diff --git a/apps/web/app/activity/page.tsx b/apps/web/app/activity/page.tsx
index c38bc88..2e8f701 100644
--- a/apps/web/app/activity/page.tsx
+++ b/apps/web/app/activity/page.tsx
@@ -11,6 +11,7 @@ import {
buildActivityFeed,
resolveActivitySortOrder,
} from "../../lib/activityFeed";
+import { attachActivityCompareLinks } from "../../lib/activityCompare";
import { collectArtifactFacets } from "../../lib/artifactFacets";
import { loadBenchmarkArtifacts, loadRunArtifacts } from "../../lib/data";
import { buildQueryString, paginateItems } from "../../lib/listPagination";
@@ -51,16 +52,20 @@ export default async function ActivityPage({
loadBenchmarkArtifacts(),
]);
const { models, presets } = collectArtifactFacets(runs, benchmarks);
- const feed = buildActivityFeed(runs, benchmarks, {
- kind,
- q: params.q,
- model: params.model,
- preset: params.preset,
- fast: params.fast,
- from: params.from,
- to: params.to,
- sort,
- });
+ const feed = attachActivityCompareLinks(
+ buildActivityFeed(runs, benchmarks, {
+ kind,
+ q: params.q,
+ model: params.model,
+ preset: params.preset,
+ fast: params.fast,
+ from: params.from,
+ to: params.to,
+ sort,
+ }),
+ runs,
+ benchmarks,
+ );
const paging = paginateItems(feed, params, {
defaultPageSize: 25,
maxPageSize: 100,
@@ -280,6 +285,21 @@ export default async function ActivityPage({
),
},
+ {
+ key: "compare",
+ label: "Compare",
+ hideOnMobile: true,
+ render: (row) => {
+ const href = (
+ row as { compareHref?: string }
+ ).compareHref;
+ return href ? (
+ Compare
+ ) : (
+ —
+ );
+ },
+ },
]}
data={paging.paged.map((entry) => ({
...entry,
@@ -289,15 +309,29 @@ export default async function ActivityPage({
`${(row as { kind: string }).kind}-${(row as { id: string }).id}`
}
renderCardActions={(row) => (
-
- Open{" "}
- {(row as { kind: string }).kind === "run"
- ? "trace"
- : "benchmark"}
-
+ <>
+
+ Open{" "}
+ {(row as { kind: string }).kind === "run"
+ ? "trace"
+ : "benchmark"}
+
+ {(row as { compareHref?: string })
+ .compareHref ? (
+
+ Compare
+
+ ) : null}
+
)}
/>
)}
diff --git a/apps/web/app/api/activity/route.ts b/apps/web/app/api/activity/route.ts
index 29e814a..61a605e 100644
--- a/apps/web/app/api/activity/route.ts
+++ b/apps/web/app/api/activity/route.ts
@@ -4,6 +4,7 @@ import {
type ActivityFeedFilters,
} from "../../../lib/activityFeed";
import { activityEntriesToCsv } from "../../../lib/activityExport";
+import { attachActivityCompareLinks } from "../../../lib/activityCompare";
import { loadBenchmarkArtifacts, loadRunArtifacts } from "../../../lib/data";
import { parseListPagination } from "../_shared/pagination";
@@ -33,7 +34,11 @@ export async function GET(request: Request) {
),
};
- const feed = buildActivityFeed(runs, benchmarks, filters);
+ const feed = attachActivityCompareLinks(
+ buildActivityFeed(runs, benchmarks, filters),
+ runs,
+ benchmarks,
+ );
const { offset, limit } = parseListPagination(url.searchParams);
const items = feed.slice(offset, offset + limit);
diff --git a/apps/web/app/api/benchmarks/[id]/route.ts b/apps/web/app/api/benchmarks/[id]/route.ts
index 9b59fd4..1495aee 100644
--- a/apps/web/app/api/benchmarks/[id]/route.ts
+++ b/apps/web/app/api/benchmarks/[id]/route.ts
@@ -1,15 +1,74 @@
-import { loadBenchmarkById } from "../../../../lib/data";
+import {
+ loadAnalysisIndex,
+ loadBenchmarkById,
+ loadBenchmarkPairsById,
+} from "../../../../lib/data";
+import {
+ buildBenchmarkRunRoster,
+ sortBenchmarkRunRoster,
+} from "../../../../lib/benchmarkRunRoster";
+import { findMostSimilarPeerRunId } from "../../../../lib/benchmarkPeers";
+import { benchmarkRunRosterToCsv } from "../../../../lib/listExport";
