Fix dashboard sub-cent costs and the most-commented files label

CHANGELOG.md

@@ -12,6 +12,11 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 - The metrics API builds with Go 1.27 (`go.mod` and the Docker builder image `golang:1.27-alpine`). Replaces Dependabot PR #5, which bumped only the Dockerfile.
 - The DCO sign-off CI job skips release pull requests into `main`. Their commits were already checked on their way into `development`, and the history includes commits from before sign-off was required.
 
+### Fixed
+
+- Dashboard: costs under one cent show as, for example, `$0.0023` instead of `$0.00`.
+- Dashboard: the "Most-commented files" count column is now labelled "Reviews". It counts reviews that commented on the file, not individual comments.
+
 ## [1.0.0] - 2026-11-01
 
 First release as PullPicket. The engine's `ModuleVersion` restarts at `1.0.0` (it was `7.0.0` under the old name). The dashboard's engine version table is unaffected: it shows the tag the pipeline checks the templates repository out at (such as `v1.0.0`), or the short commit when it points at a branch.

docs/metrics.md

@@ -124,7 +124,7 @@ Pick a window (7, 14, 30 or 90 days, or all time) at the top. The page then show
 | Est. cost per day | Spend over time, from the [pricing file](#pricing-file). |
 | Latency per day (p50 vs p95) | Time spent waiting on the model per run. A rising p95 usually means bigger pull requests or throttling. |
 | Comment severity breakdown | `CRITICAL`, `WARNING` and `SUGGESTION` counts. Mostly suggestions can mean the rules file is too loose. |
-| Most-commented files | The files the reviewer flags most, per repository. |
+| Most-commented files | The files the reviewer flags most, per repository. The count is the number of reviews that left at least one comment on the file, not the number of comments. |
 | Errors by type | Failed runs by cause, such as `llm-api-error`, `input-too-large`, `truncated-response`, `content-filtered` or `post-failed`. See [troubleshooting](troubleshooting.md#other-errors). |
 | By engine version, By model | Runs, errors, latency, tokens and cost per PullPicket version and per model deployment, to compare a change before and after. |
 

metrics/internal/summary/types.go

@@ -55,7 +55,10 @@ type severityTotals struct {
 type fileHotspot struct {
 	Repository string `json:"repository"`
 	File       string `json:"file"`
-	Comments   int    `json:"comments"`
+	// Comments counts reviews (runs) that commented on the file, not individual
+	// comments: events carry filesWithComments, a per-run list of paths. The JSON
+	// name is kept for compatibility.
+	Comments int `json:"comments"`
 }
 
 // repoFileKey is the aggregation key for fileComments -- a struct key (not a

metrics/internal/web/dashboard.html

@@ -235,7 +235,7 @@ <h1 id="title">—</h1>
       <div class="table-wrap">
         <table id="filesTable">
           <thead>
-            <tr><th>Repository</th><th>File</th><th class="num">Comments</th></tr>
+            <tr><th>Repository</th><th>File</th><th class="num" title="Reviews that left at least one comment on this file">Reviews</th></tr>
           </thead>
           <tbody></tbody>
         </table>
@@ -305,6 +305,10 @@ <h1 id="title">—</h1>
 
 function formatUSD(n) {
   const v = Number(n) || 0;
+  // A single review often costs less than a cent; two decimals would show it as $0.00.
+  if (v > 0 && v < 0.01) {
+    return v < 0.0001 ? '<$0.0001' : '$' + Number(v.toPrecision(2)).toString();
+  }
   return '$' + v.toLocaleString(undefined, { minimumFractionDigits: 2, maximumFractionDigits: 2 });
 }