ -0,0 +1,172 @@
+// Offline verification only. No browser account, uploads or publication requests.
+'use strict';
+const { test } = require('node:test');
+const assert = require('node:assert/strict');
+const fs = require('node:fs');
+const path = require('node:path');
+const vm = require('node:vm');
+const { webcrypto, createHash } = require('node:crypto');
+const { execFileSync } = require('node:child_process');
+const core = require('../browser/importer.js');
+const directory = path.join(__dirname, 'fixtures/article');
+const mdName = '原创导入测试稿.md';
+function articleFiles() {
+    return fs.readdirSync(directory).map(name => {
+        const blob = new Blob([fs.readFileSync(path.join(directory, name))], { type: name.endsWith('.png') ? 'image/png' : 'text/markdown' });
+        blob.name = name; blob.webkitRelativePath = '文章/' + name; return blob;
+    });
+}
+
+test('测试文章：标题、3 张配图、插入顺序、来源与段落完整', async () => {
+    const article = await core.prepareFiles(articleFiles());
+    assert.equal(article.title, mdName.slice(0, -3));
+    assert.deepEqual([...article.images.keys()], ['文章/image001.png', '文章/image002.png', '文章/image003.png']);
+    const output = article.render(new Map([...article.images.keys()].map((p, i) => [p, String(101 + i)])));
+    assert.deepEqual([...output.matchAll(/\[attachimg\](\d+)\[\/attachimg\]/g)].map(m => m[1]), ['101', '102', '103']);
+    assert.equal(output.includes('# 原创导入测试稿'), false);
+    const md = fs.readFileSync(path.join(directory, mdName), 'utf8');
+    const originalBody = md.split('\n').slice(1).join('\n').trim().replace(/!\[\]\(image00(\d)\.png\)/g, (_, n) => `[attachimg]${100 + Number(n)}[/attachimg]`);
+    assert.equal(output, originalBody);
+});
+
+test('缺图、远程图片、超出目录、超大图片与超长标题在上传前被拒绝', () => {
+    assert.throws(() => core.compile('# 标题\n\n![](missing.jpg)', 'a.md', new Map()), /找不到配图/);
+    assert.throws(() => core.compile('# 标题\n\n![](https://x/a.jpg)', 'a.md', new Map()), /目录内/);
+    assert.throws(() => core.imagePath('a.md', '../secret.jpg'), /超出/);
+    assert.throws(() => core.imagePath('a.md', '%2fetc/passwd'), /相对路径/);
+    assert.throws(() => core.compile('# ' + '长'.repeat(81) + '\n正文', 'a.md', new Map()), /80/);
+    assert.throws(() => core.compile('# 标题\n\n![](big.png)', 'a.md', new Map([['big.png', {size:20*1024*1024}]])), /20 MB/);
+});
+
+test('子目录、Unicode 规范化、空格编码、括号、重复图片和格式转换', () => {
+    const fileMap = new Map([['目录/图 (1).jpg', new Blob(['image'])]]);
+    const article = core.compile('# 标题\n\n**粗体**和*斜体*\n\n![](图%20(1).jpg)\n![](<图 (1).jpg> "说明")\n\n[链接](https://example.com/a)\n\n- 一\n- 二\n\n> 引用', '目录/a.md', fileMap);
+    assert.equal(article.images.size, 1);
+    assert.match(article.render(new Map([['目录/图 (1).jpg', '100']])), /\[b\]粗体\[\/b\]/);
+    assert.equal(core.normalizePath('cafe\u0301/a.jpg'), 'café/a.jpg');
+    assert.equal(core.readLink("![](child's.jpg)", 0).target, "child's.jpg");
+    assert.throws(() => core.compile('# 标题\n![][a]\n[a]: x.jpg', 'a.md', fileMap), /引用式/);
+});
+
+test('多个 Markdown 必须选择；无 Markdown 不接受', async () => {
+    const files = articleFiles();
+    const extra = new Blob(['# 另一篇\n正文']); extra.name = 'b.md'; extra.webkitRelativePath = '文章/b.md';
+    const result = await core.prepareFiles([...files, extra]);
+    assert.equal(result.candidates.length, 2);
+    assert.equal((await core.prepareFiles([...files, extra], '文章/b.md')).title, '另一篇');
+    await assert.rejects(core.prepareFiles(files.filter(f => !f.name.endsWith('.md'))), /没有/);
+});
+
+test('下载程序对接：Python 文章包与目录直接导入完全一致', async () => {
+    const data = JSON.parse(execFileSync('python3', [path.join(__dirname, '../modules/package.py'), directory], {encoding:'utf8'}));
+    const packaged = await core.preparePackage(data);
+    const direct = await core.prepareFiles(articleFiles());
+    assert.equal(packaged.title, direct.title);
+    assert.equal(packaged.body, direct.body);
+    assert.equal(packaged.images.size, 3);
+    for (const [p, blob] of packaged.images) assert.equal(blob.size, fs.statSync(path.join(directory, p)).size);
+});
+
+test('服务器返回错误、登录 HTML 或混入脚本时不猜测附件编号', () => {
+    assert.equal(core.parseAid(' 123456\n'), '123456');
+    for (const value of ['-1', '0', '<html>123</html>', '<script>123</script>', '123error', '{"aid":123}']) assert.throws(() => core.parseAid(value));
+});
+
+// Small DOM harness exercises the public API and site adapter without installing browser software.
+function harness({ failUpload = false, loseImages = false, existing = false } = {}) {
+    const ids = new Map(), callbacks = new Map(), requests = [];
+    class Element {
+        constructor(id='') { this.id=id; this.style={}; this.value=''; this.disabled=false; this.checked=false; this.hidden=false; this.listeners=new Map(); this.children=[]; }
+        attachShadow() { return this; }
+        set innerHTML(value) { this._html=value; for (const m of value.matchAll(/\bid="([^"]+)"/g)) ids.set(m[1], new Element(m[1])); }
+        get innerHTML() { return this._html || ''; }
+        getElementById(id) { return ids.get(id); }
+        appendChild(el) { this.children.push(el); return el; }
+        replaceChildren(...els) { this.children=els; }
+        addEventListener(name, fn) { this.listeners.set(name, fn); }
+        removeEventListener(name) { this.listeners.delete(name); }
+        dispatchEvent() {}
+        click() { this.onclick?.(); }
+        querySelectorAll() { return []; }
+    }
+    for (const id of ['subject', 'e_textarea', 'e_switchercheck', 'postform']) ids.set(id, new Element(id));
+    ids.get('subject').value=existing ? '原标题' : '';
+    ids.get('e_textarea').value=existing ? '原正文' : '';
+    const doc = { body:new Element(), createElement: () => new Element(), getElementById:id=>ids.get(id),
+        querySelector: s=>ids.get(s.replace(/^#/,'')), querySelectorAll:()=>[] };
+    const win = { document:doc, location:{search:'?mod=post&action=newthread&fid=95',href:'https://www.dww11.com/forum.php?mod=post&action=newthread&fid=95',origin:'https://www.dww11.com'},
+        wysiwyg:0, allowbbcode:1, allowimgcode:1, crypto:webcrypto,
+        imgUpload:{settings:{upload_url:'https://www.dww11.com/misc.php?mod=swfupload&operation=upload&fid=95',post_params:{uid:'fake-user',hash:'fake-hash',type:'image'},file_size_limit:20480}},
+        confirm:()=>true, dispatchEvent:()=>{}, addEventListener:()=>{}, editdoc:{body:new Element()} };
+    win.getEditorContents=()=>win.wysiwyg ? win.editdoc.body.innerHTML : ids.get('e_textarea').value;
+    win.switchEditor=mode=>{
+        if (mode===win.wysiwyg) return;
+        if (mode===1) win.editdoc.body.innerHTML=ids.get('e_textarea').value.replace(/\[attachimg\](\d+)\[\/attachimg\]/g, (_,aid)=>loseImages ? '' : `<img aid="attachimg_${aid}">`);
+        else ids.get('e_textarea').value=win.editdoc.body.innerHTML;
+        win.wysiwyg=mode;
+    };
+    win.updateImageList=(_action,aids)=>{ for (const id of aids.split('|')) ids.set('image_'+id,new Element()); };
+    win.fetch=async (url,options)=>{
+        requests.push({url,options});
+        return {ok:true,text:async()=>failUpload ? '-1' : String(1000+requests.length)};
+    };
+    const context = {window:win,URL,URLSearchParams,Blob,FormData,AbortController,Uint8Array,Option:Element,Event:class {},CustomEvent:class {},
+        MutationObserver:class {observe(){} disconnect(){}},atob,setTimeout,clearTimeout,console};
+    vm.runInNewContext(fs.readFileSync(path.join(__dirname, '../browser/importer.js'),'utf8'), context);
+    return {api:win.DWW11Importer, win, ids, requests};
+}
+
+test('完整导入：暂存零请求、相同图片去重、3处附件映射、富文本结果、重试复用、没有发布请求', async () => {
+    const h=harness();
+    await h.api.stageFiles(articleFiles());
+    assert.equal(h.requests.length,0);
+    const result=await h.api.importPrepared();
+    assert.equal(Object.keys(result.images).length,3);
+    const uniqueImages=new Set(fs.readdirSync(directory).filter(p=>p.endsWith('.png')).map(p=>createHash('sha256').update(fs.readFileSync(path.join(directory,p))).digest('hex'))).size;
+    assert.equal(h.requests.length,uniqueImages);
+    assert.equal(h.ids.get('subject').value,mdName.slice(0,-3));
+    assert.match(h.win.getEditorContents(),/attachimg_1001/);
+    assert.equal(h.api.getState().busy,false);
+    assert.equal(h.api.getBackup().content,'');
+    await h.api.importPrepared();
+    assert.equal(h.requests.length,uniqueImages);
+    for (const req of h.requests) {
+        assert.equal(new URL(req.url).pathname,'/misc.php');
+        assert.equal(req.options.body.get('type'),'image');
+        assert.ok(req.options.body.get('Filedata') instanceof Blob);
+    }
+});
+
+test('上传失败保留原文；富文本丢图片时回滚；取消覆盖不上传', async () => {
+    for (const options of [{failUpload:true,existing:true},{loseImages:true,existing:true}]) {
+        const h=harness(options);
+        await h.api.stageFiles(articleFiles());
+        await assert.rejects(h.api.importPrepared());
+        assert.equal(h.ids.get('subject').value,'原标题');
+        assert.equal(h.win.getEditorContents(),'原正文');
+        assert.equal(h.api.getState().busy,false);
+    }
+    const h=harness({existing:true}); h.win.confirm=()=>false;
+    await h.api.stageFiles(articleFiles());
+    assert.equal((await h.api.importPrepared()).cancelled,true);
+    assert.equal(h.requests.length,0);
+});
+
+// New integration guards must not prompt or overwrite a user's draft.
+test('自动导入：非空编辑器停止，不弹窗、不上传', async () => {
+ const h=harness({existing:true});h.win.confirm=()=>{throw new Error('unexpected prompt');};
+ await h.api.stageFiles(articleFiles());
+ await assert.rejects(h.api.importPrepared({automatic:true}),/编辑器已有内容/);
+ assert.equal(h.requests.length,0);assert.equal(h.ids.get('subject').value,'原标题');
+});
+test('自动导入：预先终止不上传，不填写', async () => {
+ const h=harness();await h.api.stageFiles(articleFiles());const c=new AbortController();c.abort(new Error('stopped'));
+ await assert.rejects(h.api.importPrepared({automatic:true,signal:c.signal}),/stopped/);
+ assert.equal(h.requests.length,0);assert.equal(h.ids.get('subject').value,'');
+});
+test('自动导入：上传中终止不会填写正文或继续下一张', async () => {
+ const h=harness();await h.api.stageFiles(articleFiles());const c=new AbortController();
+ const original=h.win.fetch;h.win.fetch=async (...args)=>{const result=await original(...args);c.abort(new Error('stopped'));return result;};
+ await assert.rejects(h.api.importPrepared({automatic:true,signal:c.signal}),/stopped/);
+ assert.equal(h.requests.length,1);assert.equal(h.ids.get('subject').value,'');
+});

