{"slug": "dockerfile-for-running-aleph-alpha-kolibri-1-on-dgx-spark", "title": "Dockerfile for running Aleph Alpha Kolibri-1 on DGX Spark", "summary": "A developer published a Dockerfile that layers Aleph Alpha's proprietary Kolibri-1 inference plugin onto a native SM121 (GB10) vLLM build for the DGX Spark, preserving the Blackwell-compiled kernels. The recipe installs aleph-alpha-inference 1.0.0 with --no-deps and pins the existing torch and transformers versions via an override file, preventing pip from swapping in PyPI wheels that would break the SM121 CUDA-13 build. A build-time self-check verifies the package and its vllm.general_plugins entry point register correctly.", "body_md": "|  | # Derived SM121 (GB10) vLLM image with the Aleph Alpha Kolibri proprietary plugin. | \n|  | # | \n|  | # Base: eugr/spark-vllm:latest — a native GB10 (SM121/12.1a) vLLM build produced | \n|  | # by eugr/spark-vllm-docker (FlashInfer + vLLM compiled for the Blackwell 40-bit SoC). | \n|  | # | \n|  | # The Aleph Alpha plugin (aleph-alpha-inference 1.0.0) is a pure-Python | \n|  | # vllm.general_plugins entry point that adds Kolibri1ForCausalLM plus the | \n|  | # kolibri1 reasoning/tool parsers. It is version-pinned to vllm>=0.29,<0.30, so | \n|  | # we install it with --no-deps to keep the SM121-compiled vLLM/torch intact and | \n|  | # only --override the declared torch/transformers bounds instead of letting pip | \n|  | # swap in PyPI wheels (which would lose the SM121 kernels / break CUDA-13 torch). | \n|  | FROM eugr/spark-vllm:latest | \n|  |  | \n|  | # Reuse the image's own pip/uv/cache conventions so installs stay consistent. | \n|  | ENV DEBIAN_FRONTEND=noninteractive | \n|  | ENV PIP_BREAK_SYSTEM_PACKAGES=1 | \n|  | ENV UV_SYSTEM_PYTHON=1 | \n|  | ENV UV_BREAK_SYSTEM_PACKAGES=1 | \n|  | ENV UV_LINK_MODE=copy | \n|  |  | \n|  | # Pin the already-installed (SM121/GB10) torch & transformers to their current | \n|  | # versions so the plugin's dependency lower-bounds cannot trigger a swap to a | \n|  | # non-SM121 PyPI wheel, then install the plugin without its vLLM dependency. | \n|  | RUN set -eux; \\ | \n|  | PINNED_TORCH=$(python3 -c \"import torch; print(torch.__version__)\"); \\ | \n|  | PINNED_TF=$(python3 -c \"import importlib.metadata as m; print(m.version('transformers'))\" 2>/dev/null \\|\\| echo \"0\"); \\ | \n|  | echo \"torch==${PINNED_TORCH}\" > /tmp/plugin-override.txt; \\ | \n|  | if [ \"$PINNED_TF\" != \"0\" ]; then echo \"transformers==${PINNED_TF}\" >> /tmp/plugin-override.txt; fi; \\ | \n|  | uv pip install aleph-alpha-inference --no-deps --override /tmp/plugin-override.txt | \n|  |  | \n|  | # Self-check: the package and its vLLM entry point are importable/registered. | \n|  | RUN python3 -c \"import importlib.metadata as m; print('aleph-alpha-inference', m.version('aleph-alpha-inference'))\" && \\ | \n|  | python3 -c \"import importlib.metadata as m; \\ | \n|  | eps=m.entry_points(); \\ | \n|  | EP=[e for e in eps.select(group='vllm.general_plugins') if 'aleph' in e.name]; \\ | \n|  | assert EP, 'aleph plugin entry point not found'; \\ | \n|  | [print('entry point:', e.name, '->', e.value) for e in EP]\" |", "url": "https://wpnews.pro/news/dockerfile-for-running-aleph-alpha-kolibri-1-on-dgx-spark", "canonical_source": "https://gist.github.com/stelterlab/8ff5d29f9d9fe10d7c6452a2b767c62b", "published_at": "2026-10-03 18:35:17+00:00", "updated_at": "2026-10-06 07:46:17.332535+00:00", "lang": "en", "topics": ["ai-infrastructure", "large-language-models", "ai-tools", "mlops"], "entities": ["Aleph Alpha", "Kolibri-1", "NVIDIA DGX Spark", "vLLM", "eugr/spark-vllm", "aleph-alpha-inference", "FlashInfer", "GB10"], "also_reported_by": [], "alternates": {"html": "https://wpnews.pro/news/dockerfile-for-running-aleph-alpha-kolibri-1-on-dgx-spark", "markdown": "https://wpnews.pro/news/dockerfile-for-running-aleph-alpha-kolibri-1-on-dgx-spark.md", "text": "https://wpnews.pro/news/dockerfile-for-running-aleph-alpha-kolibri-1-on-dgx-spark.txt", "jsonld": "https://wpnews.pro/news/dockerfile-for-running-aleph-alpha-kolibri-1-on-dgx-spark.jsonld"}}