Orion
Self-hosted monitoring, incident response and logs for personal projects and small teams.
Documentation · Quickstart · Set up with an AI assistant · orion.thesunday.studio
Run Orion on a machine you own. When something breaks you hear it first, and every byte stays on your
server. There is no hosted version and no account to create.
What you get
- Monitors for HTTP endpoints, websites, TCP ports, Docker containers, systemd services, PM2
processes, commands and resource thresholds.
- Incidents opened automatically when a monitor fails, with a timeline of what failed, who was told
and who acknowledged.
- Alerts routed by rule to webhook or email, with every delivery recorded and failed ones retried.
- Logs from your own apps over JSON or OpenTelemetry, searchable next to the incident.
- Status pages you can publish for the people who depend on you.
- One small footprint: one API process, a monitor worker, SQLite, and the bundled Console. Back it up
as ordinary files.
Why Orion?
Most uptime tools run checks from outside your network. That is useful, but it cannot see local service
state, Docker containers, systemd units, PM2 processes, command checks, or the machine's own resource
pressure.
Orion keeps that local context on the monitored host. Each Server runs next to the services it watches
and reports to a Core you control, so you can run it from Docker Compose without adopting Prometheus,
Postgres, Kubernetes or a hosted observability stack first.
Get started
About ten minutes, on the server that will host Core (any Linux machine with Docker). The full steps, with
every command, are in the Quickstart on the documentation site.
curl -fsSL https://raw.githubusercontent.com/sunday-studio/orion/main/deploy/scripts/setup-core.sh | bash
That starts Core (the API, the Console and the worker) with Docker Compose and prints the address and
your login. Then:
- Connect a server with the one-line installer the Console shows under Servers, Add server, on any Linux
or macOS host you want to watch.
- Add a monitor and an alert channel, then send a test message.
The settings are in deploy/.env.example; the compose file is
deploy/docker-compose.yml. Building Core from source on a new server:
Set up Core with Docker.
Want to see it work before installing anything? The demo starts Core, monitors a tiny Python service,
breaks it and checks recovery:
examples/python-sleep-compose/smoke.sh
Let your AI set it up
Hand the whole setup to an AI assistant. It asks you a few questions, then installs and verifies Core, connects
your servers and sends a test alert, asking before it changes anything.
Using Claude Code or another assistant that can run commands? Paste this:
Set up Orion (self-hosted monitoring) for me end to end. Read
https://raw.githubusercontent.com/sunday-studio/orion/main/docs/ai-setup.md and follow it step by step.
Ask me the questions it lists first, and ask before you change anything on a machine.
The guide is on the documentation site, and the instructions the assistant follows are in docs/ai-setup.md: what to ask, the exact
commands, how to check every step, and when to stop.
Documentation
The documentation site is at orion.thesunday.studio/docs. Start there: the quickstart, installing Core, adding a
server, sending logs, backups, the CLI and the API.
The in-repo reference, including deployment details and design notes, has its own map in
docs/README.md. The most useful pages:
Project layout
orion/
├── apps/
│ ├── node/ # Go Server daemon and orion-node compatibility CLI
│ ├── core/ # Go API server, SQLite, OpenAPI, embedded Console
│ ├── console/ # React/Vite UI source
│ └── website/ # Astro site and documentation
├── deploy/ # Docker Compose, systemd, launchd, install scripts
├── docs/ # architecture, deployment, development, milestones
├── packages/ # shared/generated package space
└── Makefile