test/serve_fixture.py

@@ -0,0 +1,19 @@
+"""Temporary local UI fixture; never contacts the news API or forum."""
+from pathlib import Path
+import sys
+import tempfile
+from http.server import ThreadingHTTPServer
+ROOT = Path(__file__).resolve().parents[1]
+sys.path.insert(0, str(ROOT))
+from app import Controller, make_handler
+with tempfile.TemporaryDirectory() as temp:
+    controller = Controller(Path(temp) / 'data', Path(temp) / 'downloads', ledger=None)
+    controller.store.enqueue(ROOT / 'test/fixtures/article')
+    server = ThreadingHTTPServer(('127.0.0.1', 8771), make_handler(controller))
+    print('UI fixture listening on 8771', flush=True)
+    try:
+        server.serve_forever()
+    except KeyboardInterrupt:
+        pass
+    finally:
+        server.server_close()

test/test_controller.py

@@ -0,0 +1,75 @@
+import json
+from pathlib import Path
+import sys
+import tempfile
+import threading
+import unittest
+from http.server import ThreadingHTTPServer
+from urllib.request import Request, urlopen
+from urllib.error import HTTPError
+
+ROOT = Path(__file__).resolve().parents[1]
+sys.path.insert(0, str(ROOT))
+from app import Controller, make_handler
+
+
+class ControllerTests(unittest.TestCase):
+    def setUp(self):
+        self.tmp = tempfile.TemporaryDirectory()
+        self.root = Path(self.tmp.name)
+        self.controller = Controller(self.root / 'data', self.root / 'downloads', ledger=None)
+        self.server = ThreadingHTTPServer(('127.0.0.1', 0), make_handler(self.controller))
+        self.thread = threading.Thread(target=self.server.serve_forever, daemon=True)
+        self.thread.start()
+        self.url = 'http://127.0.0.1:' + str(self.server.server_port)
+
+    def tearDown(self):
+        self.server.shutdown();self.server.server_close();self.thread.join();self.tmp.cleanup()
+
+    def request(self, path, body=None, token=None, origin=None, host=None):
+        headers = {}
+        if token: headers['X-Session-Token'] = token
+        if origin: headers['Origin'] = origin
+        if host: headers['Host'] = host
+        if body is not None: headers['Content-Type'] = 'application/json'
+        request = Request(self.url + path, data=None if body is None else json.dumps(body).encode(), headers=headers)
+        with urlopen(request, timeout=5) as response:
+            return response.status, response.read()
+
+    def test_auth_and_origin_boundaries(self):
+        for path, kwargs in [('/api/state', {}), ('/api/state', {'token': self.controller.bridge_token}),
+                            ('/api/state', {'token': self.controller.ui_token, 'origin': 'https://example.com'}),
+                            ('/bridge/heartbeat', {'token': self.controller.ui_token})]:
+            with self.assertRaises(HTTPError) as error:
+                self.request(path, {}, **kwargs)
+            self.assertEqual(error.exception.code, 403)
+        with self.assertRaises(HTTPError):
+            self.request('/api/session', host='attacker.example')
+        code, body = self.request('/api/state', {}, token=self.controller.ui_token)
+        self.assertEqual(code, 200)
+        self.assertNotIn(self.controller.bridge_token, body.decode())
+
+    def test_static_and_unknown_routes(self):
+        self.assertIn('新闻工作台', self.request('/')[1].decode())
+        self.assertEqual(self.request('/browser/auto-news.user.js')[0], 200)
+        with self.assertRaises(HTTPError):
+            self.request('/../store.py')
+        with self.assertRaises(HTTPError):
+            self.request('/bridge/scan', {}, token=self.controller.bridge_token)
+
+    def test_download_to_queue_without_network(self):
+        self.controller.manager.requester = lambda *_: {'code': 1, 'data': {'list': [dict(title='模拟新闻', src='模拟来源', content='<p>模拟正文</p>', url='https://example.com/news')]}}
+        self.controller.ui('download', dict(key='fake-key', channel='模拟频道', target=1))
+        self.controller.manager.thread.join(5)
+        state = self.controller.ui('state', {})
+        self.assertEqual(state['download']['saved'], 1)
+        self.assertEqual(len(state['jobs']), 1)
+        self.assertEqual(state['jobs'][0]['state'], 'queued')
+        self.assertTrue(state['jobs'][0]['source_keys'])
+        self.assertNotIn('fake-key', json.dumps(state))
+        self.controller.scan(self.controller.manager.root)
+        self.assertEqual(len(self.controller.store.state()['jobs']), 1)
+
+
+if __name__ == '__main__':
+    unittest.main()

test/test_download_core.py

@@ -0,0 +1,165 @@
+import base64
+import json
+from pathlib import Path
+import sys
+import tempfile
+import threading
+import unittest
+from http.server import ThreadingHTTPServer
+
+import requests
+sys.path.insert(0, str(Path(__file__).resolve().parents[1]))
+from app import downloader
+Downloader, make_handler, fingerprints = downloader.Downloader, downloader.make_handler, downloader.fingerprints
+from article_writer import save_article
+
+
+def article(i):
+    return dict(title=f'新闻{i}', content=f'<p>第{i}篇新闻内容</p>', url=f'https://example.com/{i}', src='测试来源', time='2026-10-31')
+
+
+class DownloadTests(unittest.TestCase):
+    def setUp(self):
+        self.temp = tempfile.TemporaryDirectory()
+        self.root = Path(self.temp.name)
+
+    def tearDown(self):
+        self.temp.cleanup()
+
+    def run_job(self, manager, target):
+        manager.start('test-key', '头条', target)
+        manager.thread.join(10)
+        self.assertFalse(manager.thread.is_alive())
+        return manager.snapshot()
+
+    def test_format_and_local_image(self):
+        data = article(1)
+        image = 'data:image/png;base64,' + base64.b64encode(b'\x89PNG\r\n\x1a\nexample').decode()
+        data['content'] += f'<img src="{image}"><img src="{image}">'
+        result = save_article(data, self.root)
+        body = Path(result.markdown).read_text()
+        self.assertTrue(body.startswith('# 新闻1\n\n第1篇新闻内容'))
+        self.assertNotIn('发布时间：2026-10-31', body)
+        self.assertNotIn('\n---\n', body)
+        self.assertTrue(body.endswith('来源：测试来源，如有侵权，请联系删除。\n'))
+        self.assertEqual(len(result.images), 1)
+        self.assertEqual(body.count('image001.png'), 2)
+        self.assertNotIn('data:image', body)
+        second = save_article(data, self.root)
+        self.assertEqual(Path(second.markdown).name, '新闻1 (2).md')
+
+    def test_bad_image_cleans_partial(self):
+        data = article(1)
+        data['content'] += '<img src="data:image/png;base64,aW52YWxpZA==">'
+        with self.assertRaises(Exception):
+            save_article(data, self.root)
+        self.assertEqual(list(self.root.iterdir()), [])
+
+    def test_duplicate_then_fill_and_restart(self):
+        items = [article(1), article(1), article(2), article(3)]
+        calls = []
+        def request(key, endpoint, params):
+            calls.append(params.copy())
+            return {'code': 1, 'data': {'list': items[params['start']:params['start']+params['num']]}}
+        manager = Downloader(self.root, request)
+        state = self.run_job(manager, 3)
+        self.assertEqual((state['status'], state['saved'], state['skipped']), ('completed', 3, 1))
+        self.assertEqual([c['start'] for c in calls], [0, 3])
+        restarted = Downloader(self.root, request)
+        state = self.run_job(restarted, 3)
+        self.assertEqual((state['status'], state['saved'], state['skipped']), ('partial', 0, 4))
+        self.assertEqual(len(list(self.root.glob('*/*.md'))), 3)
+
+    def test_offset_limit_and_partial_progress(self):
+        offsets = []
+        def request(key, endpoint, params):
+            offsets.append(params['start'])
+            return {'code': 1, 'data': {'list': [article(i) for i in range(params['start'], params['start']+params['num'])]}}
+        state = self.run_job(Downloader(self.root, request), 500)
+        self.assertEqual(state['status'], 'partial')
+        self.assertEqual(state['saved'], 440)
+        self.assertEqual(max(offsets), 400)
+
+    def test_error_and_key_redaction(self):
+        def request(*args):
+            return {'code': 10001, 'msg': 'invalid test-key'}
+        manager = Downloader(self.root, request)
+        state = self.run_job(manager, 10)
+        self.assertEqual(state['status'], 'error')
+        self.assertNotIn('test-key', json.dumps(state))
+
+    def test_stop_and_concurrent_start(self):
+        entered, release = threading.Event(), threading.Event()
+        def request(*args):
+            entered.set()
+            release.wait(2)
+            return {'code': 1, 'data': {'list': [article(1)]}}
+        manager = Downloader(self.root, request)
+        manager.start('key', '头条', 10)
+        self.assertTrue(entered.wait(2))
+        with self.assertRaises(ValueError):
+            manager.start('key', '头条', 10)
+        manager.stop.set()
+        release.set()
+        manager.thread.join(3)
+        self.assertEqual(manager.snapshot()['status'], 'cancelled')
+        self.assertEqual(manager.snapshot()['saved'], 0)
+
+    def test_failed_save_can_retry_next_job(self):
+        attempts = []
+        def writer(data, root):
+            attempts.append(1)
+            if len(attempts) == 1:
+                raise ValueError('模拟图片失败')
+            return save_article(data, root)
+        def request(key, endpoint, params):
+            return {'code': 1, 'data': {'list': [article(1)] if params['start'] == 0 else []}}
+        manager = Downloader(self.root, request, writer)
+        self.assertEqual(self.run_job(manager, 1)['failed'], 1)
+        self.assertEqual(self.run_job(manager, 1)['saved'], 1)
+
+    def test_fingerprint_content_and_tracking(self):
+        a, b = article(1), article(2)
+        b['content'] = a['content'].replace('<p>', '<div>').replace('</p>', '</div>')
+        self.assertTrue(set(fingerprints(a)) & set(fingerprints(b)))
+        b['content'] = '其他内容'
+        b['url'] = a['url'] + '?utm_source=abc#fragment'
+        self.assertTrue(set(fingerprints(a)) & set(fingerprints(b)))
+
+    def test_log_cap(self):
+        manager = Downloader(self.root)
+        for i in range(400):
+            manager.log(str(i))
+        self.assertEqual(len(manager.snapshot()['logs']), 300)
+        self.assertEqual(manager.snapshot()['logs'][0]['message'], '100')
+
+    def test_http_routes_and_access(self):
+        def request(key, endpoint, params=None):
+            return {'code': 1, 'data': {'list': ['头条', '科技']}}
+        manager = Downloader(self.root, request)
+        server = ThreadingHTTPServer(('127.0.0.1', 0), make_handler(manager, 'token'))
+        worker = threading.Thread(target=server.serve_forever, daemon=True)
+        worker.start()
+        url = f'http://127.0.0.1:{server.server_port}'
+        try:
+            self.assertIn('新闻下载器', requests.get(url).text)
+            self.assertEqual(requests.get(url+'/api/state').json()['status'], 'idle')
+            self.assertEqual(requests.post(url+'/api/channels', json={'key': 'test'}).status_code, 403)
+            self.assertEqual(requests.get(url, headers={'Host':'evil.example'}).status_code, 403)
+            headers = {'X-Session-Token': 'token'}
+            self.assertEqual(requests.post(url+'/api/channels', headers=headers, json={'key':'test'}).json()['channels'], ['头条','科技'])
+            invalid = requests.post(url+'/api/start', headers=headers, json={'key':'test', 'target':0,'channel':'头条'})
+            self.assertEqual(invalid.status_code, 400)
+            invalid = requests.post(url+'/api/debug', headers=headers, json={'key':'test','endpoint':'index','num':41,'start':0})
+            self.assertEqual(invalid.status_code, 400)
+            debug = requests.post(url+'/api/debug', headers=headers, json={'key':'test','endpoint':'channel'})
+            self.assertEqual(debug.status_code, 200)
+            self.assertNotIn('key=test', debug.text)
+        finally:
+            server.shutdown()
+            server.server_close()
+            worker.join()
+
+
+if __name__ == '__main__':
+    unittest.main()

