• Fixes missing code samples on docs-heavy sites like ai-sdk.dev by detecting when Readability strips <pre> blocks and falling back to the original DOM, so conversions now keep syntax-highlighted examples intact @src/content.js#238-320.
• Keeps extension/package metadata synchronized at v3.4.2 so downloads and builds line up with the published artifact @manifest.json#1-12, @package.json#1-9.

• update manifest version from 3.3 to 3.4
• add newline at end of file

• add mermaid SVG preprocessing and fallback handling

• Frontmatter now includes a human-friendly source (e.g., “The New York Times”) instead of only the domain/hostname.
• Frontmatter includes author directly under title for consistency.
• Saved Markdown now auto-adds a top-level # {title} heading when the converted page content doesn’t start with a heading (prevents “title-less” notes on some sites).

• Improved article parsing
• Switched content extraction to DOMParser + Readability for more reliable detection of main article content.
• Updated bundled readability.js to the latest 0.6.0 release for better handling of modern sites.
• Better Markdown generation & download flow
• Refactored the markdown conversion and download logic for improved stability and maintainability.
• Reduced edge‑case failures when triggering the “save as Markdown” flow.
• Add-on compatibility & manifest fixes
• Updated Gecko extension ID to a UUID for compliance with current Firefox add-on requirements.
• Adjusted data_collection manifest entries to match the latest schema and removed unused fields.
• Documentation
• Expanded README with more detailed usage and configuration instructions.

This version updates content parsing to use DOMParser with the latest Readability (0.6.0), and refactors the markdown conversion/download flow.

The manifest has been updated to comply with current Firefox requirements:
- New UUID-based browser_specific_settings.gecko.id.
- Adjusted data_collection fields to the current schema; no data is actually collected or transmitted.

The extension still operates entirely client-side and only processes the currently open page to generate a local Markdown file. No remote services are contacted and no telemetry is collected.