FindThatPage

Search your browser memory using natural language.

--- **You read something last week and can't find it again. FindThatPage fixes that.** FindThatPage builds a private, full-text search index of every article, doc, thread, and page you visit — and runs entirely on your device. When you need to find it again, a single keyboard shortcut opens a Spotlight-style overlay that searches your own browsing memory in milliseconds. No cloud sync. No AI API. No account. No telemetry. Your reading history never leaves your computer. --- **Why people install it** - You read a great essay on Tuesday, want to quote it on Thursday, and Google can't find it because it's buried in someone's newsletter archive. - You researched a bug last month, closed the tab, and now you vaguely remember "it had something to do with timezone offsets." - You're collecting sources for a project and want to search the stuff you've already read before searching the open web. - You don't want your browsing history in someone else's database. --- **What it does** - **Instant keyboard search.** Press `⌘⇧K` (Mac) or `Ctrl+Shift+K` (Windows/Linux) on any page. A Spotlight-style overlay appears. Start typing. - **Full-text search.** Match on titles, keywords, summaries, domains, and the actual body text of every page you've indexed. Powered by SQLite FTS5 with BM25 ranking. - **Chip-based filter stacking.** Type a term, press **Tab** to lock it in as a chip, then keep refining. `react` → Tab → `hooks` → Tab → only pages matching both. - **Smart chip prefixes.** Type `site:github.com` or `in:title react` or `in:body webassembly` — chips auto-scope to the right column for laser-precise filtering. - **Highlighted snippets.** Results show the matching phrase in context, so you can see *why* a page matched before you click. - **Deep links.** Pressing Enter opens the page scrolled directly to the matching text (uses the Chrome text-fragment API). - **Frecency ranking.** Pages you've re-read recently and often float to the top. Visit counts are displayed on each result. - **Pin important pages.** Pinned items survive retention cleanup and always appear at the top of results. A pinned/recent divider keeps the list organised. - **Undo Forget.** Accidentally deleted a result? A 30-second undo toast gives you a second chance. - **"Index this page now."** A one-click button in the popup forces indexing of the current tab — useful for pages that load content after navigation. - **Top-domain chips + one-click domain filter.** Each result has a "More from {domain}" button, and the search page surfaces frequently-matched domains as quick-filter chips. --- **Privacy is the product, not a feature** - **Everything is local.** All data is stored in SQLite (via OPFS SyncAccessHandle) and IndexedDB on your device. No server. No sync (unless you opt in to Chrome's built-in settings sync, which only covers your preferences — never pages). - **No AI APIs.** Search is keyword + synonym + BM25 ranking — not an LLM and not a vector database calling out to a third party. - **Strict content gates:** - Pages with password fields are never indexed. - Incognito windows are never indexed. - A built-in blocklist covers banking, payment, and auth domains. - Localhost, private networks (RFC 1918, link-local, IPv6 ULA), and `.internal / .corp / .home.arpa` hostnames are skipped. - **Fine-grained control.** Exclude any domain, pause indexing for 15 minutes or an hour, or disable it entirely from the popup. - **Full data portability.** Export everything as JSON. Import it back on another device. Clear everything with one click. It's your data. --- **Keyboard shortcuts** - `⌘⇧K` / `Ctrl+Shift+K` — open search overlay on the current page - `Tab` — add the current term as a filter chip (AND with existing chips) - `Backspace` on empty input — remove the last chip - `↑` / `↓` — move between results - `Enter` — open result in a new tab (scrolled to match) - `⌘+Enter` / `Ctrl+Enter` — open in a background tab (keep searching) - `Shift+Enter` — open foreground but keep the overlay open - `⌘1–⌘9` — jump directly to result 1–9 - `Esc` — close overlay Customize the shortcut at `chrome://extensions/shortcuts`. --- **Smart search prefixes** Type these inside the search input and press Tab to lock in as chips: - `site:github.com` — only pages from github.com - `domain:reddit.com` — same as site: - `in:title react` — match only in page titles - `in:body webassembly` — match only in body text - `in:summary auth` — match only in summaries - `in:keywords cdk` — match only in keywords --- **Who this is for** - Writers, researchers, students — anyone who accumulates sources. - Developers who keep bouncing between docs, GitHub issues, and Stack Overflow. - Power users who want search-your-own-history without handing it to a startup. - Anyone who's typed "that thing I read about…" into Google and hit a dead end. --- **Who this is NOT for** - If you want cloud sync of full page contents across devices — not available by design. - If you want AI-generated summaries of your reading — this uses keyword search, not an LLM. - If you only visit a handful of pages per week, a bookmarks folder might be enough. --- **How it works (technical)** - Built on WXT + React 19 + TypeScript, strict mode throughout. - SQLite (compiled to WebAssembly) with an FTS5 virtual table, running in an OPFS SyncAccessHandle-backed persistent store. - Porter tokenizer, BM25 ranking with column weighting (title ≫ keywords ≫ summary ≫ domain ≫ body), contentless FTS5 for space efficiency. - Service worker coalesces concurrent indexing writes into batched transactions. - Automatic retention prunes old non-pinned pages (default: keep 180 days or 20,000 pages, whichever hits first — configurable). - Works on long-lived SPAs via MutationObserver + pushState/popstate hooks + 10-minute incremental re-index while visible. - On-demand content-script injection via `chrome.scripting.executeScript` so the overlay and "Index this page" work on tabs that pre-date the extension install. - Firefox MV3 supported via direct-Worker fallback (no offscreen document). Release notes, issues, and support: https://github.com/asik-mydeen/find-that-page-releases ---