+import { buildBenchmarkOutlierLookup } from "../../../../lib/outlierLookup";
export async function GET(
request: Request,
context: { params: Promise<{ id: string }> },
) {
const { id } = await context.params;
+ const url = new URL(request.url);
+ const format = url.searchParams.get("format") ?? "json";
const benchmark = await loadBenchmarkById(id);
if (!benchmark) {
return Response.json({ error: "benchmark not found" }, { status: 404 });
}
- const download = new URL(request.url).searchParams.get("download") === "1";
+
+ if (format === "roster-csv") {
+ const [pairsData, index] = await Promise.all([
+ loadBenchmarkPairsById(id),
+ loadAnalysisIndex(),
+ ]);
+ const pairs =
+ pairsData.pairs.length > 0
+ ? pairsData.pairs
+ : (benchmark.payload.summary?.stability?.pairs ?? []);
+ const rosterRunIds =
+ pairsData.runIds.length > 0
+ ? pairsData.runIds
+ : (benchmark.payload.runIds ?? []);
+ const outlierLookup = buildBenchmarkOutlierLookup(index, benchmark.id);
+ const roster = sortBenchmarkRunRoster(
+ buildBenchmarkRunRoster({
+ runIds: rosterRunIds,
+ pairs,
+ modes: benchmark.payload.modes,
+ }),
+ ).map((row) => {
+ const peerRunId = findMostSimilarPeerRunId(
+ row.runId,
+ rosterRunIds,
+ pairs,
+ );
+ return {
+ ...row,
+ peerRunId,
+ peerCompareHref: peerRunId
+ ? `/runs/compare?left=${row.runId}&right=${peerRunId}`
+ : null,
+ isOutlier: outlierLookup.has(row.runId),
+ };
+ });
+ const csv = benchmarkRunRosterToCsv(roster);
+ return new Response(csv, {
+ headers: {
+ "Content-Type": "text/csv; charset=utf-8",
+ "Content-Disposition": `attachment; filename="${id}-roster.csv"`,
+ "Cache-Control": "public, max-age=60",
+ },
+ });
+ }
+
+ const download = url.searchParams.get("download") === "1";
const headers = download
? {
"Content-Disposition": `attachment; filename="${id}.json"`,
diff --git a/apps/web/app/api/questions/hub/route.ts b/apps/web/app/api/questions/hub/route.ts
new file mode 100644
index 0000000..b56af3d
--- /dev/null
+++ b/apps/web/app/api/questions/hub/route.ts
@@ -0,0 +1,61 @@
+import {
+ filterBenchmarkArtifacts,
+ filterRunArtifacts,
+ loadBenchmarkArtifacts,
+ loadRunArtifacts,
+} from "../../../../lib/data";
+import { questionHubArtifactsToCsv } from "../../../../lib/listExport";
+
+export async function GET(request: Request) {
+ const url = new URL(request.url);
+ const question = (url.searchParams.get("question") ?? "").trim();
+ if (!question) {
+ return Response.json(
+ { error: "question query parameter is required" },
+ { status: 400 },
+ );
+ }
+
+ const format = url.searchParams.get("format") ?? "json";
+ const filters = {
+ q: url.searchParams.get("q") ?? undefined,
+ model: url.searchParams.get("model") ?? undefined,
+ preset: url.searchParams.get("preset") ?? undefined,
+ fast: url.searchParams.get("fast") ?? undefined,
+ from: url.searchParams.get("from") ?? undefined,
+ to: url.searchParams.get("to") ?? undefined,
+ };
+
+ const [allRuns, allBenchmarks] = await Promise.all([
+ loadRunArtifacts(),
+ loadBenchmarkArtifacts(),
+ ]);
+ const runs = filterRunArtifacts(allRuns, filters).filter(
+ (run) => run.question === question,
+ );
+ const benchmarks = filterBenchmarkArtifacts(allBenchmarks, filters).filter(
+ (benchmark) => benchmark.question === question,
+ );
+
+ if (format === "csv") {
+ const csv = questionHubArtifactsToCsv(question, runs, benchmarks);
+ const slug = question.slice(0, 48).replace(/[^\w.-]+/g, "_");
+ return new Response(csv, {
+ headers: {
+ "Content-Type": "text/csv; charset=utf-8",
+ "Content-Disposition": `attachment; filename="question-hub-${slug || "export"}.csv"`,
+ "Cache-Control": "public, max-age=60",
+ },
+ });
+ }
+
+ return Response.json({
+ question,
+ totals: {
