Skip to main content

Welcome to Pie

Research Prototype

Pie is currently a research prototype undergoing rapid development. It is pre-release software intended for testing and research purposes only. APIs and features may change without notice.

Pie is a programmable system for LLM serving. It combines a high-performance inference engine with a WebAssembly-based runtime that lets you run custom logic—called Inferlets—directly alongside your models.

To learn more about Pie's architecture and why it exists, see the Overview page.

Getting Started​

Ready to dive in? Head to the Installation guide to set up Pie, then follow the Tutorial to run your first model and build your first Inferlet.

Learn More​

  • Overview — Understand Pie's architecture and core concepts
  • CLI Reference — Command-line interface documentation
  • Client API — Programmatic access to Pie