unstale

pypi / trafilatura

trafilatura

What changed in trafilatura 2, from its official release notes, upgrade guides and advisories. Sources checked 2026-10-21.

1 deprecations2 new APIs and options2 performance6 behavior and defaults36 fixes and other

See which of these apply to your code, ranked, in your agent:

npx @unstale/cli init

@unstale/cli 0.2.1 on npm

trafilatura 247 changes

Deprecations 1

New APIs and options 2

Performance 2

Behavior and default changes 6

  • Boilerplate pruning is more targeted to reduce over-extraction.

    2.2.0release notes

  • Subelem processing was further improved, mostly for math, code, and Markdown.

    2.2.0release notes

  • The text recovery and extraction sequence was revamped for better overall recall and forum extraction.

    2.2.0release notes

  • Title extraction was improved.

    2.2.0release notes

  • Spacing is improved in item, cell, and code blocks.

    2.1.0release notes

  • The detector supports all Justext languages known to it.

    2.1.0release notes

Fixes and other changes 36

  • Brackets inside inline code in link text are no longer escaped.

    2.3.1release notes

  • The output contains fewer duplicated segments.

    2.3.1release notes

  • Image paragraphs are no longer repeated in the output.

    2.3.1release notes

  • Text after inline code is no longer repeated in the output.

    2.3.1release notes

  • Document handling was simplified and made more correct.

    2.3.0release notes

  • Downloads and decompression were overhauled, with more robust handling of relative URLs and zstd frames.

    2.3.0release notes

  • Link discovery improved for sitemap namespaces, Atom links, and JSON feeds.

    2.3.0release notes

  • Markdown output improved for spacing, bold spans, CommonMark escaping, and LaTeX math.

    2.3.0release notes

  • Metadata handling improved for word boundaries, JSON-LD types, OpenGraph, license links, list values, canonical URLs, and TXT/Markdown headers.

    2.3.0release notes

  • Fallbacks now preserve requested images.

    2.3.0release notes

  • Scoped caches are now accepted.

    2.3.0release notes

  • Text extraction improved for forms, link farms, inline code and links, infinite scroll, boilerplate widgets, nested lists, and text after blocks.

    2.3.0release notes

  • Image extensions are matched case-insensitively.

    2.2.0release notes

  • Metadata values are quoted so the Markdown header remains valid YAML.

    2.2.0release notes

  • Angle brackets in Markdown link destinations are escaped.

    2.2.0release notes

  • A space after inline formatting is preserved in Markdown list items.

    2.2.0release notes

  • List items whose only content is a link are retained.

    2.2.0release notes

  • Superscript and subscript boundaries are preserved in Markdown output.

    2.2.0release notes

  • Processing for nested elements was improved.

    2.2.0release notes

  • Text inside nobr tags is retained.

    2.2.0release notes

  • Output formatting bugs were addressed.

    2.2.0release notes

  • Text in sub-elements is handled better.

    2.2.0release notes

  • Table processing is more accurate.

    2.2.0release notes

  • Question headers in Yoast FAQ blocks are retained.

    2.2.0release notes

  • None values in code blocks are replaced.

    2.1.0release notes

  • Fixes eager matching by the doctype regular expression.

    2.1.0release notes

  • Code is changed to handle potential empty strings.

    2.1.0release notes

  • The fencedframe element is cleaned during extraction.

    2.1.0release notes

  • The tokenizer now strips the Ideographic Full Stop.

    2.1.0release notes

  • Images are preserved when links are included.

    2.1.0release notes

  • Refines image source URL handling and fixes table extraction bugs.

    2.1.0release notes

  • Links in tables are preserved when --links is enabled.

    2.1.0release notes

  • The no_fallback setting is propagated to the fast argument.

    2.1.0release notes

  • Fixes an AttributeError that node pruning could raise in certain cases.

    2.1.0release notes

  • Code in pre-tags is detected using indicators, and code formatting issues are fixed.

    2.1.0release notes

  • Table extraction is made more robust.

    2.1.0release notes