queue: runai-model-streamer PR #2267 opened, mark in-review

Found upstream (run-ai/runai-model-streamer), filled home/repo, and recorded
the feasibility findings in the notes.

Author: luhenry

.queue.yml

@@ -5275,11 +5275,22 @@ packages:
   notes: '24 Linux wheels upstream (abi: cp310,cp311,cp312,cp313,cp314); no riscv64 on PyPI or pypi.riseproject.dev.'
 - pkg: runai-model-streamer
   version: 0.16.1
-  home: ''
-  repo: ''
-  status: claimed
-  pr: null
-  notes: '2 Linux wheels upstream (abi: py3); no riscv64 on PyPI or pypi.riseproject.dev. PyPI records no project URL; locate upstream during research.'
+  home: https://github.com/run-ai/runai-model-streamer
+  repo: https://github.com/run-ai/runai-model-streamer
+  status: in-review
+  pr: https://github.com/riseproject-dev/python-wheels/pull/2267
+  notes: 'PyPI has no project URL (home_page/project_urls both null); found via web search
+    -- github.com/run-ai/runai-model-streamer, Run:ai (NVIDIA-owned) Model Streamer, tag
+    0.16.1 matches the PyPI version exactly. Pure ctypes wrapper: setup.py bdist_wheel
+    --plat-name bundles a Bazel-built libstreamer.so (py3-none-manylinux2014_<arch> on
+    PyPI, same scheme this port reuses), no CUDA/GPU dependency at all -- it is a CPU-side
+    local-file reader; the S3/GCS/Azure backends (needing AWS/Azure/GCP C++ SDKs) are
+    separate optional PyPI packages this port does not touch. cpp/toolchain/configure.bzl
+    hardcodes ARCHITECTURES = ["x86_64", "aarch64"]; one two-line patch adds riscv64 (the
+    toolchain macros are already fully generic over the arch string). bazel has no riscv64
+    release binary, so it is bootstrapped from source (gotcha 47), same recipe as
+    build-ray.yml. torch/numpy/humanize/safetensors (test-only) are all already on the
+    registry for riscv64.'
 - pkg: tgcrypto
   version: 1.2.5
   home: https://github.com/pyrogram