Boundier

Browser extension that scores attention, emotion, framing, and source-pressure signals in web content.

Comment

Boundier

Boundier is a local-first influence-pressure analysis tool for webpages, social feeds, and video page metadata.

It analyzes visible language patterns with deterministic Rustmeter scoring and provides evidence-linked signal explanations.

Evidence model

Boundier does not only show a score. It surfaces the signals that contributed to the score, including the detected phrase or pattern, the category it affected, and a short reason.

The “Why” section explains the main evidence behind the Rustmeter score. For low scores, it may also explain which stronger pressure signals were not detected. A low score means the page may contain a few mild cues, but not enough evidence for high influence pressure.

What Boundier does

  • Scores visible page text locally by default.
  • Reports Rustmeter score and category subscores.
  • Highlights propaganda-like patterns in wording pressure.
  • Supports article, social, video, and general webpage surfaces.

What Boundier does not do

  • It does not judge objective truth.
  • It does not classify misinformation or disinformation.
  • It does not infer author intent.
  • It does not classify people.
  • It does not analyze full video or audio unless transcript or visible text is present on the page.

Extraction and limitations

  • Extraction is local and uses targeted selectors, Mozilla Readability (bundled locally), and adaptive readable-block fallback.
  • Extraction depends on visible text and can still fail on blocked pages, heavy script rendering, paywalls, or image-only pages.
  • Video support analyzes page title, metadata, description, visible text, and available transcript/page text when present.

Privacy and local-first behavior

  • Primary scoring runs locally in the extension.
  • No hosted AI API is required.
  • No telemetry or analytics is added.
  • Optional backend is for localhost experiments only.

Permissions

  • activeTab: allows analysis of the current tab when the user clicks the extension.
  • scripting: allows attaching content scripts when needed for analysis.
  • storage: stores local analysis cache and settings.

Optional backend

The backend is separate and optional for local experiments.

cd backend
pip install -r requirements.txt
python app.py

Testing

npm test
python -m pytest backend/tests/test_scoring.py

Built With

Share this project:

Updates