Notas de lançamento

WProofreader core v6.19.0 adds Japanese support to the AI writing assistant, improves Dutch spelling suggestions for users with dyslexia, enhances AI-based proofreading accuracy for German and Spanish, patches several security vulnerabilities, and adds the option to raise the rule limit per collection above 2,000 upon request

🛠 Melhorias

Localizations

• Added interface language list localization for Kurdish (Turkey) and Thai.
• Updated interface language list localization for Valencian.

Revisão baseada em IA

• Added a new German style rule for decapitalizing informal pronoun capitalizations such as Du, Dich, Ihr, and Eure when appropriate.
• Improved German AI-based proofreading to provide better performance on punctuation in large numbers, subject-verb agreement handling, pronouns in fixed expressions and gender-inclusive suggestions. Also, refined handling of words with nonbreaking hyphens.
• Enhanced Spanish AI-based proofreading to have better performance on missing question and exclamation marks, run-on sentence detection, and symbol corrections such as +/- → ±.

Gramática

• Updated LanguageTool to the October 12, 2026 version. This includes all engine changes since the September 11, 2026 update.

Ortografia

• Expanded the German dictionary with 850+ new words, covering common vocabulary, specialized fields, business, legal, military and culinary such as eigenkapitalstark, Wendeplatte, Konditoreibetrieb and Gerüstbauarbeiten.
• Expanded the Portuguese dictionary with 750+ new words covering administrative, common, and proper-name vocabulary such as Maranhão, Catarina, reexaminar and polissonografia.
• Enhanced German and Spanish medical terminology with 40+ new terms such as Katheterpflege, Kreatininwerte, NTHi.
• Enriched shared cross-language dictionaries with 60+ new proper-name entries for companies, products, and locations such as Nintendo, Panamera, Rotterdam, and iOS.
• Improved spelling suggestion prioritization for Dutch with patterns designed to support users with dyslexia, covering cases such as zecht → zegt, leew → leeuw, etc.
• Improved Dutch spellchecking for hyphenated words, such as compounds with vowel combinations or joining -s-. Correctly written words like these are no longer flagged as errors.
• Improved suggestions when two words are typed without a space, such as prepositions, conjunctions, or pronouns glued to other words. For example, tojest is now split into the correct words in Polish. This improvement covers 40 languages, including German, Spanish, Portuguese, French, Swedish, Polish, Croatian, Slovenian, Turkish, Welsh, and Irish.
• Improved suggestions from the third-party grammar engine for Greek. Spelling and style issues are now labeled and highlighted correctly, instead of all appearing as grammar issues.

Assistente de escrita com IA

• Added Japanese support to the AI writing assistant.

Guia de estilo

• Add a possibility to increase the number of rules per collection to more than 2000 rules upon request. Contact [email protected]

Idiomas

• Updated the Valencian Hunspell dictionary that follows the standard Valencian spelling by AVL convention. Dictionary maintainer: Jaume Ortolà. Version: v3.0.9 (18-05-2025). Source: https://github.com/Softcatala/catalan-dict-tools/releases/tag/v3.0.9

🐞 Correções de erros

• Style guide: Custom rules using a single punctuation character, such as & → und, only matched when spaces were added around the pattern.

🛡️ Segurança

• Updated dependencies in the AI proofreading engine’s model-serving pipeline to reduce security risk: torch (2.5.1 → 2.9.1), transformers (4.48.1 → 4.57.6), and accelerate (1.3.0 → 1.15.0).
• Reduced the attack surface of the third-party grammar engine (LanguageTool) by removing 15 unused third-party libraries, including unneeded networking, caching, and cryptography components.
• Updated Logback, used by the application server, from 1.6.3 to 1.6.5 to fix a known vulnerability: SNYK-JAVA-CHQOSLOGBACK-20824362.
• Updated Guava, used by the application server, from 33.5.0-jre to 33.7.2-jre to fix a known vulnerability: SNYK-JAVA-COMGOOGLEGUAVA-20825054.

WProofreader core v6.18.1 improves AI-based proofreading accuracy, adds visible descriptions for AI-based spelling corrections, and makes feature settings reflect actual availability for the selected language and configuration. It also expands dictionaries across multiple languages, including a specialized English veterinary dictionary, adds a German option to require the eszett (ß) instead of ss, and improves handling of hyphenated words, including words that start with a hyphen. This release also adds an “Apply all” button for accepting every suggestion at once, and fixes several security vulnerabilities.

📝 TL;DR

• Improved AI-based proofreading for Spanish and German.​
• Feature settings now reflect actual availability for the selected language and configuration.​
• Spelling issue descriptions now show a specific message for AI-based spelling corrections in English, Spanish and German.​
• Lowered AI writing assistant minimum text length to 30 characters.
• Added a German option to require the eszett (ß) instead of ss, and improved handling of hyphenated words, including words that start with a hyphen.​
• Added a specialized English veterinary dictionary, disabled by default.​
• Fixed a French bug that incorrectly suggested removing nonbreaking spaces before punctuation.​
• Added new API commands and server configuration options for administrators, including retrieving user dictionaries and checking supported languages.​
• Added an “Apply all” button to accept every suggestion at once in Proofread in dialog mode.
• Expanded dictionaries and medical terminology across multiple languages. ​
• Updated Netty, MariaDB Java client, logback-classic, and Jackson Databind to fix security vulnerabilities.

🛠 Melhorias

WProofreader JS core (v3.46.6137)

• Improved the feature settings experience. Controls now reflect actual feature availability for the selected language and configuration.
• Features that are unsupported or disabled on the backend can now appear as unavailable instead of showing as enabled.
• Added visible spelling suggestion descriptions. Classified AI-based spelling corrections now show a specific message in the UI in English, Spanish and German.
• Improved issue count accuracy in badge. Corrections disabled through the disabledRules options no longer count as mistakes.
• Added an “Apply all” button in Proofread in dialog mode that accepts every found suggestion at once, after you expand the dialog window fully.

API and configuration

• Extended the get_info API with information about feature availability, including autocomplete, autocorrect, and style, allowing integrations to better reflect backend configuration.
• Added an API getdicts action for retrieving user dictionaries, letting administrators list all dictionaries in an account. Each response returns up to 1,000 dictionaries. If more exist, a warning indicates that the list is truncated.

Localizations

• Added interface language list localization for Kurdish (Turkey) and Thai.

Servidor de aplicações

• Removed the deprecated AdditionalLang server configuration option, which no longer had any effect.
• Added support for disabling or enabling additional dictionaries individually in the server configuration, such as veterinary dictionary.

Revisão baseada em IA

• Enhanced Spanish AI-based proofreading to have better performance with underscores, unit changes and decimal separators.
• Improved German AI-based proofreading to provide better performance on texts with parentheses and non-standard quotation marks written as >><< using greater-than and less-than signs.

Gramática

• Added support for disabling individual proofreading rules or rule categories on the server side, so, for example, dash rules can be turned off correctly in every instance for cloud users.
• Improved English grammar processing for capitalization patterns in technical identifiers (such as ID casing) and common word confusion pairs such as sou versus you.
• Improved the handling of Swedish bil, art, and kol so if they are used at the end of the sentence and a following word is written in lowercase, we would suggest capitalizing that word. For example, Hon körde sin bil. vägen var nästan tom., the suggestion is Vägen.

