v41: US-NC-2 share host (code + shards + native Engram over HTTP)

v41: minimal kernel-phase staging worker

v41: stage code on CPU so GPU pods never fetch from HF

v41: stage code on CPU so GPU pods never fetch from HF

v41: B3 Engram staging worker (byte-range fetch, native layout, digests)

v41: volume-selectable CPU create + idempotent fetch

v41: volume-selectable CPU create + idempotent fetch

v41 fetch worker: config.json + tokenizer beside the GGUF

v41 launcher: selectable network volume

v41 serve worker: local volume first, HF only as fallback

v41: same-DC pre-fetch worker + authenticated runner fetch

v41: same-DC pre-fetch worker + authenticated runner fetch

v41 workers: authenticated + retrying bundle fetch (429 fix)

v41 workers: authenticated + retrying bundle fetch (429 fix)

v41 workers: authenticated + retrying bundle fetch (429 fix)

v41 serve worker: consume published GGUF, mapper patch keyed to llama.cpp naming

v41 gpu workers: --break-system-packages for PEP 668

v41 gpu workers: --break-system-packages for PEP 668

v41 launcher: shell runner that bootstraps python3/curl

v41 launcher: --no-network-volume

v41 gpu kernel worker: volume-first with HF shard fallback

v41 worker: pin bundle

v41 worker: pin bundle

v41 worker: pin bundle

v41 worker: pin bundle

v41 worker: versioned bundle + curl + naming self-check

v41 worker: versioned bundle + curl + naming self-check

v41 worker: versioned bundle + curl + naming self-check

v41 worker: versioned bundle + curl + naming self-check

v41: gpu workers

v41: gpu workers

v41: gpu workers

v41: publish worker + fixed packer bundle

v41: publish worker + fixed packer bundle

v41 worker: fail fast, always report mapping

v41 worker: export PYTHONPATH early, audit non-fatal

v41 worker: clean ERR trap

v41 worker: fix status echo unbound variable

v41 quant: headless CPU worker + runtime tree + V4.1 packer

v41 quant: headless CPU worker + runtime tree + V4.1 packer