Configuration

Almost everything is a setting in the dashboard. The first-run wizard asks for what a first deploy needs, and /settings (admin-only) lets you change any of it later, live, with no restart and no file to edit. If you installed with the one-liner, that's the whole story: it set the handful of values the container itself needs before it can start, and everything after that is the UI.

The rest of this page is for the two cases where that isn't enough: the few values that genuinely can't live in the database, and the optional config file for people who'd rather manage settings as code.

What you set in the dashboard

/settings is one page per tab, all admin-only:

Use HTTPS sets the scheme of the Dashboard URL derived from the base domain, and the helper text under it shows the resulting origin.

Sign-in methods, and the instance-wide two-factor and passkey requirements, live on their own Authentication page rather than a /settings tab, see Authentication providers. Git hosting accounts live on Git Providers, see Connecting a git provider. The built-in registry's own auth toggle and public hostname live on its own Registry page's Settings tab, see Registry. Per-account preferences (theme, accent colour) live on your profile, not here.

Secrets you enter here, SMTP passwords, OAuth client secrets, Cloudflare and Pangolin tokens, are encrypted at rest with a key derived from AUTH_SECRET.

Base domain vs. Dashboard URL

The one pair worth understanding before you type into it, because they look like the same thing and aren't:

• Base domain is the DNS suffix your deployed services are routed under, a service appears at <slug>.<base domain>. It never contains a port, because a Traefik host rule can't have one.
• Dashboard URL is where you reach Homerun itself, scheme and port included. Leave it blank to derive it from the base domain; fill it in when they differ.

They're the same in a normal deployment (example.com → https://example.com). They differ in development, where the dashboard is on http://localhost:5173 while services are routed by Traefik as <slug>.localhost. Typing a port into Base domain moves it to the Dashboard URL for you rather than corrupting your service hostnames.

What the container needs before it can start

These three can't be dashboard settings, because they're needed before there's a database to read settings from, or because changing them at runtime would break every session in flight. The installer sets all three for you; with plain docker compose you set them once in .env and never touch them again.

ORIGIN is the one that bites people, so it's worth a paragraph. It's how the app is served, which is why it's fixed for the life of the process rather than editable in the UI: the browser's cookie security depends on whether it's http or https, and flipping that at runtime would rename the session cookie and sign everyone out mid-session. Two failure modes if it's wrong:

• Leaving it at localhost while reaching the instance at anything else makes the very first sign-up fail with "Invalid origin", because that's the only origin the auth layer trusts.
• Serving over plain HTTP at a bare IP while ORIGIN says https makes sign-in appear to succeed and then hang, because the browser silently discards a Secure cookie on a non-secure origin.

Set it to exactly what you type in the address bar: http://203.0.113.10:3000, or https://homerun.example.com.

Two more exist and you can usually ignore them: PORT (default 3000, the port the app listens on inside its container) and CONFIG_FILE (default ./homerun.yaml, where the optional file below lives).

The app and the worker

Homerun's Docker access lives in a separate worker container, not the dashboard container itself: the app asks the worker for everything from a container's status to the web terminal, over a small internal HTTP API. Every compose file in this repo already wires the two together with working defaults, so you only need these if you're running them apart from each other or want more control:

Nothing deploys, no container status refreshes and the web terminal won't open while the app can't reach the worker, so if something feels stuck, check the dashboard's setup diagnostics first (see The dashboard): it reports the worker and the Docker socket it fronts as two separate checks.

Compose-only variables

Read by compose.prod.yaml itself, not by the app, so they only apply if you're running that file by hand:

The optional YAML file

homerun.yaml exists for people who'd rather keep configuration in a file than click through a UI, infrastructure-as-code setups, config management, a git- tracked server. You do not need it. Every key in it is also a /settings field, and a value stored in the database wins over the file, so a setting you change in the UI stays changed.

compose.prod.yaml doesn't mount it : the compose path needs no file beyond .env. To use one, copy homerun.example.yaml next to the compose file and add the mount yourself:

services:
  app:
    volumes:
      - ./homerun.yaml:/app/homerun.yaml:ro

The example file carries a $schema comment pointing at homerun.schema.json, generated from the same schema that validates it, so an editor with the YAML language server extension gives you linting and autocomplete. The installer (Option A) writes a homerun.yaml and mounts it for you, so there's nothing to add there.

Three things have no file form at all and are dashboard-only: orchestration mode, DNS automation (Cloudflare and Pangolin), and OAuth/OIDC sign-in providers. They hold secrets or live state that belongs in the encrypted database row rather than a plaintext file on disk. An auth.oauthProviders key left in homerun.yaml from an older version is ignored; add those providers on the Authentication page instead. A brand new instance starts in swarm mode when its Docker daemon is already a rootful swarm manager (what the installer sets up) and in standalone mode otherwise; after that the mode only changes from Settings → Docker.

The first-run wizard

Signing in for the first time drops you into a six-step wizard (Core / Docker / Traefik / Email / DNS / Review) that sets exactly the fields above, once. The DNS step turns on Cloudflare or Pangolin automation with the same fields and Test checks as Settings → Networking, including the optional Newt credentials that have Homerun run the tunnel client itself; Pangolin's target host, target port and sign-in option keep their defaults there. It calls the same code /settings does, so there's nothing it can set that you can't change afterwards, and nothing it leaves out that you have to go find in a file.

A note on lockout

Saving a broken OAuth provider (an unreachable or invalid discovery URL) used to be able to lock the entire instance out, the auth layer validates every configured provider on every request that touches auth, including a plain page load. The Authentication page now validates a discovery URL before saving anything, and session lookups degrade to "signed out" rather than a hard error on any other auth failure, so /settings stays reachable to fix whatever's wrong. If you're editing the database by hand rather than going through the UI, keep this in mind.

Next steps

• Docker Cleanup, System Logs, and what the dashboard tells you about its own health.
• Authentication providers and Users and roles, sign-in methods and roles.