Ortografia

• Expanded dictionaries with 840+ new words across English, German, Spanish, Portuguese, Italian, Dutch, French, and Greek, covering abbreviations, technology terms, common words, proper names, and linguistics terms.
• Enriched cross-language dictionaries with 150+ proper names and abbreviations, covering companies and software applications such as Etsy, Depop, WebText, and USB-C.
• Enhanced medical terminology for English, German, and Spanish with 120+ new terms, such as trichogram and hyporexia in English, anatómicamente in Spanish, and Implantatprüfung in German.
• Added an English veterinary dictionary, disabled by default, covering 180+ terms and abbreviations such as GDV, oclacitinib, and Drontal. It covers all available English dialects and administrators can enable it as needed in self-hosted set-up. For cloud users, now it can be provided upon request to [email protected]. It incorporates terms from AGROVOC (FAO) combined with additional entries.
• Extended measurement-unit abbreviation spelling suggestions, such as 2GB to 2 GB and 3GHz to 3 GHz, to uppercase and mixed-case abbreviations across many Latin-alphabet languages.
• Added a rule for German (de_DE, de_AT) that lets administrators require the eszett (ß) instead of ss. By default, both forms are accepted; disabling ALLOW_SS_INSTEAD_OF_ESZETT makes ss spellings incorrect and suggests the ß form instead, and AI suggestions follow the same setting.
• Improved handling of hyphenated words. Words that start with a hyphen, such as -anforderungen, are now recognized as correct when added to a dictionary with a leading hyphen, for most supported languages, except a few that don’t use hyphens this way, such as Arabic, Hebrew, and Japanese.

Estilo

• Enhanced style rule precision in Spanish by adding contextual awareness to foreign-word suggestions. Suggestions for terms like power now exclude medical contexts such as Doppler, based on user feedback.

Assistente de escrita com IA

• Reduced the minimum text length required to use the AI writing assistant from 100 to 30 characters, making AI-powered text actions available for shorter text.

Autocomplete

• Improved autocomplete language handling by using the detected language as a fallback when automatic language detection has insufficient context, such as with very short text.
• Improved authorization handling so autocomplete and autocorrect respect the languages enabled for a subscription.
• Updated autocomplete authorization so it’s available only when at least one applicable check type is enabled for the language.
• Improved autodetect authorization logic, so requests aren’t incorrectly rejected when an applicable check type is available for an authorized language.
• Continued improving the filtering of autocomplete suggestions that contain sensitive, unexpected, or discriminatory language.

🐞 Correções de erros

• Spelling engine (French): Suggestions incorrectly recommended removing the narrow nonbreaking space before punctuation, such as question marks and colons
• Spelling engine: Hyphenated words following a number, such as 5-wöchigen or 16-years-old, were incorrectly flagged as misspelled
• Spelling engine: Users couldn’t add abbreviations with doubled letters, such as PPP or CCCN, to custom dictionaries

🛡️ Segurança

• Updated Netty to 4.2.17.Final, addressing multiple vulnerabilities in the application server’s networking component, including CVE-2026-75595.
• Updated the MariaDB Java client from 3.4.1 to 3.4.3, addressing CVE-2026-55856, CVE-2026-55857, and CVE-2026-55858 in the application server’s database connectivity component.
• Updated logback-classic from 1.5.38 to 1.6.3 to fix a directory traversal vulnerability in the application server’s logging component.
• Updated Jackson Databind from 2.18.9 to 2.18.10, addressing multiple vulnerabilities in the application server, including unsafe reflection, untrusted deserialization, and uncontrolled resource allocation.

WProofreader core v6.17.1 fixes an issue where native browser spellcheck didn’t turn back on correctly during startup.

🐞 Correções de erros

• WProofreader JS core: When autoStartup was set to false, the restoreNativeSpellCheck option was ignored on initialization, leaving native browser spellcheck disabled. Native browser spellcheck now restores correctly in this configuration. Review configuration options para mais informações.

WProofreader core v6.17.0 introduces the German AI-based proofreading engine v3.1 with improved grammar and punctuation corrections, extends AI proofreading correction descriptions and categories to German and Spanish, expands dictionaries with 970+ new words, updates the algorithmic grammar engine with Portuguese and Italian refinements, improves AI writing assistant prompts, and addresses several security vulnerabilities.

📝 Resumo

• Released German AI-based proofreading engine (v3.1) with improved corrections
• Extended categories for AI-based suggestions to German and Spanish
• Expanded dictionaries with 970+ new words across five languages, plus 170 shared proper names and abbreviations
• Enhanced medical terminology for English and Spanish; expanded English dialect support
• Updated third-party grammar engine with refinements for Portuguese and Italian
• Security updates for Micrometer and Netty to address memory leaks and certificate validation issues

🛠 Melhorias

Revisão baseada em IA

• Released the German AI-based proofreading engine (v3.1) with improved accuracy for separable verbs, modal verbs in simple past tense, passive voice construction, noun-verb agreement, and punctuation corrections.
• Extended categories for AI-based suggestions to German and Spanish, bringing unified rule categories and descriptions to these languages. Users now receive consistent, clear explanations for suggested corrections in German and Spanish, matching the quality of English feedback.
• Refined additional improvements across languages: punctuation handling in English, quotation mark styles and en dash spacing in German, and ordinal numbers, em dashes, spacing, and colons in Spanish.

Gramática

• Updated LanguageTool to the version 2026-08-20.
• Updated Portuguese grammar logic with a modified rule. The rule handles the correct use of hyphens in words formed with prefixes under the Portuguese Orthographic Agreement. Extended the exception list to recognize the nonbreaking hyphen and U+2010 hyphen, preventing valid hyphenated forms from being incorrectly flagged.
• Updated Italian grammar logic with 6 modified rules. Fixed article rules for anglicisms where the plural can share the same written form as the singular, such as meeting, standard, and disambiguated rules for homophones, such as hai vs ai, ha vs a.

Ortografia

• Expanded dictionaries with 970+ new words across English, German, Spanish, Portuguese, and Italian, covering abbreviations, affixes, anglicisms, biology, business, common terms, economics, food, legal terms, linguistics, location, measurements, medical, neologisms, other languages, proper names, religion, science, specialized terminology, and technology.
• Enriched shared cross-language dictionaries with 170 proper names and abbreviations, such as XOF, OpenOffice, Eurocontrol, etc.
• Enhanced medical and scientific terminology for English and Spanish with 64 new words, including terms such as bromobenzoates and fCJD in English, and urotelio in Spanish.
• Expanded English dialect support with variants for medical terminology, such as bunaftine (American) and bunaphtine (British).
• Refined autocorrect with new pairs across English and Spanish, targeting frequent misspellings, such as multipanares → multiplanares.

Assistente de escrita com IA

• Improved English prompts for more accurate dialectal corrections during proofreading.
• Updated prompts across all languages to preserve emojis when applying text corrections, ensuring emoticons and emoji characters remain unchanged during proofreading operations.

