fix: use custom gptURL to verify key #815
Conversation
| @@ -13,29 +14,9 @@ export async function gptStatusCallback(status: boolean) { | |||
| const doc = dialog.window.document; | |||
|
|
|||
| try { | |||
| const models = await updateGPTModel(); | |||
There was a problem hiding this comment.
868acc5 has removed the model selection for custom model input. As a result, the logic of updateGPTModel() has become redundant.
There was a problem hiding this comment.
But authGPT() still requests /models to verify the availability of apiurl and key.
Perhaps by attempting a model chat request directly with limited tokens, it could provide more detailed error information. down to the model availability.
However, different providers currently have inconsistent HTTP status codes in their responses. For instance, when a model does not exist, OpenAI returns a 404, while One-API returns a 503.
|
Thanks. Since I no longer work on Zotero 6, I won't be able to test the z6 branch. If you confirm it's OK to merge, I'll push a new stable release. |
|
Certainly, I have been testing it on Zotero 6 for a while now. |
No description provided.