Skip to content

O05 — Hosts, discovery and tracked sources

Status: design stub. Back to the component map.

Purpose

Define the machines, profiles, launchers and browser runtimes the orchestrator tracks. Build a current inventory with enough provenance to distinguish known, running, stale and unreachable entries.

Existing foundation

Existing launchers, profile inspection and history sync provide pieces. A unified multi-host inventory/management client is proposed.

Proposed boundary

Discovery and control are different capabilities. Preserve provider/native IDs plus host/profile/source identity; the same session can appear in history and live inventory without becoming two agents. Unknown reachability is not stopped execution. The O21 host coordinator supplies unified management, foreground first and optionally daemon-hosted later; partner hosts retain local process ownership.

After failure, reconcile available records and verified live identities. Report unrecoverable or uncertain runs honestly; history-based relaunch is acceptable and does not require seamless process adoption.

Inputs and outputs: Operator-configured endpoints and source selections → host/profile/runtime/session inventory, freshness and available actions.

Candidate progression

  • Mock: Several hosts and profiles, including duplicate observations, an unavailable host and a browser-local runtime.
  • First usable slice: Configure one local management endpoint and show its actual profiles/owned runs plus supported read-only discoveries.
  • Later: Add remote hosts, source groups, configurable tracking rules and explicit connect/adopt paths where supported.

These are discussion candidates, not scheduled work or implemented capabilities.

Recorded review — 19 October 2026

Distinguish the coordinator's host inventory from browser-local test agents. The UI may display both, but browser-local entries come from its own application state, not remote administration or discovery of other tabs. Tracking scopes and refresh policy remain open questions; ask when a live discovery slice needs them. Data placement follows the investigation proposed in O01.

Your additions

  • Which scopes should be configured: individual profiles, directories, machines or saved groups?
  • Should discovery refresh continuously, on demand, or both—and how should stale results remain visible?

Additional requirements, examples and decisions: TBD with you.