What would you like to see?
Hi,
When using AnythingLLM local with the /stream-chat endpoint, the stream sends extremely small chunks (often single characters).
This causes serious rendering issues for RTL languages like Persian, resulting in broken spacing and unreadable output unless complex client-side buffering is implemented.
Is there a way to:
- Stream larger chunks?
- Or respect word boundaries for RTL languages?
Thanks!
