[ci] Clarify dashboard cohort grouping for legacy v1 records

apps/performance_dashboard/README.md

@@ -89,12 +89,15 @@ Trend charts show metric-specific axes and exact point details on hover/focus:
 - commit SHA
 - run source
 - stored status
+- canonical comparison status and reason
+- baseline status
 - baseline eligibility
 - PR number, branch, and Buildkite URL when present
 
-The latest status table uses the stored JSON `success` value. Recomputed
-baseline context applies each metric's percent and absolute regression floors
-and does not override stored status.
+The latest status table shows the stored JSON `success` value alongside the
+canonical comparison status and reason. Recomputed baseline context applies
+each metric's percent and absolute regression floors and does not override
+either stored result.
 
 ## API
 
@@ -108,7 +111,14 @@ V2 records use the same comparison cohort as CI: `workload_id`, `variant_id`,
 `benchmark_version`, `recipe_fingerprint`, `hardware_profile_id`, and
 `software_profile_id`. `model_id` and `gpu_type` remain display/filter
 metadata, so renaming either does not split history. Legacy records still group
-by `(model_id, gpu_type)`. Dashboard baselines use the latest five previous
-successful, baseline-eligible records in each group. Summary and trend filters
-match the latest display metadata after grouping, while the raw records endpoint
-continues to filter individual records.
+by `(model_id, gpu_type)` under the explicit `Legacy v1` label. Records that
+claim or partially use v2 identity without all required fields are labeled
+`Invalid v2` and remain scoped by model, GPU, and their available identity
+values. Dashboard baselines use the latest five previous successful,
+baseline-eligible records in each group.
+
+Summary rows, trend groups, and trend points expose `cohort_kind`,
+`result_schema_version`, `baseline_status`, `comparison_status`, and
+`comparison_status_reason`. Summary and trend filters match the latest display
+metadata after grouping, while the raw records endpoint continues to filter
+individual records.

apps/performance_dashboard/frontend/src/App.tsx

@@ -1,7 +1,7 @@
 import { useEffect, useMemo, useState } from "react";
 
 import { fetchSummary, fetchTrends, refreshData } from "./api";
-import type { CohortValue, RunSource, SummaryResponse, TrendGroup, TrendPoint } from "./api";
+import type { CohortKind, CohortValue, RunSource, SummaryResponse, TrendGroup, TrendPoint } from "./api";
 
 const METRIC_KEYS = ["latency", "throughput", "memory", "text_encoder_time_s", "dit_time_s", "vae_decode_time_s"];
 const RUN_SOURCES: Array<{ value: "" | RunSource; label: string }> = [
@@ -113,6 +113,7 @@ function metricLabel(metricKey: string) {
 type CohortFields = {
   model_id: string;
   gpu_type: string;
+  cohort_kind: CohortKind;
   workload_id: CohortValue;
   variant_id: CohortValue;
   benchmark_version: CohortValue;
@@ -121,50 +122,84 @@ type CohortFields = {
   software_profile_id: CohortValue;
 };
 
-function cohortValue(value: CohortValue) {
+function rawCohortValue(value: CohortValue) {
   if (value === null || value === undefined || value === "") {
-    return "legacy";
+    return "";
   }
   return String(value);
 }
 
+function displayCohortValue(value: CohortValue) {
+  return rawCohortValue(value) || "missing";
+}
+
 function shortCohortValue(value: CohortValue) {
-  const text = cohortValue(value);
-  if (text === "legacy" || text.length <= 14) {
-    return text;
-  }
-  return text.slice(0, 12);
+  const text = displayCohortValue(value);
+  return text.length <= 14 ? text : text.slice(0, 12);
 }
 
