自动抖隐私政策
1. 我们处理的信息
可选评论意向识别:首次默认勾选,可取消;已保存的关闭选择不会被覆盖,也不会在旧任务运行中因首次读取配置而开启。默认勾选不会启动自动浏览操作或发送私信。用户填写业务目标并保持开启后,在人工查看的当前聚焦视频页读取已经加载的公开评论,无需启动自动运营任务;人工浏览识别不会点击、滚动或展开评论,不读取所有后台标签页。人工浏览时新增评论经约1.5秒合并等待,或一批创建后约8秒内有新评论时分批识别,满批或切换页面时也会提交。用户主动启动自动浏览任务后,沿原流程展开评论区并按设置逐轮读取;默认每视频最多10轮,可设置1–100轮,到底后连续3轮无新评论结束,不设每视频2000条或3分钟停止限制,不进行全站抓取。自动浏览每批最多2000条,满批、约每30秒或视频结束时分批上传;人工浏览也使用最多2000条的分批上限。上传字段为作者公开标识、昵称、评论正文与标识、回复关系、来源视频标识及必要摘要、关键词、业务目标和客户需求设置,接收方为自动抖服务器。分批上传后继续识别,不因上传切换视频。未完成数据保存在本机;页面关闭、网络中断或浏览器工作进程重建后,在识别开关仍开启且账号授权有效时,可由每分钟的后台检查继续提交归属该账号的待处理批次,不恢复浏览操作或发送任务。停止浏览任务不会关闭人工浏览识别;取消总开关停止两种浏览识别的新增读取与新请求,已提交服务器的请求可能继续完成,已有客户记录保留。另行启用的视频监控按下述监控设置执行。服务器使用本地规则及本地语义模型识别,不调用外部 AI。为去重、查看依据、额度与续跑保存批次、识别结果和联系状态,可联系支持申请删除。管理员可在管理后台查看上报与识别记录;用户仅可通过授权账号读取归属当前抖音账号的客户线索及评论证据,不共享其他账号的客户列表。识别本身不会发布评论或私信,意向不等于同意被联系,使用者应确认联系范围与平台规则。
意向客户私信:识别仅入队,不自动发送。用户填写内容并主动启动后,按行轮换、同客户去重,受服务端权限、领取锁及每日额度限制。私信模板保存在本地,发送时通过抖音交付给指定收件人,不上传自动抖服务器;自动抖服务器接收任务标识与执行结果。结果不明不自动重发,需用户核实。可随时暂停,重启后不自动继续。
试用可识别并查看客户列表,仅前三位可在插件中打开或排除;服务端保存这三个客户的资格以避免重复领取试用名额。自动发送需要付费授权,并在启动及执行时复核。
- 账号与授权信息:用户主动提供的手机号、扩展生成的用户标识、抖音账号公开标识、授权令牌与授权状态,用于登录、授权校验和绑定当前账号。
- 本地设置与记录:关键词、筛选条件、互动开关、评论模板、任务进度、操作记录、防重复记录以及用户主动采集的公开视频信息。它们默认保存在浏览器本地存储中。
- 抖音公开页面信息:在用户启动任务、采集、保持意向识别开启并人工查看视频,或按用户此前主动启用并保存的视频监控配置执行巡检时,读取相关抖音网页中公开展示的视频链接、标题、作者、公开账号标识、发布时间及公开互动指标,以执行用户选择的功能。
- 私信自动回复:仅在用户主动开启后,读取已加载会话列表的昵称、消息预览、会话标识,以及打开会话后网页中已加载的聊天文本和消息标识。程序自动逐个打开一对一会话并滚动会话列表;新消息监听按启动时间和首次读取基线区分新旧,只匹配启动后的新增来信,不补回启动前历史,也不只处理最后一条。历史模式只执行一轮,逐条匹配每个会话固定范围内已加载的对方文本,包括人工或其他工具已回复过的旧消息;无需逐条勾选,不向上加载完整聊天历史。关键词规则在本地匹配;聊天文本不上传到自动抖服务器,也不作为聊天记录保存。防重复记录仅在本地保存消息/会话的摘要标识与时间,私信最多 5000 条、旧评论模块最多 500 条,卸载可删除。当前评论回复及指定账号名单入口暂时隐藏。
- AI 功能信息:仅当用户主动启用 AI 私信回复或语义规则匹配,并同意文本传输后,将必要的提示词、目标文本及最近 2–40 条已加载聊天上下文(默认 12 条,每条最多 1000 字)发送给用户选择的 AI 服务商;语义模式只由 AI 选择规则编号,回复内容使用用户配置的规则。私信具有独立的 AI 服务商、模型、API 地址和密钥设置,密钥保存在本地,用于向所选服务商请求,不上传给自动抖服务器。自动回复默认关闭,授权或配置变化、消息标签页转入后台、草稿冲突或发送结果不明时停止,不会在扩展重启后自行开启。随机间隔、巡检间隔和单次发送上限可配置;开始后不再逐条弹窗确认,用户可随时停止。
- 可选图片回复:仅在用户启用图片回复后,读取抖音网页中用户已有的收藏图片,并按指定序号或随机范围发送;自动抖不向服务器上传这些图片,也不向 AI 服务商发送图片。文字和图片发送共同计入本次发送上限,平台不允许发送或结果不明时停止。
- 标准网络信息与任务诊断:授权服务器可能接收 IP 地址、浏览器类型和请求时间。插件将任务及指定控制按钮的操作阶段、结果、错误代码、必要的页面元素诊断、版本、任务与视频标识、次数和时间在本机合并缓存,约每30秒或达到批量阈值后分批提交自动抖服务器;断网恢复后重试。记录按授权账号归属保存,管理员可用于故障排查和功能改进,服务端保留30天。诊断不包含评论正文、私信正文、密码、验证码或授权令牌;不采集普通键盘输入或无关网页点击。原有详细操作记录仍保存在本机。
- 页面适配配置:扩展会在授权后从自动抖服务获取版本化 CSS 选择器 JSON,以适配抖音网页结构变化。该请求不上传抖音页面内容、评论、视频信息或用户采集结果;配置不包含也不会执行 JavaScript、WASM、表达式或远程命令。
- 线索工作台与本机跟进记录:工作台读取的是自动抖服务器上已识别的客户线索(客户标识、昵称、意向分、状态、评论证据、来源视频标识)。客户备注、标签、人工跟进阶段、跟进队列、行业模板、来源视频标题缓存与跳转定位标记仅保存在当前浏览器本地,不上传服务器。这些本机记录不能从服务器恢复,更换设备、清理浏览器数据或卸载扩展后可能丢失。工作台的本机 JSON 导出仅包含人工标注、监控目标、运行记录、开关与业务设置,不包含服务端线索、跟进队列、行业模板或标题缓存,也没有对应的工作台备份导入功能,不能当作完整备份。清理本机记录不会删除服务端线索,服务端数据按本政策与支持申请处理。
