S03 · Doesn’t fit (interactive)

Situation. I ask for a model that doesn’t fit in free memory.

What happens. Nothing is evicted or substituted. The request is refused, with the memory needed against what’s available, the top memory holders, and my options.

What I see. An error in the client. From Phase 3 the explanation is inline; before that it’s on the menu bar and through ntfy.

How to override. Free memory (spark make-room, stop a job), pick a model that fits, or retry.

Until Phase 2 (added 2026-09-28, during Phase 1): Phase 1’s launch check refuses the same way: nothing is evicted or substituted. The client gets a plain error, and the explanation, the memory needed against what’s available, is the refused line of make status, on the Spark. There is no menu bar or ntfy yet. Task 16’s drill saw it: the coder’s request got a 500, and make status said it needed 29.0 GiB, with 45.5 GiB available and the 24 GiB reserve kept, 7.5 GiB short. (Added 2026-09-28, from Phase 1’s council: make status shows only the last refusal, and the next load that starts clears it. Earlier ones stay in llama-swap’s in-memory buffer until it restarts; Deploy the stack shows how to read them.)