Linebreaks are removed since the new Playground

Hi - I use OpenAI’s Playground to write text.

Since the new UI, line breaks have been removed, making any text generation a real problem.

I wish the formatting would stay the same, as before.

It’s a necessity for text generation.

6 Likes

Yeah, I’m having the same problem as of this morning.

Crazy, how did that pass any QA ???

1 Like

Maybe all the update requires is just to set the default model to GPT-4o to show impressive results and charge more money

and leave its users unable to use it

UP! I’m still hopeless on this one.

I’m sure I’m not the only one facing it.

Writing, coding… everything needs formating!

Now they think that we love new lines and double it for us…
It looks like OpenAI is full of researchers, or they are experimenting with Copilot … on the company product

image

i have same issue. it is annoying…!!!

Hello

For my part, when I copy and paste the content of a Playground answer, line breaks are added everywhere. This makes using the Playground totally useless on a daily basis.

The problem appeared yesterday 2024-05-20.

It would be important to have information from the OpenAI team to find out what’s going on.

In the meantime I have to use other companies’ solutions and I’m disappointed.

Thank you

2 Likes

Yep same, now when copying/pasting prompts 3x line breaks are added.

Good news that the collapsing text has stopped happening though! The Lord giveth and the Lord taketh away :wink:

1 Like

Also having the 3x problem. For vim users, you can just work around with:

'<,'>s/\n\n/\r/g

(Edit: or

'<,'>s/\n\s*\n/\r/g

)
Sure there are other ways in other text editors.
But still quite annoying!

2024/05/23
Problems remained unresolved

Same here, the bug remains

Hello

Still no improvement regarding the untimely addition of line breaks when we copy and paste the answer from the playground in chat mode, making copy and paste practically unusable, especially for coding.

I can no longer use OpenAI for my needs.

Is it possible to get a response from the OpenAI team please, if only to keep us informed? I’m new to the forum. And I’m not yet familiar with OpenAI’s commitment to dialogue with its users. Can I expect follow-ups or are we just users commenting on problems and sharing boyscout solutions to get by?

It’s very important for users to solve the problem of line returns in the Playground. At the moment, my workflow is completely destroyed by this bug. I don’t have this problem in Assistant mode or in ChatGPT, but if I use the Playground there are reasons.

Thank you and have a nice day

UPDATE: I found one way.

Right-click and select “paste while matching the style”.

It worked for me.

It’s an extra painful step, but it’s doable.

2 Likes

Your method is currently the only available solution until OpenAI thoroughly fixes this bug. Thank you for sharing it!

Any updates ? Still desperate about it. Thanks

I believe the issue arises because each new line in the playground is written using the <p> tag instead of the <br> tag. This causes some applications to insert a new line after it, as text editors typically handle paragraphs this way.
I understand this might be intentional, but I’m struggling to fully grasp the rationale behind it

The playground was screwed up by using a rich text editor called ProseMirror.

It messes with what is being copied, making even your own overriding scripts to fix this hard.

It has parameters that could potentially be tweaked to fix this abhorrent travesty.

clipboardSerializer⁠?: DOMSerializer

The DOM serializer to use when putting content onto the clipboard. If not given, the result of DOMSerializer.fromSchema will be used. This object will only have its serializeFragment method called, and you may provide an alternative object type implementing a compatible method.

clipboardTextSerializer⁠?: fn(content: Slice, view: EditorView) → string

A function that will be called to get the text for the current selection when copying text to the clipboard. By default, the editor will use textBetween on the selected range.

textBetween(from: number, to: number, blockSeparator⁠?: string, leafText⁠?: string | fn(leafNode: Node) → string | null) → string

Get all text between positions from and to. When blockSeparator is given, it will be inserted to separate text from different block nodes. If leafText is given, it’ll be inserted for every non-text leaf node encountered, otherwise leafText will be used.

2 Likes

So…we just need to wait for the OpenAI team to add the parameter for god damn ProseMirror?

Or revert it and don’t let a brochure company or whoever did the revamp use boilerplate.

This library would make it very hard to give chat completions the missing logprobs display you can get with completions.