BUG: asyncio.run() in CLI kills event loop — interview and MCP tools fail #36
No Assignees
Notifications
Due Date
No due date set.
Dependencies
No dependencies set.
Reference: Rockachopa/Timmy-time-dashboard#36
Reference in New Issue
Block a user
Delete Branch "%!s()"
Deleting a branch is permanent. Although the deleted branch may continue to exist for a short time before it actually gets removed, it CANNOT be undone in most cases. Continue?
Problem
cli.pyline 198 useslambda msg: asyncio.run(chat(msg, session_id="interview"))which callsasyncio.run()per question. This creates and destroys event loops repeatedly. MCP tool subprocesses (Gitea, Filesystem) hold async state that references the first loop — subsequent calls hit "Event loop is closed."Result: ~50% of interview questions fail. Any CLI command that makes multiple async calls is unreliable.
Root Cause
asyncio.run()is designed to be called once per process. The interview calls it N times in a loop. MCP stdio transports keep references to the first loop.Files
src/timmy/cli.py— lines 184, 198src/timmy/tools.py— line 538 (same pattern in delegate_task)Acceptance Criteria
timmy interviewcompletes all questions without "Event loop is closed"timmy chat "multi sentence question"works reliablyasyncio.run()called more than once per CLI invocationApproach
Use a single persistent event loop for the CLI process. Either
asyncio.get_event_loop().run_until_complete()or refactor to a singleasyncio.run(main())that dispatches all calls.Priority: HIGH — blocks all other testing