System¶

/system is the operational dashboard. Five sticky sections in the left sidebar (hidden on mobile, where they collapse into stacked cards): Doctor, Token usage, Catalog, Team history, Maintenance.

Doctor¶

A live view of the /doctor checks:

Status summary — total checks, passed, failed, overall OK / Fail
Per-check rows: name, status badge (pass / fail), detail message, remediation hint
Skip network toggle — bypass live HTTP / SQL probes for an offline run
Re-run button

CLI equivalent: /doctor (with optional --skip-network, --debug). See /doctor.

Token usage¶

The accounting surface for everything AMX has spent on LLM calls.

Window picker at the top — Today / 24 h / 7 d / 30 d / All time (the selection is sticky across page visits)
Per-(provider, model) table — Provider, Model, Runs, Input tokens, Output tokens, Total tokens, Cost (USD). Sortable, totals row at the bottom
A Counted N runs in this window line below the table

Older runs that pre-date the live-pricing system contribute zero cost in this view. Use the ↻ button in the top bar to refresh the pricing cache; future runs will fold the up-to-date rates in.

CLI equivalent: /usage [TODAY|24h|7d|30d|all] and /usage --live to recompute against current rates.

Catalog¶

The status of the catalog index that backs Browse and /ask:

Ready / not initialised
Entity counts (schemas, tables, columns reachable in scope)
Description counts and coverage %
Index type (SQLite, Chroma)
Refresh button — triggers a sync job

Active sync jobs render below as a small queue with per-job progress.

CLI equivalent: /sync [--db-profile NAME…] and /rebuild.

Team history¶

The on/off toggle for the shared run-history backend:

Enabled — every /run, /run-apply, and /ask is dual-written to a shared backend the team owns. Other maintainers' runs show up in Runs and Audit
Disabled — local SQLite only (~/.amx/history.db)

A separate Placeholder cleanup button removes temporary placeholder columns that some backends create during dry-runs.

CLI equivalent: /history-store status / enable / disable / migrate-from-local / flush-pending / dump-ddl. See Shared history store.

Maintenance¶

Backend-specific admin actions, surfaced when they're available for the active profile (e.g. Databricks-only PAT rotation hints, MySQL information_schema reload). The exact buttons here depend on the profile in scope.