更新日期 / Last updated: 2026-11-20
开发者 / Developer: 谷粒(kuhung)
支持邮箱 / Support: [email protected]
简体中文
Baku 静貘在 X(Twitter)网页端隐藏或遮罩低质量帖子和回复。扩展会读取页面上可见的帖子及相关引用和回复上下文,包括正文、作者用户名和显示名称、帖子编号、来源链接、发布时间和认证状态等可见信息,用于判断页面上哪些内容需要过滤。
扩展在当前浏览器的本地扩展存储中保存设置、黑白名单、近期拦截记录、统计和同步状态。你配置的服务令牌或模型 API 密钥也保存在本地,仅用于向相应服务发起请求。卸载扩展可以删除这些本地数据。你自己导出的文件由你管理。
共享过滤默认开启。需要远程判别时,扩展会将待判别帖子、相关上下文、扩展随机生成的安装标识和信任名单发送到由开发者运营的共享服务。你执行屏蔽、放行等操作时,扩展会发送对应的反馈和相关帖子或账号信息。服务使用 Cloudflare Workers 和 D1 处理请求与存储数据,并可将帖子及上下文交由 Google 模型服务分类。服务会使用请求 IP 的哈希值执行限流和防滥用,并保留分类结果、用于维护共享名单的反馈样本和运行日志。记录按类型和用途保留,部分记录按容量清理。
你可以在管理台关闭共享过滤,只使用本地名单。如果配置了自建服务,相关的同步、分类或反馈会发往你选择的服务。如果启用自有模型密钥,扩展会将需要分类的帖子及上下文和用于认证的密钥发送至你选定的模型接口。你可以分别关闭共享过滤和模型直连。关闭后续请求不会自动删除之前由远程服务接收的数据。
我们仅为过滤、同步、处理反馈和保护服务使用这些数据。我们不出售用户数据,不将其用于广告画像、信用评估或贷款。扩展不读取 X 的密码、登录令牌或私信。查阅或请求删除共享服务中与你有关的数据,或反馈误拦,可以联系 [email protected]。
English
Baku hides or blurs low-quality posts and replies on X (Twitter). It reads visible posts and related quote or reply context, including text, author handles and display names, post identifiers, source links, timestamps, and verification indicators, to determine which content to filter.
The extension stores preferences, allow and block lists, recent filtering records, counts, and sync state in the current browser’s local extension storage. Service tokens or model API keys you configure are stored locally and used to authenticate requests to the corresponding service. Uninstalling the extension removes local extension data. You manage any files you export.
Shared filtering is enabled by default. When remote classification is needed, the extension sends candidate posts and their context, a randomly generated installation identifier, and the trusted account list to the shared service operated by the developer. Actions such as blocking or restoring a post can send feedback and associated post or account information. The service uses Cloudflare Workers and D1 to process requests and store data, and may send posts and context to a Google model service for classification. The service hashes request IP addresses to limit abuse and retains classification results, feedback samples used to maintain shared lists, and operational logs. Retention varies by record type and purpose, and some records are pruned by capacity.
You can turn off shared filtering in the dashboard and use local lists. If you configure a self-hosted service, applicable sync, classification, and feedback requests go to the service you choose. If you enable direct model access, the extension sends posts and context that need classification and your authentication key to the model endpoint you selected. You can turn off shared filtering and direct model access independently. Turning off future requests does not automatically remove data previously received by a remote service.
We use this data for filtering, syncing, handling feedback, and protecting the service. We do not sell user data or use it for advertising profiles, creditworthiness, or lending. The extension does not read X passwords, login tokens, or private messages. To ask about or request removal of data associated with you from the shared service, or to report a filtering error, contact [email protected].