Repeated/broken text in response; frontend desync

[me-eujene]

Hey all,
First of all - it is a brilliant product. Thanks to everyone involved!
Did anyone encounter an issue where the front end gets de-synchronised with the actual LLM response? It shows as

• Repeating/broken text, which is fixed by refreshing the page
• The answer is not loading until I refresh the page.
  this is how it looks like:

This is what I see in logs:

[ResponseAborted: ]
 ⨯ unhandledRejection:  [ResponseAborted: ]
TypeError: Invalid state: WritableStream is closed
    at EventEmitter.<anonymous> (.next/server/app/api/chat/route.js:1:3008)
    at z.handleStream (.next/server/chunks/911.js:62:2926) {
  code: 'ERR_INVALID_STATE'

Browser logs
Error: Promised response from onMessage listener went out of scope

Any idea what might be causing this?

[delfireinoso]

Wich models are you using? Local or remote?

[me-eujene]

Remote openAI Op di 10 jun 2025, 23:41 schreef Delfi Reinoso ***@***.***>:

…

Wich models are you using? Local or remote? — Reply to this email directly, view it on GitHub <#800 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AS6ILOH7FENFAJBOTYXRO633C5GI5AVCNFSM6AAAAAB7AW6UW2VHI2DSMVQWIX3LMV43URDJONRXK43TNFXW4Q3PNVWWK3TUHMYTGNBSGYYDONA> . You are receiving this because you authored the thread.Message ID: ***@***.***>

[delfireinoso]

You should try use the hugginface embedding model and after this another remote model (Gemini, for example)

Some models vive strange results