01-setup.md

Tutorial 1: Setup & Build

Welcome to vllm-lite! In this tutorial we'll go from a fresh clone to a working build.

Prerequisites

• Rust 1.88+ (we use edition 2024 — matches rust-version in the root Cargo.toml and the rust-toolchain.toml pinned by the repo). Install via rustup:

  curl --proto '=https' --tlsv1.2 -sSf https://sh.rustup.rs | sh

• Git — for cloning and version control

• ~5 GB disk — for Rust toolchain, dependencies, and build artifacts

• (Optional) CUDA 11.8+ — only needed for GPU inference (Qwen3, Llama with CUDA kernels). CPU-only inference works without it.

Clone & Build

git clone https://github.com/pplmx/vllm-lite.git
cd vllm-lite
cargo build --workspace

The first build downloads ~300 crates and takes 5-15 minutes. Subsequent builds are incremental (seconds).

Verify the Build

# Run all unit + integration tests (skips #[ignore] slow tests)
just nextest

# Or with cargo directly
cargo test --workspace --no-fail-fast

Expected: all tests pass, 0 failures. The exact count changes every batch — run just nextest for the current number; don't trust hard-coded numbers in older docs.

Install Just (Task Runner)

We use just (a Make alternative) for common workflows:

cargo install just --locked

Verify: just --version → 1.x+

Code Style

Before committing, run:

just fmt-check    # cargo fmt --all --check
just clippy       # cargo clippy with workspace lints
just doc-check    # cargo doc --no-deps (0 warnings)

All three must pass for CI to merge your PR.

Next Steps

→ Tutorial 2: Load a Model