By default a translation is saved as soon as the engine returns a well-formed answer. The plugin
\nchecks that the answer is valid JSON and that its fields match the ones that were sent \u2014 it does not
\nread the text. If the model mistranslates a term, the mistranslation goes live.<\/p>\n
The Typical uses:<\/p>\n What it cannot do.<\/strong> These are text checks, not comprehension. A translation that reverses the It runs for post translations, taxonomy term translations and interface string batches, with every In delta mode<\/strong> (“Translate only changed strings”, PRO) the filter only ever sees the fields that The source text of a field:<\/p>\n Return the array with whatever changes you want saved.<\/p>\n Do not remove a field to skip it.<\/strong> The translated post is rebuilt from the original, so a title, Return a Anything other than an array is treated as a rejection too, so always return A rejected job is not retried.<\/strong> It stays failed until someone translates the post again, and What it does By default a translation is saved as soon as the engine returns a well-formed answer. The plugin checks that the answer is valid\u2026<\/p>\n","protected":false},"author":1,"featured_media":0,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"_latwm_translated_slug":"","footnotes":""},"categories":[17],"tags":[],"class_list":["post-99","post","type-post","status-publish","format-standard","hentry","category-latw-for-polylang"],"_links":{"self":[{"href":"https:\/\/docs.latw.ai\/pl\/wp-json\/wp\/v2\/posts\/99","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/docs.latw.ai\/pl\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/docs.latw.ai\/pl\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/docs.latw.ai\/pl\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/docs.latw.ai\/pl\/wp-json\/wp\/v2\/comments?post=99"}],"version-history":[{"count":0,"href":"https:\/\/docs.latw.ai\/pl\/wp-json\/wp\/v2\/posts\/99\/revisions"}],"wp:attachment":[{"href":"https:\/\/docs.latw.ai\/pl\/wp-json\/wp\/v2\/media?parent=99"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/docs.latw.ai\/pl\/wp-json\/wp\/v2\/categories?post=99"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/docs.latw.ai\/pl\/wp-json\/wp\/v2\/tags?post=99"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}latwaitp_translated_strings<\/code> filter is the point where your site can look at the finished
\ntranslation before<\/em> it is written, and either correct it or reject it. It is part of the free
\nplugin.<\/p>\n\n
\nmeaning of a sentence while keeping every number, tag and glossary term intact will pass every
\ndeterministic rule you write here. Catching that needs a second AI call, which this filter lets you
\nmake, but does not make for you.<\/p>\nWhere it runs<\/h2>\n
engine answers \u2192 JSON parsed \u2192 latwaitp_translated_strings \u2192 post, term or strings saved<\/code><\/pre>\n
\nengine and in every processing mode, in the same background process that saves the translation.<\/p>\n
\nwere actually sent to the AI. The unchanged parts of the post keep their existing translation and
\nnever reach your code \u2014 so a check that needs to see the whole post should load the translated post
\nitself rather than assume $strings<\/code> is complete. An update that turned out to need no translation
\nat all does not call the filter.<\/p>\nThe filter<\/h2>\n
add_filter( 'latwaitp_translated_strings', function ( $strings, $translation ) {\n return $strings;\n}, 10, 2 );<\/code><\/pre>\n$strings<\/code> is an array of field name => translated text<\/code>. The field names are the plugin’s own:
\npost_title<\/code>, post_excerpt<\/code>, content_1<\/code>, meta_1<\/code>, builder_1<\/code> for posts (pll_1<\/code> and so on on
\nPolylang Pro), name<\/code> and description<\/code> for terms, string_\u2026<\/code> for interface strings.<\/p>\n$translation<\/code> is the job’s row in the plugin’s queue:<\/p>\n\n\n