- 视频监控(可选):默认关闭,需用户主动启用并添加已启用的视频目标后生效。开关、目标与排期保存在本机;浏览器或扩展重新启动后,可能按已保存且已启用的配置恢复巡检。后台每分钟检查开关和目标到期时间,按用户设置的排期(15/30/60 分钟,可能因其他任务忙碌而延后)在不抢焦点的独立临时窗口中打开视频标签页,读取目标视频公开评论增量,提交至与“可选评论意向识别”相同的服务器识别流程;提交或本轮处理结束后,关闭扩展自己创建且仍处于抖音页面的临时窗口。本机保存监控配置、去重索引和最近运行结果;上传后的识别与线索数据由服务器保存。关闭监控后不再启动新的巡检,但每分钟检查闹钟仍保留,已经开始的一轮可能继续读取、提交并结束,不会立即取消。监控与主运营任务独立,停止运营任务不会自动关闭监控。监控不读取私信内容,不代替用户发送消息。
2. 信息用途
我们仅将信息用于提供登录授权、执行用户发起的关键词筛选/互动/数据采集、保存任务状态与记录、导入导出配置、提供可选 AI 回复、保障服务安全以及处理支持请求。
自动抖对用户数据的使用遵守 Chrome Web Store 用户数据政策,包括有限使用(Limited Use)要求。数据仅用于用户明确选择的功能,不用于广告、出售数据或与产品用途无关的用户画像。
3. 数据共享与第三方服务
我们不出售个人信息,不用于个性化广告,也不将数据用于与扩展单一用途无关的信用评估或用户画像。以下情况可能发生必要的数据传输:
- 授权校验与页面适配配置请求通过 HTTPS 发送至自动抖服务(shinewaycargo.com)。
- 用户启用 AI 功能时,必要文本和请求凭据发送至用户选择的 DeepSeek、Moonshot、OpenAI、OpenRouter、Google Gemini、自定义兼容接口或本地 Ollama;其处理受相应服务商政策约束。
- 法律要求、保护用户或服务安全时,可能依法披露必要信息。
4. Chrome 权限用途
- storage / unlimitedStorage:保存设置、任务状态、采集结果、防重复记录与操作记录。
- tabs / scripting:定位用户当前打开的抖音标签页,并在用户发起操作时执行页面交互或读取公开页面信息;用户启用视频监控后,用于按设定间隔创建巡检用临时标签页、定位其页面并只关闭该扩展自己创建的这一页。
- sidePanel:在浏览器侧边栏提供控制界面。
- notifications / alarms:显示任务结果或异常提醒,管理用户设置的任务时限与私信队列唤醒。意向识别有待处理批次且总开关开启时,另设每分钟一次的续传检查;没有待处理批次或关闭总开关时清除该续传闹钟,不会因此启动浏览或发送。监控后台注册每分钟一次的开关/排期检查,仅在用户已启用监控且目标到期时启动巡检;浏览器或扩展重新启动后可能按本机保存的已启用配置恢复巡检。关闭监控不会清除该检查闹钟,但会阻止新的巡检,不会立即取消已开始的一轮。
- 主机权限:访问抖音网页、自动抖授权服务,以及用户选择使用的 AI 服务接口。
5. 保存、安全与用户控制
本地数据保留到用户在扩展中清除、浏览器清除扩展数据或卸载扩展为止。授权服务器仅在提供授权与安全所需期间保留账号记录。数据传输使用 HTTPS(用户自行配置的本地服务除外)。配置导出文件可能包含用户设置的 API 密钥,用户应妥善保管,不要公开分享。
用户可以随时停止任务、关闭功能、清除操作/采集/防重复记录、删除本地设置或卸载扩展。如需查询、更正或删除服务器端账号信息,请通过下方邮箱联系我们。
6. 儿童、变更与联系
本服务不面向未满 18 周岁的儿童。政策发生重大变化时,我们将更新本页的日期。问题或数据权利请求请联系:kan831019@gmail.com。
Dy Auto Privacy Policy
Dy Auto is a user-controlled browser extension for workflows on public Douyin web pages. It is not an official Douyin product.
Optional lead recognition: Preselected on first use and can be unchecked. A previously saved off choice is preserved, and first-time initialization does not enable it during an already-running task. Preselection does not start automated browsing or messaging. Once a business goal is configured and the option remains enabled, recognition reads already-loaded public comments on the video page you are viewing in the focused tab, without starting an operations task, clicking, scrolling, opening comments or reading all background tabs. Manual browsing batches wait approximately 1.5 seconds after new comments, or are submitted when new comments are present at approximately eight seconds