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
{{ message }}
Repository navigation
Bug: Response text disappears after streaming completes (regression in latest update) #37075
When the assistant finishes generating a response, the streamed text disappears immediately after completion, leaving the chat area blank/empty. The content is not actually lost — refreshing the app restores it — but the UI fails to retain/render the finished message.
Environment
OpenCode version: v1.18.1
Operating System: Windows
Interface: Desktop app
Current Behavior
While the response is streaming (i.e., during "thinking"/generation), the text appears normally. However, as soon as generation finishes, the response disappears and the chat area becomes blank with no visible text.
Expected Behavior
The completed response should remain visible in the chat once streaming/generation finishes, without requiring a manual refresh.
Steps to Reproduce
Open the OpenCode desktop app on Windows (v1.18.1)
Start a new chat and send a prompt
Watch the response stream in normally while generating
Wait for the response to finish generating
Observe that the chat area goes blank immediately after completion
Workaround
Refreshing the app restores the response, confirming the content is generated and saved correctly — this appears to be a rendering/display bug rather than data loss.
Impact
⚠️Priority: High — This is a regression introduced in the latest update. It makes the chat effectively unusable without a workaround, since users must refresh after every response to see the answer.
Additional context
This did not happen prior to this update — it appears to be a new regression rather than a pre-existing issue.
Having partially similar problem - on my end some (totally random) parts of text never renders, but the text is existing and still there (switching conversation to another one and then back makes it OK).
Similar to this is problem when the thinking never renders, and loading the session does not fix it. Using OpenCode Zen DeepSeek V4 Flash Free, but it is random (= at some time, often on the beginning of the session, it just stops rendering); however I think it's model-dependent, as using NVIDIA Nemotron 3 Ultra 550B A55B from NVIDIA NIM does not have this problem (maybe had it once in like 20 sessions, with DeepSeek it's every session).
Using 1.18.1 CLI in VSCode terminal; however I have this problem for at least a weak (I guess since 1.17.X).
Having partially similar problem - on my end some (totally random) parts of text never renders, but the text is existing and still there (switching conversation to another one and then back makes it OK). Similar to this is problem when the thinking never renders, and loading the session does not fix it. Using OpenCode Zen DeepSeek V4 Flash Free, but it is random (= at some time, often on the beginning of the session, it just stops rendering); however I think it's model-dependent, as using NVIDIA Nemotron 3 Ultra 550B A55B from NVIDIA NIM does not have this problem (maybe had it once in like 20 sessions, with DeepSeek it's every session). Using 1.18.1 CLI in VSCode terminal; however I have this problem for at least a weak (I guess since 1.17.X).
Edit: looks like at least something might be a problem between OpenCode and VSCode terminal, as in the same folder in classic Powershell the thinking is there even for DeepSeek.
Activity
github-actions commentedon Jul 15, 2026
This issue might be a duplicate of existing issues. Please check:
DenisSilent commentedon Jul 15, 2026
Having partially similar problem - on my end some (totally random) parts of text never renders, but the text is existing and still there (switching conversation to another one and then back makes it OK).
Similar to this is problem when the thinking never renders, and loading the session does not fix it. Using OpenCode Zen DeepSeek V4 Flash Free, but it is random (= at some time, often on the beginning of the session, it just stops rendering); however I think it's model-dependent, as using NVIDIA Nemotron 3 Ultra 550B A55B from NVIDIA NIM does not have this problem (maybe had it once in like 20 sessions, with DeepSeek it's every session).
Using 1.18.1 CLI in VSCode terminal; however I have this problem for at least a weak (I guess since 1.17.X).
DenisSilent commentedon Jul 15, 2026
Edit: looks like at least something might be a problem between OpenCode and VSCode terminal, as in the same folder in classic Powershell the thinking is there even for DeepSeek.
github-actions commentedon Sep 14, 2026
To stay organized issues are automatically closed after 60 days of no activity. If the issue is still relevant please open a new one.