OpenAI model selection option instead of hardcoded gpt-3.5-turbo
Refresh model button to manually fetch a list of current models from OpenAI
Filtering to exclude preview and specialised models
Editable ChatGPT prompt with testing facility under the tools menu at /admin.php?tags-test-openai/
Improved error handling for API failures
Fallback logic for when a selected model is no longer available
Tag rule filtering for when GPT returns invalid data (with error logging)
The model we're using on AVForums is gpt-5-chat-latest.
Note that it's possible to completely screw up the results if you break the prompt, so edit it with extreme care and use the testing tool to review the resulting tags.
If Tag Essentials is installed and the forum used while Content Tagging is disabled, it caused permission errors.
(Yes, can you believe, someone actually did that?)
So now Tag Essentials does not run its JS when Content Tagging is disabled.
Thanks for reporting the issues. I'm still not able to run 2.3 yet, so I haven't been able to do the usual testing myself on AVForums, but the coder has tested it on his local installation, so it should be ok.
Fixed a rare bug when updating the wiki and too many characters are returned. We now limit the number of characters taken.
Fixed an unrelated 2.3. bug we found.
After further testing, we've taken it out of beta.
This version is not compatible with Xenforo 2.2 and we will not be releasing further versions for 2.2.
We now have two versions of Tag Essenials. We have the version for Xenforo 2.2 which will have no further updates.
And we have this new version with Javascript updates to remove reliance on jQuery for Xenforo 2.3.x.
This version being for 2.3 is in beta. We will do a final release after Xenforo 2.3 goes gold.