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
Context overflow from vLLM is not handled -> no auto compaction #17762
There are already many patterns for detecting context overflow, for different providers. However, there are no dedicated patterns for vLLM yet. This means that when using a vLLM instance as a provider, auto compaction is not trigger when running into an error due to context overflow.
Plugins
No response
OpenCode version
1.2.26
Steps to reproduce
Serve some model with vLLM
configure that vLLM instance as a provider (OpenAI compatible)
send a message to vLLM that is too large for the model's context window
Observe that there is an error, but no auto compaction
Regarding #17746: It's slightly different. vLLM does not return context_length_exceeded yet. See vllm-project/vllm#37011. I intentionally reported this separately and created separate PRs, to keep the changes small and isolated.
Activity
github-actions commentedon Mar 16, 2026
This issue might be a duplicate of existing issues. Please check:
context_length_exceededcode path, which may also address vLLM's responses depending on how vLLM formats its errors. It may be worth consolidating these or clarifying how vLLM's error format differs from what Auto compaction is not triggered by code: context_length_exceeded in API call errors #17746 already addresses.fix(core): detect vLLM context overflow errors
ghost91- commentedon Mar 16, 2026
Regarding #17746: It's slightly different. vLLM does not return
context_length_exceededyet. See vllm-project/vllm#37011. I intentionally reported this separately and created separate PRs, to keep the changes small and isolated.