\n \nKey<\/th>\n Meaning<\/th>\n<\/tr>\n<\/thead>\n \n id<\/code><\/td>\nRow id, as shown on the History<\/strong> screen.<\/td>\n<\/tr>\n \n element_type<\/code><\/td>\npll_post_job<\/code> for posts, term<\/code> for taxonomy terms, pll_string_job<\/code> for interface strings.<\/td>\n<\/tr>\n\n source_post_id<\/code><\/td>\nThe original post, or the original term for term<\/code>. 0<\/code> for interface strings.<\/td>\n<\/tr>\n\n source_lang<\/code><\/td>\nLanguage slug translated from, e.g. en<\/code>.<\/td>\n<\/tr>\n\n target_lang<\/code><\/td>\nLanguage slug translated into, e.g. de<\/code>.<\/td>\n<\/tr>\n\n model<\/code>, provider<\/code><\/td>\nThe model and engine this job was actually sent to.<\/td>\n<\/tr>\n \n translation_type<\/code><\/td>\non-demand<\/code> or auto<\/code>, with -delta<\/code> or -delta-context<\/code> for an update that sent only what changed.<\/td>\n<\/tr>\n\n source_strings_json<\/code><\/td>\nJSON of the fields that make up the job, each with the source<\/strong> text in data<\/code>. Not set for terms.<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n$source = json_decode( (string) $translation['source_strings_json'], true );\n$text = isset( $source[ $field ]['data'] ) ? $source[ $field ]['data'] : '';<\/code><\/pre>\nReturning a corrected translation<\/h3>\n
\/\/ Editors kept fixing this one word by hand after every translation.\nadd_filter( 'latwaitp_translated_strings', function ( $strings, $translation ) {\n if ( $translation['target_lang'] !== 'de' ) {\n return $strings;\n }\n\n foreach ( $strings as $field => $translated ) {\n $strings[ $field ] = str_replace( 'Bodensatz', 'Sediment', $translated );\n }\n\n return $strings;\n}, 10, 2 );<\/code><\/pre>\n
\nexcerpt, paragraph or Elementor text missing from the array is written back in the source
\nlanguage<\/strong>. Fields you add are ignored \u2014 there is no place in the post to put a field that was never
\npart of the job.<\/p>\nRejecting a translation<\/h3>\n
WP_Error<\/code> \u2014 or an empty array \u2014 and nothing is saved. The job is marked failed<\/strong>, the
\npost keeps whatever translation it had before, and the reason is stored on the job and written to
\nthe log.<\/p>\n$strings<\/code> from a
\nbranch that does not apply.<\/p>\n\/\/ Words that must never appear, per language.\nadd_filter( 'latwaitp_translated_strings', function ( $strings, $translation ) {\n $banned = array(\n 'de' => array( '\/bBodensatzb\/u' ),\n 'it' => array( '\/bspremiagrumib\/iu' ),\n );\n\n if ( empty( $banned[ $translation['target_lang'] ] ) ) {\n return $strings;\n }\n\n foreach ( $strings as $field => $translated ) {\n foreach ( $banned[ $translation['target_lang'] ] as $pattern ) {\n if ( preg_match( $pattern, $translated ) ) {\n return new WP_Error(\n 'mysite_banned_term',\n sprintf( 'Banned term in field \"%s\" (%s)', $field, $pattern )\n );\n }\n }\n }\n\n return $strings;\n}, 10, 2 );<\/code><\/pre>\n\/\/ Every measurement in the source must appear in the translation.\nadd_filter( 'latwaitp_translated_strings', function ( $strings, $translation ) {\n $source = json_decode( (string) $translation['source_strings_json'], true );\n $measurement = '\/d+(?:[.,]d+)?s?(?:mm|cm|m|ml|l|g|kg|%)b\/u';\n\n foreach ( $strings as $field => $translated ) {\n $text = isset( $source[ $field ]['data'] ) ? $source[ $field ]['data'] : '';\n\n preg_match_all( $measurement, wp_strip_all_tags( $text ), $in_source );\n preg_match_all( $measurement, wp_strip_all_tags( $translated ), $in_translation );\n\n if ( count( $in_source[0] ) !== count( $in_translation[0] ) ) {\n return new WP_Error(\n 'mysite_measurements_lost',\n sprintf(\n 'Field \"%s\": %d measurements in the source, %d in the translation.',\n $field,\n count( $in_source[0] ),\n count( $in_translation[0] )\n )\n );\n }\n }\n\n return $strings;\n}, 10, 2 );<\/code><\/pre>\n
\nthat sends the same prompt again: nothing tells the model what you objected to, so a model that made
\na deliberate word choice will usually make it again \u2014 at your provider’s expense. Rejection is best
\nused for mistakes a model makes occasionally<\/em> (a dropped number, a truncated field). For a word the
\nmodel keeps choosing, correct it in the filter, add the preferred term to the glossary, or say so in
\nthe translation prompt.<\/p>\nWhere to check on it<\/h2>\n
\n
failed<\/code>, with your message in the Error<\/strong> column.<\/li>\nResponse processing failed<\/code> with the error
\nmessage.<\/li>\n<\/ul>\n","protected":false},"excerpt":{"rendered":"