Repository navigation
Releases
Web-Obs 0.3.3
Publish no longer refuses two kinds of report it should create. Both were reported by @gcollins808.
Fixed
- Crawl traps no longer block Publish. The "Possible crawl traps" finding stored a path such as
/calendarwhere a full address belongs, and the report check refused it. New captures store the full address. - The site's own words are not treated as promises. A brand name, page title or page address containing a word like "guarantee" stopped the report. The promise check now covers only what the report itself says.
Updating
- Captures you already have publish as they are. No new crawl is needed.
- Nothing changes in how the map, the audit or the capture engine look or work.
Full Changelog: 0.3.2...0.3.3
Web-Obs 0.3.2
Web-Obs 0.3.1
Web-Obs 0.3.0
What's Changed
- Wait for the map to go quiet before the render test counts by @AgriciDaniel in #7
- Clusters stability and Road scaling from gcollins808, with two follow-ups by @AgriciDaniel in #8
- Scale the Road view to the site and let Road and Clusters zoom out fully by @gcollins808 in #5
- Keep the Clusters layout stable on densely linked sites by @gcollins808 in #4
- Fix clean CI installs and browser regression timing by @AgriciDaniel in #12
New Contributors
- @gcollins808 made their first contribution in #5
Full Changelog: 0.2.2...0.3.0
Web-Obs 0.2.2
Web-Obs 0.2.2
Safer page reading, cleaner logos in notes, and a proper cancel on Windows.
Fixed
- Closing tag text inside a page is left alone. 0.2.1 deleted the text
</body>or</html>wherever it appeared, for example in a title such as "How to close tags". Only the page's real closing tags are moved now. - A logo inside a link stays an image in the note, instead of turning into text in the link label.
- Cancelling a capture on Windows now lets the engine clean up, and ends its workers and browser too if it does not stop in time. The engine also stops if Obsidian quits during a capture.
Changed
- A more careful explanation of the 0.2.1 fix: on GitHub's Windows runner, lxml 6.1.3 with its bundled libxml2 2.11.9 dropped content after
</html>. The earlier notes blamed every libxml2 before 2.14, which is broader than what was shown. - Every automated check now passes on Windows, as on Linux and macOS.
Needs Obsidian 1.5.0 or newer on desktop, and Python 3.14 or newer (see the README for the short setup).
Install Download main.js, manifest.json and styles.css from this release into YOUR-VAULT/.obsidian/plugins/web-obs/. Web-Obs unpacks its capture engine on first start. The zip holds the same files if you prefer one download.
Full details are in the changelog.
Web-Obs 0.2.1
Web-Obs 0.2.1
Windows captures now see the whole page.
Fixed
- Links, images and text that a page places after its closing
</html>tag are captured on every system, as browsers show them. On Windows they were lost, because the lxml package there bundles an older libxml2 that drops them. This was the known Windows image problem.
Changed
- The README shows the Windows setup: the PowerShell commands and the
Scripts\python.exepath for the settings.
Needs Obsidian 1.5.0 or newer on desktop, and Python 3.14 or newer (see the README for the short setup).
Install Download main.js, manifest.json and styles.css from this release into YOUR-VAULT/.obsidian/plugins/web-obs/. Web-Obs unpacks its capture engine on first start. The zip holds the same files if you prefer one download.
Full details are in the changelog.
Web-Obs 0.2.0
Web-Obs 0.2.0
A much deeper audit, still fully local.
New
- Around 40 new page checks: thin content, readability, duplicate meta tags, mixed content, render-blocking resources, DOM size, URL quality, social tags, basic accessibility and more.
- A site audit: duplicate and near-duplicate pages, a fix list for internal links, internal link opportunities, conflicting canonical and noindex signals, pages competing for one topic, crawl traps, stale pages, trust pages and structured data checked against Google's rich result requirements.
- Certificate expiry, HSTS, Content Security Policy and cookie checks. Optional email DNS checks (SPF, DMARC, DKIM, CAA).
- A browser sample of up to five pages: lab timings, automated accessibility rules (axe-core), screenshots, JavaScript dependence and third parties on first load.
_audit.md,_findings.csv, a Page Audit section in every note, and_tasks.md, a checklist that remembers your ticks and closes tasks the next capture no longer finds.- A command to compare two captured sites side by side.
- Notes that mirror the page: header, navigation and footer kept as labelled regions, readable media and link sections, and a full-page snapshot of the most important pages at the top of each note.
- A calmer interface: a slim toolbar, tabbed Site Intel and page panels, charts in the Site audit, and a 3D map with fewer, non-overlapping labels.
- The site's logo at the centre of the map, ringed in its brand colour, including on sites you captured before.
- An honest crawler: the user agent is now
Mozilla/5.0 (compatible; WebObs/0.2.0; +https://github.com/AgriciDaniel/web-obs/blob/main/docs/CRAWLER.md), robots.txtCrawl-delayis honoured, and the capture backs off when a site answers 429 or sendsRetry-After. Site owners can read what WebObs sends.
Optional packages dnspython, pyspellchecker and html5lib unlock email DNS, spell check and HTML validation. Web-Obs works without them.
Needs Obsidian 1.5.0 or newer on desktop, and Python 3.14 or newer (see the README for the short setup).
Install Download main.js, manifest.json and styles.css from this release into YOUR-VAULT/.obsidian/plugins/web-obs/. Web-Obs unpacks its capture engine on first start.
Full details are in the changelog.
Web-Obs 0.1.1
Web-Obs 0.1.1
Accuracy fixes from a full capture of a live WordPress site.
Fixed
- Sites that use trailing slashes (like WordPress) no longer show nearly every page as a redirect. Pages are fetched and saved at the address the site publishes.
- Pages missing from the sitemap are counted even when they also redirect.
- A sitemap that several addresses lead to is counted once.
Changed
- Response time now measures the final page request only, so redirects no longer inflate it.
- Empty alt text on decorative images is no longer an error. Only a missing alt is.
- Heading level skips in headers, menus and footers are counted apart from the page content.
- A broken image or page found during a capture is listed as a finding and no longer marks the whole capture as partial.
- Internal links written without the site's trailing slash are listed on the page that contains them.
Needs Obsidian 1.5.0 or newer on desktop, and Python 3.14 or newer (see the README for the short setup).
Install Download main.js, manifest.json and styles.css from this release into YOUR-VAULT/.obsidian/plugins/web-obs/. Web-Obs unpacks its capture engine on first start. The zip holds the same files if you prefer one download.
A site first captured with 0.1.0 may show its addresses as changed on the first recapture. Full details are in the changelog.
Web-Obs 0.1.0
Web-Obs 0.1.0
First release. Type a website address and Web-Obs saves its pages as Markdown notes in your vault, then draws the whole site as an interactive 3D globe.
Highlights
- A clean note for every page, with images saved locally
- Globe, Road and Clusters views of the site structure
- A health card (crawlable, searchable, AI ready, secure, links, speed signals) and a checklist of discovery files such as sitemap, robots.txt and llms.txt
- A site audit with charts and findings that link back to the source page
- A built-in diagnostics command that checks your setup and explains how to fix problems
- Everything runs locally. No account, no tracking, no AI service, no API key.
Needs Obsidian 1.5.0 or newer on desktop, and Python 3.14 or newer (see the README for the short setup).
Install Download main.js, manifest.json and styles.css from this release into YOUR-VAULT/.obsidian/plugins/web-obs/. Web-Obs unpacks its capture engine on first start. The zip holds the same files if you prefer one download.
Full details are in the changelog.