test/test_queue.py

@@ -0,0 +1,155 @@
+import hashlib
+import json
+from pathlib import Path
+import shutil
+import sys
+import tempfile
+import unittest
+
+ROOT = Path(__file__).resolve().parents[1]
+sys.path.insert(0, str(ROOT))
+from articles import inspect_article, short_title
+from store import Store
+
+
+class QueueTests(unittest.TestCase):
+    def setUp(self):
+        self.tmp = tempfile.TemporaryDirectory()
+        self.root = Path(self.tmp.name)
+        self.article = self.root / 'article'
+        self.article.mkdir()
+        self.md = self.article / 'a.md'
+        self.md.write_text('# 测试新闻\n\n正文内容。\n\n来源：测试来源，如有侵权，请联系删除。\n')
+        self.store = Store(self.root / 'queue.sqlite3')
+        self.store.configure(dict(fid=125, minimum=1, maximum=1, interval=0, history_checked=True))
+        self.owner = dict(owner='tab-a', account=dict(uid='123', name='测试账号'), extension={'idle': True})
+
+    def tearDown(self):
+        self.tmp.cleanup()
+
+    def begin(self):
+        self.store.enqueue(self.article)
+        self.store.control('start')
+        return self.store.next(self.owner)['job']
+
+    def transition(self, job, state, **extra):
+        return self.store.event(dict(owner='tab-a', id=job['id'], state=state, epoch=self.store.run['epoch'], **extra))
+
+    def submit(self, job):
+        for state in ('importing', 'formatting', 'awaiting_user_review'):
+            self.transition(job, state)
+        self.transition(job, 'user_submitting', snapshot={'title': job['title']})
+
+    def accept(self, job):
+        return self.store.accept(dict(owner='tab-a', id=job['id'], post_url='https://www.dww11.com/thread/12345.html',
+            author_uid='123', fid=125, title=job['title'], state='accepted_pending_moderation', evidence={'test': 'fixture only'}))
+
+    def test_stable_identity_and_original_preserved(self):
+        before = self.md.read_bytes()
+        item = inspect_article(self.article)
+        copy = self.root / 'renamed'
+        shutil.copytree(self.article, copy)
+        (copy / 'a.md').write_text(self.md.read_text().replace('测试新闻', '改标题'))
+        self.assertEqual(inspect_article(copy)['article_key'], item['article_key'])
+        self.assertEqual(self.md.read_bytes(), before)
+        self.assertEqual(item['source'], '测试来源')
+
+    def test_title_does_not_break_emoji(self):
+        short = short_title('😀' * 50)
+        self.assertTrue(short.endswith('...'))
+        self.assertLessEqual(len(short.encode('utf-16-le')) // 2, 80)
+        self.assertEqual(short_title('短标题'), '短标题')
+
+    def test_unsupported_images_stop_before_queue(self):
+        (self.article / 'a.webp').write_bytes(b'webp')
+        with self.assertRaisesRegex(ValueError, '暂不支持'):
+            self.store.enqueue(self.article)
+        self.assertEqual(self.store.state()['jobs'], [])
+
+    def test_missing_and_outside_image_rejected(self):
+        for image in ('missing.png', '../outside.png', 'https://example.com/a.png'):
+            self.md.write_text('# 标题\n\n正文 ![](' + image + ')\n\n来源：测试，如有侵权，请联系删除。')
+            with self.assertRaises(ValueError):
+                self.store.enqueue(self.article)
+
+    def test_no_duplicate_queue_or_cross_tab_claim(self):
+        job = self.begin()
+        self.assertIn('已在队列', self.store.enqueue(self.article))
+        self.assertEqual(self.store.next({**self.owner, 'owner': 'tab-b'})['action'], 'other_tab')
+        self.assertEqual(self.store.next(self.owner)['job']['id'], job['id'])
+
+    def test_stop_invalidates_old_continuation(self):
+        job = self.begin()
+        epoch = self.store.run['epoch']
+        self.store.control('stop')
+        self.store.control('start')
+        with self.assertRaisesRegex(ValueError, '过期'):
+            self.store.event(dict(owner='tab-a', id=job['id'], state='importing', epoch=epoch))
+
+    def test_restart_pauses_and_preserves_claim(self):
+        job = self.begin()
+        restarted = Store(self.root / 'queue.sqlite3')
+        self.assertFalse(restarted.run['running'])
+        self.assertEqual(restarted.run['active'], job['id'])
+        self.assertEqual(restarted.run['batch']['target'], 1)
+
+    def test_accept_requires_manual_submission(self):
+        job = self.begin()
+        with self.assertRaisesRegex(ValueError, '人工提交'):
+            self.accept(job)
+
+    def test_accept_is_atomic_and_idempotent(self):
+        job = self.begin()
+        self.submit(job)
+        self.accept(job)
+        self.accept(job)
+        self.assertEqual(self.store.run['batch']['count'], 1)
+        self.assertIn('已受理', self.store.enqueue(self.article))
+        with self.assertRaisesRegex(ValueError, '已受理'):
+            self.store.resolve(dict(id=job['id'], confirmed_not_accepted=True, retry=True))
+
+    def test_unknown_never_requeues_automatically(self):
+        job = self.begin()
+        self.submit(job)
+        self.transition(job, 'submit_unknown')
+        self.assertEqual(self.store.next(self.owner)['job']['state'], 'submit_unknown')
+        self.assertEqual(self.store.run['batch']['count'], 0)
+        self.assertIsNone(self.store.run['switch'])
+
+    def test_migrate_real_ledger_shape(self):
+        record = dict(markdown_sha256=hashlib.sha256(self.md.read_bytes()).hexdigest(), block_repost=True,
+                      markdown_path=str(self.md), status='accepted_pending_moderation', post_id='12345')
+        self.store.import_history([record])
+        self.assertIn('已受理', self.store.enqueue(self.article))
+
+    def test_switch_only_after_target_and_with_more_work(self):
+        job = self.begin()
+        self.submit(job)
+        self.accept(job)
+        other = self.root / 'other';shutil.copytree(self.article, other)
+        (other / 'a.md').write_text(self.md.read_text().replace('正文内容', '另一篇内容'))
+        self.store.enqueue(other)
+        self.assertEqual(self.store.next(self.owner)['action'], 'switch')
+        self.assertEqual(self.store.next(self.owner)['action'], 'switch')
+        self.store.switch_begin({'owner': 'tab-a'})
+        with self.assertRaisesRegex(ValueError, '禁止重复'):
+            self.store.switch_begin({'owner': 'tab-a'})
+        self.assertEqual(self.store.next(self.owner)['action'], 'switch_wait')
+        next_job = self.store.next({**self.owner, 'account': {'uid': '456', 'name': '新测试账号'}})
+        self.assertEqual(next_job['action'], 'prepare')
+        self.assertEqual(self.store.run['batch']['count'], 0)
+
+    def test_no_switch_when_queue_done(self):
+        job = self.begin();self.submit(job);self.accept(job)
+        self.assertEqual(self.store.next(self.owner)['action'], 'idle')
+        self.assertIsNone(self.store.run['switch'])
+
+    def test_bad_result_does_not_increment(self):
+        job = self.begin();self.submit(job)
+        with self.assertRaises(ValueError):
+            self.store.accept(dict(owner='tab-a', id=job['id'], post_url='https://example.com/thread/12345.html'))
+        self.assertEqual(self.store.run['batch']['count'], 0)
+
+
+if __name__ == '__main__':
+    unittest.main()