🛡 Segurança

• Micrometer upgrade: Updated Micrometer from 1.15.12 to 1.16.7 to address two Snyk-reported vulnerabilities. The update fixes a high-severity memory leak in metric collection that could affect registry stability. Also updated the Prometheus integration dependency to 1.16.7 for compatibility.
• Netty upgrade: Updated netty-handler from 4.1.136.Final to 4.1.137.Final to address two Snyk-reported vulnerabilities. The fixes address a high-severity hostname mismatch issue in certificate validation and a critical exception handling flaw. This update benefits all users relying on secure network communication.
• VAPT: Updated the report v6.17.0

O WProofreader core v6.16.0 introduz um novo sistema de licenciamento com concessões de produto, limites de utilização e restrições de versão, melhora a fiabilidade das correções gramaticais em inglês, atualiza o motor gramatical de terceiros, corrige uma vulnerabilidade de segurança de gravidade elevada e amplia os dicionários ortográficos com mais de 1 960 novas palavras em nove idiomas.

📝 Resumo

• Foi introduzido um novo formato de licença v4 com concessões de produto, limites de utilização e restrições de versão
• Os clientes com instalação própria devem reativar a sua licença após a atualização para a versão 6.16.0 (ver alterações significativas abaixo)
• Corrigiu-se uma vulnerabilidade de segurança de gravidade elevada, CVE-2026-40984, numa dependência do LanguageTool
• Foi adicionado o novo license_status comando e relatórios e avisos melhorados sobre o fim da validade das licenças
• Atualizou-se o motor de gramática de terceiros (LanguageTool) para a versão de 03/08/2026
• Dicionários ortográficos ampliados com mais de 1 960 novas palavras em nove línguas

⚠️ Alteração importante para clientes com alojamento próprio

• Os clientes com instalação própria devem reativar a sua licença após cada atualização de versão do WProofreader, a partir da v6.16.0. Isto confirma que a versão atualizada cumpre as restrições de versão da licença. Reative a sua licença imediatamente após a atualização para evitar a interrupção do serviço. Consulte a documentação sobre a ativação da licença no Linux, Windows ou Docker para obter mais informações.

🛠 Melhorias

Licenciamento

• Foi introduzido um novo formato de licença v4 que suporta concessões de produtos, limites de utilização e restrições de versão. As licenças definem agora diretamente os direitos de acesso aos produtos e a elegibilidade para atualizações.
• Foi adicionado o novo license_status Comando da API para recuperar informações sobre a licença. Revisão Documentação relativa ao estado da licença para mais informações.
• Melhoria na apresentação de informações sobre o prazo de validade da licença, de modo a indicar o tempo exato restante em dias, horas, minutos e segundos. As mensagens de estado utilizam agora a formulação mais clara «A licença expira em ».
• Foi adicionado um aviso de expiração da licença nas últimas 24 horas antes da data de expiração, servindo como um lembrete adicional para renovar a licença antes da interrupção do serviço.

Implementação

• Por vezes, os contentores Docker não conseguiam ativar a licença após o reinício com um novo bilhete de licença e um ficheiro license.xml vazio. Agora isto funciona corretamente, inclusive em configurações em que o diretório da licença é armazenado num volume montado.

Gramática

• Atualizou-se o motor de gramática de terceiros (LanguageTool) para a versão de 03/08/2026.
• Regras gramaticais de inglês melhoradas, com correções mais claras e menos falsos positivos:
  • Corrigiu uma sugestão errada de vírgula antes e em frases que utilizavam a expressão fixa «de vez em quando»
  • Deixou de assinalar incorretamente a palavra «pretrain» como se faltasse um infinitivo, uma vez que esta pode funcionar como substantivo em textos especializados
  • Deixou de assinalar incorretamente o uso de «am» em frases que incluem um nome ou título no meio, como em «Eu, John Smith, Diretor de Marketing da XXX, escrevo para confirmar…»
  • Foram adicionadas sugestões para escrever com maiúscula e pontuar as interjeições no início de uma frase: ahh finalmente → Ahh, finalmente
  • Foram adicionadas sugestões para pontuar frases que começam com uma interjeição seguida de «that»: «wow, isso foi por pouco » → «Wow, isso foi por pouco»

Ortografia

• Dicionários ampliados com mais de 1 960 novas palavras em inglês, alemão, espanhol, árabe, norueguês, italiano, albanês, grego e português. As novas palavras abrangem vocabulário quotidiano, científico e especializado.
• Enriquecemos os dicionários multilingues partilhados com 26 novos nomes próprios e abreviaturas, abrangendo termos das áreas do software, jurídica, médica e tecnológica, tais como AgentCore, GPLv2 e eSIM.
• Vocabulário médico alargado para o inglês, o alemão e o espanhol, com mais de 820 novos termos, abreviaturas e nomes de doenças, tais como «cryofixation» (inglês), «Leukozyten» (alemão) e «sonografía» (espanhol).
• Ampliação da cobertura dos dialetos ingleses com 10 novos pares de palavras, tais como «conization » e «conisation».
• Correção automática e sugestões ortográficas melhoradas para o árabe, o norueguês, o italiano, o espanhol, o alemão e o inglês, com foco em erros ortográficos comuns e variantes dialetais, como «trøt» → «trøtt» em norueguês.

🛡 Segurança

• Atualizado micrometer-core da versão 1.7.12 para a 1.15.12, para corrigir CVE-2026-40984, uma vulnerabilidade de negação de serviço de gravidade elevada que afeta as dependências do LanguageTool incluídas no WProofreader.
• Também mudei para micrometer-registry-prometheus-simpleclient para preservar a compatibilidade com a integração existente com o Prometheus.

WProofreader core v6.15.0 adds spelling correction support for Thai, introduces the applyAllSuggestions method for applying every suggested correction in one call, and improves suggestion descriptions across the grammar and AI-based proofreading engines. This release also expands dictionaries with 1,200+ new words and terms across nine languages, updates the German spelling, and includes Docker build improvements and fixes for multiple security vulnerabilities.

📝 Resumo

• Added spelling correction support for Thai.
• Introduced the applyAllSuggestions method for applying every suggested correction in a document with a single call.
• Improved suggestion classification across the grammar and AI proofreading engines, including new style-issue detection.
• Expanded dictionaries with 1,200+ new words and terms across nine languages.
• Missing-period suggestions are now disabled by default in headings and list items.
• Improved Docker build compatibility across Kaniko, Buildah, and the classic Docker builder.
• Security fixes. Updated the Rangy, jackson-databind, and ch.qos.logback libraries to address CVE-2023-26102, CVE-2026-54515, CVE-2026-9828, and CVE-2026-10532.

WProofreader JS core (v3.44.6097)

• Added a new applyAllSuggestions command that accepts every suggested correction in one go, instead of applying them one by one. It’s available now for developers to add to their setup; a ready-to-use button in the interface is coming in a future release.
• Missing-period reminders no longer appear in headings and list items by default, cutting down on suggestions that don’t apply in these contexts. Review disabledRulesPerEnvironment no configuration documentation for more information.
• Improved how German abbreviations with spaces and capital letters (such as z. B.) are recognized, so they no longer trigger incorrect capitalization suggestions after them.
• Improved handling of the “Ignore domain names” option. Domain names are now consistently skipped, even in cases where they were previously still flagged as spelling mistakes. For more details, review Spelling ignore options documentation.

