\n\n in the beginning of the outputs

[ruby shell]

hi, whenever i request https://api.openai.com/v1/completions
{
"model": "text-davinci-003",
"prompt": "what date today",
"max_tokens": 256,
"temperature": 0.7,
"top_p": 1,
"n": 1
}

i got extra \n\n

{
"id": "cmpl-72zbXBGbdV42HxWkyeacV9K5ykFgP",
"object": "text_completion",
"created": 1680947819,
"model": "text-davinci-003",
"choices": [
{
"text": "\n\nToday is August 13, 2020.",
"index": 0,
"logprobs": null,
"finish_reason": "stop"
}
],
"usage": {
"prompt_tokens": 3,
"completion_tokens": 9,
"total_tokens": 12
}
}

other example

[celest skiff]

can you not simply strip out any new line /carriage returns from the front of the string?

[ruby shell]

i tried, but requires additional step which is inefficient
is the \n\n default addition from the output?

[ruby shell]

so only me got the \n\n?

[celest skiff]

output from the Ai should be parsed and cleaned by your application

many languages include a .trim() function for this

[ruby shell]

Yes i know, so its by design right?

[celest skiff]

no, it's just that a large language model can output anything, so you have to do some text conditioning of the result

[ruby shell]

Since i tried in postman and all of my prompt gives extra /n/n

[celest skiff]

or... ask the AI to output it's results in JSON format and then desterilise it

[celest skiff]

I mean... it could be a postman thing, I don't use it much

[ruby shell]

Ok so it must be me, thanks

[ruby shell]

Actually not only in postman, i use power automate http call it it return the same thing

[celest skiff]

interesting, well I use the trim function in my c++ apps so.. I never encounter this kind of thing

[ruby shell]

i know the solution is to trim it, but i think its unnecessary to be generated by the engine

from the ai itself hahahah

[celest skiff]

You should treat all output as unclean and trim it, as best practice