Offline Web

Offline Web is a free, cross-platform (macOS / Linux / Windows) desktop app that mirrors a web page — or a whole website — to a local folder so you can read it offline. It is free, with no subscription, but requires signing in with a free InterlinedList account (interlinedlist.com) to unlock scraping. The login is a pure access gate: sign in once and every feature is unlocked.

Screenshot placeholder. App screenshots (sign-in, New scrape, live Progress, Results with capture report) go here once captured.

What it does

• Save one page or a whole site — "This page only" (a page plus its assets) or "Whole site" (a bounded, polite crawl with depth and domain scope).
• Honest static snapshots — captures images, CSS, fonts, and scripts and rewrites links so the copy opens offline from your disk. Some dynamic features (logins, live feeds, search boxes, streaming) won't work offline, and the app says so clearly.
• Safe, polite defaults — respects robots.txt, fetches slowly (1 req/s), stays same-domain, and stops at finite caps (500 pages / 2 GB / 30 min). You can't produce an abusive crawl without deliberately opening Advanced.
• Legible, resumable jobs — live progress with Pause / Resume / Stop and a live rate control; jobs survive quit, crash, network loss, and session expiry.
• Trustworthy results — a capture report showing captured vs. skipped (with reasons) and honest fidelity notes.

Quick start

Installers are built by CI. Two channels (see .github/workflows/): dev — every push to main publishes unsigned installers to a rolling latest prerelease (build.yml); stable — pushing a vX.Y.Z tag publishes a signed + notarized, auto-updatable per-version release (release.yml). Both cover macOS .pkg/.dmg, Windows .msi/.exe, Linux .AppImage/.deb/.rpm. Signing/notarization and auto-update activate once their secrets/certs are set (they're gated, so builds stay green until then). You can also run from source, below.

Prereqs: Node.js (npm) and the Rust toolchain. Plus each platform's build deps for the desktop shell — Xcode Command Line Tools on macOS; webkit2gtk / libssl / build-essential on Linux; MSVC + WebView2 on Windows.

npm install

# Desktop app (real sign-in + real scraping):
npm run tauri dev

# Browser-only UI preview (no sign-in, no scraping):
npm run dev        # serves the UI at http://localhost:1420

Sign in with your InterlinedList account email and password. Signing in and scraping only work in the desktop app — the browser preview lets you click through the screens but can't reach the keychain or fetch pages.

New users: start with docs/user-guide.md.

Documentation

• User guide — install, sign in, make your first mirror, whole-site & advanced options, watching a job, reading results.
• Acceptable use & your responsibilities — what the app is for, how it protects sites by default, and the copyright / personal-data principles you're asked to follow.

How it works & fidelity

The app fetches pages over HTTPS, saves the assets needed to render them, and rewrites in-page links and references so the result opens from file:// with no network. The output is a browsable folder tree with an index.html entry point:

~/Offline Web/<host>/
  index.html          # rewritten entry page, opens offline
  assets/             # downloaded images / CSS / fonts / JS
  ...                 # additional captured pages (whole-site jobs)

Because it's a static snapshot, anything that needs a live server or a running app — search, sign-in areas, live/streamed content, some interactive JavaScript — won't function offline. The capture report tells you specifically what was skipped and what likely won't work. Your scraped content never leaves your device; the only server the app talks to is InterlinedList's sign-in.

Tech stack

• Tauri v2 desktop shell (Rust backend + native webview).
• Vite + vanilla TypeScript frontend — no UI framework, minimal deps.
• Rust scraping: reqwest (rustls) + scraper + url.
• Credentials in the OS keychain (macOS Keychain / Windows Credential Manager / Linux Secret Service) — never in a plaintext file. Your password is used only for the sign-in exchange and is never written to disk or logs.

Status

Milestones are incremental vertical slices (see docs/plan.md §6).

Milestone	Scope	Status
Auth	Real email + password sign-in to interlinedlist.com; token in OS keychain; session-expiry auto-pause + resume	Shipped
M0	Walking skeleton: launch → login gate → single-page capture → open in browser	Shipped
M1	Whole-site crawl: depth, domain scope, dedupe, safety caps, robots.txt, rate limiting, backoff	Shipped
M2	Long-job UX: live Progress, Pause / Resume / Stop / live Rate, persisted crash-survivable resume	Shipped
M3	Results & capture report: captured vs. skipped, fidelity notes, inline fixes, Re-scrape / Delete, recovery states	Shipped
M4	Opt-in JavaScript rendering (drives system Chrome over CDP, no bundled Chromium) for JS-only pages	Shipped
M5	Settings tabs, first-run ToS acknowledgment, native menus/notifications, accessibility, packaged installers, auto-update	Mostly shipped — settings, ToS, menus, notifications, a11y, and CI-built installers are done. Code signing and auto-update are fully wired in CI (release.yml, gated on secrets); they activate once the signing key/certs are added as repo secrets

Honest limitations today

• JavaScript rendering is opt-in (M4). Pages that build their content with JavaScript are flagged Needs JavaScript; you can opt into rendering them by driving your installed system Chrome over CDP (no Chromium is bundled). Off by default.
• Dev-channel installers are unsigned. Rolling main builds (build.yml) aren't code-signed, so Gatekeeper (macOS) / SmartScreen (Windows) warn on first launch. Tagged releases (release.yml) sign + notarize once the certs are added as repo secrets — until then they publish unsigned too.
• Auto-update goes live once the signing key is set. The updater plugin, public key, and GitHub-Releases endpoint are wired; tagged releases produce the signed latest.json the app reads. It only delivers updates for versions newer than what's installed, so bump the version before tagging.
• One job at a time. Concurrent jobs are deferred beyond v1.

Branding

Brand values (name, wordmark, logo, colors) are placeholders derived from interlinedlist.com and centralized for a one-edit swap when official assets land:

• src/styles/brand.css — colors, radius, fonts (CSS custom properties).
• src/brand.ts — product name, wordmark, and inline SVG logo.