DeWash 是一款专为浏览器设计的智能防护扩展。它能实时帮您甄别网页虚假新闻、高亮警示隐晦的主观感情色彩词汇与“春秋笔法”带节奏言论，过滤恶意仇恨，捍卫独立思考的权利。
DeWash is a smart browser extension designed to protect your cognitive sovereignty. It highlights fact-twisting framing, warns you of subjective narrative manipulations, filters hostile hate speech, and empowers critical thinking in real time.
在昨天的交通事故通报中，警方表示双方车主已达成初步和解协议。
In yesterday's traffic report, the police stated that both car owners had reached a preliminary settlement agreement.
根据目击者的指认，当彭某某被物业要求移车时，彭某某 态度竟然极其恶劣嚣张地 声称自己出差在外。但随后被警方证实，彭某某当时其实一直在家休息。
According to eyewitnesses, when Peng was requested to move her car, she shockingly displayed an extremely arrogant attitude and lied that she was out of town, which was later proven false.
另外，同小区的业主闵某在此前也曾因为类似的通道问题， 蛮横无理地阻拦过消防车， 引发了周边邻里的强烈反弹。
In addition, Min, another owner in the same community, had previously unreasonably blocked the fire engine during a similar passage dispute, provoking intense neighborhood outrage.
全面自主掌控认知安全防护网
Complete control over your cognitive safety net
本扩展属于无云端服务架构。您的 API Key 仅安全存储于您本机的 chrome.storage 环境中。不经过任何云端转发，绝对保障您个人数据隐私和秘钥安全。
Zero cloud-server mapping. Your API Key is strictly stored in your local chrome.storage. No third-party servers involved, ensuring absolute privacy for your browsing text.
支持“按需手动核验”、“自动白名单”及“全局扫描”三种模式。您可以细粒度限制在哪些指定域名下才发起扫描，完美防止大模型 Token 与流量隐性消耗。
Supports "On-demand Manual", "Auto Whitelist", and "Global Scan". Fine-tune which sites trigger the analysis to prevent accidental token consumption.
无论是配置页还是 Panel 面板上都集成了 Trust Score 警醒滑块。只有当 AI 计算出的信任分低于此预警分值时，才会在网页高亮划线，彻底告别噪音干扰。
Control sensitivity with an alert threshold slider on both the settings tab and popup panel. Highlights text only when its trust score falls below your setting.
全面兼容 OpenAI、Google Gemini 以及 DeepSeek 等大模型。系统完全支持自定义 Base URL (端点基址) 与模型名称，灵活对接您的中转或私有模型服务。
Fully compatible with OpenAI, Google Gemini, DeepSeek, and custom LLMs. Supports custom Base URL endpoints and specific model name variables.
对偏低分数的存疑段落，AI 将语义化提取一到两个最精准的核心实体名词。用户点击诊断框可直接携带实体一键跳转搜索引擎，进行快速交叉对比验证。
For questionable facts, the AI extracts semantic entities. Click the link in the diagnostic box to jump straight to Google, Bing, or Baidu for search checking.
自动放行所有使用的是中立语调的客观事实、时间地点或警方通报陈述。只对具有强烈主观道德评判修饰词、偏颇形容词或刻意带节奏话术的语句拉响划线警报。
Automatically allows neutral factual statements to reduce noise. Flags only emotionally biased words, moral judgments, and loaded phrasing designed to spin the story.
零数据上传承诺，捍卫您的数字隐私尊严
Zero server transit. We protect your privacy sovereignty.
DeWash 是一款完全运行于您本机客户端的隐私友好型工具。我们不运行任何云端中转服务器。本扩展绝对不会搜集、抓取、留存或向任何第三方传输/转售您的个人数据、阅读偏好、网页浏览记录以及页面扫描文本。
DeWash operates strictly client-side. We do not host or run any server-side databases. We never collect, inspect, harvest, store, transmit, or sell your personal data, reading habits, browsing history, or scanned text.
为了触发事实与话术分析，您提供的 API 密钥（如 OpenAI, Gemini, DeepSeek）将仅加密存储在本机浏览器环境（chrome.storage.local）中。HTTPS 连接直接建立在您的客户端与大模型官方服务器之间，不经过任何中介代理。
To power the AI cognitive check, your own API Keys (e.g. OpenAI, Gemini, DeepSeek) are stored strictly in your local browser sandbox (chrome.storage.local). HTTPS queries are dispatched directly from your device to the official AI servers.
本插件仅向浏览器申请了最精简的 `storage`（本地存储）与 `activeTab`（当前标签页访问）权限。前者用于保存白名单与阈值，后者仅用于在您点击弹出面板时临时获取当前域名以做白名单检测。本插件物理清除了高危的常驻注入或 `scripting` 权限。
DeWash only requests `storage` (to save whitelists and preferences locally) and `activeTab` (to safely read the current URL for whitelist checking when clicking popup). Elevated, constant spying permissions and `scripting` are completely stripped from our code.