test/验证记录.md

@@ -0,0 +1,22 @@
+# 本轮验证记录
+
+日期：2026-10-31。测试均未使用真实账号，未向新闻 API 请求内容，未向论坛上传或发布。
+
+| 范围 | 结果 |
+| --- | --- |
+| Python 下载器、队列与 HTTP 接口 | 27/27 通过 |
+| JavaScript 导入器与停止/覆盖保护 | 11/11 通过 |
+| 连接桥旧响应与终止竞态 | 1/1 通过 |
+| 本地前端 | 配置保存、开始/停止、桌面与手机布局通过，无页面脚本错误、无手机横向溢出 |
+| 模拟编辑器 | 导入→排版→待人工审核；来源唯一、重复排版保持一致；程序无自动发帖点击 |
+| 模拟人工发表与未知结果 | 真实测试点击被记录；候选结果页未适配时暂停，不调用受理接口 |
+| 模拟换号插件 | 启动请求尚未返回时终止，返回后调用暂停，不执行登录；状态明确 paused |
+| JavaScript 语法 | 合并脚本、连接脚本、更新后的插件 content.js 通过 |
+
+浏览器使用本机 Chrome 的独立临时配置。对论坛域名的所有请求均被测试 HTML 拦截。后端使用临时目录和 SQLite，不写入正式任务库。
+
+页面截图：`artifacts/desktop.png`、`artifacts/mobile.png`。
+
+测试中曾发现模拟网页未声明 UTF-8，导致中文登录入口无法匹配；已修正测试响应编码并重新通过。真实网页编码未由此假定。
+
+尚未验证：用户浏览器中油猴本地通信权限、真实网站当前的导入与附件转换、真实换号账号链路、结果页受理/审核/公开状态。按用户决定，结果页 HTML 待提供，自动核对暂不实现；因此当前不宣称全流程批量处理已完成在线验收。

