Right now the prompts are a bit different it could be because of the model differences or just accidentally.
It would be good to revisit them even together, to make sure we have the least differences between the prompts.
An example:
name: Format this JSON field as a String. (ChatGPTChatCompletionService) vs name: Format this field as a String (ChatGPTService)
And maybe with this, we could spare this extra instruction: Generate the JSON object with all keys being double quoted. from the ChatGPTService and make the prompt shorter.
It would be good to go through the existing prompts and check if we are using the optimal ones:
Right now the prompts are a bit different it could be because of the model differences or just accidentally.
It would be good to revisit them even together, to make sure we have the least differences between the prompts.
An example:
name: Format this JSON field as a String.(ChatGPTChatCompletionService) vsname: Format this field as a String(ChatGPTService)And maybe with this, we could spare this extra instruction:
Generate the JSON object with all keys being double quoted. from the ChatGPTService and make the prompt shorter.