Skip to content

Getting Started

This section walks you through everything you need to go from a fresh checkout to running your first LLM inference with ZigLlama.

Prerequisites at a Glance

Requirement Minimum Recommended
Zig compiler 0.13.0 latest 0.13.x
Operating system Linux, macOS, or Windows Linux x86-64
RAM 4 GB 16 GB+ (for larger models)
CPU Any x86-64 or AArch64 AVX2-capable x86-64

No additional system libraries are required. BLAS back-ends (OpenBLAS, MKL, Accelerate) are optional and covered in the installation guide.

Section Contents

  • :material-download: Installation

    Install the Zig toolchain, clone the repository, verify your setup, and optionally configure hardware-accelerated BLAS libraries.

  • :material-play-circle: Quick Start

    Run the bundled demos, write your first inference program, execute layer-by-layer tests, and explore the 12 example files.

  • :material-hammer-wrench: Building from Source

    Understand build.zig, the available build targets and modes, cross-compilation, and CI/CD integration.

  • :material-file-tree: Project Structure

    Navigate the repository: source code layout, the 6-layer mapping, test organisation, examples, and documentation.

If you are new to ZigLlama, work through the pages in the order listed above:

  1. Installation -- make sure zig build test passes.
  2. Quick Start -- run the demos and try the minimal code sample.
  3. Building from Source -- learn how to customise builds.
  4. Project Structure -- orient yourself in the codebase before diving into the layer-by-layer documentation.

After completing this section, continue to the Architecture overview or jump directly to Layer 1: Foundations.