llmgw/internal/web/templates/admin_usage.html

{{/* The usage tab of the people page. */}}
{{define "usage_body"}}

{{if .Error}}<p class="error">{{.Error}}</p>{{end}}

{{/* One row of controls above the charts, which is where a reader looks for
     them. Links rather than a script: the choice is in the URL, so a view worth
     coming back to can be bookmarked or pasted to somebody else. */}}
<div class="filters">
  <span class="filter-label">Group by</span>
  {{range .Groups}}<a class="chip{{if .Selected}} on{{end}}"
     href="/admin/usage?by={{.Key}}&amp;window={{$.Window}}">{{.Label}}</a>{{end}}
  <span class="filter-label">Window</span>
  {{range .Windows}}<a class="chip{{if .Selected}} on{{end}}"
     href="/admin/usage?by={{$.By}}&amp;window={{.Key}}">{{.Label}}</a>{{end}}
</div>

<div class="chart-row">
  {{template "tschart" dict "t" .Requests "title" "Requests"
     "note" "Requests served in each slice, counted where the key belongs now rather than where it was at the time."}}
  {{template "tschart" dict "t" .Tokens "title" "Tokens"
     "note" "Prompt and completion together, which is the figure a person recognises as what they consumed."}}
</div>

<div class="chart-row">
  {{if .Priced}}
  {{template "tschart" dict "t" .Spend "title" "Spend"
     "note" "In whole units, the same ones a budget is written in."}}
  {{else}}
  {{/* Said rather than drawn. A spend chart flat along the bottom reads as a
       quiet month, and this is not that. */}}
  <div class="chart-wrap">
    <h3 class="chart-title">Spend</h3>
    <p class="empty">No model has a price, so every request costs zero and this
    chart would be a flat line however busy the cluster is. Set prices in the
    <code>pricing</code> block of the server configuration.</p>
  </div>
  {{end}}
</div>

{{if .Rows}}
<table class="rows">
  <tr><th>Group</th><th class="num">Requests</th><th class="num">Tokens</th><th class="num">Spend</th><th class="num">Errors</th>
    {{if .HasStats}}<th class="num" title="Failed requests, refusals by a limit included">Error rate</th>
    <th class="num" title="Refused by a rate, token or budget limit">Limited</th>
    <th class="num" title="Time to first token, for streamed requests">First token p50 / p95</th>
    <th class="num">Latency p95</th>{{end}}</tr>
  {{range .Rows}}
  <tr>
    <td>{{.Name}}</td>
    <td class="num">{{compact .Requests}}</td>
    <td class="num">{{compact .Tokens}}</td>
    <td class="num">{{.Spend}}</td>
    <td class="num">{{if .Errors}}{{.Errors}}{{else}}—{{end}}</td>
    {{if $.HasStats}}{{with .Stats}}
    <td class="num{{if ge .ErrorRate 5.0}} q-warn{{end}}">{{if .Requests}}{{pct1 .ErrorRate}}%{{else}}—{{end}}</td>
    <td class="num{{if .Limited}} q-warn{{end}}">{{if .Limited}}{{compact .Limited}}{{else}}—{{end}}</td>
    <td class="num">{{pctile .TTFT 50}} / {{pctile .TTFT 95}}</td>
    <td class="num">{{pctile .Latency 95}}</td>{{end}}{{end}}
  </tr>
  {{end}}
</table>
{{else}}
<p class="empty">Nothing recorded in this window.</p>
{{end}}

{{with .GPUHours}}
<h2>GPU time, by owner</h2>
{{/* What each owner's models held, whether or not anybody asked them
     anything: a model kept loaded holds its cards all the same, and that is
     the time a GPU quota is about. */}}
<div class="chart-row">
  {{template "tschart" dict "t" $.GPUHeld "title" "GPUs held, by owner" "note" "The cards held by the replicas of each owner's models: models whose owner is a part of the quota tree."}}
</div>
<table class="rows">
  <tr><th>Owner</th><th class="num">GPU-hours</th><th class="num">Held now</th></tr>
  {{range .}}<tr><td>{{.Owner}}</td><td class="num">{{printf "%.1f" .Hours}}</td><td class="num">{{printf "%.0f" .Now}}</td></tr>{{end}}
</table>
<p class="note">From the metrics store, which keeps its history for as long as it is set to (fifteen days by default); a model with no owner is counted under “no owner”.</p>
{{end}}

{{/* The charts fold their tail; the table does not. Both are here because they
     answer different questions, and the one that is easy to get wrong is
     thinking the four coloured lines are all of it. */}}
<p class="note">The charts draw the four busiest groups by total and sum the
rest into “other”, so they always add up to the whole cluster. The table below
is every group, in the same order.</p>

<p class="note">A person is not a level of the quota tree — one person holds
keys in several projects — so grouping by person crosses the tree rather than
naming a level of it. Quotas are written on the other three, on the
<a href="/admin/quotas">Limits</a> tab.</p>

{{end}}