 function cohortKey(cohort: CohortFields) {
   return [
+    cohort.cohort_kind,
     cohort.model_id,
     cohort.gpu_type,
-    cohortValue(cohort.workload_id),
-    cohortValue(cohort.variant_id),
-    cohortValue(cohort.benchmark_version),
-    cohortValue(cohort.recipe_fingerprint),
-    cohortValue(cohort.hardware_profile_id),
-    cohortValue(cohort.software_profile_id)
+    rawCohortValue(cohort.workload_id),
+    rawCohortValue(cohort.variant_id),
+    rawCohortValue(cohort.benchmark_version),
+    rawCohortValue(cohort.recipe_fingerprint),
+    rawCohortValue(cohort.hardware_profile_id),
+    rawCohortValue(cohort.software_profile_id)
   ].join("|");
 }
 
 function cohortTitle(cohort: CohortFields) {
-  const workload = cohortValue(cohort.workload_id);
-  const variant = cohortValue(cohort.variant_id);
-  const version = cohortValue(cohort.benchmark_version);
-  const versionLabel = version === "legacy" ? version : `v${version}`;
-  return `${workload} / ${variant} / ${versionLabel}`;
+  if (cohort.cohort_kind === "legacy_v1") {
+    return "Legacy v1";
+  }
+
+  const workload = displayCohortValue(cohort.workload_id);
+  const variant = displayCohortValue(cohort.variant_id);
+  const version = displayCohortValue(cohort.benchmark_version);
+  const versionLabel = version === "missing" ? version : "v" + version;
+  const identity = [workload, variant, versionLabel].join(" / ");
+  return cohort.cohort_kind === "invalid_v2" ? "Invalid v2: " + identity : identity;
 }
 
 function cohortDetail(cohort: CohortFields) {
+  if (cohort.cohort_kind === "legacy_v1") {
+    return "";
+  }
   return [
-    `recipe ${shortCohortValue(cohort.recipe_fingerprint)}`,
+    "recipe " + shortCohortValue(cohort.recipe_fingerprint),
     shortCohortValue(cohort.hardware_profile_id),
     shortCohortValue(cohort.software_profile_id)
   ].join(" | ");
 }
 
