kshitijk4poor f84eb589ea fix(local-runtime): a crashed router's empty group is released before the restart backoff ...

The watcher noticed a crashed router, waited out the restart backoff (up to 60 s) and only then
called stop_owned_group_survivors, which signals a group because its number is still in the
registered set. When the crash left nothing in the group, that number was free for reuse for the
whole backoff while still registered, so a stranger's group that got it would be sent TERM and then
KILL. The watcher now releases an already-empty group as soon as it sees the crash; a group with
survivors stays registered, and its number cannot be reused while they run.

Test: a router stand-in that exits at once is no longer registered when the backoff starts; fails
without the release.

2026-10-29 16:58:09 +05:30

..

__init__.py

…

binaries.py

…

bootstrap.py

fix(local-runtime): price a drafted model once in the residency cap

2026-10-17 00:48:10 -04:00

capabilities.py

fix(local): keep long local-model turns alive on laptops

2026-10-16 15:21:07 -04:00

catalog.json

fix(local-runtime): point Nemotron 3.5 Bolt at its renamed repositories

2026-10-17 00:48:10 -04:00

catalog.py

feat(local-runtime): list Nemotron 3.5 Bolt as an opt-in model

2026-10-16 02:52:31 -04:00

context_policy.py

…

detect.py

…

devices.py

…

endpoint.py

refactor(deadline): one poll_until for check-until-deadline loops

2026-10-27 16:30:14 +05:30

estimator.py

Merge remote-tracking branch 'origin/main' into feat/local-bolt-dflash

2026-10-16 23:26:46 -04:00

gguf.py

feat(local-runtime): read recurrent-state size from GGUF headers

2026-10-16 02:31:02 -04:00

growth.py

fix(local-runtime): fit drafts to free memory and keep update restarts

2026-10-16 12:09:47 -04:00

hardware.py

fix(local-runtime): keep the prompt cache beside a one-slot draft

2026-10-16 23:30:02 -04:00

hf_browse.py

…

load_progress.py

fix(local): keep long local-model turns alive on laptops

2026-10-16 15:21:07 -04:00

presets.py

fix(local-runtime): price a drafted model once in the residency cap

2026-10-17 00:48:10 -04:00

processes.py

refactor(process): spawn_owned is the one way to own a child; own/release replace the verb string

2026-10-29 16:58:09 +05:30

recovery.py

…

supervisor.py

fix(local-runtime): a crashed router's empty group is released before the restart backoff

2026-10-29 16:58:09 +05:30