After the upgrade to 2024.2 and setting up generative AI (SQLSERVER_SF), I tried to test AI code review. Now for some templates I get a response (see below), but most of the time I get no comment and no error. So I think there is something going wrong(?).
When I look in the debug tab or start the task with the process flow monitor (Windows), I can't find any information.
What could be the reason and where can I find it?
Best answer by Mark Jongeling
The problem is identified. OpenAI now starts the returned result with
```jsonand that was not expected. This causes the extraction of comments to fail and no comments are placed due to that. We will resolve this issue this sprint (ends 18 october).
+5Hi Andre,
When there are no errors and no comments, please submit a ticket. This shouldn’t happen.
Under the hood, this AI code review uses an enrichment.
You trace what happened via the enrichment run data of the enrichment ‘Template code review’. To dive even deeper, the information is stored in ‘enrichment_run_ai_data’ and ‘model_update’ entities of the Software Factory database.
I created a ticket for this question. It seems that OpenAI returns an answer that can't be parsed in 2024.2.
+23The problem is identified. OpenAI now starts the returned result with
```jsonand that was not expected. This causes the extraction of comments to fail and no comments are placed due to that. We will resolve this issue this sprint (ends 18 october).
Enter your E-mail address. We'll send you an e-mail with instructions to reset your password.