Application Server

• Removed obsolete hardware validation messages from application server logs. Hardware hasn’t been used for license validation for some time, but the related messages remained and could cause confusion during troubleshooting.

Implementação

• Improved Docker build reliability across Kaniko, Buildah, and the classic Docker builder, including setups that override values with --build-arg. Build arguments now carry through properly to every stage of a multi-stage build, fixing cases where later stages couldn’t see them.

Revisão baseada em IA

• Dash-related suggestions from the AI proofreading engine are now easier to identify and turn off, which is useful if your style guide restricts or avoids dashes. For more information on rules, review information on English AI proofreading categories and rules.
• In addition to spelling and grammar, AI proofreading suggestions can now also be classified as style issues, so they’re labeled more accurately. Review English AI proofreading style categories and rules.
• Added curated descriptions for English AI-based proofreading corrections. Our linguistic team prepared these descriptions, making it easier to understand why a suggestion was made.

Gramática

• Suggestions from the third-party grammar engine are now correctly labeled as grammar, spelling, or style issues, so they’re highlighted the right way. Previously, everything from this engine showed up as a grammar issue, even when it wasn’t.
• Updated LanguageTool to the 2026-07-14 release.
• Improved recognition of German words that can function as either a verb or an adjective, such as strukturierte. Previously, some of these were only recognized as a verb, which could cause the comma rule to misread the sentence structure and incorrectly suggest removing a valid comma.
• Improved consistency across rules names so that turning a specific grammar rule on or off now applies the same way, no matter which checking engine handles it.

Ortografia

• Dutch now recognizes words with apostrophes, such as ‘s-Hertogenbosch, as correctly spelled.
• Improved spelling suggestions for German abbreviations that need a space between their two parts, such as correcting z.B. to z. B.
• Added 640+ of new words and terms across English, German, Spanish, Ukrainian, Bulgarian, Dutch, Thai, Greek, and Portuguese. They are covering everyday vocabulary as well as specialized fields like science, business, and technology.
• Expanded the shared cross-language dictionary with 30+ new proper names and abbreviations, including companies, products, and organizations, such as Canon, Ltd, and Rakuten. Additionally, added it as an additional dictionary for Greek spell check.
• Added 550 new medical terms for English, German, and Spanish, including abbreviations and procedure names, such as hepatoscopy, Unkovertebralgelenke and USG.
• Expanded English dialect support with 17 new dialectal pairs, such as hematoporphyrin (American) and haematoporphyrin (British).
• Improved 40+ spelling and autocorrect suggestions across English, Spanish, Ukrainian, Bulgarian, Catalan, and the shared cross-language dictionary. Targeting common misspellings, proper names, and abbreviations, such as llm → LLM.

Autocomplete

• Autocomplete now stops any output suggestions when the typed text contains words that can be considered strong swear language.

Idiomas

• Implemented spelling correction support for the Thai language.

🛡 Segurança

• Rangy library: We backported the upstream patch for CVE-2023-26102 into our forked Rangy library used in SCAYT3, closing the vulnerability while keeping it compatible with our fork.
• jackson-databind: We updated jackson-databind to 2.18.9 to fix CVE-2026-54515. This affects components that depend on jackson-databind, directly or transitively.
• ch.qos.logback: We updated ch.qos.logback from 1.5.25 to 1.5.35 to fix two deserialization vulnerabilities, CVE-2026-9828 and CVE-2026-10532. This affects components that depend on ch.qos.logback, directly or transitively.

🐞 Correções de erros

• Spelling engine: Elongated words, such as loooooooooooong, produced “No suggestions”
• Spelling engine: Unintelligible letter strings, such as fjkdkdjshkhfdk, produced unrelated real-word suggestions

O WProofreader core v6.14.0 lança o motor de revisão baseado em IA v3.1.0 para o espanhol, alarga o suporte ao assistente de escrita com IA ao grego, melhora a contagem de palavras multilingue, amplia a cobertura dos dicionários com mais de 7 800 novas palavras, aperfeiçoa a validação de dicionários personalizados e o tratamento do apóstrofo e inclui atualizações de segurança para as dependências do Jackson.

📝 Resumo

• Foi lançada a versão 3.1.0 do motor de revisão baseado em IA para o espanhol, com uma qualidade de correção melhorada em termos de gramática, uso de palavras, formas verbais, preposições, grafia dos números e confusões comuns entre palavras
• Foi adicionado o suporte ao grego ao assistente de escrita com IA
• Assistente de escrita em espanhol otimizado com IA, através da migração a partir dos modelos da Anthropic
• Dicionários ampliados com mais de 2 700 novas palavras em várias línguas, incluindo abreviaturas, nomes próprios, termos técnicos e termos linguísticos
• Melhoria na contagem e no processamento de palavras em vários idiomas, incluindo o japonês e o chinês
• Foi adicionado o suporte a hífens nos nomes dos dicionários e foi alargado o tratamento dos apóstrofos nos dicionários personalizados e dos utilizadores
• Atualizei o LanguageTool para a versão de 12 de junho de 2026
• Maior abrangência do dicionário de inglês dinamarquês e neozelandês
• Atualizações de segurança. Atualizámos o jackson-core e o jackson-databind para corrigir as vulnerabilidades CWE-770 e CVE-2026-54513

⚠️ Alteração importante

• Contagem de palavras melhorada para vários idiomas, incluindo aqueles sem espaços. Anteriormente, as palavras eram identificadas com base nos espaços e na pontuação, o que fazia com que frases inteiras em idiomas como o japonês e o chinês fossem contadas como uma única palavra. O WProofreader utiliza agora a segmentação baseada no ICU para contar morfemas nestes idiomas, proporcionando contagens de palavras e estatísticas linguísticas mais precisas. Por exemplo, «わかんない» é agora contado como 3 morfemas («わか» / «ん» / «ない»), em vez de 1.

🛠 Melhorias

Núcleo do WProofreader JS (v3.43.6079)

• Foi adicionada uma versão em grego do aviso legal do assistente de escrita com IA.
• Atualização da validação das palavras do dicionário do utilizador e do dicionário personalizado, de modo a suportar todos os caracteres de apóstrofo permitidos.

Revisão baseada em IA

• Foi lançada a versão 3.1.0 do motor de revisão baseado em IA para o espanhol. Foi melhorada a qualidade da correção no que diz respeito ao modo subjuntivo, aos tempos verbais, aos verbos reflexivos, às preposições, à concordância de adjetivos e advérbios, à concordância do particípio passado, às construções partitivas, à grafia dos números e às confusões comuns entre palavras, incluindo «ser» e «haber». Consulte a documentação para obter mais detalhes sobre as melhorias na revisão do espanhol.
• Melhoria do desempenho da revisão baseada em IA, em todas as línguas, em textos que contenham formatação com expoentes e subíndices.

Gramática

• Atualizei o LanguageTool para a versão de 12 de junho de 2026.

Assistente de escrita com IA

• Foi adicionado o suporte ao grego ao assistente de escrita com IA, utilizando o modelo Google Gemma 3 27B.
• Assistente de escrita em espanhol baseado em IA otimizado através da migração de prompts do Anthropic Claude Haiku 4.5, melhorando o desempenho geral.

