combine the search archive tools

Author: cig13zs

.github/workflows/test.yml

@@ -1,11 +1,25 @@
 name: test
-on: [push, pull_request]
+
+on:
+  push:
+  pull_request:
+
+permissions:
+  contents: read
+
 jobs:
   test:
     runs-on: ubuntu-latest
     steps:
       - uses: actions/checkout@v4
       - uses: actions/setup-node@v4
         with:
-          node-version: '20'
-      - run: node core.test.js
+          node-version: 22
+      - uses: actions/setup-python@v5
+        with:
+          python-version: "3.12"
+      - run: node tests/core.test.js
+      - run: node tests/content.test.js
+      - run: python scripts/package.py
+      - run: python tests/site.test.py
+      - run: python tests/package.test.py

.gitignore

@@ -1 +1,5 @@
-*.zip
+dist/
+__pycache__/
+*.py[cod]
+.DS_Store
+Thumbs.db

MAINTENANCE.md

@@ -0,0 +1,46 @@
+# Maintenance notes
+
+Search Restore changes a page it does not control. Keep these assumptions in the
+release checklist.
+
+## Current Google result structure
+
+On August 23, 2026, a read-only browser check of
+`https://www.google.com/search?q=example&hl=en&gl=us&pws=0` returned:
+
+- one `#rso` result container;
+- seven `h3` elements with a closest external link;
+- one `.MjjYud` wrapper per linked result;
+- no unusual-traffic or captcha page.
+
+The extension uses `#rso` and linked `h3` elements as its main result signals.
+`.MjjYud` is only the preferred pagination wrapper. A generic single-heading
+ancestor is the fallback because generated Google class names can change.
+
+The production `core.js` and `content.js` files were also run in a fresh
+logged-out tab. They added seven archive rows and one load control. One click
+fetched `start=10`, found nine new result blocks, raised the linked-title count
+from 7 to 16 and changed the button to `Load results 21-30`.
+
+That test found a false block signal in the first detector: Google's normal page
+scripts contain literal `/sorry/index` checks. `looksBlocked()` now strips script
+and style blocks before using its HTML fallback. The response status and an
+actual redirect to `/sorry/` remain the primary checks in `content.js`.
+
+## Known failure cases
+
+The extension intentionally does nothing when `/search` has no `#rso` block.
+This includes some AI Mode, consent and zero-result layouts. A Google captcha or
+HTTP 429 stops pagination and leaves a status message. The extension does not
+attempt to bypass those checks.
+
+If Google removes `start=` pagination, the load-more feature cannot reconstruct
+pages that Google no longer serves. If result titles stop using linked `h3`
+elements, archive rows will need a new reviewed selector.
+
+## Release check
+
+Use an unsigned, logged-out browser profile. Confirm `#rso`, linked `h3` titles,
+archive-row placement and one next-page fetch. Check a normal web query and a
+query with an AI Overview. Record limitations instead of marking an untested
+layout as working.

PRIVACY.md

@@ -0,0 +1,27 @@
+# Privacy
+
+Search Restore does not collect, sell or transmit personal data. It has no
+analytics, advertising, account system or remote server.
+
+The extension reads result titles and destination URLs already displayed on a
+supported Google Search page. It uses them to add archive links and does not
+keep a search history. The load-more feature requests the next Google results
+page only after the user presses its button.
+
+The extension stores the `archiveLinks` and `loadMore` preferences with
+`chrome.storage.local`. Chrome keeps these settings on the device instead of
+placing them in Chrome Sync.
+
+## Chrome Web Store Limited Use
+
+Search Restore's use of information received from Chrome APIs follows the
+Chrome Web Store User Data Policy, including its Limited Use requirements.
+Search result titles and destination URLs are used only to provide the visible
+archive and result-loading features the user installed. They are not retained,
+sold, used for advertising, or transferred to another party except when the
+user deliberately opens an archive link.
+
+Archive sites receive a result URL only when the user chooses an archive link.
+Those sites have their own privacy policies.
+
+Effective August 23, 2026.

