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
My understanding (based on reading the other issues) the llm is returning this corrupt JSON which stops OpenCode execution. ?
I tested qwen 3.6 27b and 30b, so these models should / can not be used in OpenCode, at least not reliably ?
My understanding (based on reading the other issues) the llm is returning this corrupt JSON which stops OpenCode execution. ? I tested qwen 3.6 27b and 30b, so these models should / can not be used in OpenCode, at least not reliably ?
The first object is cut off at "logpro (should be "logprobs":null,...). Then data: appears as literal text, followed by a valid second chunk.
I've added detailed analysis and a fix pattern in #23442 (comment) — including a splitMalformedSSE pre-processor that splits on data: boundaries when the primary parser finds zero complete objects.
This confirms the bug is not provider-specific — it affects Z.AI GLM-5.1 and Ollama/LM Studio with Qwen 3.6-27b. Likely a shared upstream SSE library issue or a common proxy/load-balancer buffering problem.
Activity
github-actions commentedon May 1, 2026
This issue might be a duplicate of existing issues. Please check:
If your issue is distinct (e.g. a different reproduction path or different root cause), please add more context to differentiate it.
BastianZuehlke commentedon May 1, 2026
My understanding (based on reading the other issues) the llm is returning this corrupt JSON which stops OpenCode execution. ?
I tested qwen 3.6 27b and 30b, so these models should / can not be used in OpenCode, at least not reliably ?
water-in-stone commentedon May 2, 2026
I have submmited a MR to fix this bug.
darkhipo commentedon May 14, 2026
Same root cause as #23442
This is the same server-side SSE buffering bug. Both chunks concatenated on one line, first truncated mid-key:
The first object is cut off at
"logpro(should be"logprobs":null,...). Thendata:appears as literal text, followed by a valid second chunk.I've added detailed analysis and a fix pattern in #23442 (comment) — including a
splitMalformedSSEpre-processor that splits ondata:boundaries when the primary parser finds zero complete objects.This confirms the bug is not provider-specific — it affects Z.AI GLM-5.1 and Ollama/LM Studio with Qwen 3.6-27b. Likely a shared upstream SSE library issue or a common proxy/load-balancer buffering problem.
github-actions commentedon Jul 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.