Ortografia

• Dicionários ampliados com mais de 2 700 novas palavras em inglês, alemão, espanhol, ucraniano, dinamarquês e português, abrangendo abreviaturas, termos tecnológicos, palavras comuns, nomes próprios e termos linguísticos. 
• Os dicionários alemães também passaram a incluir mais de 4 600 novas formas substantivas com um -s de ligação e formas combinadas com um hífen, como Fort-. 
• Terminologia médica alargada para o inglês, o alemão e o espanhol, com 495 novos termos, abrangendo adjetivos médicos, nomes de doenças e abreviaturas, tais como palmitoilcarnitina, teste de Finkelstein e distrófica.
• Melhoria na cobertura dialetal do inglês com novos pares de palavras, tais como «preprosthetic» vs «pre-prosthetic». 
• Melhorou o processamento de texto em alemão, permitindo o uso de hífens no início das palavras e restringindo o uso de dois pontos no interior das palavras, como por exemplo «-konzepte».
• Regras de ortografia aperfeiçoadas e sugestões de correção automática para erros ortográficos comuns e fonéticos em inglês e espanhol, tais como «Afroasiatic» → «Afro-Asiatic».

Dicionários personalizados e de utilizador

• Validação melhorada dos nomes dos dicionários do utilizador. Os nomes dos dicionários suportam agora hífens e aceitam letras latinas ASCII, algarismos, sublinhados e hífens. As regras de validação e as mensagens de erro foram atualizadas para um tratamento mais consistente dos nomes inválidos.
• Os dicionários personalizados e de utilizador já permitem a adição de palavras com diferentes tipos de apóstrofo, tais como l’acqua, l’acqua, etc .

Estilo

• Continuámos a melhorar a filtragem das sugestões de preenchimento automático que contêm linguagem sensível, inesperada ou discriminatória.
• Foi adicionado o suporte a sugestões de estilo e padrões que começam com um espaço, como no alemão*in → (m/w/d), de modo a ter um único padrão para *Arzt*in → Arzt (m/w/d) e Professor*in → Professor (m/w/d).

Idiomas

• Modifiquei o dicionário Hunspell dinamarquês para reduzir os falsos positivos em palavras compostas. Fonte: WebSpellChecker hunspell-dictionaries-modifications
• Ampliou-se o vocabulário do inglês da Nova Zelândia através da integração do dicionário Hunspell v4.0.7 do inglês australiano, da autoria de Marco Pinto, para proporcionar uma cobertura adicional do vocabulário comum.
  Fonte: https://github.com/marcoagpinto/aoo-mozilla-en-dict 

🛡 Segurança

• Atualização do jackson-core: Atualizou-se o jackson-core para a versão 2.18.7 para resolver CWE-770. Âmbito: componentes que dependem do jackson-core (diretamente ou de forma transitiva) no pacote langtool/libs.
• Atualização do jackson-databind: Atualizou-se o jackson-databind para a versão 2.18.8 para resolver CVE-2026-54513. Âmbito: componentes que dependem do jackson-databind (diretamente ou de forma transitiva) no pacote langtool/libs.

🐞 Correções de erros

• Núcleo do WProofreader JS: Aparecem marcadores pretos inesperados na interface do WProofreader (WordPress 7.0, plugin Guttenberg). Resolveu-se um conflito de estilo que fazia com que os marcadores de lista predefinidos fossem apresentados nos componentes da interface do WProofreader.

WProofreader core v6.13.1 improves compatibility with editors hosted inside Shadow DOM containers by fixing automatic editor detection, initialization, and auto-search functionality, strengthens the security of the WProofreader Docker image with OpenNLP and Netty vulnerability fixes, and updates the VAPT report.

📝 Resumo

• Improved compatibility with editors hosted inside Shadow DOM containers, including automatic editor detection, initialization, and style loading
• Fixed auto-search functionality in Shadow DOM environments
• Updated OpenNLP and Netty components in the WProofreader Docker image to address seven reported vulnerabilities
• Updated the VAPT report to version 6.13.1 and improved vulnerability scanner compliance

🛠 Melhorias

WProofreader JS core (v3.42.6067)

• Improved Shadow DOM support. Fixed automatic editor detection and proofreader initialization for editable elements hosted inside Shadow DOM containers. WProofreader now correctly detects active editors, preserves instances when editors remain connected to the page, and loads required UI and markup styles within Shadow DOM environments. 

🛡 Segurança

• OpenNLP security fixes: Integrated a patched OpenNLP 1.9.4 build with backported fixes for CVE-2026-40682, CVE-2026-42027, and CVE-2026-42440, preserving LanguageTool compatibility while addressing reported vulnerabilities. Scope: OpenNLP components used transitively by LanguageTool and packaged in the WProofreader Docker image. 
• Netty upgrade: Updated bundled Netty from 4.1.133.Final to 4.1.135.Final to address CVE-2026-44249, CVE-2026-45416, CVE-2026-45674, and CVE-2026-47691. Scope: Netty components used transitively by LanguageTool and packaged in the WProofreader Docker image.
• VAPT: Update the report v6.13.1

WProofreader core v6.13.0 migrates AI writing assistant operations to new models and restores its Google Docs support, improves AI-based proofreading and autocomplete quality, expands multilingual dictionaries with 1400+ words, introduces limits for custom dictionaries and style guide collections, updates Ukrainian and Finnish dictionaries, and delivers a security update.

📝 Resumo

• Migrated AI writing assistant models across multiple languages to Amazon and Google alternatives following the deprecation of Anthropic Claude models and to optimize performance
• Updated the Ukrainian and Finnish dictionaries 
• Expanded dictionaries with 1,400+ new words across English, German, Spanish and Portuguese with common words, medical terminology, additional proper names and abbreviations
• Improved spelling quality with better handling of Greek capitalized words
• Improved filtering of autocomplete suggestions containing sensitive or discriminatory language
• Updated style guidance with new Spanish anglicism rules
• Improved AI proofreading quality across all supported models and enhanced Spanish AI proofreading v3.0 for more reliable corrections.
• Restored AI writing assistant compatibility with Google Docs
• Security update. Updated OpenTelemetry to address CVE-2026-45292

🛠 Melhorias

WProofreader JS core (v3.42.6057)

• Improved Google Docs compatibility. Fixed issues preventing AI writing assistant from working correctly in Google Docs.

Revisão com IA

• Improved AI proofreading quality across English, German, and Spanish in punctuation and spacing handling, including decimal commas, en dashes, and punctuation around corrected text.
• Improved Spanish correction accuracy by preventing redundant pronouns, unnecessary punctuation, degree symbols, square brackets, and duplicate suggestions.
• Improved suggestion quality in German by restricting word removal at the end of German sentences.

Assistente de escrita com IA

• Migrated the English AI writing assistant from Anthropic Claude Haiku 3 following its deprecation. For more information, review documentation current English models.
• Migrated selected Norwegian, Italian, Spanish, and Danish AI writing assistant workflows from Anthropic Claude Sonnet 4 following its deprecation.
• Updated prompts and migrated the German and Ukrainian AI writing assistant from Anthropic Claude Haiku 4.5 to boost performance.
• Improved AI writing assistant prompts across Dutch, Swedish, Portuguese, and Polish for more consistent rewriting, tone adjustment, shortening, and expansion results.

