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
Server becomes unresponsive after external client disconnects during SSE streaming #15149
When running opencode serve with external clients (e.g., via curl or API client), if the client disconnects abruptly while connected to SSE endpoints (/event or /global/event), the server enters a corrupted state where subsequent requests fail with "Unexpected EOF" errors.
If your issue is specifically about the server becoming unresponsive (rather than just a resource leak), it may still be worth keeping open as a distinct symptom — but please review #14091 first to see if it covers your case.
Activity
github-actions commentedon Feb 26, 2026
This issue might be a duplicate of or closely related to existing issues. Please check:
serveandwebcommands leak child processes on exit; SSE heartbeat interval leaks on server-initiated close #14091: Directly related — describes the SSE/eventendpoint leak where Hono'sonAbortonly fires on client disconnect, not on server-initiated close. This is likely the root cause of the corrupted state you're seeing after an external client disconnects abruptly.If your issue is specifically about the server becoming unresponsive (rather than just a resource leak), it may still be worth keeping open as a distinct symptom — but please review #14091 first to see if it covers your case.
neocody commentedon Mar 6, 2026
Seeing this in production with CodeNomad (desktop client for OpenCode).
OpenCode
servecloses the/eventSSE endpoint after ~7 seconds:After this, the SSE stream never reopens server-side. The desktop client loses all events and the UI goes blank across all workspace instances.
OpenCode 1.2.18, macOS 15.7.4 ARM64. Multiple workspaces open simultaneously.
github-actions commentedon Jul 4, 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.