README.md

@@ -1,81 +1,84 @@
-# Return Google Cache
+# Search Restore
 
-Google removed the Cached link from search results in 2024. This puts a working
-one back on every result, pointed at the Wayback Machine and archive.today.
+Search Restore combines two Google Search repairs in one Chrome extension:
+archive links under normal results and a button that loads later result pages
+into the current list.
 
-[![Ko-fi](https://img.shields.io/badge/Ko--fi-buy_me_a_coffee-FF5E5B?style=flat-square&logo=ko-fi&logoColor=white)](https://ko-fi.com/jju1s)
-[![License](https://img.shields.io/badge/license-MIT-blue?style=flat-square)](LICENSE)
-[![Tests](https://img.shields.io/github/actions/workflow/status/cig13zs/return-google-cache/test.yml?style=flat-square&label=tests)](https://github.com/cig13zs/return-google-cache/actions)
+[![Ko-fi](https://img.shields.io/badge/Ko--fi-support-FF5E5B?style=flat-square&logo=ko-fi&logoColor=white)](https://ko-fi.com/jju1s)
+[![License](https://img.shields.io/badge/license-MIT-17324d?style=flat-square)](LICENSE)
+[![Tests](https://img.shields.io/github/actions/workflow/status/cig13zs/search-restore/test.yml?style=flat-square&label=tests)](https://github.com/cig13zs/search-restore/actions)
 
-**[cig13zs.github.io/return-google-cache](https://cig13zs.github.io/return-google-cache/)**
+## Features
 
-For about twenty years every Google result had a Cached link. One click to see
-the page as Google last saw it, which was the thing you reached for when a page
-was down, paywalled, changed or gone. Google retired it. This adds it back using
-the Internet Archive.
+Archive links can open the latest Wayback Machine capture, check archive.today
+or save the live page. The load-more button requests one Google result page per
+click and stops near 100 results.
 
-Under each result you get a small row:
+Each feature has its own switch. The switches default to on and are stored with
+`chrome.storage.local`.
 
-```
-↻ Cached · archive.today · save now
-```
+## Install
 
-Cached is the latest Wayback Machine snapshot, archive.today is its independent
-mirror, and save now archives the live page immediately.
+1. Download the release ZIP and extract it.
+2. Open `chrome://extensions` and turn on Developer mode.
+3. Choose Load unpacked and select the extracted folder.
+4. Open a supported Google search page.
 
-## No server, no tracking
+The ZIP puts `manifest.json` at its root, so the extracted folder is the one to
+select. Chrome, Edge, Brave and Opera can load MV3 extensions this way.
 
-It reads the result links already rendered on your screen and builds archive
-URLs locally. It opens no connection of its own, has no analytics, and declares
-no `permissions` key. It runs on Google search pages and can do nothing else.
+## Scope and privacy
 
-## Install
+The manifest declares only the `storage` permission. It has no
+`host_permissions` key. Its content script is limited to the HTTPS `/search`
+path on the Google domains listed in `manifest.json`, with a second runtime
+check that requires `location.pathname === '/search'`.
 
-Not on the Chrome Web Store yet, so load it unpacked. Works in Chrome, Edge,
-Brave and Opera.
+Chrome may describe that content-script scope as permission to read and change
+data on the listed Google sites. That access is what lets Search Restore inspect
+the visible result links and add controls to the search page; it does not cover
+other sites or other Google paths.
 
-1. Download the latest zip from [Releases](https://github.com/cig13zs/return-google-cache/releases) and unzip it.
-2. Open `chrome://extensions`, turn on Developer mode.
-3. Load unpacked, then pick the `extension` folder.
-4. Search Google. The Cached links are there.
+There is no analytics or extension server. The two switches stay in local
+extension storage. Google receives a request only when the user presses the
+load-more button. An archive service opens only after the user chooses its link.
+See [PRIVACY.md](PRIVACY.md) for the full policy.
 