Ortografia

• Improved handling of Greek all-caps words, like ΕΙΔΗΣΗ.
• Expanded dictionaries with 1,000+ new words across English, German, Spanish, and Portuguese, covering abbreviations, affixes, biology, common words, dialectal variations, food, linguistics, music, proper names, slang, and technology.
• Enriched shared cross-language dictionaries with 30+ proper names and abbreviations, like Googlebot and ISO.
• Enhanced English and Spanish medical terminology with 471 new words covering medical terms, legal terminology, abbreviations, and related vocabulary.
• Expanded English dialect coverage with new dialectal word pairs, such as radiolabeled vs radiolabelled.
• Refined suggestion rules for common English and Spanish misspellings, such as gritings → greetings.

Custom dictionary

• Added a limitation on the number of custom dictionaries depending on the subscription plan. Up to 250 for self-hosted.

Estilo

• Added a limitation on the number of style guide collections depending on the subscription plan. Up to 250 for self-hosted.
• Made ‘collection’ parameter required for any actions with style guide rules using API
• Updated Spanish style guide with new anglicism rules, such as download, shopping center, etc.
• Added the possibility to use part from the pattern field in suggestion filed. For example, Lehrer → Lehrer*in

Autocomplete

• Improved filtering of autocomplete suggestions containing sensitive, unexpected, or discriminatory language, such as war, drug, etc.

Idiomas

• Updated the Ukrainian (uk_UA) Hunspell dictionary to v6.8.0 (Apr 23, 2026) by Rysin, A. and Starko, V., which included an expansion of the dictionary with 12,000+ entries. Source: https://github.com/brown-uk/dict_uk/releases#release-v6.8.0 
• Updated the Finnish (fi_FI) Hunspell dictionary to v0.2 (Dec 6, 2023) by Voiko, significantly expanding vocabulary coverage across technical, medical, and proper names. Source: https://github.com/fginter/hunspell-fi/releases/tag/0.2 

🛡 Segurança

• OpenTelemetry upgrade: Updated opentelemetry to 1.62.0 to address CVE-2026-45292. Scope: components depending on opentelemetry (direct or transitive) packaging of langtool/libs.

🐞 Correções de erros

• WProofreader JS core: AI writing assistant doesn’t work in Google Docs
• AI writing assistant stopped working for free extension for auto-detect language

WProofreader v6.11.0 simplifies self-hosted deployment by removing domain name requirements, introduces new JS core options for scaled layouts and RTL badge control, changes the default English check kit to ai_lt_hs1 for cloud users, expands dictionaries with 9,700+ new entries, updates the grammar engine with major Catalan improvements, and patches a security vulnerability.

Consulte o resumo abaixo.

📝 Resumo

• [Cloud] Default text-checking configuration for English changed to ai_lt_hs1. Combines the AI proofreading engine with the algorithmic grammar and spelling engines for improved accuracy
• Self-hosted installation no longer requires a domain name — simplified setup for Docker, Windows, and Linux
• New enableScale and disableBadgeRTLMirroring options in WProofreader JS core
• Expanded dictionaries with 9,700+ new entries across English, German, Spanish, Ukrainian, and Russian
• Extended German dictionary with 13,550+ compound word forms for feminine nouns with Fuge-s
• Updated third-party grammar engine with major Catalan improvements and rule refinements for German, Spanish, Portuguese, Ukrainian, and English
• Custom user dictionaries now accept words that already exist in core spelling dictionaries
• Security fix. Updated jackson-core to address GHSA-72hv-8253-57qq

🛠 Melhorias

WProofreader JS core (v3.39.6037)

• Introduced a new enableScale option. Enables support for CSS scale transformations applied to parent elements. When enabled, WProofreader detects the scale value of the parent element and adjusts UI elements accordingly, ensuring that highlighting and interface elements remain correctly aligned when the surrounding content is scaled. Disabled by default.
• Introduced a new disableBadgeRTLMirroring option. Disables automatic badge position mirroring for right-to-left (RTL) languages such as Arabic. When enabled, the badge keeps its configured position instead of being automatically mirrored according to text direction. Manual badge positioning options remain unchanged. Disabled by default.

Implementação

• Removed the requirement to specify a domain name during installation for the self-hosted version. This simplifies initial setup across Docker, Windows, and Linux environments:
  • Docker. Removed the WPR_DOMAIN_NAME parameter. Added support for running the container from the root path (/) via WPR_VIRTUAL_DIR. Learn more about the Docker configuration.
  • Helm chart. Updated to v1.3.1. Allowed setting empty virtualDir (/), removed WPR_DOMAIN_NAME from the extraEnv example, and updated appVersion para 6.11.0.0. Learn more about the Helm chart.
  • Windows and Linux installers. The web server now listens on any domain by default, removing a common setup step.

Servidor de aplicações

• [Cloud] Changed the default text-checking configuration for English to ai_lt_hs1. This configuration combines the AI proofreading engine with the algorithmic grammar engine (LanguageTool) and the algorithmic spelling engine (Hunspell) with spelling suggestions for improved accuracy. Previously, the default configuration (ai_lt_hs0) included Hunspell for error detection but without spelling suggestions. If needed, the previous behavior can be restored by setting the checkKit option to ai_lt_hs0.
• Improved error logging. The application server now provides more detailed messages when Hunspell dictionaries fail to load, when an AI language model is missing, or when an n-gram model isn’t found. This makes it easier to diagnose configuration issues during deployment.
• Added the ability to configure the maximum sequence length in the AI model configuration.
• Improved file path handling across the application server to use absolute paths instead of searching in the current working directory. This resolves issues when the application server runs as a systemd service on Linux.

Gramática

• Updated the third-party grammar engine with improvements across multiple languages.
  • Catalan. Major update including a new proper noun recognition system that reduces false spelling errors for names seen earlier in text, improved verb replacement suggestions with better pronoun handling, new multitoken speller dictionary for better multi-word spelling suggestions, added support for causal constructions (al no + infinitive → com que…), updated Catalan POS dictionary to v3.3, expanded spelling and multiword dictionaries, improved sentence segmentation with the fund. abbreviation for Catalan and Spanish, and various disambiguation rule refinements.
  • German. Improved grammar rules for the nach zu geben construction and added rules for handling German number formatting with dot separators (e.g., 100.000).
  • Spanish. Added an antipattern for volares as an adjective to prevent incorrect subjunctive suggestions.
  • Portuguese. Significant disambiguation improvements including better handling of pela/pelas as prepositions, refined verb-to-noun/adjective disambiguation, new context-based tagging for semantic categories, enabled the redundant comma after dash rule, expanded confused words rules, added antipatterns for K2 mountain and Samsung Galaxy references, and various style rule refinements for Brazilian Portuguese.
  • Ukrainian. Updated morphological dictionary to v6.7.5, expanded abbreviation handling (e.g., крим, пом, суч), improved word tokenizer patterns, added numerous soft replacement suggestions, expanded derivation rules, and improved disambiguation for vocative case and other grammar constructs.
  • English. Added misgendered, misgender, misgenders, misgendering to the spelling dictionary and improved prefix handling in spelling suggestions.

