- Zig 87%
- Python 11%
- Shell 2%
- Docker <1%
prefect-server #
Zig implementation of Prefect server, for orchestrating workflows with Prefect clients and workers. supports SQLite or PostgreSQL storage and in-memory or Redis messaging. API coverage and compatibility work are tracked in the docs.
live: prefect-server.waow.tech
install #
with Docker:
docker run --rm -p 4200:4200 atcr.io/zzstoatzz.io/prefect-server:latest
when remapping the host port, also set PREFECT_UI_URL to that address
(e.g. -p 4201:4200 -e PREFECT_UI_URL=http://localhost:4201) so the startup
link opens the published port.
or build from source with Zig 0.17 as required by build.zig.zon:
zig build -Doptimize=ReleaseFast
./zig-out/bin/prefect-server
use #
the server starts the API and background services together, using SQLite and in-memory messaging by default. point your Prefect client or worker at it:
export PREFECT_API_URL=http://localhost:4200/api
the Docker image includes the Prefect UI. when running a source build, set
PREFECT_UI_STATIC_DIR to a built UI bundle to enable it.
design #
- client compatibility — reuse Prefect SDKs and workers against a smaller server implementation.
- separate API and services — scale request handling while managing background work independently.
- dual storage and messaging backends — support local development and persistent deployments with the same API.
develop #
Agent instructions live in AGENTS.md. Shared operator skills live
in .agents/skills/, with .claude/skills symlinked there for Claude Code.
just provides the development recipes; uv runs the Python test scripts.
just build-release # optimized build
zig build test # unit tests
just test # API tests; requires a running server
just --list # all recipes, including deployment and benchmarks
docs #
docs/ covers compatibility, architecture, configuration, deployment, and benchmark results. implementation details and historical design notes live there.