+function statusLabel(value: string, fallback: string) {
+  if (!value) {
+    return fallback;
+  }
+  return value
+    .toLowerCase()
+    .split("_")
+    .map((part) => part.charAt(0).toUpperCase() + part.slice(1))
+    .join(" ");
+}
+
+function comparisonStatusLabel(status: string, cohortKind: CohortKind) {
+  return statusLabel(status, cohortKind === "legacy_v1" ? "Historical" : "Unknown");
+}
+
+function comparisonStatusClass(status: string, cohortKind: CohortKind) {
+  const normalized = status.toLowerCase().replaceAll("_", "-");
+  return normalized
+    ? "comparison-" + normalized
+    : cohortKind === "legacy_v1"
+      ? "comparison-historical"
+      : "comparison-unknown";
+}
+
 function formatMetricValue(metricKey: string, value: number | null | undefined, tooltip = false) {
   const definition = METRIC_DEFINITIONS[metricKey];
   if (!definition) {
@@ -273,7 +308,9 @@ function TrendChart({ group, metricKey }: { group: TrendGroup; metricKey: string
         {chartPoints.map((point) => {
           const pointLabel = `${metricLabel(metricKey)} ${formatMetricValue(metricKey, point.value, true)} at ${formatTime(
             point.point.timestamp
-          )}, commit ${shortSha(point.point.commit_sha)}, ${runSourceLabel(point.point.run_source)}`;
+          )}, commit ${shortSha(point.point.commit_sha)}, ${runSourceLabel(
+            point.point.run_source
+          )}, comparison ${comparisonStatusLabel(point.point.comparison_status, point.point.cohort_kind)}`;
           return (
             <g
               key={`${point.plotIndex}-${point.value}-${point.point.commit_sha ?? ""}`}
@@ -307,6 +344,9 @@ function TrendChart({ group, metricKey }: { group: TrendGroup; metricKey: string
           {metric?.secondary ? <span>{metric.secondary(activePoint.value)}</span> : null}
           <span>{shortSha(activePoint.point.commit_sha)}</span>
           <span>{runSourceLabel(activePoint.point.run_source)}</span>
+          <span>
+            {comparisonStatusLabel(activePoint.point.comparison_status, activePoint.point.cohort_kind)}
+          </span>
         </div>
       ) : null}
       <div className="point-tooltip" aria-live="polite">
@@ -318,6 +358,13 @@ function TrendChart({ group, metricKey }: { group: TrendGroup; metricKey: string
         <span>Commit {shortSha(selectedPoint.point.commit_sha)}</span>
         <span>{runSourceLabel(selectedPoint.point.run_source)}</span>
         <span>{selectedPoint.point.success ? "Stored status: pass" : "Stored status: fail"}</span>
+        <span>
+          Comparison: {comparisonStatusLabel(selectedPoint.point.comparison_status, selectedPoint.point.cohort_kind)}
+        </span>
+        <span>Baseline status: {statusLabel(selectedPoint.point.baseline_status, "Unavailable")}</span>
+        {selectedPoint.point.comparison_status_reason ? (
+          <span className="tooltip-wide">{selectedPoint.point.comparison_status_reason}</span>
+        ) : null}
         <span>{selectedPoint.point.baseline_eligible ? "Baseline eligible" : "Not baseline eligible"}</span>
         {selectedPoint.point.pr_number ? <span>PR #{selectedPoint.point.pr_number}</span> : null}
         {selectedPoint.point.branch ? <span>Branch {selectedPoint.point.branch}</span> : null}
@@ -486,6 +533,7 @@ export default function App() {
             <table>
               <thead>
                 <tr>
+                  <th>Comparison</th>
                   <th>Stored Status</th>
                   <th>Recomputed</th>
                   <th>Model</th>
@@ -506,6 +554,16 @@ export default function App() {
               <tbody>
                 {latestRows.map((row) => (
                   <tr key={cohortKey(row)}>
+                    <td>
+                      <div className="status-cell">
+                        <span className={`badge ${comparisonStatusClass(row.comparison_status, row.cohort_kind)}`}>
+                          {comparisonStatusLabel(row.comparison_status, row.cohort_kind)}
+                        </span>
+                        {row.comparison_status_reason ? (
+                          <span className="status-reason">{row.comparison_status_reason}</span>
+                        ) : null}
+                      </div>
+                    </td>
                     <td>
                       <span className={`badge ${row.status}`}>{row.status}</span>
                     </td>
@@ -519,7 +577,7 @@ export default function App() {
                     <td>
                       <div className="cohort-cell">
                         <strong>{cohortTitle(row)}</strong>
-                        <span>{cohortDetail(row)}</span>
+                        {cohortDetail(row) ? <span>{cohortDetail(row)}</span> : null}
                       </div>
                     </td>
                     <td>{shortSha(row.commit_sha)}</td>
@@ -566,7 +624,7 @@ export default function App() {
                     <p>
                       {group.model_id} | {group.gpu_type}
                       <span>{cohortTitle(group)}</span>
-                      <span>{cohortDetail(group)}</span>
+                      {cohortDetail(group) ? <span>{cohortDetail(group)}</span> : null}
                     </p>
                   </div>
                   <TrendChart group={group} metricKey={metricKey} />

apps/performance_dashboard/frontend/src/api.ts

@@ -23,6 +23,15 @@ export type ComparisonCohort = {
   hardware_profile_id: CohortValue;
   software_profile_id: CohortValue;
 };
+export type CohortKind = "v2" | "legacy_v1" | "invalid_v2";
+
+export type DashboardRecordMetadata = {
+  cohort_kind: CohortKind;
+  result_schema_version: CohortValue;
+  baseline_status: string;
+  comparison_status: string;
+  comparison_status_reason: string;
+};
 
 export type SummaryRow = {
   model_id: string;
@@ -45,7 +54,7 @@ export type SummaryRow = {
   build_id: string;
   job_id: string;
   metrics: Record<string, MetricValue>;
-} & ComparisonCohort;
+} & ComparisonCohort & DashboardRecordMetadata;
 
 export type RunSource = "pr" | "local" | "scheduled_main" | "unknown";
 