发布流程操作记录与自动化规范.md

@@ -0,0 +1,167 @@
+# 广谈大湾网：下载、导入排版、人工发布与换号
+
+本文中的帐号名称和用户编号已匿名化；真实记录保存在本地排除目录中。
+
+记录日期：2026-10-31（北京时间）。本文依据用户指导、Chrome 实际操作和现有源码检查编写。它是后续实现的流程依据；需求不等于功能已经实现。
+
+## 1. 不可改变的边界
+
+- **“发表帖子”只能由用户亲手点击。** 自动化只下载、选择、上传、填写、排版、校验、提示审核、观察提交结果和记录计数。禁止模拟点击发布、调用提交函数、直接发送发帖请求或通过快捷键提交。
+- 标题被截断、正文或来源被修改后，都必须重新进入人工审核。发生标题过长错误后，“重新尝试发表”也由用户再次点击。
+- 用户决定分区和文章根目录；“理财财经”只是本次例子，不能成为所有文章的固定分区。
+- 全流程页面常驻“终止流程”按钮。点击后立即取消后续动作、重试、队列领取和自动换号，不自动恢复。已经到达网站的请求不能撤回，需保留记录并核对。
+- 已发表或已被网站接收待审核的文章不能再次排入发布队列，换账号也不能绕过去重。
+
+## 2. 本次演示记录
+
+| 项目 | 结果 |
+| --- | --- |
+| 站点 | https://www.dww11.com/ |
+| 当前账号 | 示例用户，页面用户编号 <已匿名> |
+| 分区 | 理财财经，fid=125 |
+| 入口 | 首页右上角“免费发布信息”→ 分区页“理财财经” |
+| 分区页 | https://www.dww11.com/special/post/ |
+| 新帖页 | https://www.dww11.com/forum.php?mod=post&action=newthread&fid=125 |
+| 标签行为 | 点击分区后新开标签，原分区页保留 |
+| 测试标题 | 刚靠新电影赚23亿票房，周星驰就卖掉内地影院 |
+| 原标题长度 | 22 个 Unicode 码点；本篇无需截断。网站计数方式尚未实测 |
+| 目录 | `<项目根目录>/tanshuapi_news_downloader/downloads/刚靠新电影赚23亿票房，周星驰就卖掉内地影院` |
+| 目录内容 | 1 个 Markdown + 26 张 JPG，共 27 个文件 |
+| 来源 | 洲洲影视娱评（取文章末尾来源，不取正文“信息来源”截图） |
+| 导入 | 已完成；面板提示“已填写标题、正文和 26 张配图” |
+| 排版 | 已完成，正文 4 号黑色，尾注 1 号银色 |
+| 核对 | 编辑器 DOM 中图片 26 张，最终来源尾注仅 1 条 |
+| 发布 | 用户已亲手点击；网站已受理，详情页显示“审核中” |
+| 帖子链接 | https://www.dww11.com/thread/23697411.html |
+| 帖子标识 | 主题 ID 23697411；首帖 PID 238766 |
+| 数量 | 已明确受理 1 篇；已确认审核通过并公开 0 篇 |
+| 发布后图片 | 正文 21 张 + 附件区 5 张，共 26 张 |
+| 换号 | 未执行；本次未设置最小值和最大值 |
+
+本次页面没有“终止流程”按钮。用户明确同意本次演示暂用浏览器控制条停止，常驻页面按钮仍是后续自动化的必备条件。该例外只适用于本次演示。
+
+本次已验证导入、排版、用户亲手发表以及网站受理后的帖子详情页；页面明确显示“审核中”。尚未验证审核通过后的访客可见性、标题超限错误或换号链路。用户称发表后自动回首页，但本次实际读取的原编辑器标签已导航至帖子详情页，不能将“提交后必然自动回首页”写成网站行为。应在结果登记后主动回首页并核对。
+
+## 3. 用户操作步骤与自动化对应
+
+1. 启动前让用户选择分区、下载目录、待处理文章范围，并填写每个账号发布数量的最小值和最大值。展示当前账号。先初始化常驻终止按钮与持久任务记录。
+2. 从下载器接收完整落盘的文章目录。检查正文、图片引用、来源和去重记录；保存中或缺文件的目录不进入导入队列。
+3. 点击“免费发布信息”，展示实际可用分区，由用户选择。等待新标签打开并核对分区名称及 fid；不能继续操作旧标签或仅凭坐标判断。
+4. 在右下角“本地文章导入 · 1.0”选择“选择文章目录”。日常演示选择单篇目录，目录应包含 Markdown 和全部配图；不要只选 `.md`。如果选择根目录导致出现多个 Markdown，必须明确选中哪一篇。
+5. 在系统文件选择器进入用户指定根目录，再选择对应文章文件夹。系统“Upload”是把所选文件提供给页面读取，不是发表帖子；浏览器另有目录文件数量确认。
+6. 阅读导入面板标题、正文预览、图片数量和告警。点击“上传配图并填写”后才开始向网站上传正文引用的图片。等待明确完成提示，不能用固定等待时间代替完成判断。
+7. 读取文章最后的标准尾注 `来源：名称，如有侵权，请联系删除。`，只取名称。本篇是“洲洲影视娱评”；用户先前茅台文章例子是“酒业内参”，不能硬编码。
+8. 点击编辑工具栏绿色“一键排版”（不是网站自带“自动排版”，也不是“一键导入微信公众号并排版”）。实际弹窗要求输入**来源公众号名称**，并非标题。本篇填“洲洲影视娱评”，确认后会提示排版完成。
+9. 标题由本地导入器填入页面标题框；仍需核对完整标题。当前“一键排版”不弹标题输入框。如果用户说“输入标题”，应在标题框填写和确认，而不是填到来源弹窗。
+10. 校验标题、正文、图片数量和顺序、尾注唯一性、账号和分区。保留用户原文；本篇的发布时间、“打开网易新闻 查看精彩图片”等原始文字未擅自清理。需要清理时应成为用户可见的规则。
+11. 页面提示“已完成，等待人工审核，请自行点击发表帖子”。自动化停止修改编辑器，保留终止按钮，等待用户操作；不自动保存草稿或发表。网站原有自动保存仍可能运行。
+12. 用户自行点击“发表帖子”。自动化只观察结果。若标题超长，按第 5 节修正，重新提示审核，等待用户再次点击。
+13. 收集帖子链接/编号、作者、分区和结果证据后记账。审核中单独记录并阻止重发；结果不明时先核对，不能自动重试发布。
+14. 已明确成功后先持久保存发布记录与计数，再回到首页。达到当前账号目标篇数且没有未决提交时，点击页面顶部“登陆下一帐号”，核对登录成功后再处理下一篇。
+
+## 4. 页面适配依据与当前缺口
+
+以下选择器来自本次可见页面或本地源码，未来运行时须重新检查，不能保证网站改版后仍成立。
+
+| 控件 | 当前依据 | 注意事项 |
+| --- | --- | --- |
+| 当前用户 | `#mumucms_username` | 同时读取页面用户名与用户链接编号 |
+| 下一账号 | `#dww11-extension-next` | 完成记录后才允许换号；核对成功，不以按钮点击作为成功 |
+| 标题 | `#subject` | 本次读取到 maxLength=-1，不能只依赖 HTML maxlength |
+| 可视化正文 | `iframe#e_iframe` 内的 body | 图文与尾注检查应在 iframe 内进行 |
+| 纯文本开关 | `#e_switchercheck` | 不将 Markdown 直接当网站 BBCode |
+| 一键排版 | `#dww11-format-oneclick` | 原生 prompt 输入来源，随后 alert 提示完成 |
+| 发表 | `#postsubmit` | 仅用于识别状态和观察人工提交，禁止自动触发 |
+| 导入面板 | 油猴脚本面板，包含 choose/import 等控件 | 面板可能位于 Shadow DOM；普通页面选择器未必能定位，优先按可见控件操作 |
+
+本次实际遇到的兼容问题：
+
+- 浏览器连接器直接设置目录文件被“允许访问文件网址”权限阻止。本次没有更改扩展权限，改用 Chrome 原生系统选择器完成。后续实现应优先复用本地文章包接口，减少对文件对话框的依赖。
+- macOS 路径含中文，逐键输入曾丢失中文；通过文本框完整赋值成功定位。输入后必须核对完整路径，不能只看末级数字或英文。
+- 导入器能完成 26 张图片上传，并临时禁用发表按钮；上传完成后恢复。上传过程中不允许换号。
+- **来源重复：** 一键排版保留了导入正文中的原尾注，又追加银色尾注。本次在编辑器中删除原来的那一条，保留银色尾注，复查最终只出现一次。后续应在排版适配器中精确处理末尾标准尾注，不能全局删除正文中的“来源”。
+- 页面“还可输入 80 个字符”在导入后仍显示 80，不能将这个显示值当作可靠的标题长度校验结果。
+- 页面有“上次未提交成功的数据”恢复提示；本次没有点击恢复，避免把旧稿混入新稿。后续遇到已有正文或恢复冲突必须暂停处理。
+
+## 5. 标题超过 80 字符
+
+保留 `original_title`，派生 `submitted_title`。超过站点限制时截取前缀并追加字面量 `...`，总长不超过 80；即长度允许时保留前 77 个字符再加 3 个点。示例中的“标题文本...”表示实际标题前缀和省略号，不额外加入中文括号。
+
+当前站点的“字符”计数方式尚未通过超长标题实测。实现时兼容中文和 emoji，不能切断代理对；先按保守规则校验，再根据站点明确错误调整。若站点按字节计数，则按其实际字节预算截取，不能反复盲目减少。
+
+现有导入器在读取阶段对超过 80 个 Unicode 码点的标题直接报错。后续需把截断放在导入前，否则到不了人工审核阶段。截断不覆盖原 Markdown 标题，也不改变文章去重身份。
+
+如果用户点击发表后才收到标题超限提示：确认网站明确未接受 → 修正标题 → 显示新旧标题 → 重新人工审核 → 用户再次点击。任何自动修正都不等于用户已批准发表。
+
+## 6. 发布记录、去重与恢复
+
+建议独立 SQLite 发布任务库，与下载去重库分开；下载成功不能视为发布成功。本次先保存演示任务 JSON，尚未实现自动数据库或网站历史帖子同步。
+
+稳定文章身份优先采用下载器原始文章 ID 或规范化源 URL，同时保存正文与图片内容指纹用于发现重新下载/改名副本。禁止只按本地目录名、截断标题或当前账号去重。默认整个站点跨账号、跨版块不重复发表同一篇；如以后需要重发，必须由用户显式决定。
+
+建议字段：article_key、source_url/source_id、content_hash、原始/最终标题、目录、来源、状态、账号 UID、fid、附件映射、排版版本、人工提交时间、提交尝试编号、帖子 ID/URL、网站受理/审核/公开状态、错误、创建和更新时间。日志不保存密码、Cookie 或上传凭证。
+
+主状态：
+
+```text
+downloaded → queued → importing → formatting → awaiting_user_review
+  → user_submitting → published / accepted_pending_moderation
+```
+
+分支：`stopped`、`needs_attention`、`submit_unknown`、`rejected`。未决提交不得重新上传或领取同一文章。单批次串行，跨标签领取加持久锁；刷新后先恢复状态，不自动启动。
+
+去重记录需要覆盖用户已发表帖子：优先导入原有记录，或从“修改/删除信息”进入本人帖子列表进行核对。仅标题相同只能标记疑似重复，不能据此自动认定是同一正文。本次未扫描历史帖子，后续批量运行前必须完成历史对账；本次文章已记为“网站已受理、审核中”，已有帖子 ID 与 URL，必须阻止重发；不能误写为等待用户点击，也不能写为审核通过。
+
+成功记录与计数使用同一个事务：帖子唯一键（站点+帖子 ID）防止页面刷新重复加一。先记录再返回首页。遇到网络超时、页面关闭或不明确提示，记录 `submit_unknown`，人工/页面核对后再决定，禁止自动重复发表。
+
+网站明确受理待审核时，记 `accepted_pending_moderation` 并锁住文章。本规范建议把明确受理计入账号本轮提交数量一次，另存公开发表数量；后续审核通过只改变公开状态，不再加本轮数量。本次已观察到网站审核中状态，因此本地记录明确受理数为 1、已确认公开数为 0；批量自动记账尚未实现。驳回不自动补发或重投。
+
+## 7. 每个账号随机篇数与换号
+
+用户填写整数 `min`、`max`，要求 `1 ≤ min ≤ max`。首次为当前账号创建发布批次及每次确认换号成功后，从闭区间 `[min,max]` 均匀抽取整数 n，保存目标和当前计数。min=max 时固定该数量。不要按每篇重新抽 n，刷新、暂停或重启也不重抽。
+
+- 本轮计数只含当前发布批次明确成功/明确受理的唯一文章，不含上传、排版、等待人工审核、失败和结果不明。
+- 达到 n 后先保存记录与 `switch_pending`，回首页，核对没有待提交草稿，再点击“登陆下一帐号”。
+- 观察并确认实际 UID/用户名变化且登录成功，再建立下一账号批次、重置批次计数并抽新 n。不可在不确定结果时重复点击换号。
+- 验证码、账号异常、频率限制、登录失败或账号池无可用账号时暂停，保留当前文章与计数；不把受限文章换账号重投。
+- 用户手动改变账号时立即暂停，重新关联批次与上传附件；不能沿用旧账号附件上下文。
+- 队列已经处理完时直接结束，不为“凑满 n”重复文章，也不无意义换号。
+
+本次没有给定 min/max，所以不随机设置目标，也不执行换号。账号扩展现有自己的抽号/登录持久记录，应复用其登录流程，不另存账号密码。
+
+## 8. 常驻终止按钮的实现要求
+
+按钮在首页、分区页、新帖页、结果页、换号页面保持可见，不依赖折叠的导入面板。跨标签共享同一批次停止状态，导航后先恢复按钮和停止状态再执行任务。
+
+点击立即写入持久 `stopped` 并提升运行版本号，取消定时器、队列和可取消的上传/下载请求；每个异步步骤返回时检查停止标志和版本，过期回调不得再填写、跳转或换号。浏览器自带模态弹窗会阻挡网页按钮，因此自动流程应把来源/审核提示改成页面内面板，保留停止入口。
+
+已被服务器接收的图片/帖子不能被停止按钮撤回。停止后允许被动记录已在途请求的最终结果，但不得继续业务动作；需要清楚显示“已停止，已发出的请求可能仍已完成”。人工终止后必须由用户明确重新启动，不能因刷新或服务重启自动恢复。
+
+## 9. 后续实现与验收清单
+
+1. 优先实现常驻终止按钮、持久队列、站点级去重和历史发布对账，再接批量导入。
+2. 改造导入前标题截断、唯一来源尾注和可重复执行排版；保留原文与本地文件。
+3. 建立本地任务服务与浏览器执行器之间的配对通信，只传当前文章所需数据，不传下载 API Key 或账号库。
+4. 人工发布边界验收：没有用户点击时，等待任意时间都不会发表；标题修正后仍等待新的用户点击。
+5. 验证缺图、上传部分失败、重复来源、长标题、emoji、多 Markdown、非空编辑器、人工修改、刷新/断网、多标签竞争。
+6. 用户实际发布后，补录成功页或审核中页面证据、帖子 URL/ID，并验证重复打开不会增加计数。
+7. 给定 min/max 后验证 n 范围、每账号仅抽一次、达到目标换号、核对新 UID、恢复不丢计数。
+8. 在下载、上传、排版、等待审核、观察结果和换号各阶段点击终止，确认不再触发后续动作。
+
+## 10. 当前交接
+
+用户已亲手点击发表。原编辑器标签现为帖子详情页，标题与作者“示例用户”一致，面包屑为理财财经，显示“审核中”。已保存主题 ID、首帖 PID、URL、受理数量与去重记录，禁止再次发同一文章。登记后已点击网站首页入口，并核对 URL 为 `https://www.dww11.com/`，当前账号仍为示例用户（UID <已匿名>）；没有设置 min/max，因此未换号。
+
+发布结果检查发现正文内有 21 张图片，另外 5 张在附件区：image009.jpg、image014.jpg、image016.jpg、image025.jpg、image026.jpg，总数 26 张；来源尾注仅一条。发布后的正文与初次待审稿存在差异（例如原发布时间及“打开网易新闻 查看精彩图片”文字已不在结果正文中）。本次未记录用户审核期间的具体编辑动作，无法将差异归因于用户修改、编辑器同步或网站转换；后续需在用户点击的时刻保存最终待提交快照，再与结果对照。不得自动编辑或重发这篇已受理的帖子。
+
+截图：`test/history/周星驰文章_待人工审核.jpg`（历史待审稿）、`test/history/周星驰文章_网站审核中.jpg`（网站受理结果）。演示任务：`test/history/演示任务记录.json`；站点级去重登记：`发布去重记录.json`。记录保存本次结果，不代表完整历史同步或批量程序已经实现。
+
+
+## 11. 发布结果判断与本次补录
+
+- 用户点击后，不按“回到首页”判断成功；先记录实际跳转 URL 和页面结果。
+- 本次 URL 为 `/thread/23697411.html`，页面标题、理财财经面包屑、楼主示例用户及“审核中”共同确认网站已受理。DOM 首帖容器为 `#postmessage_238766`。
+- 网站审核中并非失败，不能为追求公开状态而再次提交；也不能仅凭当前登录用户能看到详情，宣称所有访客可见。
+- 刷新、重复读取或后来审核通过时，都用站点+主题 ID 保证受理数量只计一次。
+- 本次没有出现标题超限，相关处理仍是待验证需求。
+- 完整历史对账、常驻终止按钮、随机数量和自动换号仍属后续实现及验收范围。

