Welcome to Gyoza

Gyoza is a self-hosted container execution ecosystem. Write your logic in aFaaS (Function as a Service) style, package it once as a Docker image, and let the gyoza ecosystem schedule and execute it across any number of workers, with automatic state tracking, retry handling, and real-time progress monitoring.

Get started

📦 Install Gyoza Install the gyoza package and set up the environment variables. 🚀 Quickstart Write your first @gyoza_op function and run it from the command line.

Learn the concepts

🥟 Gyoza Ops The main unit of work, a containerised function with a standard I/O contract. 🧠 Gyoza Server The central control plane that stores definitions, accepts runs, and schedules work. ⚙️ Gyoza Worker The execution layer that pulls containers and runs your ops.

For users

▶️ Generate an OpRun Submit a run for a deployed op via the REST API. 📊 Monitor an OpRun Track progress in real time, retrieve outputs, and retry failed runs.

For developers

🔨 Build an Op Write the Python op, create a Dockerfile, and build the container image with gyoza build. 🧪 Test an Op Run your op locally from source or inside its Docker container before deploying it. 🥟 Deploy an Op Push the image to a registry and register the OpDefinition on the server with gyoza deploy. 🖥️ Deploy the Server Launch the gyoza server with Docker Compose and configure it with environment variables. ⚙️ Deploy a Worker Connect workers to the server, enable GPU support, and scale horizontally across machines.