after batch creation, when full, or on a page transition. If you actively start an automated browsing task, the existing flow opens comments and reads successive rounds: 10 rounds per video by default, configurable from 1 to 100, ending after three bottom-of-list rounds without new comments. There is no 2000-comment or three-minute stop per video and no site-wide crawl. Automated browsing batches of up to 2000 comments are submitted when full, approximately every 30 seconds while collection is active, or when the video ends. Manual browsing uses the same 2000-comment batch limit. Uploaded fields include public author identifiers and nicknames, comment text and identifiers, reply relationships, source video identifiers and limited summaries, keywords, business goals and configured customer needs; the recipient is the Dy Auto server. Collection continues after a batch upload without switching videos. Pending batches persist locally. After page closure, network failure or browser worker restart, a once-per-minute check may resume uploading batches belonging to the authorized account while recognition remains enabled; this does not resume browsing actions or messaging. Stopping an operations task does not disable manual recognition. Unchecking stops new reading and new requests for both browsing modes; already-submitted server requests may finish and existing leads remain. Separately enabled video monitoring follows its settings below. Server-local rules and semantic inference identify leads without external AI calls. Batches, recognition results, evidence and contact status are retained for deduplication, quotas and task resumption; contact support to request deletion. Recognition alone does not post comments or send messages, and detected intent is not consent to contact.
Lead contact messages: Recognition only queues leads. Sending requires the user to configure outgoing text and actively start the task, subject to server-side authorization, claim locks and daily quotas. Message templates and outgoing text remain in the browser and are sent to the selected recipient through Douyin, not to the Dy Auto server; the server receives task identifiers and execution results. Ambiguous delivery is not automatically retried. Sending can be paused and does not automatically resume after restart. Trial users can view leads but may open or exclude only the first three eligible customers; the server retains those trial qualifications. Automatic sending requires paid authorization, checked at start and during execution.
