🇧🇪Belgium @arwillame

arwillame → created an issue.

I found an issue with this logic.
I have cases where the result is indeed an array because multiple value are possible, but if there is only one result anyway, i want to keep the same logic as it was multiple.
I think the best would be to just remove the code that makes the "reset($array)" to avoid this issue.
The handling of an array of a single value should not be handled in code but in the field mapping.
I created a new MR for this proposal.

arwillame → created an issue.

Correct, i create a custom_field "Expertise Item field" that have a "firstname" and "lastname" field.
Then i can map then like : firstname => $.*.firstName
This will populate the "Expertise Item field" with multiple items with each have "firstname" and "lastname" filled in.

arwillame → made their first commit to this issue’s fork.

arwillame → created an issue.

Hi,
I have the same issue and it persists when i list the external entities.
It seems the entity does not have an id when i try to list them.
To avoid that i propose to set the entity id as you do in the "RestClient->save(), but in the externalEntityStorage->mapRawDataToExternalEntities()

Something like this at the end of the method maybe :

   if ($entities[$id]->id() === NULL) {
        $entities[$id]->set('id', $id);
    }

arwillame → created an issue.

arwillame → created an issue.

arwillame → created an issue.

Hi,
Thanks for the feedback.
I merged the main branch into my MR and made the requested changes.

It seem in certain cases, the media are not copied correctly, still need a bit of work to have a good solution.
Also not a big fan of how the undefinded languages are an exception.
if ($referencedEntity->language()->getId() === LanguageInterface::LANGCODE_NOT_SPECIFIED) {
This is most probably the remaining issue i got.
I'll need a bit more time to work on this.

Side node, i still need to change the weight as it is in MR.
I believe it's because otherwhise "language_access" module take the lead on the route and dont take this one into account.

Hi, merged the last version of the code and tested it again and it seem to be working fine for me now !
Thanks for your contribution on this issue !

Hi, Thanks to have reviewed this, i tested it on my issue and it's working fine for me.

arwillame → created an issue.

arwillame → created an issue.

Hi,
I added the missing schema file in the module.
Best regards.

Hi, i made the little change in the schema file.
Best regards.

arwillame → made their first commit to this issue’s fork.

arwillame → made their first commit to this issue’s fork.

arwillame → created an issue.

Hi, I just checked the code a bit and test it again.

I think my issue comes because of my data structure.
I have a node with a field that reference a paragraph. Then this paragraph references a bunch of paragraph wich hav they strings fields.

I think the code only check for the string on the first layer of paragraph.

I think we should dive into all possible subparagraphs with a recursive method in https://git.drupalcode.org/project/ai/-/blob/1.0.x/modules/ai_content_su...

I Updated my MR ( 395) to integrate the fact that an entity reference can reference an other entity and it has to be translated ortakethe value from the default language as well.

So i created a recursive to go through all the references.
It translate any text that has been found by the extractor and copy any reference from the main language if no translation found.

We have the same issue with the entity references.

Example, if a node has an entity reference field that references a paragraph, and that paragraph references a media, the media will not be copied to the translation.

At the end it i just added the missing type "webform_flexbox" in the type to ignore

arwillame → created an issue.

with Merge request !395, if no translation found for the entity reference, we just copy the reference value on the translated entity.
This way, image and all other reference from a source language will be ported on the translation.
This feel more natural for editors since the all page looks the same instead of just having only the texts translated.

I, Thanks for the the idea, it seems nice indeed.

I made a quick check with one project where language_access module is used.
First it just did not seemed to work, so i updated the weight .
Then i had a php error when testing it :

TypeError: array_splice(): Argument #1 ($array) must be of type array, null given in array_splice() (regel 177 van /var/www/html/web/modules/contrib/ai/modules/ai_translate/src/Form/AiTranslateForm.php).

So i would say it would need a bit more of testing/work.

My example contain webform, but i had the same issue with a simple media with file stored in the private part. I was not able to access it neither. So i dont think it's linked to webform.
I had the issue localy as well so i assume it woul not be related to a load balancer issue.

arwillame → created an issue.