+ runs: runs.length,
+ benchmarks: benchmarks.length,
+ },
+ runs,
+ benchmarks,
+ });
+}
diff --git a/apps/web/app/api/routes.test.ts b/apps/web/app/api/routes.test.ts
index 5287980..5e959c7 100644
--- a/apps/web/app/api/routes.test.ts
+++ b/apps/web/app/api/routes.test.ts
@@ -32,6 +32,7 @@ import { GET as getOutliers } from "./outliers/route";
import { GET as getCatalog } from "./catalog/route";
import { POST as postAnalysisRebuild } from "./analysis/rebuild/route";
import { GET as getQuestions } from "./questions/route";
+import { GET as getQuestionsHub } from "./questions/hub/route";
import { GET as getPairs } from "./pairs/route";
import { GET as getErrors } from "./errors/route";
@@ -2314,6 +2315,165 @@ describe("web api routes", () => {
expect(await csvResponse.text()).toContain("Shared topic");
});
+ it("exports question hub artifacts as CSV", async () => {
+ const dir = await makeTempDir();
+ process.env.RUNS_DIR = dir;
+ await writeFile(
+ join(dir, "run_a.json"),
+ JSON.stringify({
+ kind: "run",
+ id: "run_a",
+ question: "Hub topic",
+ metadata: {
+ createdAt: "2026-01-02T00:00:00.000Z",
+ model: "gpt-a",
+ pipelinePreset: "standard",
+ fastMode: false,
+ },
+ run: { id: "run_a", finalAnswer: "A", steps: [], metrics: {} },
+ }),
+ "utf-8",
+ );
+ await writeFile(
+ join(dir, "benchmark_a.json"),
+ JSON.stringify({
+ kind: "benchmark",
+ id: "benchmark_a",
+ question: "Hub topic",
+ metadata: {
+ createdAt: "2026-01-01T00:00:00.000Z",
+ model: "gpt-a",
+ pipelinePreset: "standard",
+ fastMode: false,
+ },
+ payload: {
+ runs: 2,
+ modeCount: 1,
+ modeSizes: [2],
+ divergenceEntropy: 0.2,
+ summary: { stability: { pairwiseMean: 0.9, pairs: [] } },
+ },
+ }),
+ "utf-8",
+ );
+
+ const response = await getQuestionsHub(
+ new Request(
+ "http://localhost/api/questions/hub?question=Hub%20topic&format=csv",
+ ),
+ );
+ expect(response.status).toBe(200);
+ const csv = await response.text();
+ expect(csv).toContain("Hub topic");
+ expect(csv).toContain("run_a");
+ expect(csv).toContain("benchmark_a");
+ });
+
+ it("filters runs API by pipeline errors", async () => {
+ const dir = await makeTempDir();
+ process.env.RUNS_DIR = dir;
+ await writeFile(
+ join(dir, "run_ok.json"),
+ JSON.stringify({
+ kind: "run",
+ id: "run_ok",
+ question: "Alpha",
+ metadata: {
+ createdAt: "2026-01-01T00:00:00.000Z",
+ model: "gpt-a",
+ pipelinePreset: "standard",
+ fastMode: false,
+ },
+ run: { id: "run_ok", finalAnswer: "A", steps: [], metrics: {} },
+ }),
+ "utf-8",
+ );
+ await writeFile(
+ join(dir, "run_fail.json"),
+ JSON.stringify({
+ kind: "run",
+ id: "run_fail",
+ question: "Beta",
+ metadata: {
+ createdAt: "2026-01-02T00:00:00.000Z",
+ model: "gpt-a",
+ pipelinePreset: "standard",
+ fastMode: false,
+ },
+ run: {
+ id: "run_fail",
+ finalAnswer: "B",
+ steps: [
+ {
+ agentName: "Solver",
+ role: "solver",
+ createdAt: "2026-01-02T00:00:00.000Z",
+ error: "timeout",
+ },
+ ],
+ metrics: {},
+ },
+ }),
+ "utf-8",
+ );
+
+ const response = await getRuns(
+ new Request("http://localhost/api/runs?errors=true"),
+ );
+ expect(response.status).toBe(200);
+ const json = (await response.json()) as {
+ filtered: number;
+ items: Array<{ id: string }>;
+ };
+ expect(json.filtered).toBe(1);
+ expect(json.items[0].id).toBe("run_fail");
+ });
+
+ it("exports benchmark roster CSV", async () => {
+ const dir = await makeTempDir();
+ process.env.RUNS_DIR = dir;
+ await writeFile(
+ join(dir, "benchmark_roster.json"),
+ JSON.stringify({
+ kind: "benchmark",
+ id: "benchmark_roster",
+ question: "Roster topic",
+ metadata: {
+ createdAt: "2026-01-01T00:00:00.000Z",
+ model: "gpt-a",
+ pipelinePreset: "standard",
+ fastMode: false,