@@ -79,13 +88,13 @@ export type TrendPoint = {
   build_id: string;
   job_id: string;
   metrics: Record<string, number | null>;
-} & ComparisonCohort;
+} & ComparisonCohort & DashboardRecordMetadata;
 
 export type TrendGroup = {
   model_id: string;
   gpu_type: string;
   points: TrendPoint[];
-} & ComparisonCohort;
+} & ComparisonCohort & DashboardRecordMetadata;
 
 export type TrendsResponse = {
   groups: TrendGroup[];

apps/performance_dashboard/frontend/src/styles.css

@@ -191,7 +191,7 @@ h3 {
 
 table {
   width: 100%;
-  min-width: 1260px;
+  min-width: 1500px;
   border-collapse: collapse;
 }
 
@@ -233,6 +233,19 @@ td {
   font-size: 0.72rem;
 }
 
+.status-cell {
+  display: grid;
+  gap: 4px;
+  max-width: 280px;
+}
+
+.status-reason {
+  color: #607080;
+  font-size: 0.72rem;
+  line-height: 1.35;
+  white-space: normal;
+}
+
 .badge {
   display: inline-flex;
   align-items: center;
@@ -243,18 +256,43 @@ td {
   font-size: 0.75rem;
   font-weight: 800;
   text-transform: uppercase;
+  white-space: nowrap;
 }
 
-.badge.pass {
+.badge.pass,
+.badge.comparison-pass {
   color: #166534;
   background: #dcfce7;
 }
 
-.badge.fail {
+.badge.fail,
+.badge.comparison-regression,
+.badge.comparison-infra-error {
   color: #991b1b;
   background: #fee2e2;
 }
 
+.badge.comparison-calibration-needed {
+  color: #92400e;
+  background: #fef3c7;
+}
+
+.badge.comparison-recipe-mismatch {
+  color: #1d4ed8;
+  background: #dbeafe;
+}
+
+.badge.comparison-quality-blocked {
+  color: #6b3e00;
+  background: #ffedd5;
+}
+
+.badge.comparison-historical,
+.badge.comparison-unknown {
+  color: #475569;
+  background: #e2e8f0;
+}
+
 .badge.muted {
   opacity: 0.78;
 }
@@ -403,6 +441,10 @@ td {
   font-size: 0.9rem;
 }
 
+.point-tooltip .tooltip-wide {
+  grid-column: 1 / -1;
+}
+
 .point-tooltip a {
   color: #0f6b8f;
   font-weight: 700;

docs/contributing/performance_benchmarks.md

@@ -424,6 +424,13 @@ main/full-suite uploads and remain eligible for rolling baselines.
 Current `perf_*.json` artifacts that lack the v2 comparison identity are
 normalized for reporting but skip rolling-baseline comparison and are not marked
 baseline eligible.
+Dashboard reports keep historical v1 records readable by grouping them under
+the explicit `Legacy v1` label by `(model_id, gpu_type)`. Complete v2 cohorts
+remain separate and use the six-field comparison identity. Records that claim
+or partially use v2 identity without all required fields are labeled
+`Invalid v2` and remain scoped by their display metadata and available identity
+fields. Stored comparison status, comparison reason, and baseline status are
+shown when present; old records remain readable when those fields are absent.
 
 New v2 records compare only against the same `workload_id`, `variant_id`,
 `benchmark_version`, `recipe_fingerprint`, `hardware_profile_id`, and
@@ -487,7 +494,9 @@ When the rolling-baseline phase runs, it emits:
   per-benchmark row with current vs. baseline values for latency, throughput,
   memory, text encoder time, DiT time, and VAE decode time.
 * **Plotly dashboard** — `dashboard_<sha>_<ts>.html` showing time-series for
-  each metric grouped by comparison cohort.
+  each metric grouped by comparison cohort. Chart titles distinguish
+  `Legacy v1` and incomplete `Invalid v2` records, and point hover data includes
+  stored comparison and baseline status metadata.
 * **Normalized records** — `normalized_perf_*.json`, one per benchmark.
   Useful as input to the
   [`reseed-performance-baseline`](https://github.com/hao-ai-lab/FastVideo/blob/main/.agents/skills/reseed-performance-baseline/SKILL.md)