fix: coherent 1-bit decode shard 3 (GGUF-convention weight-fold fix)

fix: coherent 1-bit decode shard 2 (GGUF-convention weight-fold fix)

fix: coherent 1-bit decode shard 1 (GGUF-convention weight-fold fix)

fix: coherent 1-bit decode shard 0 (GGUF-convention weight-fold fix)

op-packages ship with the app, not the download (bundle policy refuses executables)

Bonsai-27B 1-bit v81 bundle (4 decode shards + 1-bit lmhead + fp16 embed + tokenizer + op-package stub)

Upload README.md with huggingface_hub

initial commit