+ },
+ payload: {
+ runs: 2,
+ modeCount: 1,
+ modeSizes: [2],
+ divergenceEntropy: 0.2,
+ runIds: ["run_a", "run_b"],
+ modes: [{ members: [0, 1] }],
+ summary: {
+ stability: {
+ pairwiseMean: 0.5,
+ pairs: [{ i: 0, j: 1, similarity: 0.5 }],
+ },
+ },
+ },
+ }),
+ "utf-8",
+ );
+
+ const response = await getBenchmarkById(
+ new Request(
+ "http://localhost/api/benchmarks/benchmark_roster?format=roster-csv",
+ ),
+ { params: Promise.resolve({ id: "benchmark_roster" }) },
+ );
+ expect(response.status).toBe(200);
+ const csv = await response.text();
+ expect(csv).toContain("run_a");
+ expect(csv).toContain("avgSimilarity");
+ });
+
it("returns filtered activity feed as JSON and CSV", async () => {
const dir = await makeTempDir();
process.env.RUNS_DIR = dir;
diff --git a/apps/web/app/api/runs/route.ts b/apps/web/app/api/runs/route.ts
index c5bc604..b01bf95 100644
--- a/apps/web/app/api/runs/route.ts
+++ b/apps/web/app/api/runs/route.ts
@@ -1,23 +1,43 @@
-import { filterRunArtifacts, loadRunArtifacts } from "../../../lib/data";
+import { filterRunArtifacts, loadAnalysisIndex, loadRunArtifacts } from "../../../lib/data";
import {
resolveRunSortOrder,
sortRunArtifacts,
} from "../../../lib/artifactSort";
import { runArtifactsToCsv } from "../../../lib/listExport";
+import { buildOutlierRunIdSet } from "../../../lib/outlierLookup";
+import {
+ applyRunListExtraFilters,
+ buildPipelineErrorRunIdSet,
+ parseRunListExtraFilters,
+} from "../../../lib/runListFilters";
import { parseListPagination } from "../_shared/pagination";
export async function GET(request: Request) {
const url = new URL(request.url);
const format = url.searchParams.get("format") ?? "json";
- const runs = await loadRunArtifacts();
- const filtered = filterRunArtifacts(runs, {
- q: url.searchParams.get("q") ?? undefined,
- model: url.searchParams.get("model") ?? undefined,
- preset: url.searchParams.get("preset") ?? undefined,
- fast: url.searchParams.get("fast") ?? undefined,
- from: url.searchParams.get("from") ?? undefined,
- to: url.searchParams.get("to") ?? undefined,
+ const [runs, index] = await Promise.all([
+ loadRunArtifacts(),
+ loadAnalysisIndex(),
+ ]);
+ const extraFilters = parseRunListExtraFilters({
+ outlier: url.searchParams.get("outlier") ?? undefined,
+ errors: url.searchParams.get("errors") ?? undefined,
});
+ const filtered = applyRunListExtraFilters(
+ filterRunArtifacts(runs, {
+ q: url.searchParams.get("q") ?? undefined,
+ model: url.searchParams.get("model") ?? undefined,
+ preset: url.searchParams.get("preset") ?? undefined,
+ fast: url.searchParams.get("fast") ?? undefined,
+ from: url.searchParams.get("from") ?? undefined,
+ to: url.searchParams.get("to") ?? undefined,
+ }),
+ extraFilters,
+ {
+ outlierRunIds: buildOutlierRunIdSet(index),
+ errorRunIds: buildPipelineErrorRunIdSet(runs),
+ },
+ );
const sort = resolveRunSortOrder(url.searchParams.get("sort") ?? undefined);
const { offset, limit, page } = parseListPagination(url.searchParams);
const sorted = sortRunArtifacts(filtered, sort);
diff --git a/apps/web/app/benchmarks/[id]/page.tsx b/apps/web/app/benchmarks/[id]/page.tsx
index 93b9c36..dbbbb1e 100644
--- a/apps/web/app/benchmarks/[id]/page.tsx
+++ b/apps/web/app/benchmarks/[id]/page.tsx
@@ -143,6 +143,13 @@ export default async function BenchmarkDetailPage({
href={`/api/benchmarks/${benchmark.id}?download=1`}
filename={`${benchmark.id}.json`}
/>
+
+ Export roster CSV
+
Pipeline timing
+
+ Runs with errors
+
diff --git a/apps/web/app/outliers/page.tsx b/apps/web/app/outliers/page.tsx
index 77ae2f3..16c0662 100644
--- a/apps/web/app/outliers/page.tsx
+++ b/apps/web/app/outliers/page.tsx
