johnmaeda/js-localpage

LocalPage is a command-line tool that captures web pages from your browser, converts them to Markdown, and optionally runs them through a locally installed AI model for summarizing or question-answering. The central design principle is that nothing leaves your machine: no cloud services, no external AI APIs, and no data collection of any kind. The tool is aimed at people who work with authenticated web pages, such as internal dashboards, private documents, or tools that require login. Because those pages often contain sensitive information, the author built LocalPage to be entirely local. A browser automation library called Playwright drives your existing Edge installation to load the page, including any session cookies from your logged-in account. The page is then converted to Markdown using two standard open-source libraries. If you want to ask questions or clean up the Markdown, those requests go to Ollama, which runs AI models on your own hardware. Workflow commands include logging into sites to store session credentials, capturing a URL, listing previous captures, printing or searching them, reformatting the Markdown for readability, and asking free-form questions about the page content. The AI model used for those last two steps is strictly a text-in, text-out tool: it has no ability to call functions, access the internet, or take actions. The README is explicit that this is a deliberate design choice, not a limitation. Setup requires Node.js, Playwright (for browser control), and Ollama with at least one downloaded model. Captured pages are stored in a hidden folder in your home directory. A configuration file controls which model size to use for different tasks, ranging from a smaller faster option to a larger one for more complex pages. The project is written in TypeScript and includes a security policy file that actively warns if any configuration would attempt to send data off the machine.