feat(dashboard): Tags tab — group runs by any promptfoo tag key; remove legacy manual tags (av-qsxw)

apps/cli/src/commands/results/combine-run.ts

@@ -56,7 +56,6 @@ export interface CombineRunSource {
   readonly displayName: string;
   readonly manifestPath: string;
   readonly experiment: string;
-  readonly tags?: readonly string[];
 }
 
 export interface DuplicateConflict {
@@ -111,7 +110,6 @@ export interface CombineRunResult {
   readonly duplicateConflicts: readonly DuplicateConflict[];
   readonly testCount: number;
   readonly targetCount: number;
-  readonly tags: readonly string[];
 }
 
 function parseJsonlLine(line: string): ResultManifestRecord {
@@ -538,7 +536,6 @@ export function buildCombineRunSources(
   options?: {
     ids?: readonly string[];
     displayNames?: readonly string[];
-    tags?: readonly string[][];
   },
 ): CombineRunSource[] {
   return sourcePaths.map((sourcePath, index) => {
@@ -561,7 +558,6 @@ export function buildCombineRunSources(
       displayName: options?.displayNames?.[index] ?? path.basename(runDir),
       manifestPath,
       experiment: normalizeExperimentName(experiment),
-      tags: options?.tags?.[index],
     };
   });
 }
@@ -632,7 +628,6 @@ export function combineRunSources(options: CombineRunOptions): CombineRunResult
   const summaryPath = path.join(runDir, 'summary.json');
   writeJson(summaryPath, summaryWithMetadata);
 
-  const tags = [...new Set(loadedSources.flatMap((source) => source.tags ?? []))].sort();
   return {
     runDir,
     runId: toRunId(options.cwd, runDir),
@@ -644,6 +639,5 @@ export function combineRunSources(options: CombineRunOptions): CombineRunResult
     duplicateConflicts: conflicts,
     testCount: rows.length,
     targetCount: new Set(results.map((result) => result.target ?? 'unknown')).size,
-    tags,
   };
 }

apps/cli/src/commands/results/eval-runner.ts

@@ -35,7 +35,6 @@ import {
   normalizeExperimentName,
 } from '../eval/result-layout.js';
 import { findRepoRoot } from '../eval/shared.js';
-import { normalizeTags, writeRunTags } from './run-tags.js';
 
 // ── In-memory run tracker ────────────────────────────────────────────────
 
@@ -146,7 +145,6 @@ interface RunEvalRequest {
   test_ids?: string[];
   target?: string;
   experiment?: string;
-  tags?: string[];
   threshold?: number;
   workers?: number;
   /** Resume an interrupted run: skip already-completed tests and append results to `output`. */
@@ -176,24 +174,12 @@ function validateResumeOptions(req: RunEvalRequest): string | undefined {
   return undefined;
 }
 
-function parseInitialTags(value: unknown): string[] {
-  if (value === undefined) return [];
-  if (!Array.isArray(value)) {
-    throw new Error('tags must be an array of strings');
-  }
-  return normalizeTags(value);
-}
-
-function normalizeRunMetadata(req: RunEvalRequest): { experiment: string; tags: string[] } {
+function normalizeRunMetadata(req: RunEvalRequest): { experiment: string } {
   const experiment = normalizeExperimentName(req.experiment);
-  const tags = parseInitialTags(req.tags);
   if ((req.resume || req.rerun_failed) && req.experiment?.trim()) {
     throw new Error('experiment cannot be changed when resuming an existing run');
   }
-  if ((req.resume || req.rerun_failed) && tags.length > 0) {
-    throw new Error('initial tags can only be set when creating a new run');
-  }
-  return { experiment, tags };
+  return { experiment };
 }
 
 function buildCliArgs(req: RunEvalRequest, experiment?: string): string[] {
@@ -331,12 +317,6 @@ function openConsoleLogStream(outputDir: string): WriteStream | undefined {
   }
 }
 
-function writeInitialRunTags(outputDir: string, tags: readonly string[]): void {
-  if (tags.length === 0) return;
-  mkdirSync(outputDir, { recursive: true });
-  writeRunTags(path.join(outputDir, RESULT_INDEX_FILENAME), tags);
-}
-
 // ── Route registration ───────────────────────────────────────────────────
 