@@ -79,6 +79,12 @@ export default async function OutliersPage({
All benchmarks
+
+ Outlier runs list
+
diff --git a/apps/web/app/questions/view/page.tsx b/apps/web/app/questions/view/page.tsx
index 8c05f5d..b11b9d1 100644
--- a/apps/web/app/questions/view/page.tsx
+++ b/apps/web/app/questions/view/page.tsx
@@ -309,6 +309,23 @@ export default async function QuestionHubPage({
Compare benchmarks (picker)
) : null}
+
+ Export hub CSV
+
diff --git a/apps/web/app/runs/page.tsx b/apps/web/app/runs/page.tsx
index c083b2a..a9a828c 100644
--- a/apps/web/app/runs/page.tsx
+++ b/apps/web/app/runs/page.tsx
@@ -17,6 +17,12 @@ import {
} from "../../lib/data";
import { buildIndexRunLookup } from "../../lib/indexRunLookup";
import { buildOutlierRunIdSet } from "../../lib/outlierLookup";
+import {
+ applyRunListExtraFilters,
+ buildPipelineErrorRunIdSet,
+ hasActiveRunListExtraFilters,
+ parseRunListExtraFilters,
+} from "../../lib/runListFilters";
import {
resolveRunSortOrder,
sortRunArtifacts,
@@ -35,6 +41,8 @@ type RunsSearchParams = {
fast?: string;
from?: string;
to?: string;
+ outlier?: string;
+ errors?: string;
sort?: string;
page?: string;
pageSize?: string;
@@ -50,18 +58,24 @@ export default async function RunsPage({
loadBenchmarkArtifacts(),
loadAnalysisIndex(),
]);
+ const params = await searchParams;
const indexLookup = index ? buildIndexRunLookup(index) : null;
const outlierRunIds = buildOutlierRunIdSet(index);
+ const errorRunIds = buildPipelineErrorRunIdSet(runs);
+ const extraFilters = parseRunListExtraFilters(params);
const { models, presets } = collectArtifactFacets(runs, benchmarks);
- const params = await searchParams;
- const filtered = filterRunArtifacts(runs, {
- q: params.q,
- model: params.model,
- preset: params.preset,
- fast: params.fast,
- from: params.from,
- to: params.to,
- });
+ const filtered = applyRunListExtraFilters(
+ filterRunArtifacts(runs, {
+ q: params.q,
+ model: params.model,
+ preset: params.preset,
+ fast: params.fast,
+ from: params.from,
+ to: params.to,
+ }),
+ extraFilters,
+ { outlierRunIds, errorRunIds },
+ );
const sort = resolveRunSortOrder(params.sort);
const sorted = sortRunArtifacts(filtered, sort);
const paging = paginateItems(sorted, params, {
@@ -120,6 +134,24 @@ export default async function RunsPage({
className="input"
title="Created at or before"
/>
+
+
diff --git a/apps/web/lib/activityCompare.test.ts b/apps/web/lib/activityCompare.test.ts
new file mode 100644
index 0000000..faf1d0f
--- /dev/null
+++ b/apps/web/lib/activityCompare.test.ts
@@ -0,0 +1,83 @@
+import { describe, expect, it } from "vitest";
+import type { BenchmarkArtifact, RunArtifact } from "./data";
+import { buildActivityFeed } from "./activityFeed";
+import { attachActivityCompareLinks } from "./activityCompare";
+
+function sampleRun(
+ id: string,
+ question: string,
+ createdAt: string,
+): RunArtifact {
+ return {
+ kind: "run",
+ id,
+ question,
+ metadata: {
+ createdAt,
+ model: "gpt-test",
+ pipelinePreset: "standard",
+ fastMode: false,
+ },
+ run: {
+ id,
+ finalAnswer: "answer",
+ steps: [],
+ metrics: {},
+ },
+ };
+}
+
+function sampleBenchmark(
+ id: string,
+ question: string,
+ createdAt: string,
+): BenchmarkArtifact {
+ return {
+ kind: "benchmark",
+ id,
+ question,
+ metadata: {
+ createdAt,
+ model: "gpt-test",
+ pipelinePreset: "standard",
+ fastMode: false,
+ },
+ payload: {
+ runs: 3,
+ modeCount: 2,
+ modeSizes: [2, 1],
+ divergenceEntropy: 0.5,
+ },
+ };
+}
+
+describe("attachActivityCompareLinks", () => {
+ it("adds run compare links against the latest same-question peer", () => {
+ const runs = [
+ sampleRun("run_old", "Shared?", "2026-01-01T00:00:00.000Z"),
+ sampleRun("run_new", "Shared?", "2026-01-02T00:00:00.000Z"),