I created a MR to add a try -> catch arround the entityReference->save() so in case an error like that occur, the translation still continue and the error is logged as an error in the watchdog.

arwillame → created an issue.

I created a MR that will avoid infinite loop.
The MR add a parameter "Maximum Reference Depth" that allow the user to limit the reference depth.
Default value is set to 1 so it translate only the first referenced element.

Remaining task :
It seem that even if the depth is larger than 1, the system extract correctly all the data but only translate the first level entity.
If the referenced entity reference an other entity again, the second entity will not be translated correctly.
This still need to be fixed.

I'm not able to reproduce it anymore.
Maybe the code changes have solved the issue, or an open ai update do not generate that type of content anymore.
I'll close this thicket.

I fixed the phpcs et cspell issues.

I made more intensive testing and the MR seems fine for me for a "classic" install as specified in my last comment.

For the layout builder configured to be asymmetric (with Layout Builder Asymmetric Translation → ), i created a contrib module to handle this case :
https://www.drupal.org/project/ai_translate_lb_asymmetric →

With this module, you'll be able to translate you layout builder page with the following config:

- Translatable Fields: Ensure that the fields referencing blocks (layout) are configured to be translatable.
- Non-Translatable Block Fields: Make sure that the block fields themselves are not set to be translatable.

I tested as well and it seem to work in the specific default case :
- The content type has a node "layout_builder__layout" wich is NOT TRANSLATABLE
- All the blocks used in Layout builder need to be translatable and have text fields translatable.

So in this configuration it will work as expected, it will just create a translation for the node, where the blocks ids are kept from the source language and translated.

It's the same idea as for the paragraph, this is a "symmetric" setup.

For my use case i would need to be able to have an "asymmetric" translation of the blocks for layout_builder (Dupplicate the block from source language, translate it and attach it to node translation).

I'll work on a module as i made for the asymmetric paragraphs : https://www.drupal.org/project/ai_translate_paragraph_asymetric →

Handle the "allowed_languages" module as well in the last commit of the MR.
I would prefer to use a more global method to fiw this but i dont for now.
If anyone has a better approach?

The issue was in fact linked to the fact we are using revisions in our site.

Inspecting the logic of the content translation controller

I discover that there is a check of the revision to be sure to load the right entity in the right revision.

I added this check in my merge request and it solves my issue.

arwillame → created an issue.

arwillame → created an issue.

arwillame → created an issue.

arwillame → created an issue.

Loading the entity is not working in every cases.
It seems it could be linked to the workflow module. Still investigating.z

arwillame → created an issue.

It is the case, i enabled gitlab ci and the phpcs job runs without issues

Thanks for you review,

I updated the .info file to add the missing line.

For the ai_translate_paragraph_asymetric/src/Plugin/FieldTextExtractor/TextFieldParagraphAsymetricExtractor.php not using the right alphabetical order in the "use", i believe it's fine, i had the issue before and it was because my "Drupal/coder" was not up to date.
Maybe you have the same issue?

For the Readme, is it mandatory to have that line size? it seems a bit too small if i respect this.

Plus, it seems to pass the default phpcs in the pipeline.

arwillame → created an issue.

I created a MR where I altered the prompt to explicitly instructs the summarization function to avoid returning the original text or restating it word-for-word. This should lead to more useful and concise summaries.

arwillame → created an issue.

Thanks for your remark, i changed the branch name to 1.0.x

arwillame → created an issue.

arwillame → created an issue.

arwillame → created an issue.

arwillame → created an issue.

I had some issues testing your proposition.
The form configuration do not allowed me to have multiple values in the "entity_autocomplete".

So i created a patch that allow the user to add multiple entities using the "#tags" parameter.
I also add a check to only do that when the related view allow ultiple elements.

Finally i made a little change in the execute method to be able to use the multiple values in the context filter itself.

Please feel free to comment my proposition.

arwillame → created an issue.

I got the same issue, the configuration is well recorded in stabase but i got the main front page on every domain..

I found that in contrib/domain/domain_config/src/DomainConfigOverrider.php

$config_override_exists never goes to TRUE because "$GLOBALS['config']" variable does not include my configurations.

Here is a patch proposal.