On this page
Binary plomid (crates/cli, name=plomid-cli). Global connection options (most commands): --host (PLOMID_HOST, 127.0.0.1), --port (PLOMID_PORT, 5432), --user (PLOMID_USER, plomid), --password (PLOMID_PASSWORD, secret, hidden), --database (PLOMID_DATABASE, plomid), --data-dir (PLOMID_DATA_DIR, ./data).
plomid server {start|stop|status}#
start [conn]:create_dir_all; pid-exists → error; locate siblingplomid-server[.exe]; spawn with env; write pid;✓ PASS server started (pid N).startalso accepts--profile {edge|laptop|server|auto}(forwarded asPLOMID_PROFILE; validated by the server at startup). Local Unix development only.stop [conn]: read pid; Unixkill -TERM, elsesupported on Unix hosts only; remove pid;✓ PASS stop signal sent.status [conn]: 2 s TCP connect;✓ PASS server <addr> accepts TCP connectionsor unavailable.
plomid db {readiness|health|info|diagnostics} [--json] [conn]#
readiness:SELECT 1 → 1probe; text table + latency +READY/NOT READY(exit 1 if not); JSON{status, checks[], latency_ms}. Write/TransactionNOT SUPPORTED; PersistenceWARNING.health: +database_bytes+ live counters fromplomid_stat_counters()and load gauges fromplomid_htap_stats();HEALTHY/DEGRADED; memory/CPUunavailable.info(local): version/platform/arch/data dir/database+WAL bytes/tables (unavailablewhen uncountable).diagnostics: + configuration check + recommendations; errors when not ready.
plomid config|version|logs|trace|metrics#
config (password configured (not displayed)), version (plomid <ver>), logs/trace hints, metrics [--json] (probe latency plus live server counters from plomid_stat_counters() when reachable).
plomid benchmark run [--json] [conn]#
--operation select (only), --iterations 100, --concurrency 1 (≠1 errors), --duration-seconds, --dataset-size (recorded). Loops SELECT 1 over one connection; avg/p50/p95/p99/throughput; persists <data_dir>/benchmarks/<run-id>.json.
plomid test {list|run [name]}#
list: connectivity/read/write/persistence/recovery/performance. run with no name defaults to connectivity; run connectivity|read → readiness probes. run write|persistence|recovery|performance are hard errors (unknown test module), not notices.
plomid-server offline tools#
The server binary doubles as the backup tool (never starts a listener): backup --data <DIR> --to <DIR>, restore --from <DIR> --to <DIR>, backup-inspect --backup <DIR>, and help-backup (prints usage, exit 0). Both --flag value and --flag=value forms work. Mutating commands take a lock on the target directory and refuse live targets. See Backup and restore.
WireClient: simple Q + T/D/C/E/Z, 16 MiB backend limit, 5 s timeouts. No extended/SCRAM.