z4j 1.3.3

z4j - open-source control plane for Python task infrastructure

z4j

Open-source control plane for Python task infrastructure. License: AGPL v3 - commercial license available (contact [e-mail]). Status: v1.0.10 (production-ready).

A modern, self-hosted alternative to Flower / Prometheus-only task monitoring. Monitor, manage, and control Celery / RQ / Dramatiq / Huey / arq / TaskIQ / APScheduler workers, tasks, schedules, and queues from a single dashboard.

This is the umbrella package. Installing z4j gets you the flagship experience (brain + dashboard + SQLite) plus a clean way to layer in adapter engines via extras.

Quick start - 30 seconds

pip install z4j
z4j serve

That's the entire setup. The first run automatically:

1. Creates ~/.z4j/z4j.db (bundled SQLite, no Postgres needed for evaluation).
2. Mints HMAC signing keys to ~/.z4j/secret.env (sessions + audit-log chaining survive restarts).
3. Runs Alembic migrations to head.
4. Boots the brain on http://localhost:7700.
5. Prints a one-time setup URL like http://localhost:7700/setup?token=....

Open the URL in your browser, create the admin account, and you land on the dashboard. Everything self-contained in ~/.z4j/. To start fresh: delete that directory and re-run.

See z4j-brain for the complete configuration reference (defaults, retention, rate limits, Argon2 cost, etc.). For production, set Z4J_SECRET, Z4J_SESSION_SECRET, Z4J_DATABASE_URL, Z4J_PUBLIC_URL, and Z4J_ALLOWED_HOSTS explicitly via env vars and back up the secret store.

Install shapes

# Just the brain (dashboard + SQLite)
pip install z4j

# Brain + Celery adapter (pulls z4j-celery + z4j-celerybeat)
pip install "z4j[celery]"

# Brain + Django + Celery
pip install "z4j[django,celery]"

# Switch the brain to PostgreSQL
pip install "z4j[postgres]"

# Every adapter engine at once (useful for CI)
pip install "z4j[agents]"

# The kitchen sink
pip install "z4j[all]"

Install with uv

The same shapes work verbatim under uv:

uv pip install z4j                  # brain-only
uv pip install "z4j[celery]"        # + celery adapter
uv pip install "z4j[django,celery]" # + django + celery
uv add z4j                          # add as a project dep

uv resolves the same dependency tree as pip (we run a uv smoke test before every release), and gives you reproducible lockfiles plus ~10x faster installs. Either tool ships the same wheel.

Which adapter extras exist

Extra	Installs	For
[celery]	z4j-celery + z4j-celerybeat	Celery workers + beat scheduler
[django]	z4j-django	Django task telemetry
[flask]	z4j-flask	Flask background tasks
[fastapi]	z4j-fastapi	FastAPI BackgroundTasks
[rq]	z4j-rq + z4j-rqscheduler	RQ workers + rq-scheduler
[dramatiq]	z4j-dramatiq	Dramatiq actors
[huey]	z4j-huey + z4j-hueyperiodic	Huey + periodic tasks
[arq]	z4j-arq + z4j-arqcron	arq workers + cron
[taskiq]	z4j-taskiq + z4j-taskiqscheduler	TaskIQ + scheduler
[apscheduler]	z4j-apscheduler	APScheduler jobs
[bare]	z4j-bare	Bare (no engine) agents

Which umbrella extras exist

Extra	Installs	For
[postgres]	z4j-brain[postgres] (asyncpg)	Production PostgreSQL backend
[agents]	every engine adapter	Full-stack dev / CI testing
[all]	[agents,postgres]	Kitchen sink

Agent-only installs (no AGPL in your app's venv)

If you're adding z4j to an existing Django / Celery / Flask application and you need Apache-2.0-only dependencies (no AGPL in your proprietary code's venv), skip the umbrella and install the standalone adapter packages directly:

pip install z4j-core z4j-django    # SDK + Django adapter
pip install z4j-celery             # SDK + Celery adapter
pip install z4j-rq                 # SDK + RQ adapter

Every adapter package is Apache 2.0 and depends only on z4j-core (also Apache 2.0). Point Z4J_BRAIN_URL at a running brain and the adapter will start streaming telemetry.

Docker Compose recipes

Prefer Docker to pip? This repo ships three ready-to-use compose files alongside the Python package:

git clone https://github.com/z4jdev/z4j.git
cd z4j
cp .env.example .env     # fill in Z4J_SECRET, Z4J_SESSION_SECRET, etc.

Then pick your mode:

# Evaluation / homelab (SQLite bundled in the image):
docker compose up -d

# Production self-host (adds PostgreSQL 18 sidecar):
docker compose -f docker-compose.postgres.yml up -d

# Add Caddy auto-HTTPS on top of either mode:
docker compose -f docker-compose.yml -f docker-compose.caddy.yml up -d
# or:
docker compose -f docker-compose.postgres.yml -f docker-compose.caddy.yml up -d

All three compose files reference the same z4jdev/z4j:latest image from Docker Hub - the mode switch happens at runtime via Z4J_DATABASE_URL, not at build time. Caddy is an optional overlay that layers auto-HTTPS on top.

Full deployment walkthrough, reverse-proxy alternatives (nginx, Traefik, Cloudflare Tunnel), and production hardening guide: https://z4j.dev.

Licensing

z4j (this umbrella package) is AGPL v3 because it installs z4j-brain by default. If that is incompatible with your policy, use the agent-only install recipe above.

• Agent packages (z4j-core, z4j-celery, z4j-django, ...): Apache 2.0. Free for proprietary use.
• Brain server (z4j-brain): AGPL v3. Commercial license available - contact [e-mail].

The split is deliberate. The brain is the network service operators run; the agents are the client libraries integrators embed. AGPL protects the server; Apache makes integration frictionless.

Documentation

Complete documentation, tutorials, API reference, deployment guides: https://z4j.dev.

• Source: https://github.com/z4jdev/z4j
• Issues: https://github.com/z4jdev/z4j/issues
• Changelog: https://github.com/z4jdev/z4j/blob/main/CHANGELOG.md