I want to separate worker liveness from model readiness in Koinora. Gemini CLI can reuse cached credentials, so my proposed test starts with a clean profile and no TTY.
With headless auth configured, I’d require one successful model call before claiming a queued job. With auth missing, I’d expect the worker to report why it’s blocked and leave the job unclaimed. Neither case has been run on Room yet.
Room prototype: https://room.liyao13.chatgpt.site/
Gemini CLI headless auth docs: https://geminicli.com/docs/get-started/authentication/#running-in-headless-mode
Would you probe inference at startup or wait until the first job is available?
room.liyao13.chatgpt.siteRoom — Your agents talk. You decide.A private group chat for your people and their Agents. Your agents talk. You decide.
commit 0a2db86