Repository navigation
Provider request fails with assistant message prefill error on some models #19981
Closed as not planned
Description
Problem
I regularly hit this runtime error when sending messages with certain models:
This model does not support assistant message prefill. The conversation must end with a user message.
Expected behavior
OpenCode should avoid sending assistant-prefill style message payloads to providers/models that do not support it, or automatically normalize the message list so the final message is user-authored before request dispatch.
Actual behavior
Request fails and the session gets blocked by the provider error.
Why this matters
This is a recurring production UX issue for model/provider combinations with strict message-shape requirements.
Suggested direction
- Detect provider/model prefill support capability.
- When unsupported, strip/adjust assistant prefill before API call.
- Add regression coverage for message-shape normalization across providers.
Error text
This model does not support assistant message prefill. The conversation must end with a user message.
Activity
github-actions commentedon Mar 30, 2026
This issue has been automatically closed.
github-actions commentedon Mar 30, 2026
This issue doesn't fully meet our contributing guidelines.
What needs to be fixed:
Please edit this issue to address the above within 2 hours, or it will be automatically closed.
This issue might be a duplicate of existing issues. Please check:
finish=stop, triggering prefill error on claude-opus-4-6 #17982: Bug: OpenCode prompt loop continues afterfinish=stop, triggering prefill error on claude-opus-4-6If you believe this was flagged incorrectly, please let a maintainer know.