Different Chat GPT results API vs. Workflow

[lisustar]

Hi,
I tried two different ways to integrate my ChatGPT prompt in my app and the results differ greatly.
I’ve integrated my ChatGPT prompt directly in my workflow and I’ve used the same prompt when initializing the API Call and then integrate that specific API. The results in the second case, when using the prompt in the API, are much more valuable and precise than the first option. I know that ChatGPT is generating every time different results, but the quality of the results are totally different. Even though the settings like temperature etc. are totally the same, when using the prompt in the workflow the results are totally random and not even close to the other way.
Does anyone know why this is the case? And how can I have the same quality of results when integrating the prompt directly in the workflow?

Option 1

Option 2

Thanks for your support
Katrin

[boston85719]

temperature is what sets the level of ‘creativity’ the AI has in drafting a response.

Change your temperature from 0.88 (which is High) lower, toward 0.

If that alone doesn’t work, then try changing your prompt.