Data handled
We process a phone number, extension/user identifier, public Douyin account identifier, authorization token/status, locally stored settings and task records, and public video/profile information only when needed for features the user starts, manual viewing while lead recognition remains enabled, or monitoring previously enabled and saved by the user. Administrators can view upload and recognition records in the admin console; authorized users can read only leads and comment evidence belonging to their current Douyin account, not other accounts' customer lists. Optional AI features send only the necessary prompt and related text to the provider selected by the user. AI API keys are stored locally in the browser.
Lead workspace: identified leads (customer identifier, nickname, intent score, status, comment evidence, source video identifier) are read from the Dy Auto server. Customer notes, tags, manual follow-up stage, follow-up queue, industry templates, cached source-video titles and jump markers are stored only in the current browser and are not uploaded. These local records cannot be restored from the server and may be lost after switching devices, clearing browser data or uninstalling. The workspace's local JSON export contains only manual annotations, monitoring targets, run records, feature flags and business settings. It excludes server-side leads, the follow-up queue, industry templates and title cache; there is no corresponding workspace backup import, so it is not a complete backup. Clearing local records does not delete server-side leads; contact support for server-side data requests.
Video monitoring (optional, off by default): the user must actively enable it and add an enabled video target. Flags, targets and schedules persist locally; after a browser or extension restart, checks may resume according to saved enabled settings. A once-per-minute alarm checks the flag and target due time. Due targets use a 15/30/60-minute schedule, which may be delayed while other tasks are busy. Each run opens a video tab in a separate temporary window without taking focus, reads incremental public comments, submits them to the same server-side recognition flow above, then closes the extension-created window if its page is still on Douyin. Monitoring settings, deduplication indexes and recent run results stay local; uploaded recognition and lead data are stored on the server. Disabling monitoring prevents new runs but does not clear the once-per-minute alarm or immediately cancel an in-progress run, which may finish reading and uploading. Monitoring is independent of the main operations task: stopping that task does not disable monitoring. Monitoring never reads private messages or sends messages.
Use and sharing
After you actively start private-message automation, the extension reads rendered conversation metadata, automatically scrolls the conversation list, opens one-to-one chats, and matches their loaded text. New-message listening uses the start time and first-read baseline to match new incoming messages only, not earlier history or just the last bubble. History mode processes each conversation's fixed range of loaded incoming text in one pass, including older messages previously answered manually or by other tools, without individual selection. It does not load the entire chat history. Chat text is not uploaded to Dy Auto or saved as chat history. Duplicate prevention retains up to 5000 local hashed message/conversation identifiers and timestamps for private messages (500 for the inactive comment module). Comment and specified-account controls are currently hidden.
AI generation and semantic rule matching send only the necessary prompt, target text and 2–40 recent loaded context messages (default 12; up to 1000 characters each) to your chosen AI provider after your consent. Semantic classification selects a configured rule; it does not invent its reply. Private-message AI provider, model, endpoint and API key are configured independently and stored locally. Keys are not sent to Dy Auto. Private-message automation is off by default and never resumes after restart. Authorization/configuration changes, a hidden message tab, draft conflicts, uncertain delivery or AI timeouts stop the task. Random delays, polling intervals and per-run send caps are configurable; there are no repeated confirmation dialogs, and you can stop at any time.
Data is used only for authorization, user-initiated automation and collection, task continuity, records, optional AI replies, security, and support. The extension retrieves versioned CSS selector JSON after authorization to remain compatible with Douyin page changes. That request does not upload Douyin page content or collection results, and the configuration contains no JavaScript, WASM, expressions, or executable commands. We do not sell personal data or use it for personalized advertising.
Optional image replies read your existing Douyin favorite images only after you enable the feature, then send the selected index or a random item in your configured range. Images are not uploaded to Dy Auto or AI providers. Text and images share the per-run send cap; platform restrictions or uncertain delivery stop the task.
Dy Auto's use of user data adheres to the Chrome Web Store User Data Policy, including the Limited Use requirements. Data is used only for user-selected functionality, not advertising, data sales or unrelated profiling.
Task diagnostics: operation stages, selected control-button actions, outcomes, error codes, limited element diagnostics, extension version, task/video identifiers, counts and times are merged and buffered locally, then sent in batches to the Dy Auto server approximately every 30 seconds or at the batch threshold. Failed uploads are retried after connectivity returns. Records belong to the authorized account, are available to administrators for troubleshooting and product improvement, and are retained for 30 days. Diagnostics exclude comment and private-message text, passwords, verification codes and authorization tokens. Ordinary keystrokes and unrelated page clicks are not collected. Existing detailed operation records remain local.
Control and contact
Users can stop tasks, clear local records and settings, or uninstall the extension. To request access, correction, or deletion of server-side account data, email kan831019@gmail.com.