+ ];
+ const feed = buildActivityFeed(runs, [], { kind: "run" });
+ const linked = attachActivityCompareLinks(feed, runs, []);
+
+ expect(linked[0]?.compareHref).toBe(
+ "/runs/compare?left=run_new&right=run_old",
+ );
+ expect(linked[1]?.compareHref).toBe(
+ "/runs/compare?left=run_old&right=run_new",
+ );
+ });
+
+ it("adds benchmark compare links when another benchmark shares the question", () => {
+ const benchmarks = [
+ sampleBenchmark("bench_a", "Shared?", "2026-01-01T00:00:00.000Z"),
+ sampleBenchmark("bench_b", "Shared?", "2026-01-02T00:00:00.000Z"),
+ ];
+ const feed = buildActivityFeed([], benchmarks, { kind: "benchmark" });
+ const linked = attachActivityCompareLinks(feed, [], benchmarks);
+
+ expect(linked[0]?.compareHref).toBe(
+ "/benchmarks/compare?left=bench_b&right=bench_a",
+ );
+ });
+});
diff --git a/apps/web/lib/activityCompare.ts b/apps/web/lib/activityCompare.ts
new file mode 100644
index 0000000..81c8ac2
--- /dev/null
+++ b/apps/web/lib/activityCompare.ts
@@ -0,0 +1,68 @@
+import type { BenchmarkArtifact, RunArtifact } from "./data";
+import type { ActivityEntry } from "./activityFeed";
+
+function latestPeerId(
+ entries: Array<{ id: string; createdAt: string }>,
+ anchorId: string,
+): string | null {
+ const anchor = entries.find((entry) => entry.id === anchorId);
+ if (!anchor) return null;
+
+ const peer = entries
+ .filter((entry) => entry.id !== anchorId)
+ .sort((a, b) => b.createdAt.localeCompare(a.createdAt))[0];
+
+ return peer?.id ?? null;
+}
+
+export function attachActivityCompareLinks(
+ entries: ActivityEntry[],
+ runs: RunArtifact[],
+ benchmarks: BenchmarkArtifact[],
+): ActivityEntry[] {
+ const runsByQuestion = new Map();
+ for (const run of runs) {
+ const bucket = runsByQuestion.get(run.question) ?? [];
+ bucket.push(run);
+ runsByQuestion.set(run.question, bucket);
+ }
+
+ const benchmarksByQuestion = new Map();
+ for (const benchmark of benchmarks) {
+ const bucket = benchmarksByQuestion.get(benchmark.question) ?? [];
+ bucket.push(benchmark);
+ benchmarksByQuestion.set(benchmark.question, bucket);
+ }
+
+ return entries.map((entry) => {
+ if (entry.kind === "run") {
+ const peers = runsByQuestion.get(entry.question) ?? [];
+ const peerId = latestPeerId(
+ peers.map((run) => ({
+ id: run.id,
+ createdAt: run.metadata.createdAt,
+ })),
+ entry.id,
+ );
+ if (!peerId) return entry;
+ return {
+ ...entry,
+ compareHref: `/runs/compare?left=${entry.id}&right=${peerId}`,
+ };
+ }
+
+ const peers = benchmarksByQuestion.get(entry.question) ?? [];
+ const peerId = latestPeerId(
+ peers.map((benchmark) => ({
+ id: benchmark.id,
+ createdAt: benchmark.metadata.createdAt,
+ })),
+ entry.id,
+ );
+ if (!peerId) return entry;
+ return {
+ ...entry,
+ compareHref: `/benchmarks/compare?left=${entry.id}&right=${peerId}`,
+ };
+ });
+}
diff --git a/apps/web/lib/activityExport.ts b/apps/web/lib/activityExport.ts
index 4a29f65..b4e33df 100644
--- a/apps/web/lib/activityExport.ts
+++ b/apps/web/lib/activityExport.ts
@@ -18,6 +18,7 @@ export function activityEntriesToCsv(entries: ActivityEntry[]): string {
"fastMode",
"detail",
"href",
+ "compareHref",
];
const rows = entries.map((entry) =>
[
@@ -30,6 +31,7 @@ export function activityEntriesToCsv(entries: ActivityEntry[]): string {
entry.fastMode ? "true" : "false",
entry.detail,
entry.href,
+ entry.compareHref ?? "",
]
.map((value) => escapeCsv(String(value)))
.join(","),
diff --git a/apps/web/lib/activityFeed.ts b/apps/web/lib/activityFeed.ts
index 829f17f..cbd4207 100644
--- a/apps/web/lib/activityFeed.ts
+++ b/apps/web/lib/activityFeed.ts
@@ -14,6 +14,7 @@ export type ActivityEntry = {
fastMode: boolean;
