alt-glitch a0f0ab8f36 refactor(platform): one native_arch for local-runtime assets, desktop PE gate, managed Node ...

The three architecture readers could disagree under emulation.
One shared native-architecture result prevents wrong asset selection.

2026-09-22 01:13:23 -07:00

..

__init__.py

…

binaries.py

refactor(platform): one native_arch for local-runtime assets, desktop PE gate, managed Node

2026-09-22 01:13:23 -07:00

bootstrap.py

fix(local-runtime): cap resident models by the hardware budget, not a count

2026-09-20 20:42:05 -07:00

capabilities.py

…

catalog.json

…

catalog.py

…

context_policy.py

…

detect.py

…

endpoint.py

fix: trim llamacpp salvage to invariant tests, drop redundant config load, document fixed-port servers

2026-09-20 10:03:32 -07:00

estimator.py

…

gguf.py

fix(local-runtime): name the MXFP4 row in the GGML size table

2026-09-20 15:29:34 -07:00

growth.py

…

hardware.py

…

hf_browse.py

…

load_progress.py

…

presets.py

fix(local-runtime): cap resident models by the hardware budget, not a count

2026-09-20 20:42:05 -07:00

processes.py

fix: scrub credentials at the llama-server spawn only, not the generic probe spawner (review follow-up)

2026-09-20 11:50:48 -07:00

recovery.py

…

supervisor.py

fix: scrub credentials at the llama-server spawn only, not the generic probe spawner (review follow-up)

2026-09-20 11:50:48 -07:00