v1.0.5 — GPU single-worker images + attached release binaries

v1.0.5 — GPU single-worker images + attached release binaries

What this release adds (core-node / consensus payload unchanged from v1.0.4)

GPU single-worker container images (GHCR)

• Hopper / H100 (vLLM v0.19): ghcr.io/tensorcash/simple-worker:vllm-v019-h100
  docker pull ghcr.io/tensorcash/simple-worker:vllm-v019-h100
• Blackwell (vLLM v0.19, arm64, source build):
  ghcr.io/tensorcash/vllm-backend:blackwell-arm64-pow-v19-src
  docker pull ghcr.io/tensorcash/vllm-backend:blackwell-arm64-pow-v19-src
• All service images (core-node, vllm-backend, verification-api, miner-proxy,
  ipfs-models, llama-backend) are published to GHCR.

Release binaries

• llama-server-darwin-arm64.tar.gz — standalone PoW-enabled llama.cpp server (macOS, Apple Silicon).
• miner-proxy-darwin-arm64.tar.gz — miner proxy (macOS, Apple Silicon).
• TensorMiner-1.0.5-arm64.zip — TensorMiner desktop miner app (macOS, Apple Silicon).
  (Linux llama-server and the desktop tarballs build in CI.)

Running the macOS artifacts (ad-hoc signed)

The macOS binaries are ad-hoc signed and NOT notarized, so Gatekeeper will
quarantine them on download. Clear the quarantine attribute after unpacking:

TensorMiner app

unzip TensorMiner-1.0.5-arm64.zip
xattr -dr com.apple.quarantine TensorMiner.app
open TensorMiner.app

llama-server

tar xzf llama-server-darwin-arm64.tar.gz
xattr -dr com.apple.quarantine llama-server libs chat-templates 2>/dev/null
chmod +x llama-server
./llama-server --help

miner-proxy

tar xzf miner-proxy-darwin-arm64.tar.gz
xattr -dr com.apple.quarantine miner-proxy
chmod +x miner-proxy

If macOS still reports the app "cannot be opened", right-click the app and
choose Open once, or re-run the xattr command on the extracted path.

Notes

The core-node image and consensus rules are identical to v1.0.4 (same bcore
revision); this release is additive — worker images, standalone binaries, and
the desktop miner. No mandatory operator action.