-## Files
+## Build and test
 
-```
-extension/
-  manifest.json   MV3, runs only on google search pages
-  core.js         cacheLinks() + enhance(), works in browser and Node
-  content.js      injects styling, runs enhance(), re-runs as results stream in
-  popup.html      toolbar popup
-  icons/
-core.test.js      node core.test.js
-```
+Node and Python are enough for the release checks.
 
-```bash
-node core.test.js
+```text
+node tests/core.test.js
+node tests/content.test.js
+python scripts/package.py
+python tests/site.test.py
+python tests/package.test.py
 ```
 
-The result walk is anchored on each result's `<h3>` title and then its link,
-rather than on Google's build-hashed class names, so a redesign doesn't silently
-break it.
-
-## Limits
+`scripts/package.py` writes a sorted ZIP with fixed timestamps, then writes its
+SHA-256 checksum. `tests/package.test.py` builds twice in temporary directories
+and compares the bytes.
 
-Covers the main Google TLDs (com, co.uk, ca, de, fr, in, br, ph, jp and others).
-If your local Google domain isn't in the manifest, add one line and reload.
+## Google markup limits
 
-Archives don't have every page. If the Wayback Machine never captured a URL its
-snapshot page says so, and you can use save now to capture it going forward.
+This extension is a page repair, not an official Google API client. The archive
+feature looks for result title links under `#rso`. Pagination uses the `start`
+query parameter and extracts result blocks from the next page. Google can change
+either behavior without notice.
 
-Reads organic result links only. Ads, Google's own properties and image
-thumbnails are skipped.
+A read-only check on August 23, 2026 found one `#rso` container and seven linked
+`h3` result titles for the query `example`. The same check found no captcha. See
+[MAINTENANCE.md](MAINTENANCE.md) for the selectors and failure cases that need
+watching.
 
-## More tools
+## Files
 
-- [Carryover](https://github.com/cig13zs/carryover), AI chat context transfer for ChatGPT, DeepSeek and Grok
-- [Invisibles](https://github.com/cig13zs/invisibles), reveal and strip hidden Unicode from text
-- [Rinse](https://github.com/cig13zs/rinse), see the GPS in a photo and wash it off
-- [Return 100 Results](https://github.com/cig13zs/return-100-results), browse ~100 Google results as one page
+```text
+extension/           Chrome MV3 package source
+scripts/package.py   deterministic ZIP and checksum builder
+store-assets/        Chrome Web Store artwork and screenshots
+tests/               core, content, manifest, page and package checks
+```
 
-Not affiliated with Google or the Internet Archive. MIT licensed.
-[Ko-fi](https://ko-fi.com/jju1s) if you want to.
+Not affiliated with Google, the Internet Archive or archive.today. MIT licensed.

STORE-LISTING.md

@@ -0,0 +1,47 @@
+# Chrome Web Store listing
+
+## Name
+
+Search Restore
+
+## Summary
+
+Add archive links to Google results and load later result pages into the same list.
+
+## Description
+
+Search Restore puts two optional controls on supported Google Search pages.
+
+Archive links appear under normal web results. They open the latest Wayback
+Machine capture, archive.today or the Internet Archive save page.
+
+The load-more button requests the next Google results page and places its normal
+results below the current list. It loads one page per click and stops near 100
+results, when no more results are found or when Google asks for verification.
+
+Both features have separate on/off switches. Preferences stay in local Chrome
+extension storage. Search Restore has no analytics, advertising, account system
+or extension server.
+
+Google page markup changes over time, so some result layouts may not be handled.
+The project documents its current selectors and tests in the public repository.
+
+Search Restore is not affiliated with Google, the Internet Archive or
+archive.today.
+
+## Category
+
+Tools
+
+## Permission justification
+
+`storage` saves the two feature switches on the device. No other named
+permission is requested. The content script is declared only for the listed
+HTTPS Google Search paths and the manifest has no separate `host_permissions`.
+Chrome may still describe that declared content-script scope as access to the
+listed Google sites; reading and updating the visible results page is the
+extension's single purpose.
+
+## Privacy policy
+
+https://cig13zs.github.io/search-restore/privacy.html