href: string;
detail: string;
+ compareHref?: string;
};
export type ActivityFeedFilters = {
diff --git a/apps/web/lib/listExport.test.ts b/apps/web/lib/listExport.test.ts
index 0be7d6f..f7b37c5 100644
--- a/apps/web/lib/listExport.test.ts
+++ b/apps/web/lib/listExport.test.ts
@@ -4,6 +4,7 @@ import {
agentStatsToCsv,
agentTimingToCsv,
benchmarkArtifactsToCsv,
+ benchmarkRunRosterToCsv,
confidenceDriftToCsv,
counterfactualExplorerToCsv,
evidenceExplorerToCsv,
@@ -13,6 +14,7 @@ import {
presetLeaderboardToCsv,
qualityRunsToCsv,
questionGroupsToCsv,
+ questionHubArtifactsToCsv,
runArtifactsToCsv,
searchResultsToCsv,
} from "./listExport";
@@ -286,4 +288,32 @@ describe("listExport", () => {
expect(csv).toContain("missing evidence");
expect(csv).toContain("run_1");
});
+
+ it("exports benchmark roster rows", () => {
+ const csv = benchmarkRunRosterToCsv([
+ {
+ runIndex: 0,
+ runId: "run_a",
+ avgSimilarity: 0.42,
+ modeIndex: 1,
+ peerRunId: "run_b",
+ peerCompareHref: "/runs/compare?left=run_a&right=run_b",
+ isOutlier: true,
+ },
+ ]);
+ expect(csv).toContain("run_a");
+ expect(csv).toContain("isOutlier");
+ expect(csv).toContain("true");
+ });
+
+ it("exports question hub artifacts", () => {
+ const csv = questionHubArtifactsToCsv(
+ "Shared topic?",
+ [sampleRun],
+ [sampleBenchmark],
+ );
+ expect(csv).toContain("Shared topic?");
+ expect(csv).toContain("run_1");
+ expect(csv).toContain("bench_1");
+ });
});
diff --git a/apps/web/lib/listExport.ts b/apps/web/lib/listExport.ts
index 82387ad..2723267 100644
--- a/apps/web/lib/listExport.ts
+++ b/apps/web/lib/listExport.ts
@@ -1,3 +1,4 @@
+import type { BenchmarkRunRosterRow } from "./benchmarkRunRoster";
import type { AgentStatRow } from "./agentStats";
import type { CatalogStats } from "./catalogStats";
import type { ConfidenceDriftRow } from "./confidenceDrift";
@@ -553,6 +554,87 @@ export function evidenceExplorerToCsv(
return sections.join("\n");
}
+export function benchmarkRunRosterToCsv(
+ rows: Array<
+ BenchmarkRunRosterRow & {
+ peerRunId?: string | null;
+ peerCompareHref?: string | null;
+ isOutlier?: boolean;
+ }
+ >,
+): string {
+ const header = [
+ "runIndex",
+ "runId",
+ "modeIndex",
+ "avgSimilarity",
+ "peerRunId",
+ "peerCompareHref",
+ "isOutlier",
+ ];
+ const csvRows = rows.map((entry) =>
+ row([
+ entry.runIndex,
+ entry.runId,
+ entry.modeIndex ?? "",
+ entry.avgSimilarity ?? "",
+ entry.peerRunId ?? "",
+ entry.peerCompareHref ?? "",
+ entry.isOutlier ? "true" : "false",
+ ]),
+ );
+ return [header.join(","), ...csvRows].join("\n");
+}
+
+export function questionHubArtifactsToCsv(
+ question: string,
+ runs: RunArtifact[],
+ benchmarks: BenchmarkArtifact[],
+): string {
+ const sections: string[] = [];
+ sections.push(`question,${escapeCsv(question)}`);
+ sections.push("");
+ sections.push(
+ "section,id,createdAt,model,preset,fastMode,runs,modeCount,entropy,finalAnswerPreview",
+ );
+
+ for (const run of runs) {
+ sections.push(
+ row([
+ "run",
+ run.id,
+ run.metadata.createdAt,
+ run.metadata.model,
+ run.metadata.pipelinePreset,
+ run.metadata.fastMode,
+ "",
+ "",
+ "",
+ run.run.finalAnswer.slice(0, 200),
+ ]),
+ );
+ }
+
+ for (const benchmark of benchmarks) {
+ sections.push(
+ row([
+ "benchmark",
+ benchmark.id,
+ benchmark.metadata.createdAt,
+ benchmark.metadata.model,
+ benchmark.metadata.pipelinePreset,
+ benchmark.metadata.fastMode,
+ benchmark.payload.runs,
+ benchmark.payload.modeCount,
+ benchmark.payload.divergenceEntropy,
+ "",
+ ]),
+ );
+ }
+
+ return sections.join("\n");
+}
+
export function catalogStatsToCsv(stats: CatalogStats): string {
const sections: string[] = [];
