Cluster observability

The in-process panel shows the live feed for its own node. For fleet-wide live telemetry — every node's HTTP and SQL events aggregated in one place — Orbit ships three sibling modules, releasable independently:

Module	Role
orbit/proto	Connect-RPC contract + generated stubs.
orbit/agent	In-process agent that ships events to an admin server.
orbit/server	Standalone admin server that receives them.

Most applications need none of this. The root orbit module's in-process panel works on its own, and a single-process deployment can aggregate across nodes with just the Redis relay (cluster_* in Configuration). Reach for the agent/server fleet when you want a dedicated, always-on observability server independent of any single application node.

How the pieces fit

  app node 1 ──[orbit/agent]──┐
  app node 2 ──[orbit/agent]──┼──> [orbit/server] ──> admin UI
  app node N ──[orbit/agent]──┘     (Connect-RPC bidi stream)

Each agent embeds in a framework process and streams observability events over a single Connect-RPC bidirectional stream to the standalone server, which fans them out to connected admin UIs. The wire contract between all three is orbit/proto.