Ortografia

• Expanded dictionaries with 9,700+ new entries across English, German, Spanish, Ukrainian, and Russian, covering abbreviations, affixes, biology, chemistry, common terms, dialectal variations, medical, military, neologisms, proper names, technology, and more.
• Extended the German dictionary with 13,550+ Fuge-s compound word forms for feminine nouns combined with a hyphen (e.g., Präventions-).
• Enriched shared cross-language dictionaries with 125 new common proper names and abbreviations.
• Added new Russian language resources.
• Refined word lists by removing 500+ invalid or incorrect entries across English, Spanish, and German, targeting nonexistent words, incorrect medical terms, invalid abbreviation and capitalized forms, and non-dialectal entries.
• Optimized split patterns with 20 new rules for English and Spanish covering reflexive verbs, common words, pronouns, and adverbs to prevent incorrect word splitting. Improved suggestion quality by excluding offensive terms in certain contexts and removing invalid suggestions across multiple categories.
• Enhanced medical terminology for English and Spanish with 255+ new entries covering Latin terms, medical adjectives, disease names, and abbreviations. Refined medical word lists by removing 230+ invalid entries targeting incorrect words, abbreviation forms, and capitalized forms.
• Added support for two types of apostrophes in Catalan (U+2019 and U+0027).

Correção automática

• Refined prioritization rules for common English and Spanish misspellings. Enhanced prioritization in the shared cross-language dictionary.
• Improved correction logic by removing invalid English and Spanish autocorrect entries targeting incorrect suggestions.

Estilo

• Updated style rules for English with new rules addressing slurs and race and ethnicity insults.

Custom dictionary

• Starting with v6.11.0, words that exist in core spelling dictionaries can be added to custom user dictionaries as well.

🛡 Segurança

• Updated the vulnerable jackson-core library (used by the grammar engine) to version 2.18.6 to address GHSA-72hv-8253-57qq (number length constraint bypass in async JSON parser, which could lead to a potential denial of service condition).

🐞 Correções de erros

• WProofreader JS core: Highlighted suggestions are shifted when the Zoom plugin is used in CKEditor
• WProofreader JS core: Inconsistent behavior of the keyboard navigation in the AI writing assistant dialog in mixed left-to-right/right-to-left environments
• Spelling engine: Words spelled using typographic apostrophes aren’t recognized as correct in Catalan
• Spelling engine: Split rules containing numbers aren’t allowed
• Application server: Allowed rules functionality doesn’t work as expected
• Application server: Hunspell fails if n-gram model isn’t found
• Style guide: Core style guide rules can’t be read
• Style guide: Relative paths are used for style guide files

WProofreader v6.10.0 introduces a major upgrade to spelling suggestion ranking with trigram-based n-gram prioritization for English, German, and Spanish, adds autocomplete keyboard hints, updates the grammar engine, expands language support with Greek and Czech resources, and patches a security vulnerability.

Consulte o resumo abaixo.

📝 Resumo

• New trigram-based spelling suggestion prioritization for English, German, and Spanish — the correct suggestion now ranks first ~80% of the time, up from ~62% without prioritization
• Expanded dictionaries with 2,000+ new entries and 600+ medical terms across multiple languages
• Autocomplete suggestions now show a keyboard hint (Tab / →) for accepting completions
• 80+ new English style rules addressing profanity, slurs, and inclusive language
• Improved AI proofreading reliability by disabling 15 problematic rules across English, German, and Spanish
• Added Greek and Czech language resources; updated Czech Hunspell dictionary
• AI writing assistant disclaimer localized for Danish and Norwegian
• Security fix. Updated lucene-core to address PRISMA-2021-0081

🛠 Melhorias

WProofreader JS core (v3.38.6019)

• Added a keyboard hint next to autocomplete suggestions that shows how to accept them (Tab / →). The hint is visible wherever autocomplete suggestions are shown.
• Added “Powered by WProofreader” branding to the suggestion pop-up for the browser extension, WordPress plugin, and Drupal integration.
• Redesigned the promotional banner in the free SCAYT plugin for CKEditor 4. The banner is now smaller and less intrusive, with a link to the website for more details.

✨ Assistente de escrita com IA

• Localized the AI disclaimer for Danish and Norwegian.

📚 Motor de correção ortográfica

• Introduced trigram-based n-gram suggestion prioritization for English, German, and Spanish. This replaces the previous bigram-based approach and improves the accuracy of spelling suggestions by considering three-word sequences instead of two. When multiple spelling candidates are equally plausible — e.g., store, shore, stare — the model uses surrounding context to rank the most relevant suggestion higher. In benchmarks on American English, the correct suggestion now ranks first approximately 80% of the time, compared to 71% with the previous approach. Each language model requires approximately 100 MB of RAM. Currently, only suggestions from the main algorithmic spelling engine are reordered by the n-gram model.
• Expanded dictionaries with 2,000+ new entries across multiple languages (including English, German, Spanish, Ukrainian, Danish, Portuguese and Finnish). This covers a broad spectrum of categories: abbreviations (shortened forms), affixes, biology, business, chemistry, common terms, culinary, dialectal variations, legal, locations, neologisms, proper names, religion, science, slang, sport, and technology.
• Refined word lists by removing invalid dialectal entries, nonexistent words generated by compounding mechanisms, and incorrect casing.
• Enhanced medical terminology across English, Spanish, and German with 600+ new entries covering medical abbreviations, adjectives, disease names, and measurements.
• Optimized split patterns by adding 20+ rules for common words, adverbs, and conjunctions to prevent incorrect word splitting, while removing outdated or less restrictive patterns.
• Improved suggestion quality by excluding offensive terms in certain contexts and removing company names that overlap with common English words.

⚙️ Motor gramatical

• Updated the third-party grammar engine, LanguageTool, to version 2026-02-13 with improvements across multiple languages.
  • Portuguese. Extensive updates including new academic and formal style rules, improved handling of agreement, contractions, homonyms, and commas, disambiguation refinements for rare verbs and part-of-speech tagging, and numerous false-positive reductions. Portuguese (Portugal) received additional style improvements for foreign words, gerund avoidance, and vocabulary corrections.
  • Catalan. New grammar rules, significant dictionary expansion, case checking improvements, and refinements to previously disabled rules. Enhanced specific rules such as ESCOLTAR_SENTIR and RESPONDRE_A.
  • Spanish. Improved sentence tokenizer accuracy, added antipatterns for agreement, accent, date, and month rules, and enhanced medical-context language support with new medical terms.
  • English. Maintenance-oriented improvements including rule precision fixes, match position corrections, case sensitivity adjustments, and prevention of incorrect word synthesis.
  • German. Incremental rule improvements and refinements.
  • Ukrainian. Dictionary update, tokenization improvements, and a new grammar rule.

💡 Autocorrect

• Refined prioritization rules for 120+ common misspellings across English, German, Spanish, Ukrainian, French, and Portuguese.
• Enhanced correction logic by removing 360+ invalid autocorrect entries for Spanish, specifically targeting incorrect suggestions and capitalized forms that were previously triggered by mistake.

📝 Style

• Updated style rules for English with 80+ new rules addressing strong profanity, slurs, and language related to race, ethnicity, gender, and ableism.