diff --git a/apps/web/lib/runListFilters.test.ts b/apps/web/lib/runListFilters.test.ts
new file mode 100644
index 0000000..4489cbf
--- /dev/null
+++ b/apps/web/lib/runListFilters.test.ts
@@ -0,0 +1,104 @@
+import { describe, expect, it } from "vitest";
+import type { RunArtifact } from "./data";
+import {
+ applyRunListExtraFilters,
+ buildPipelineErrorRunIdSet,
+ hasActiveRunListExtraFilters,
+ parseRunListExtraFilters,
+} from "./runListFilters";
+
+function sampleRun(
+ id: string,
+ overrides: Partial = {},
+): RunArtifact {
+ return {
+ kind: "run",
+ id,
+ question: "Sample?",
+ metadata: {
+ createdAt: "2026-01-01T00:00:00.000Z",
+ model: "gpt-test",
+ pipelinePreset: "standard",
+ fastMode: false,
+ },
+ run: {
+ id,
+ finalAnswer: "answer",
+ steps: [],
+ metrics: {},
+ },
+ ...overrides,
+ };
+}
+
+describe("runListFilters", () => {
+ it("parses outlier and error flags", () => {
+ expect(parseRunListExtraFilters({})).toEqual({
+ outlierOnly: false,
+ errorsOnly: false,
+ });
+ expect(
+ parseRunListExtraFilters({ outlier: "true", errors: "true" }),
+ ).toEqual({
+ outlierOnly: true,
+ errorsOnly: true,
+ });
+ });
+
+ it("detects active extra filters", () => {
+ expect(
+ hasActiveRunListExtraFilters({
+ outlierOnly: false,
+ errorsOnly: false,
+ }),
+ ).toBe(false);
+ expect(
+ hasActiveRunListExtraFilters({
+ outlierOnly: true,
+ errorsOnly: false,
+ }),
+ ).toBe(true);
+ });
+
+ it("builds a set of run ids with pipeline errors", () => {
+ const runs = [
+ sampleRun("run_ok"),
+ sampleRun("run_fail", {
+ run: {
+ id: "run_fail",
+ finalAnswer: "answer",
+ steps: [
+ {
+ id: "step_1",
+ agentName: "Solver",
+ role: "solver",
+ createdAt: "2026-01-01T00:00:00.000Z",
+ error: "timeout",
+ },
+ ],
+ metrics: {},
+ },
+ }),
+ ];
+
+ expect([...buildPipelineErrorRunIdSet(runs)]).toEqual(["run_fail"]);
+ });
+
+ it("filters runs by outlier and error flags", () => {
+ const runs = [
+ sampleRun("run_a"),
+ sampleRun("run_b"),
+ sampleRun("run_c"),
+ ];
+ const filtered = applyRunListExtraFilters(
+ runs,
+ { outlierOnly: true, errorsOnly: true },
+ {
+ outlierRunIds: new Set(["run_a", "run_b"]),
+ errorRunIds: new Set(["run_b", "run_c"]),
+ },
+ );
+
+ expect(filtered.map((run) => run.id)).toEqual(["run_b"]);
+ });
+});
diff --git a/apps/web/lib/runListFilters.ts b/apps/web/lib/runListFilters.ts
new file mode 100644
index 0000000..142eee5
--- /dev/null
+++ b/apps/web/lib/runListFilters.ts
@@ -0,0 +1,53 @@
+import type { RunArtifact } from "./data";
+import { buildPipelineErrorRows } from "./pipelineErrors";
+
+export type RunListExtraFilterParams = {
+ outlier?: string;
+ errors?: string;
+};
+
+export type RunListExtraFilterFlags = {
+ outlierOnly: boolean;
+ errorsOnly: boolean;
+};
+
+export function parseRunListExtraFilters(
+ params: RunListExtraFilterParams,
+): RunListExtraFilterFlags {
+ return {
+ outlierOnly: params.outlier === "true",
+ errorsOnly: params.errors === "true",
+ };
+}
+
+export function hasActiveRunListExtraFilters(
+ flags: RunListExtraFilterFlags,
+): boolean {
+ return flags.outlierOnly || flags.errorsOnly;
+}
+
+export function buildPipelineErrorRunIdSet(runs: RunArtifact[]): Set {
+ const ids = new Set();
+ for (const row of buildPipelineErrorRows(runs)) {
+ ids.add(row.runId);
+ }
+ return ids;
+}
+
+export function applyRunListExtraFilters(
+ runs: RunArtifact[],
+ flags: RunListExtraFilterFlags,
+ context: {
+ outlierRunIds: Set;
+ errorRunIds: Set;
+ },
+): RunArtifact[] {
+ let result = runs;
+ if (flags.outlierOnly) {
+ result = result.filter((run) => context.outlierRunIds.has(run.id));
+ }
+ if (flags.errorsOnly) {
+ result = result.filter((run) => context.errorRunIds.has(run.id));
+ }
+ return result;
+}