bd config get|set|list|show|unset|set-many|validate|apply|drift — configuration, with three subcommands that are unusual enough to note.
bd config showprints “all effective configuration with provenance” — which value won, and from where. In a tool with per-repo config, git config, environment variables, and CLI flags, “why is this setting what it is?” is a real question and most tools cannot answer it.bd config driftdetects “config-vs-reality inconsistencies”, andbd config apply“reconciles system state to match configuration”. That is a declarative-convergence loop — the config is the desired state, drift reports divergence, apply fixes it — which is infrastructure-tool thinking applied to a CLI’s own setup.
bd context (effective backend identity and repository context), bd info (database information, including --whats-new), bd ping (connectivity), bd version, and bd metrics (anonymous usage metrics: on/off/example, honouring DO_NOT_TRACK) round out the introspection surface.
Compare seeds-config, which publishes a JSON Schema for its config so an external UI can render a form. Beads goes the other direction — richer imperative surface, provenance and drift instead of a machine-readable contract. Both are reasonable; the schema is easier to integrate against, the drift loop is easier to operate.
Maps to no canonical SDLC stage.
See Also
- beads-doctor — the health command drift feeds.
- beads-repo — the
routing.*keys. - seeds-config — the cross-store counterpart and its contrasting posture.