🤖 AI proofreading

• Improved AI engine reliability by disabling 15 problematic rules that caused logic loops, pattern shifting, and nonexistent word suggestions across English, German, and Spanish.
• Resolved bugs where the AI would incorrectly flag unknown proper names or offer shifted suggestions.

🌐 Languages

• Added new language resources for Greek, including disabled problematic grammar rules for unpaired brackets. Added new Czech language resources with spelling prioritization.
• Updated Czech Hunspell dictionary. Source: Czech CC0 dictionaries by Štěpán Škorpil.

Servidor de aplicações

• Updated third-party libraries:
  • TensorFlow from 2.11 to 2.18 (Linux and Windows)
  • rust_swig from 0.4.0 to 0.10.0
  • tokenizers from 0.15.1 to 0.22.2

🛡 Segurança

• Updated the vulnerable lucene-core library used by the grammar engine to address PRISMA-2021-0081.

A versão 6.9.0 do WProofreader core inclui dicionários ortográficos alargados e terminologia médica, regras gramaticais e de estilo aperfeiçoadas em várias línguas, melhorias na revisão por IA e no assistente de escrita por IA para o dinamarquês, atualizações nos rótulos da interface do utilizador, correções de segurança e novas demonstrações do editor.

Consulte o resumo abaixo.

📝 Resumo

• Dicionários ortográficos alargados, com mais de 2 200 novas palavras em 12 línguas, e terminologia médica melhorada em cinco línguas.
• Melhorias na autocorreção, nos guias de estilo, na gramática e na qualidade da revisão por IA em vários idiomas.
• Foi adicionado o suporte ao dinamarquês no assistente de escrita com IA e foram atualizados os modelos de IA subjacentes.
• Atualização dos rótulos da interface do utilizador, da nomenclatura linguística e das localizações.
• Foi corrigida uma vulnerabilidade de segurança e foram adicionadas novas demonstrações do Slate e do Draft.js.
• Foram corrigidos problemas que afetavam o comportamento da correção automática, a divisão de frases e a utilização do assistente de escrita com IA no Help Scout.

🛠️ Melhorias

Núcleo do WProofreader JS

• Versão atualizada v3.37.6005
• Nomes de etiquetas atualizados e as respetivas traduções no Geral secção de definições:
  • Guia de estilo > Estilo
  • Corrigir ortografia automaticamente > Correção automática
  • Sugestões de preenchimento automático > Preenchimento automático de texto

Ortografia

• Dicionários ampliados com mais de 2 200 novas palavras em 12 línguas, abrangendo abreviaturas (incluindo formas abreviadas), afixos, biologia, negócios, química, expressões coloquiais, termos comuns, palavras compostas, variações dialetais, vocabulário de ficção e poético, gastronomia, terminologia jurídica, linguística, localizações, matemática e lógica, terminologia militar, música, neologismos, nomes próprios, ciência, gíria, terminologia especializada, desporto, tecnologia e anglicismos, incluindo termos de outras línguas e palavras arcaicas.
• Terminologia médica melhorada em inglês, alemão, italiano, espanhol e francês, com base em dicionários criados pelos utilizadores (mais de 330 entradas no total), abrangendo abreviaturas médicas, doenças, adjetivos médicos, unidades de medida, nomes de medicamentos e termos médicos relacionados.
• Atualizou-se o dicionário de norueguês (nb_NO) para corrigir problemas com palavras compostas que terminam em -dom, -sel e -ion.
• Foram adicionadas regras de priorização para as sugestões ortográficas em dinamarquês.
• Foram adicionadas regras relativas a palavras excluídas e a palavras ofensivas excluídas nas sugestões ortográficas em polaco.

Correção automática

• Aperfeiçoou as regras de correção automática, removendo entradas inválidas e adicionando prioridade aos erros ortográficos mais comuns em inglês, alemão, espanhol, ucraniano e dinamarquês.

Estilo

• Guias de estilo ucranianos melhorados, com adições e algumas correções relativas a formas contrárias ao surzhyk e a anglicismos.
• Corrigiu-se uma descrição de palavra que induzia em erro no guia de estilo ucraniano contra o «surzhyk».
• Foram removidas entradas dos guias de estilo alemães em que tanto a correspondência como a sugestão são variantes válidas ou são utilizadas em contextos diferentes.

Gramática

• Atualizou-se o motor gramatical de terceiros, o LanguageTool, para a versão 2026-01-19. Melhorou-se a revisão do catalão com dicionários alargados, melhor reconhecimento de expressões compostas, sugestões de pronomes aperfeiçoadas e ativação da validação das formas verbais.
• Melhorámos as regras gramaticais do inglês, removendo regras fixas das listas de regras desativadas.
• Regras específicas melhoradas para o espanhol, o português e o neerlandês, incluindo o tratamento de parênteses desparelhados, concordância, acentuação, segmentação, pronomes clíticos, hifenização de pronomes sujeitos, antipadrões para construções específicas e uso de maiúsculas.
• Foram alteradas as regras para evitar que as abreviaturas em italiano e espanhol desencadeiem o início de frases.

Revisão com IA

• Desativei regras problemáticas que causavam erros, sugestões inexistentes, padrões alterados e loops para o inglês, o alemão e o espanhol.

✨ Assistente de escrita com IA

• Foi adicionado suporte para operações de texto em dinamarquês, incluindo nomes de operações traduzidos na interface do utilizador. As operações de texto em dinamarquês utilizam o Claude Haiku 4.5, com exceção da função «Resumir», que utiliza o Sonnet 4.

Importante

• O Claude Haiku 3.5 foi substituído pelo Haiku 4.5 devido à sua obsolescência. O Haiku 3.5 era anteriormente o modelo predefinido para a maioria das operações de texto em línguas que não o inglês. O acesso alargado ao público está disponível até 19 de março de 2026, estando o fim da vida útil do modelo previsto para 19 de junho de 2026. Encontre mais detalhes na documentação sobre o ciclo de vida dos modelos do AWS Bedrock.

Recomendamos a atualização para versões mais recentes da aplicação, que utilizam instruções compatíveis com os modelos suportados.

Servidor de aplicações

• Atualizou-se a localização da lista de idiomas, adicionaram-se as traduções em falta, padronizou-se a nomenclatura dos idiomas para «Idioma (País)» (por exemplo, Inglês (Reino Unido)) e renomeou-se a opção «Detecção automática do idioma» para «Detecção automática».

🛡️ Segurança

Atualizou-se a biblioteca ch.qos.logback, que apresentava uma vulnerabilidade e era utilizada pelo componente de gramática, para corrigir: CVE-2026-1225

Demonstrações

• Adicionámos algumas novas demonstrações que mostram o verificador ortográfico e gramatical WProofreader integrado nos editores Slate e Draft.js. 

Experimente as demonstrações:

• Versão de demonstração do WProofreader para a Slate 
• Versão de demonstração do WProofreader para o Draft.js

🐞 Correções de erros

• Núcleo do WProofreader: Não é possível aceitar as sugestões do assistente de escrita com IA utilizando o botão «Substituir» no Help Scout
• Correção automática ortográfica: as palavras da lista de sugestões excluídas são apresentadas como sugestões de correção automática
• Núcleo do WProofreader: a divisão incorreta da frase em «e.g.» ativa uma regra em que faltam chaves