vllm dependencies

Vllm is used in Python projects. A high-throughput and memory-efficient inference and serving engine for LLMs It has 72 direct runtime dependencies. Check its dependency graph on PyDeps to understand the full transitive dependency tree, reverse dependents, known CVEs, and license compatibility before installing.

What is vllm?

A high-throughput and memory-efficient inference and serving engine for LLMs

What are the dependencies of vllm?

vllm declares 72 direct runtime dependencies on PyPI. Each one is resolved into the full dependency tree below:

regex dependencies
cachetools dependencies
psutil dependencies
sentencepiece dependencies
numpy dependencies
requests dependencies
tqdm dependencies
blake3 dependencies
py-cpuinfo dependencies
transformers dependencies
tokenizers dependencies
safetensors dependencies
protobuf dependencies
fastapi dependencies
starlette dependencies
aiohttp dependencies
openai dependencies
pydantic dependencies
prometheus_client dependencies
pillow dependencies
prometheus-fastapi-instrumentator dependencies
tiktoken dependencies
lm-format-enforcer dependencies
llguidance dependencies
outlines_core dependencies
diskcache dependencies
lark dependencies
xgrammar dependencies
typing_extensions dependencies
filelock dependencies
partial-json-parser dependencies
jsonschema dependencies
pyzmq dependencies
msgspec dependencies
mistral_common dependencies
opencv-python-headless dependencies
pyyaml dependencies
six dependencies
setuptools
einops dependencies
compressed-tensors dependencies
depyf dependencies
cloudpickle dependencies
watchfiles dependencies
python-json-logger dependencies
ninja dependencies
pybase64 dependencies
cbor2 dependencies
ijson dependencies
setproctitle dependencies
openai-harmony dependencies
anthropic dependencies
model-hosting-container-standards dependencies
mcp dependencies
opentelemetry-sdk
opentelemetry-api
opentelemetry-exporter-otlp
opentelemetry-semantic-conventions-ai
numba dependencies
torch dependencies
torchaudio dependencies
torchvision dependencies
flashinfer-python dependencies
flashinfer-cubin dependencies
apache-tvm-ffi dependencies
tilelang dependencies
nvidia-cudnn-frontend
fastsafetensors dependencies
nvidia-cutlass-dsl
quack-kernels dependencies
tokenspeed-mla dependencies
humming-kernels dependencies

vllm transitive dependencies

Beyond its direct dependencies, vllm can pull in further packages through its dependency tree. PyDeps resolves the entire chain from PyPI and deps.dev so you can see every transitive (nested) dependency of vllm, expand any node on demand, and understand the full set of code that ships when you run pip install vllm.

Does vllm have known vulnerabilities (CVEs)?

PyDeps checks vllm and every package in its dependency tree against the OSV vulnerability database in real time. For each CVE you can see the severity, the affected version ranges, and the first fixed version, so you know exactly which vllm version is safe to install before you ship.

What license does vllm use?

vllm is distributed under the Apache-2.0 license. PyDeps also shows the license of every dependency in the tree so you can audit license compatibility across your whole vllm install, not just the top-level package.

How to install vllm with all dependencies

Install from PyPI with pip install vllm. For offline or air-gapped environments, PyDeps can download vllm together with every resolved dependency as wheel files in a single bundle, matched to your target Python version and operating system.

Which packages depend on vllm?

Switch to the dependents view to see the reverse dependencies of vllm — the PyPI packages that list vllm as a requirement. Reverse dependencies are a strong signal of how widely a package is trusted and how disruptive a breaking change would be.