实施计划.md

@@ -0,0 +1,178 @@
+# 新闻下载、导入排版、人工发表与换号实施计划
+
+本文中的帐号名称和用户编号已匿名化；真实记录保存在本地排除目录中。帖子编号仍用于说明既有受理证据，不代表当前公开状态。
+
+更新日期：2026-10-31（北京时间）。本版整体替换早期自动发帖设想，依据当前源码、[发布流程操作记录与自动化规范](发布流程操作记录与自动化规范.md)、[演示任务记录](演示任务记录.json)和[发布去重记录](发布去重记录.json)重新梳理。
+
+本次进行了本地源码与记录核对，未重新操作线上页面、下载新闻、上传图片或发表帖子。下文“演示已验证”引用既有操作记录；“源码已有”不等于最新版本已在线上重新验收。
+
+## 1. 当前目标与确定规则
+
+目标是：自动下载、选文、上传、填写、排版和校验，等待用户亲手发表，然后观察结果、记录数量，达到账号目标后换号，继续下一篇。
+
+- **发表及重试发表必须由用户亲手点击。** 不模拟点击、不调用提交函数、不直接发送发帖请求、不通过快捷键提交。自动化没有发布接口。
+- 用户选择分区、文章根目录、处理范围及每账号篇数 min/max；理财财经 fid=125 是演示样例，不硬编码为全部任务的目标。
+- 标题、正文或来源修改后重新进入人工审核；等待审核期间自动化不继续修改编辑器。
+- 全流程常驻“终止流程”按钮，跨标签、导航和换号保持生效；停止后不因刷新或服务重启自动恢复。
+- 同一文章在整个站点跨账号、跨版块去重。已公开发表或已受理待审核均阻止再次入队；结果不明时先核对。
+- 新下载文章不额外插入接口发布时间和分隔线。历史文件及源正文中原有的时间、文字不因此自动删除。
+
+结论：工作流可行，单篇人工链路已有实机记录，但持久队列、本地桥接、统一终止、自动计数和与换号扩展联动仍需实现。实施重点从“寻找全部页面元素”转为“连接已有组件并保证停止、审核、去重和结果核对”。
+
+## 2. 本轮变化与现状
+
+| 模块 | 当前确认 | 后续工作 |
+| --- | --- | --- |
+| 排版脚本 `dww11-wechat-oneclick.user.js` v1.0.9 | 新增独立“一键排版”；`existingSource()` 识别文末来源/作者及版权行；`runFormat()` 取出原尾注后交给 `formatBody()` 重新排版；无来源才弹窗补充 | 提供可等待、可取消、无模态弹窗的程序接口；验证复杂尾注与编辑器转换后重复执行 |
+| 下载保存 `article_writer.py` | 标题后直接接正文；不再插入接口 time；移除无文字且无图片的空标题节点；末尾仍生成标准来源尾注 | 输出持久文章清单与结构化来源/源网址/指纹，建立完整保存后的入队衔接 |
+| 下载调度 `app.py` | 频道、分页、下载去重、进度、停止；历史库仍只有指纹到目录映射；状态接口含汇总、日志、输出目录 | 独立发布队列、每篇完成登记、重启对账；不能拿日志当可靠文章队列 |
+| 导入器 v1.0.0 | `stagePackage()`、`importPrepared()`、附件上传及登记、编辑器填写校验 | 接入取消信号、无弹窗冲突状态、持久恢复；导入前处理长标题 |
+| `dww11_make_package.py` | 已能生成 v1 Markdown/base64 图片包 | 包大小限制、当前任务按需传输、引用完整性校验 |
+| 换号扩展 v3.0.4 | 自有账号库、抽号、登录、暂停与持久流程；下一账号按钮已有 ID | 接收统一终止、输出可核对的换号结果，与发文批次计数协调；不能假定网站脚本能直接调用扩展内部路由 |
+| `auto_news_upload/` | 操作规范和两份 JSON 记录已存在 | 尚无完整自动执行程序或 SQLite 发布任务库 |
+
+新版排版的“保留来源”指保留识别到的尾注文字；内部会规范化空白并重建为纯文本格式节点，不保证原 HTML、链接或空格逐字不变。自动化应展示提取结果，疑似误识别时暂停。
+
+旧操作记录中“每次输入来源”和“排版后手工删除重复来源”属于当时演示行为。当前 v1.0.9 的独立排版入口已针对已有来源提供处理，不能再把这一功能整体列为未实现；仍需回归验证。微信 URL 采集入口仍走另一条流程，本项目只接独立排版逻辑。
+
+下载器测试文件已含不插入发布时间的断言，本次未重跑测试。新下载文件的时间格式已明确，不再列为待确认项；这不授权清理历史稿件。
+
+## 3. 已有实机证据与尚未解决的问题
+
+既有演示记录：账号示例用户（UID <已匿名>），理财财经 fid=125，文章“刚靠新电影赚23亿票房，周星驰就卖掉内地影院”，来源“洲洲影视娱评”。用户亲手发表后，网站返回主题 23697411、首帖 PID 238766，记录时显示审核中。
+
+- 已明确受理 1 篇；已确认公开 0 篇。不得根据旧记录推断现在已审核通过，也不得重发这篇文章。
+- 导入阶段编辑器有 26 张图片；结果页正文 21 张、附件区 5 张。总数齐全不能证明正文位置和顺序正确。
+- 附件区图片为 image009.jpg、image014.jpg、image016.jpg、image025.jpg、image026.jpg。需要以附件 ID 与正文引用逐一对照，不只比较文件名或总数量。
+- 结果正文与待审稿有文字差异；记录未覆盖用户审核期间编辑动作，不能归因于网站转换或用户操作。
+- 本次未重新读取线上结果页；原演示未验证审核后的访客可见性、超长标题报错、随机数量与换号链路。
+
+后续准备完成时保存基线快照，用户亲手点击时再捕获最终标题、编辑器正文、图片引用和附件映射，记录内容修订号与提交尝试。持久化可能来不及完成时要能识别缺失证据；不能为补记状态而代替用户重新提交。发布后分别检查受理状态与内容完整性，内容异常只标记待处理，不自动编辑或重发。
+
+## 4. 目标工作流与架构
+
+```text
+用户设置分区、根目录、处理范围、min/max
+  → 初始化终止按钮与持久批次
+  → 历史去重对账
+  → 已完整下载文章校验 / 持续接收新文章
+  → 单篇领取与文章包准备
+  → 确认新帖标签、账号、分区
+  → stagePackage → importPrepared
+  → 独立一键排版 → 标题/正文/图片/尾注检查
+  → 等待人工审核与用户亲手点击发表
+  → 只观察提交结果
+  → 原子登记受理记录、去重与本轮计数
+  → 回首页并核对
+  → 未达目标：下一篇；达到目标且仍有任务：换号并核对
+  → 队列结束：结束批次
+```
+
+建议继续采用“本地 Python 任务服务 + 浏览器执行器 + 网站适配层”，复用导入器与换号扩展。第一步验证油猴特权请求与回环服务通信；若目标浏览器不兼容，再改为扩展桥接，不同时维护多套主链路。
+
+本地服务仅监听回环地址，使用配对凭据和限定接口；只向网站传当前文章内容，不传下载 API Key 或账号库。不能把现有下载器 `/api/session` 直接当作对任意网页开放的任务授权入口。导入器是 `@grant none`，排版器有特权权限，必须实测页面与脚本隔离上下文间的调用，不能简单假定合并后仍可访问原来的 `window` 对象。
+
+文件传递优先复用文章包，不依赖自动操作系统目录选择器。原生选择器可作为人工导入备选。同一版块的点击可能打开新标签，任务必须绑定实际标签及导航版本，不能在旧分区页继续执行。
+
+电脑、本地服务及浏览器需保持运行；没有新增定时无人值守发表需求。发布始终需要用户点击。新任务的下载 Key 仍按现有方式在内存中使用，不默认落盘。
+
+## 5. 数据接入、格式与来源
+
+### 完整落盘与元数据
+
+复用下载器现有保存及去重逻辑，新增每篇成功结果清单，至少保存稳定身份候选、原始标题、来源、源 URL、目录、Markdown 路径、图片清单和校验值。新发布记录写入失败时可通过清单对账补录，不应为回滚发布登记而删除完整文章。
+
+当前保存器先创建最终目录，再逐个移动文件，扫描到目录并不代表下载结束。新下载采用成功回调/完成标记与清单协调入队；旧目录在确认非保存中且引用完整后显式导入。下载历史库失败时原下载器自有清理行为，与未来发布登记失败处理分开。
+
+### 标题
+
+保留 `original_title`，派生 `submitted_title`。超过 80 字符时按允许的计数预算截取前缀并追加字面量 `...`，不能截断 emoji 代理对；站点实际按码点、UTF-16 或字节计数仍需实测。当前页面曾读到 maxLength=-1，“还可输入80个字符”也未随填入更新，不能作为可靠校验。
+
+截断必须发生在导入器 `compile()` 之前。现有 `preparePackage()` 已接受 `data.title`，可以由适配层提供派生标题，不改原 Markdown；现有打包器本身尚未生成这个覆盖值。显示新旧标题并进入人工审核。若人工提交后收到明确超限拒绝，修正后仍由用户再次点击；结果不明时不改稿重投。
+
+### 来源与正文
+
+复用 v1.0.9 的 `existingSource()` 与 `formatBody()`，保留 4 号黑色正文、1 号银色尾注、图文间隔及图片适配。识别末尾来源时保留整段来源/版权文字，不只抽出名称再拼回默认句子。没有来源则在页面内要求补充，不能让原生 prompt 阻挡终止按钮。
+