 // biome-ignore lint/suspicious/noExplicitAny: Hono Context generic varies by route
@@ -400,7 +380,7 @@ export function registerEvalRoutes(
       return c.json({ error: resumeError }, 400);
     }
 
-    let metadata: { experiment: string; tags: string[] };
+    let metadata: { experiment: string };
     try {
       metadata = normalizeRunMetadata(body);
     } catch (err) {
@@ -440,7 +420,6 @@ export function registerEvalRoutes(
     activeRuns.set(runId, run);
 
     try {
-      writeInitialRunTags(outputDir, metadata.tags);
       const child = spawn(cliPaths.binPath, [...cliPaths.args, ...args], {
         cwd,
         stdio: ['ignore', 'pipe', 'pipe'],
@@ -576,7 +555,7 @@ export function registerEvalRoutes(
       return c.json({ error: 'Invalid JSON body' }, 400);
     }
 
-    let metadata: { experiment: string; tags: string[] };
+    let metadata: { experiment: string };
     try {
       metadata = normalizeRunMetadata(body);
     } catch (err) {
@@ -636,7 +615,7 @@ export function registerEvalRoutes(
       return c.json({ error: resumeError }, 400);
     }
 
-    let metadata: { experiment: string; tags: string[] };
+    let metadata: { experiment: string };
     try {
       metadata = normalizeRunMetadata(body);
     } catch (err) {
@@ -671,7 +650,6 @@ export function registerEvalRoutes(
     activeRuns.set(runId, run);
 
     try {
-      writeInitialRunTags(outputDir, metadata.tags);
       const child = spawn(cliPaths.binPath, [...cliPaths.args, ...args], {
         cwd,
         stdio: ['ignore', 'pipe', 'pipe'],
@@ -775,7 +753,7 @@ export function registerEvalRoutes(
     } catch {
       return c.json({ error: 'Invalid JSON body' }, 400);
     }
-    let metadata: { experiment: string; tags: string[] };
+    let metadata: { experiment: string };
     try {
       metadata = normalizeRunMetadata(body);
     } catch (err) {

apps/cli/src/commands/results/manifest.ts

@@ -28,6 +28,8 @@ export interface ResultManifestRecord {
   readonly suite?: string;
   readonly category?: string;
   readonly experiment?: string;
+  /** promptfoo-shaped tag map (`Record<string,string>`), e.g. `{experiment, team, env}`. */
+  readonly tags?: Record<string, string>;
   readonly target?: string;
   readonly variant?: string;
   readonly score: number;
@@ -315,6 +317,8 @@ export interface LightweightResultRecord {
   readonly target?: string;
   readonly variant?: string;
   readonly experiment?: string;
+  /** promptfoo-shaped tag map from the JSONL row's `tags` field. */
+  readonly tags?: Record<string, string>;
   readonly score: number;
   readonly scores?: readonly Record<string, unknown>[];
   readonly executionStatus?: string;
@@ -324,6 +328,24 @@ export interface LightweightResultRecord {
   readonly runtimeSource?: RunRuntimeSourceMetadata;
 }
 
+/**
+ * Coerce a raw JSONL `tags` value into a `Record<string,string>`, dropping
+ * non-string values. Returns undefined when the map is absent or empty so the
+ * lightweight record stays sparse for old runs that never wrote a tags map.
+ */
+export function normalizeTagMap(value: unknown): Record<string, string> | undefined {
+  if (!value || typeof value !== 'object' || Array.isArray(value)) {
+    return undefined;
+  }
+  const entries: [string, string][] = [];
+  for (const [key, raw] of Object.entries(value as Record<string, unknown>)) {
+    if (typeof raw === 'string') {
+      entries.push([key, raw]);
+    }
+  }
+  return entries.length > 0 ? Object.fromEntries(entries) : undefined;
+}
+
 export function loadLightweightResults(sourceFile: string): LightweightResultRecord[] {
   const resolvedSourceFile = resolveRunManifestPath(sourceFile);
   const content = readFileSync(resolvedSourceFile, 'utf8');
@@ -335,6 +357,7 @@ export function loadLightweightResults(sourceFile: string): LightweightResultRec
     target: record.target,
     variant: record.variant,
     experiment: record.experiment,
+    tags: normalizeTagMap(record.tags),
     score: record.score,
     scores: record.scores,
     executionStatus: record.execution_status,