You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Describe the bug
On iOS browsers, while streaming a ChatGPT response, the window starts jumping when the text streamed is below the visible viewport. This continues even after scrolling manually.
This makes it impossible to read the response while it is being generated. For long GPT-4 answers, this behavior unfortunately makes the app unusable.
Where is it happening?
Which device [Mobile, iOS]:
Which browser: Safari, Chrome, Firefox
Which website:
To Reproduce
On iOS browser, prompt a ChatGPT answer longer than the available viewport. Jumping will start as soon as the viewport is filled up.
I have not tested for other LLMs.
Expected behavior
Min expected behavior to fix this bug:
The viewed screen stays fixed after scrolling manually once. The user can then read the text and scroll uninterrupted. This is the behavior of openai.com
Optimal expected behavior:
The viewed screen stays fixed automatically as soon as the streamed response exceeds the viewport. The user has to scroll down automatically, when he wants to see more of the response.
The text was updated successfully, but these errors were encountered:
LennardSchwarz
changed the title
[BUG] Viewport jumps while streaming GPT-4 response on iOS
[BUG] Viewport jumps while streaming long ChatGPT responses on iOS
Dec 29, 2023
Describe the bug
On iOS browsers, while streaming a ChatGPT response, the window starts jumping when the text streamed is below the visible viewport. This continues even after scrolling manually.
This makes it impossible to read the response while it is being generated. For long GPT-4 answers, this behavior unfortunately makes the app unusable.
Where is it happening?
To Reproduce
On iOS browser, prompt a ChatGPT answer longer than the available viewport. Jumping will start as soon as the viewport is filled up.
I have not tested for other LLMs.
Expected behavior
Min expected behavior to fix this bug:
The viewed screen stays fixed after scrolling manually once. The user can then read the text and scroll uninterrupted. This is the behavior of openai.com
Optimal expected behavior:
The viewed screen stays fixed automatically as soon as the streamed response exceeds the viewport. The user has to scroll down automatically, when he wants to see more of the response.
Screenshots / context
See attached video.
https://github.com/enricoros/big-AGI/assets/40561724/dee04725-3d22-46a6-a5f3-274ecf7787bf
The text was updated successfully, but these errors were encountered: