refactor(vllm): retire l4t13 pyproject.toml in favor of requirements-*.txt

pyproject.toml only existed because uv pip install -r requirements.txt
doesn't honor [tool.uv.sources]. The previous commit dropped [tool.uv.
sources] (PyPI now serves the aarch64 + cu130 wheels directly), so the
file no longer carries any logic the requirements-*.txt path can't.

Replace with the same two-file pattern every other build profile uses:

  - requirements-l4t13.txt       (accelerate / torch / transformers /
                                  bitsandbytes - matches cublas13's split)
  - requirements-l4t13-after.txt (vllm; runs after the base resolve so
                                  the cu130 torch wheel lands first)

install.sh's whole l4t13 elif branch goes away; libbackend.sh's
installRequirements already handles the requirements-install.txt build-
deps pass, the C_INCLUDE_PATH export for PORTABLE_PYTHON, and the
runProtogen call, so falling through to the standard else: branch
produces identical install behavior with less surface area.

No functional change at install time - same wheels, same order.

Assisted-by: Claude:claude-opus-4-7 [Read] [Edit] [Write] [Bash]
Signed-off-by: Ettore Di Giacinto <mudler@localai.io>

backend/python/vllm/install.sh

@@ -47,12 +47,7 @@ fi
 # (torch 2.11+ ships aarch64 + cu130 manylinux wheels and vllm 0.20+ ships
 # an aarch64 wheel pinned to that torch). They're cp312-only, so bump the
 # venv Python accordingly. JetPack 6 keeps cp310 + USE_PIP=true.
-#
-# l4t13 still drives the install through pyproject.toml (see the elif
-# branch below) so the requirements-install.txt build-deps pass runs
-# first; the historical [tool.uv.sources] / jetson-ai-lab pinning was
-# dropped after that mirror started shipping ABI-mismatched torch / vllm
-# pairs. See backend/python/vllm/pyproject.toml for the full story.
+# https://pytorch.org/blog/vllm-and-pytorch-work-together-to-improve-the-developer-experience-on-aarch64/
 if [ "x${BUILD_PROFILE}" == "xl4t12" ]; then
     USE_PIP=true
 fi
@@ -105,25 +100,6 @@ if [ "x${BUILD_TYPE}" == "xintel" ]; then
         export CMAKE_PREFIX_PATH="$(python -c 'import site; print(site.getsitepackages()[0])'):${CMAKE_PREFIX_PATH:-}"
         VLLM_TARGET_DEVICE=xpu uv pip install ${EXTRA_PIP_INSTALL_FLAGS:-} --no-deps .
     popd
-# L4T arm64 (JetPack 7): drive the install through pyproject.toml so the
-# requirements-install.txt build-deps pass (pybind11 for fastsafetensors,
-# etc.) can run before the main resolve under --no-build-isolation. Bypasses
-# installRequirements because requirements.txt doesn't carry that separate
-# pass natively. See backend/python/vllm/pyproject.toml for the full
-# rationale on why the jetson-ai-lab mirror was retired in favor of PyPI.
-elif [ "x${BUILD_PROFILE}" == "xl4t13" ]; then
-    ensureVenv
-    if [ "x${PORTABLE_PYTHON}" == "xtrue" ]; then
-        export C_INCLUDE_PATH="${C_INCLUDE_PATH:-}:$(_portable_dir)/include/python${PYTHON_VERSION}"
-    fi
-    pushd "${backend_dir}"
-        # Build deps first (matches installRequirements' requirements-install.txt
-        # pass — fastsafetensors and friends need pybind11 in the venv before
-        # their sdists can build under --no-build-isolation).
-        uv pip install ${EXTRA_PIP_INSTALL_FLAGS:-} -r requirements-install.txt
-        uv pip install ${EXTRA_PIP_INSTALL_FLAGS:-} --requirement pyproject.toml
-    popd
-    runProtogen
 # FROM_SOURCE=true on a CPU build skips the prebuilt vllm wheel in
 # requirements-cpu-after.txt and compiles vllm locally against the host's
 # actual CPU. Not used by default because it takes ~30-40 minutes, but

backend/python/vllm/pyproject.toml

@@ -1,53 +0,0 @@
-# L4T arm64 (JetPack 7 / sbsa cu130) install spec for the vllm backend.
-#
-# Since PyTorch 2.11 (April 2026) PyPI publishes aarch64 + cu130 manylinux
-# wheels directly for torch / torchvision / torchaudio, and vllm 0.20+ ships
-# an aarch64 wheel whose Requires-Dist pins those exact versions. uv's
-# resolver therefore locks an ABI-consistent set without any custom index.
-# https://pytorch.org/blog/vllm-and-pytorch-work-together-to-improve-the-developer-experience-on-aarch64/
-#
-# Historically this file pinned torch / vllm / flash-attn / torchvision /
-# torchaudio to pypi.jetson-ai-lab.io's SBSA cu130 mirror via
-# [tool.uv.sources]. That mirror drifted out of sync (it published torch
-# 2.11.0 next to a vllm wheel still built against torch 2.10's c10 ABI,
-# producing `undefined symbol: _ZN3c1013MessageLoggerC1EPKciib` at import
-# time). Moving to PyPI eliminates that drift class entirely.
-#
-# flash-attn is intentionally dropped: PyPI ships no aarch64 wheel for it,
-# but vLLM 0.20+ already bundles its own vllm_flash_attn (fa2 + fa3)
-# inside the main wheel, so the Dao-AILab package is not required at
-# runtime.
-#
-# pyproject.toml (rather than requirements.txt) is still used on l4t13 so
-# the build deps pass in requirements-install.txt - fastsafetensors's sdist
-# needs pybind11 in the venv before --no-build-isolation can succeed - can
-# run first; install.sh's l4t13 branch invokes `uv pip install --requirement
-# pyproject.toml` after that pre-pass.
-[project]
-name = "localai-vllm-l4t13"
-version = "0.0.0"
-requires-python = ">=3.12,<3.13"
-dependencies = [
-    # Mirror of requirements.txt - kept in sync manually for now since the
-    # l4t13 path bypasses installRequirements (see install.sh).
-    "grpcio==1.80.0",
-    "protobuf",
-    "certifi",
-    "setuptools",
-    "pillow",
-    "charset-normalizer>=3.4.7",
-    "chardet",
-    # Accelerator stack from PyPI (aarch64 + cu130 wheels). vllm's
-    # Requires-Dist locks torch==2.11.0 / torchvision==0.26.0 /
-    # torchaudio==2.11.0, so listing them unpinned here just lets the
-    # resolver echo those exact versions back.
-    "torch",
-    "torchvision",
-    "torchaudio",
-    "vllm",
-    # PyPI-resolvable packages that complete the runtime - accelerate,
-    # transformers, bitsandbytes carry their own wheels for aarch64.
-    "accelerate",
-    "transformers",
-    "bitsandbytes",
-]

backend/python/vllm/requirements-l4t13-after.txt

@@ -0,0 +1,4 @@
+# vLLM 0.20+ ships an aarch64 manylinux wheel on PyPI whose Requires-Dist pins
+# torch==2.11.0 / torchvision==0.26.0 / torchaudio==2.11.0, locking an ABI-
+# consistent set with the cu130 torch wheel installed above.
+vllm

backend/python/vllm/requirements-l4t13.txt

@@ -0,0 +1,8 @@
+# JetPack 7 / L4T arm64 + CUDA 13. Since PyTorch 2.11 (April 2026), PyPI ships
+# aarch64 + cu130 manylinux wheels for torch/torchvision/torchaudio directly,
+# so we no longer need a custom --extra-index-url for the L4T mirror.
+# https://pytorch.org/blog/vllm-and-pytorch-work-together-to-improve-the-developer-experience-on-aarch64/
+accelerate
+torch
+transformers
+bitsandbytes