Extend search in automatic suggestion tab from translation memory only to weblate internal machinery, as well

[rhofer]

Situation

During translation, the "Automatic suggestions" tab provides powerful tooling and according machinery results, which are giving a major contribution to translators efficiency. It provides a search field in order to play with deviations / alternatives to the given source string.

Goal

As a translator, I want to use the search field in the "Automatic suggestions" tab in order to benefit from the "Copy" or "Copy and save" actions based on the provided results. Therefore, I expect to receive search results from any configured machinery, not limited to translation memory.

Problem

The search field only provides results given by Weblate Translation Memory machinery. This might be a relict from the times where we had a Machine translation tab and a Translation memory tab which were merged to Automatic suggestion tab.

Solution Approach

Extend this search to include any configured machinery. May to limit inclusion to weblate machinery (to avoid unforseeable external costs of other machines as DeepL et.al.)

Optional:

• offer options to search in TM only or to include other machineries
• could be realized with checkbox, or, two buttons search all and search TM
• Question is, how fine granular the selection of option shall be offered. I.e. shall it be possible to select/deselect any machinery specifically? > this can result in many checkboxes.

Describe alternatives you've considered

No response

Screenshots

No response

Additional context

No response

[rhofer]

/cc @KadAnna

[nijel]

This is not relict, it's caused by the behavior - you can not "search" in the machine translation, you can search only in memory like services.

[rhofer]

But when you click on "Automatic suggestion" tab, then results are received from any machinery. Why can the same not be provoked when entering a search string in the local search field belonging to the "Automatic suggestion" tab?

[nijel]

Yes, but the search can be for any term, it doesn't have to be term present in the current results.

With memory it is possible to search for anything and return matching phrases.

With machine translation, we prefer not to offer free form interface for translating - it usually has attached costs.

[github-actions]

This issue has been put aside. It is currently unclear if it will ever be implemented as it seems to cover too narrow of a use case or doesn't seem to fit into Weblate.

Please try to clarify the use case or consider proposing something more generic to make it useful to more users.

[rhofer]

@nijel thanks for your clarification. Would it then be possible to include weblate translation memory and weblate machinery in addition, but still skipping external ones as deepL?

[nijel]

Yes, that could be the way to go.