The Overview tab

The Overview tab is the first thing you see when you open Trove. It answers one question at a glance: is telemetry flowing, and where is it going? It pairs a health summary with a live data-flow chart that traces every signal from the AI coding tools on your machine, through the bundled collector, out to the observability backends you configured.

Health summary

The top of the tab is a status band that rolls up the whole pipeline:

• Collector state — whether the bundled OpenTelemetry Collector sidecar is running, starting, or has crashed. The Rust core supervises the sidecar and restarts it on crash, so a transient blip self-heals; a persistent red state points at a config or binding problem.
• Active harnesses — how many of your detected harnesses are enabled and emitting.
• Active platforms — how many backends are enabled and receiving, with their aggregate health.
• Signals received / sent — running counts of OTLP signals the collector has accepted from harnesses (received) and forwarded to your backends (sent).

Note

“Received” and “Sent” count total signals (spans, metrics, and logs), not spans alone — so a healthy session shows both numbers climbing together.

The data-flow chart

The centerpiece is an animated graph of your live pipeline:

  harnesses  ──▶  Trove collector (localhost)  ──▶  your backends
  (sources)        normalize · tag · fan out        (destinations)

Each enabled harness appears as a source node on the left. Activity halos pulse on a node whenever telemetry is actively flowing from that tool, so you can watch a claude -p turn or a Cursor agent response light up the graph in real time. The collector sits in the middle; your enabled backends fan out on the right, each carrying the same health pill you see on the Platforms tab.

The Orbital Hub

With three or fewer enabled harnesses, each renders as its own labeled node. The moment you cross four or more, the source nodes collapse into a single animated cluster — the Orbital Hub — so a busy machine with a dozen tools stays legible instead of turning into a tangle. The hub keeps the per-source activity halos: you still see which tools are firing, just grouped into one orbiting constellation rather than a row of separate boxes.

Tip

If you want to confirm a specific tool is emitting and the Orbital Hub has it clustered, the Harnesses tab shows each row’s telemetry pill individually, and the Logs tab tails the raw collector output.

Collector and sidecar diagnostics

The Overview tab is also where the collector’s own health surfaces. The sidecar is a supervised child process of the Trove core: it is spawned at startup, monitored continuously, restarted with backoff on unexpected exit, and shut down cleanly when you quit the app. Overview reflects that lifecycle:

• A healthy collector means the OTLP receiver is up on 127.0.0.1 and the health endpoint is responding.
• A starting state appears briefly after a config change — applying new harness or platform settings regenerates the collector config and triggers a short reload (a 1–3 second blip is normal).
• A crashed or failed state is your cue to open the Logs tab, where the collector’s stderr tail usually names the cause (a bad endpoint, a missing binary, a port conflict).

Caution

The collector binds to localhost only (127.0.0.1). It receives OTLP from your machine and forwards exclusively to the backends you configured — never to Trove or Intevity. See The collector for the binding and forwarding model.

Where to go next

Harnesses Enable the AI coding tools that feed the Overview graph.

Platforms Configure the backends that appear as destination nodes.

Logs Tail the live collector log when a node turns red.

The collector How the bundled OTel Collector sidecar receives, normalizes, and exports.