YiPie

YiPie sends the page you are reading to an AI model you configure yourself, and gives you a summary you can read straight away in a floating panel at the top right of the page. You can keep asking about the page afterwards. No account and no server in between: requests go straight from your browser to the service you entered, billed to your own account. YOUR OWN MODEL • Works with any OpenAI Chat Completions–compatible service, in the cloud or running on your own computer • One click fills in the address of common services; "Test connection" reads the model list first, spending no tokens • Save several models and switch between them in the panel FAST, AND FAITHFUL TO THE PAGE • Answers stream in; a reasoning model's thinking is folded away above the answer • Citations: click a marker in the answer and the page scrolls to the quoted passage and highlights it • Select some text and right-click "Summarize the selection with YiPie" to summarize just that • Reload the page and open the panel again: the conversation comes back • Long pages are trimmed the way you choose; token usage is shown as reported, and clearly marked as an estimate when the service does not report it THE WAY YOU LIKE IT • Edit or add prompts, with variable hints and a preview • Chinese or English interface; choose the answer language, light or dark theme, and font size • Import and export models and prompts; exports leave out API keys unless you ask • Trigger it from the toolbar icon, the context menu or Alt+Shift+Y; everything works from the keyboard PRIVACY • Asks for no site access at install; it reads the current page only when you click the icon, use the context menu or press the shortcut • Your API key stays in your browser, out of reach of web pages • No analytics or reporting of any kind; images and other external resources in answers are never loaded • Open source (MIT): https://github.com/lonemoonspace/YiPie Before first use, enter your service's address, API key and model name in the extension's settings.