← Documentation

The report

Part of team usage — the header counts, each account and window line, the pace and ages, the per-team rows, and the why-lines.

usage on this machine, 1 team, 2 labs, 1 account

The header counts the machine: every team the walk read — a store whose record cannot be read is one of them — every lab any team's file puts in use (a lab in use with no account declared for it yet is one), and every account any file declares or any state holds a reading for. The counts are whole-machine in both views, whatever the caller's own team turns out to carry; the blocks under them are not: the restricted view renders the labs the caller's own team carries, and a lab only other teams put in use is counted and renders for the owner alone. Every block is a lab and its accounts, in lab name order; an account belongs to the lab its own team's rule gives it — the vendor of the first seat that spends it, else the account's own name — and one account keeps one lab whoever else maps it.

Under a lab, one line per account and window, grouping every team's own row for that window:

openai  session  left 40%  used 60%  resets in 44m  -  read 2m ago  check  fresh

The figure is the newest counted reading of that account and window on this machine, whichever team took it, and the line is status's own row shape (budgetLine) except for where the figure came from and how the line closes:

  • The caller's own team's reading prints as its own seat and moment — scout changed 4m ago, or - where no seat is recorded.
  • Another team's reads read by another team, 4m ago in the restricted view and read by <team>, 4m ago in the full one. On equal moments the caller's own team stands first, so a machine line says least about another team when the figure could be the caller's.
  • A reader whose team's watch is not recording is told so in a closing clause: (no watch is recording for <project>) for the caller's own team, (no watch is recording for it) for another team's reading the restricted view cannot name, and (not known whether a watch is recording) when that team's state cannot be read. A watch that is recording is not announced.
  • A line with nothing counted for it reads unknown and carries no figure. A figure inside its reserve says inside reserve 20%, and a figure that did not come from the first source its account names is status line (fallback) — the same two marks status prints.
  • A line with a figure ends it with the pace the figure is moving at — pace 5.3%/h, the used points per hour the window's last two readings measure, one decimal and signed — or with pace not known where there is no pair to measure. The rows carry the word too. A spend line, a spend account's row, and a machine line with nothing counted for it carry no word at all. A restricted machine line whose newest figure is another team's carries none either: that pace is that team's, withheld rather than unknown, and nothing stands in its place — the line passes straight from its state to its watch clause.

Pace is read from the pair the state keeps and nothing else: the reading before this one, its figure and the moment it moved (the point the watch writes beside a reading whenever one was in the state before it). It is the difference of the two figures over the hours between their moments, one decimal, signed — a window filling back up reads below zero, and a figure that did not move across its pair reads pace 0.0%/h. It reads pace not known where the state holds no usable pair: a reading stored before the point was written, a point whose reading is older than the section's stale_after, a pair whose two moments are the same, and a window whose reset changed since the pair was taken. Pace is a passenger — it moves no figure, no state, no mark and no other command's table, and status, the gate and the watch print the same with it or without it.

Ages are exact in both views — 2m, 40m, 1h30m — the moment a reader can compute the figure's age from; no view rounds them.

Under each line come that account's rows: the caller's own team's only in the restricted view, every team's in the full one. A row repeats the line's own fields — what is left and used, when it resets, the seat the figure came from, how long ago, its source and its state — after its team's name where status writes the indentation:

beacon  daily  left 70%  used 30%  resets unknown  claude-keeper  last seen 40m ago  status line (fallback)  stale

status line (fallback) marks a figure that did not come from the first source its account names, inside reserve 20% a figure inside its reserve, and last seen a stale figure with no known reset: the same words, from the same table, that status prints.

A row prints only while the reading behind it is bound: its account must be one the team's file in force names — a budget account, or the account a seat in force's own account:/vendor: resolves to — and its window and source must be ones this tool writes: session, daily, weekly; check, status_line. One that is not is not rendered at all, and one fixed line per kind says so, under the caller's own block:

note: a stored reading names an account this team's file does not: not shown
note: a stored reading carries a window or source this tool does not write: not shown

The team in force is the approved copy once a verified approval is in force, never the live file: a name only the live file writes — a vendor: edited after approval, a seat it added — binds no stored reading, and the standing's own why-line already says the file differs. A copy that cannot be read binds nothing at all: no block of the caller's own carries a row, the figures of its budgets included, and the one line

note: the approved copy of the team file cannot be read: nothing is shown by name

stands for them — mine is null in --json, and the watch clause names it rather than any project, since the name too is that copy's.

An account nothing can read carries the tool's own why-line under its rows — doctor's and the gate's own words, never a message a read produced:

  (beacon: no pattern can read this account)

one of no pattern can read this account, its check is unapproved; that account reads unknown, first sight only, not yet counted (doctor.ts:154, :192, gate.ts:90). A why prints only under an account that reads unknown everywhere in the report: an account whose newest counted reading is another team's says nothing about the team whose own pattern cannot read it. The restricted view prints another team's why as (another team: …) when the account is one the caller's own file names.

A file that names no account gets the tool's own sentence in the same place — not known: this team's file declares no account, so nothing is counted — and a file whose budgets are not the ones in force gets the why-line status prints for that standing: the one above for a file never approved, the legacy record's repair line, or the one that says the file differs from the approved copy and names the section — budgets. Those are why-lines, not refusals: the exit is 0 either way.

A lab whose accounts the caller's own file does not name is counted, never named or given a figure:

anthropic: 2 other accounts

and a lab with nothing declared for it yet says so:

anthropic  not known (no team declares an account for it yet)

The why-line for a file never approved — the “one above” this names — is on the reads-and-writes page.