+需要验证：标准下载尾注、多行版权说明、重复尾注、正文中出现“来源”、来源后有图片、嵌套 FONT、纯文本换行、编辑器来回切换、重复排版。不得全局删除正文中的来源文字；原稿本身已有多条来源时应提示核对，而非宣称新函数能语义去重。
+
+正文清理只沿用已确认规则；不自动删除历史发布时间、“打开网易新闻 查看精彩图片”等文字。下载器新生成格式与旧文件格式并存，文章身份不能仅靠整个 Markdown 的精确哈希判断。
+
+### 图片与编辑器
+
+下载器支持 WebP/AVIF/SVG，打包器和导入器仍只支持 JPG/JPEG/PNG/GIF。先在入队前拦截不兼容格式，不能等打包时静默漏掉；如后续增加转码，生成派生副本并重写引用，保留原文件。上传大小同时遵守导入器 20 MB 和网站当前限制。
+
+校验顺序为：本地引用完整 → 上传附件返回编号 → 网站附件登记 → BBCode/可视化图片关联 → 排版前后附件引用保持 → 人工提交时最终快照 → 结果页正文与附件区对照。现有缓存仅在页面内有效，持久复用必须绑定账号与上传上下文并重新验证。
+
+遇到旧稿恢复提示、非空编辑器或上传期间人工修改，暂停并保留现场。导入器现有 confirm 弹窗需要转换为页面内状态；不能自动确认覆盖。
+
+## 6. 持久状态、发布去重与观察
+
+独立 SQLite 任务库与下载历史库分开。启动前先迁入两份现有 JSON 的已受理记录，保证主题 23697411 不再入队；保留原 JSON 作为证据。再核对原有发布记录或本人历史帖子列表。标题相同只标记疑似重复，不能据此确认同一正文。完整历史对账尚未完成，不能把这一个样本当全量历史。
+
+稳定身份优先关联原始文章 ID/规范化源 URL，同时保存正文与图片指纹、文件版本哈希和身份别名。当前下载历史只有指纹与目录，原始 URL/来源并未形成完整持久清单；历史稿无法恢复的信息要标为未知，不能猜造。跨账号、跨版块去重，改名、标题截断、删除后重新下载均不改变已受理身份。
+
+建议记录：article_key、身份别名、source_url/source_id、内容及图片指纹、原始/提交标题、来源尾注、目录、账号 UID、fid、附件映射、排版版本、内容修订号、任务状态、批次 ID、领取租约/执行版本、尝试 ID、用户点击时间、最终快照、帖子 ID/URL、受理/审核/公开状态和错误。
+
+```text
+downloaded → queued → importing → formatting → awaiting_user_review
+  → user_submitting → published / accepted_pending_moderation
+```
+
+异常分支为 `needs_attention`、`submit_unknown`、`rejected`、`stopped`。其中 `published` 仅在有公开发表证据时使用；提交意图或点击事件不是网站受理证据。人工修改后记录新修订，继续等待人工审核。
+
+- 自动化只监听用户操作及网站结果，不主动触发提交。点击前预登记候选尝试，点击时关联最终快照；即使快照写入失败，也不能自动补交。
+- 结果需综合帖子 ID/URL、作者、分区、标题和受理/审核提示；跳转到首页、HTTP 200 或按钮变灰均不能单独证明成功。
+- 以“站点+帖子 ID”唯一键登记受理记录，并在同一事务更新去重与批次计数。重复打开、刷新及后续审核通过不重复加一。
+- 审核中计入明确受理数一次，公开数量另记；驳回不自动补发。结果页内容异常单独记录，不撤销受理去重。
+- 网络中断、关闭标签或证据不足进入 `submit_unknown`，先对账；租约过期也不能重投未决文章。
+- 多标签单篇互斥；刷新或服务重启先显示恢复状态，等待用户明确继续。已停止批次不得自动恢复。
+
+## 7. 账号目标与统一终止
+
+用户填写整数 `1 ≤ min ≤ max`。首次建立当前账号批次、以及确认换号成功后，从闭区间均匀抽取目标 n 并持久化。每账号批次只抽一次；暂停、刷新和重启不重抽。没有 min/max 时不能擅自设值或启动自动换号。
+
+本轮只计明确受理的唯一文章，导入、排版、等待审核、失败和结果不明不计。达到 n 后，先完成记账，确认没有未决提交，回首页并核对，再触发现有“登陆下一帐号”。核对扩展流程结束和实际账号 UID/用户名；无法确认时暂停，不连续点换号。账号池只有一个可用账号等导致 UID 未变化时不能假装换号成功，应显示原因等待处理。
+
+队列空时结束，不为凑数重发，也不无意义换号。账号手动变化、验证码、限制或登录异常时暂停，不带着旧附件上下文继续。账号密码继续由现有扩展管理。
+
+统一终止必须覆盖首页、分区页、新帖页、结果页和换号页面。点击先持久写入 stopped 并更新执行版本；立即禁止新领取、下一张上传、排版写入、跳转和换号，取消可取消请求。所有异步返回重新检查停止状态，过期回调不得继续。
+
+**现有组件的停止并不贯通：** 下载器停止是等待当前请求/保存结束；导入器没有对外取消接口；换号扩展有独立流程；排版仍使用模态弹窗。仅在外层增加按钮不能满足全流程终止。需要给这些模块增加取消/暂停衔接，尤其要在扩展开始退出/登录前以及后续各步骤检查共享停止状态。不能把扩展内部消息接口当作已公开的油猴 API。
+
+停止不撤回已经发出的请求，允许被动核对其最终结果，但不继续业务动作。原演示使用浏览器控制条停止的例外不延续至正式实现。
+
+## 8. 页面适配依据与剩余信息
+
+| 环节 | 已有依据 | 仍需核对 |
+| --- | --- | --- |
+| 分区入口 | `/special/post/`；演示选择后新开标签 | 用户实际目标分区及各区必填项，不固定 fid=125 |
+| 当前账号 | `#mumucms_username`，用户名/用户链接编号 | 登录失效与换号成功的稳定判定 |
+| 标题/正文 | `#subject`、`#e_textarea`、`#e_iframe`、`#e_switchercheck` | 当前网站编辑器同步与标题计数 |
+| 导入 | `window.DWW11Importer`，面板在 Shadow DOM | 实际浏览器中程序接口与停止信号 |
+| 排版 | `#dww11-format-oneclick` | v1.0.9 来源识别与发布后往返结果；目前没有公开排版 API |
+| 发表 | `#postsubmit`、`#postform` | 只观察人工操作；失败、超限、审核及成功提示适配 |
+| 受理结果 | `/thread/23697411.html`；样本首帖 `#postmessage_238766` | 从当前结果提取动态帖子 ID/PID，不硬编码样本 |
+| 换号 | `#dww11-extension-next` | 扩展完成/暂停状态及全局终止联动 |
+| 历史对账 | “修改/删除信息”进入本人帖子列表 | 列表 URL、分页、审核项与重复证据 |
+
+无需重新提供已记录的完整单篇演示。下一批所需输入主要是：实际分区与文章范围、min/max、其他版块必填项、历史帖子入口及异常提示样本。标题计数、5 张图片落到附件区的原因、审核后的公开状态需要针对性验证。不需要密码、Cookie 或 API Key 值进入文档。
+
+## 9. 实施顺序与验收
+
+| 阶段 | 工作 | 完成条件 |
+| --- | --- | --- |
+| 1：基础状态与停止 | 独立任务库、现有 JSON 迁入、历史对账、统一终止状态和常驻按钮；确定各组件取消协议 | 已受理样本不可重入队；跨标签互斥；刷新不自动运行；停止后无后续动作 |
+| 2：文章衔接 | 完整落盘清单、文章身份、包传输、图片检查、导入前标题派生、本地配对通信 | 新旧格式均可处理；半成品不入队；长标题不改原稿；缺图和不兼容格式明确报错 |
+| 3：排版与人工审核 | 封装 v1.0.9 独立排版接口，替换模态提示，接入导入取消和编辑冲突检查 | 尾注文字保留、标准来源只出现一次；反复排版及编辑器转换无累积问题；附件引用与顺序不变 |
+| 4：人工提交观察 | 提交快照、动态结果解析、审核/公开区分、唯一受理计数、正文与附件核对 | 不经用户点击永不发表；用户修改后重新审核；刷新不重复计数；结果不明不重投；定位图片落附件区问题 |
+| 5：批次与换号 | min/max 持久目标、记账后回首页、复用登录流程、账号核对、终止联动 | 每批仅抽一次目标；达到目标且队列非空才换号；异常不连续换号；无旧账号附件串用 |
+| 6：小批实机验收 | 覆盖故障恢复与多个页面场景 | 下载/上传/排版/待审/结果/换号各阶段终止均有效；断网、重启、多标签和人工改稿状态正确 |
+
+重点回归场景：普通/多图/重复图片、空标题节点、多行尾注、来源后图片、重复排版、长标题与 emoji、非空编辑器、旧稿恢复、部分上传失败、上传超时、用户提交后响应丢失、审核中、驳回和换号中终止。先模拟验证，再做用户亲手提交的少量实机验收；不以已受理样本重发来测试。
+
+## 10. 目录与本次交付范围
+
+新实现仍放在 `auto_news_upload/`，后续规划为 `service/`、`browser/`、`tests/`、`data/` 和启动说明；当前这些实现尚未建立。保留原下载目录、原历史库及现有演示记录。复用源码时记录版本，明确安装时启用的脚本组合，避免重复工具争用编辑器。
+
+本次仅重写本实施计划，未修改业务脚本和历史记录。历史操作规范用于追溯当时实机行为；当前组件能力以本轮核对的源码为准，人工发表、去重、终止和换号规则继续遵循操作规范。

