Closed (fixed)
Project:
Universal AI Provider
Version:
1.0.0-beta2
Component:
Code
Priority:
Normal
Category:
Bug report
Assigned:
Reporter:
Created:
15 Jul 2026 at 06:46 UTC
Updated:
29 Jul 2026 at 13:05 UTC
Jump to comment: Most recent
Problem/Motivation
ai_provider_universal returns array
Steps to reproduce
when i try to configure search server and i want to use embedings engine configured in ai_provider_universal the dropdown shown "Universal|Array" and this breaks indexing as no valid embedings engine can be configured.
| Comment | File | Size | Author |
|---|---|---|---|
| universal_array.png | 50.79 KB | persuader |
Comments
Comment #2
leofishman commentedI am working on a fix, will rewrite part of the module so it will not require any patch to ai_core or ai_search, new version with the fix will be released today.
Thanks for reporting and for your patiente
Comment #4
leofishman commentedThanks for the report, @persuader.
Cause: getConfiguredModels() returned models nested by server (for optgroups). ai_search builds the embeddings dropdown with a flat foreach and concatenates the label, so PHP rendered "Universal | Array".
Fix in 1.0.0-beta2:
- Model options are always a flat model_id => label map (labels like "Server: model").
- Model entity IDs use a dot separator (server.model) so AI core's provider__model parsing and ai_search work without AI core patches.
Upgrade from beta1: drush updb (update 10102 rewrites stored model IDs) then drush cr. Re-open the Search API server form and pick the embeddings engine again if needed.
Release: https://www.drupal.org/project/ai_provider_universal/releases/1.0.0-beta2
Please confirm on beta2 and re-open if anything still fails.