Home

Trove Wiki

Trove is a read-only service registry for homelabs, small teams, and self-hosted infrastructure.

Agents run beside the systems you already operate, collect a snapshot of what exists, and push that snapshot to a central server. The server stores the catalog in SQLite, serves the dashboard, exposes read-only APIs, checks image freshness, tracks agent health, and sends alerts or email digests when something changes.

The important bit: Trove observes. It does not manage, restart, deploy, mutate, or control anything.

What Trove solves

Most homelabs end up with services spread across a few different places:

• Docker hosts
• Kubernetes clusters
• Proxmox VMs and LXC containers
• bare-metal/systemd services
• old boxes you forgot were still doing something useful

Each platform has its own UI and its own idea of health. Trove gives you one catalog view across all of them.

Use Trove when you want to answer questions like:

• What services are running across my whole environment?
• Which host does this thing live on?
• Which containers are stopped, unhealthy, stale, or outdated?
• Did an agent stop reporting?
• What changed recently?
• Are my running images behind their registry tags?

Core design

Trove is intentionally boring:

• Agents push reports to the server.
• The server never connects back to agents.
• Agents use read-only platform APIs.
• The database is a single SQLite file.
• The dashboard and read APIs can be protected with OIDC.
• The dashboard and API are read-only.
• Alerts are based on recorded state transitions.

That makes it NAT-friendly, easy to run, and much safer than a management plane.

Start here

• Quickstart - get the server and first agent running.
• Architecture - how Trove works end to end.
• Server - server process, HTTP routes, background loops, and configuration.
• Agents - how agents collect and push full-state reports.
• Alerts-and-Digest - instant alerts, cooldowns, and email summaries.
• Image-Freshness - how registry digest checks work.
• Data-Model - SQLite schema, migrations, and report ingest semantics.
• Operations - running, upgrading, backing up, and troubleshooting Trove.
• Authentication - OIDC login, Authentik setup, logout, and API tokens.
• Security-Model - what Trove can and cannot touch.
• Development - build, test, release, and contributor notes.

What Trove is not

Trove is not a replacement for Portainer, Kubernetes, Proxmox, or Uptime Kuma.

It does not:

• deploy workloads
• restart containers
• change Kubernetes resources
• modify Proxmox guests
• run commands on hosts
• replace uptime checks from outside the host
• expose a public management interface

That boundary is deliberate. Trove is the inventory and observation layer.