流程检查与Git文件清单.md

@@ -0,0 +1,150 @@
+# dww11 自动化流程检查与 Git 文件清单
+
+检查日期：2026-10-31（北京时间）。本轮仅检查本地说明、源码和离线测试；没有访问网站、消耗下载 API 额度、登录、上传或发表，最初检查时未执行 Git 操作。用户随后已确认仓库范围并授权本地初始化和首次提交；下面保留检查发现，并补充本次整理结果。
+
+## 一、检查结论
+
+当前是可以人工衔接的组件集合，尚不是已实现的端到端自动流程。
+
+```text
+探数 Key → 下载 Markdown 和图片
+  → 手工选择文章目录，或运行打包器生成文章包
+  → 已登录的论坛新帖页面导入并上传图片
+  → 点击独立“一键排版”并核对标题、正文、配图和来源
+  → 用户审核并亲手点击发表
+  → 核对受理/审核结果，登记去重，再决定下一篇或换号
+```
+
+发表与再次尝试发表均由用户亲手点击，这是现有流程规范要求。下载器、打包器、导入器、排版器和换号扩展之间没有完整的任务调度连接。
+
+## 二、逐环节检查
+
+| 环节 | 主要文件 | 当前能力 | 缺口或注意事项 |
+| --- | --- | --- | --- |
+| 下载 | `tanshuapi_news_downloader/app.py`、`article_writer.py`、`static/` | 频道、分页、正文图片落盘、下载历史去重、进度和停止 | 下载历史仅保存指纹与目录，没有供发布队列可靠消费的完整文章清单和完成标记；停止需要等待当前请求/保存结束 |
+| 打包 | `dww11_make_package.py` | 将目录转换为 v1 Markdown/base64 图片包；可直接作为 Python 函数调用 | 不会自动把文章送进浏览器；仅收集 JPG/JPEG/PNG/GIF，没有完整的 Markdown 引用预检或整包大小上限 |
+| 导入 | `dww11-article-importer.user.js` | 选目录或文章包、预览、按正文引用上传、附件登记、填写校验与失败回滚 | 标题超过 80 个 Unicode 码点直接拒绝；没有对外取消接口；非空编辑器确认仍使用模态弹窗 |
+| 排版 | `dww11-wechat-oneclick.user.js` v1.0.9 | 独立“一键排版”，识别文末来源并重建尾注；无来源时要求补充 | 没有公开可等待/取消的排版接口，仍有 prompt/alert；复杂尾注、重复执行和发布后的图片位置仍需浏览器回归 |
+| 发布 | `auto_news_upload/` 两份流程文档和两份记录 JSON | 已有单篇人工操作的历史记录 | 本目录没有完整发布执行程序、持久任务库、自动受理计数、全量历史对账或统一终止功能 |
+| 换号 | `dww11-extension/` v3.0.4 | CSV 导入、扩展 IndexedDB 帐号库、抽号与登录、暂停恢复 | 未与文章任务的 min/max、受理计数和统一终止联动；扩展内部路由不能直接视作油猴接口 |
+
+### 优先处理的问题
+
+1. **图片格式衔接不一致。** 下载器可保存 WebP/AVIF/SVG，打包器和导入器只支持 JPG/JPEG/PNG/GIF。打包器会跳过不支持格式，导入时可能报缺图；应在打包/入队前给出明确错误，或另外实现保留原文件的转码。
+2. **长标题衔接未实现。** 规范要求保留原题并派生截断标题，现导入器却在编译时直接拒绝超过 80 字的标题。需在导入前生成并展示提交标题，站点真实计数方式仍需实测。
+3. **缺少可靠调度和统一终止。** 下载完成不等于可以发布；目录存在也不等于完整保存。现有模块停止行为不贯通，不能直接启动无人看管的批量链路。
+4. **发布去重仍是手工记录。** 下载去重不能替代跨帐号、跨版块的发布去重。需要持久队列、未决结果核对、唯一受理登记及恢复逻辑。
+5. **历史样本的图片位置有异常。** 原记录显示 26 张图片中正文 21 张、附件区 5 张，总数相同不能证明位置正确。应按附件 ID 对照最终待提交快照与结果页。
+6. **旧路径已整理。** 原导入测试与打包说明引用了搬迁前目录；首次提交准备时改用 `tests/fixtures/article/` 的自造 Markdown 和 PNG，消除 legacy 依赖。
+7. **扩展测试资料缺失。** 扩展 README 提到 `tests/core.test.mjs`、`tests/browser-integration.mjs`、`tests/browser-restart.mjs` 和 `测试说明.md`，当前项目中未找到，不能把文档中的历史测试声明当成本轮复验结果。
+
+历史记录中的主题 23697411 已被网站受理并显示审核中，必须保留其本地去重证据。本轮没有核实现在是否已经公开。
+
+## 三、运行时需要帐号或凭据的文件
+
+下表区分“程序需要凭据才能运行”和“源码里保存真实凭据”。程序需要登录不代表源码不能进入 Git。
+
+| 文件 | 所需帐号/凭据 | 使用和保存方式 | Git 建议 |
+| --- | --- | --- | --- |
+| `tanshuapi_news_downloader/app.py` | 探数 API Key | 调用新闻 API；Key 从页面传入，在运行内存使用 | 源码可纳入；不写入真实 Key |
+| `tanshuapi_news_downloader/static/app.js`、`static/index.html` | 用户填写探数 API Key | 页面输入和内存变量，经本机会话接口传给服务；源码未配置固定 Key | 可纳入 |
+| `dww11-article-importer.user.js` | 已登录的大湾网帐号及页面上传凭据 | 从当前页面读取上传参数，使用当前登录会话上传图片；无须写入密码 | 可纳入源码 |
+| `dww11-wechat-oneclick.user.js` | 发布页面的登录会话 | 独立排版入口在当前编辑器操作；微信 URL 入口属于另一条采集路径 | 可纳入源码 |
+| `dww11-extension/manager.js`、`manager.html`、`core.js` | 帐号 CSV | 选择、解析、确认导入帐号和密码 | 可纳入程序及虚拟模板，排除真实 CSV |
+| `dww11-extension/db.js`、`service.js` | 扩展帐号库 | 明文密码保存在浏览器扩展来源的 IndexedDB；列表返回时去除密码 | 可纳入程序；排除浏览器资料及数据库备份 |
+| `dww11-extension/background.js`、`content.js` | 当前抽取的帐号和密码 | 扩展消息传递，并填入目标网站登录表单 | 可纳入程序 |
+| `today-jobs.user.js` | 可访问招聘管理列表的登录帐号/权限 | 使用当前页面和同源会话读取职位 | 辅助功能，按需纳入 |
+| `dww-today-threads.user.js` | 论坛后台管理权限与页面 formhash | 使用当前管理会话执行统计查询 | 辅助功能，按需纳入 |
+| `legacy/acc_switch_lite_python_legacy/server.py`、两个 lite 目录的 `import_excel.py` 与换号油猴脚本 | 旧版帐号文件、Excel 和网站登录 | 旧版实现；与当前扩展分开 | 不属于当前主链路；归档时排除真实数据 |
+| `legacy/acc_switch_heavy_sqlite_legacy/init_db.py`、`import_excel.py` | 旧版帐号 Excel/数据库 | 可生成并导入保存密码的 `accounts.db` | 不属于当前主链路；数据库不纳入 |
+
+`dww11_make_package.py` 仅处理本地文件，不需要帐号、API Key 或网站 Cookie。
+
+## 四、已经存在的帐号数据文件：排除真实版本
+
+| 文件 | 本轮确认 | 建议 |
+| --- | --- | --- |
+| `账号密码-去重可导入.csv` | 表头外 1152 条数据行、3 列；为真实帐号导入文件 | 本地保留，不纳入 Git |
+| `legacy/acc_switch_lite/accounts.json` | 169 条帐号记录，169 个非空 password 字段 | 本地保留，不纳入 Git |
+| `legacy/acc_switch_lite_python_legacy/accounts.json` | 169 条帐号记录，169 个非空 password 字段 | 本地保留，不纳入 Git |
+
+以上仅统计结构和数量，不展示用户名、密码或密钥。没有执行帐号有效性验证，也没有核对 CSV 行与 JSON 记录之间的对应关系。
+
+当前扩展运行帐号库在浏览器内，不在 `dww11-extension/` 源码目录。不要把浏览器配置目录、Cookie 导出文件或 IndexedDB 备份作为源码提交。`example.csv` 是示例模板，需保持为虚拟凭据。
+
+## 五、拟纳入 Git 的文件清单
+
+### A. 当前主链路源码与说明
+
+```text
+dww11_make_package.py
+dww11-article-importer.user.js
+dww11-wechat-oneclick.user.js
+本地文章导入脚本使用说明.md
+tanshuapi_news_downloader/app.py
+tanshuapi_news_downloader/article_writer.py
+tanshuapi_news_downloader/requirements.txt
+tanshuapi_news_downloader/start.command
+tanshuapi_news_downloader/README.md
+tanshuapi_news_downloader/.gitignore
+tanshuapi_news_downloader/static/index.html
+tanshuapi_news_downloader/static/app.js
+tanshuapi_news_downloader/static/style.css
+tanshuapi_news_downloader/tests/test_downloader.py
+dww11-extension/manifest.json
+dww11-extension/background.js
+dww11-extension/content.js
+dww11-extension/core.js
+dww11-extension/db.js
+dww11-extension/service.js
+dww11-extension/manager.html
+dww11-extension/manager.js
+dww11-extension/manager.css
+dww11-extension/example.csv
+dww11-extension/README.md
+auto_news_upload/实施计划.md
+auto_news_upload/发布流程操作记录与自动化规范.md
+auto_news_upload/流程检查与Git文件清单.md
+```
+
+两份流程文档的帐号名称、UID 和电脑绝对路径已匿名化，原版保存在被排除的 `auto_news_upload/local-records/`。历史帖子编号与链接保留为受理证据说明。本文不含密码，但也引用了本地敏感文件名和数量，可按仓库公开范围进一步精简。
+
+### B. 修复或检查后再决定
+
+- `test-dww11-importer.cjs`：已改用仓库内自造测试稿与三张 PNG，一并纳入首次提交。
+- `auto_news_upload/演示任务记录.json`、`发布去重记录.json`：属于真实运行状态。建议保留本地，另提供匿名结构模板；不因 Git 排除而删除原去重记录。
+- `dww11-extension-3.0.4.zip`：本轮逐文件比对，与当前扩展目录的 11 个文件一致。可作为发行附件；日常优先跟踪源码，避免双份维护。
+- `today-jobs.user.js`、`dww-today-threads.user.js`：辅助工具，不是四步主流程的必需文件，按你的范围选择。
+- `.wechat-script-test/`：本地排版验证页、源码副本、截图和结果，需整理为可复现测试后再选择纳入。
+- `legacy/`：默认不进入主链路首批提交；若保留历史实现，必须逐项排除帐号数据、虚拟环境、缓存及下载文章。
+- `credits-video/`：滚动字幕视频工具，与本次流程无直接关系，默认不纳入本批。
+
+### C. 建议排除的本地产物
+
+真实帐号 CSV/JSON、下载文章及图片、下载历史数据库、真实发布记录与截图、文章包 `*.dww11.json`、浏览器配置和 Cookie、`.venv/`、`__pycache__/`、`*.pyc`、`.DS_Store`、视频输出和临时诊断文件。
+
+首次提交准备时已增加根目录 `.gitignore`，覆盖真实帐号、运行记录、下载目录及本批范围之外的组件；按白名单添加源码和说明。
+
+## 六、本轮验证结果
+
+| 检查 | 结果 |
+| --- | --- |
+| Git 状态 | 当前目录不是 Git 仓库；本轮没有新建仓库 |
+| 原始导入测试 `node --test test-dww11-importer.cjs` | 8 项中 3 项通过、5 项因旧素材目录不存在失败 |
+| 只在内存中修正素材路径后执行同一测试 | 8/8 通过，覆盖打包与直接导入一致、缺图/长标题拦截、上传失败保留原文、富文本丢图回滚、无发表请求；磁盘测试文件未修改 |
+| 下载器单元测试 | 默认 Python 3.15 与 Python 3.12 均因缺少 `requests` 无法加载测试；未安装依赖，不能认定通过或认定下载逻辑失败 |
+| JavaScript 语法 | 6 个扩展 JS 按 ES 模块检查通过；4 个根目录油猴脚本检查通过 |
+| Python 语法 | 打包器、下载器 app 与保存器均通过语法解析 |
+| 扩展 ZIP | 11 个文件与源码目录逐一一致 |
+| 在线验收 | 本轮未执行；历史文档中的单篇演示不等于当前版本全链路验收 |
+
+## 七、首次提交整理范围
+
+用户已授权新建本地仓库并完成 initial commit，范围为总流程文档、当前换号扩展、新闻下载器、文章打包/导入/排版及相关说明与测试。
+
+- 增加根目录 README 与 `.gitignore`。
+- 导入测试改用项目自有测试稿及三张程序生成的 PNG，修正打包说明路径。
+- 原始流程说明备份到本地排除目录，再匿名化仓库文档中的帐号名称、UID 与电脑路径。
+- 真实帐号、下载历史、演示 JSON、截图及原始说明均保留本地，不纳入提交。
+- 下载器测试仍需先安装依赖，扩展缺失的测试文件尚未补齐；不以首次提交表示全流程已经完成。
+- 本次只建立本地版本历史，未配置远程仓库或推送。