feat(eval): read-only batch-evaluation TUI (list + get)

src/components/BatchEvaluationPicker.tsx

@@ -0,0 +1,88 @@
+import type { BatchEvaluationSummary } from "@aws-sdk/client-bedrock-agentcore";
+import { useNavigate } from "react-router";
+import type { ScreenProps } from "../handlers/types";
+import { coreOptsFromCtx } from "../handlers/utils";
+import { formatTimestamp } from "./formatTimestamp";
+import { PaginatedTablePicker } from "./PaginatedTablePicker";
+import type { DataTableColumn } from "./ui/data-table";
+
+// BatchEvaluationRow is the flat, display-ready shape the table renders. It also
+// satisfies DataTable's `T extends Record<string, unknown>` constraint, which the
+// SDK's BatchEvaluationSummary interface does not. The list API returns summary
+// fields only; per-session results come from GetBatchEvaluation.
+interface BatchEvaluationRow extends Record<string, unknown> {
+  batchEvaluationId: string;
+  name: string;
+  status: string;
+  updatedAt: string;
+}
+
+export const batchEvaluationColumns = [
+  { key: "name", header: "name", flex: true },
+  { key: "status", header: "status", width: 22 },
+  {
+    key: "updatedAt",
+    header: "updated UTC",
+    width: 16,
+    render: formatTimestamp,
+  },
+] satisfies DataTableColumn<BatchEvaluationRow>[];
+
+function toRow(summary: BatchEvaluationSummary): BatchEvaluationRow {
+  const id = summary.batchEvaluationId ?? "";
+  return {
+    batchEvaluationId: id,
+    name: summary.batchEvaluationName ?? id,
+    status: summary.status ?? "-",
+    updatedAt: summary.updatedAt?.toISOString() ?? "-",
+  };
+}
+
+export interface BatchEvaluationPickerProps extends ScreenProps {
+  breadcrumb: string[];
+  description?: string;
+  onSelect: (batchEvaluationId: string) => void;
+  onEscape?: () => void;
+}
+
+/**
+ * Fetches the caller's batch evaluations and renders them as a navigable table.
+ * The shared body of every "pick a batch evaluation" screen. Esc returns to the
+ * parent menu derived from the breadcrumb unless a host supplies its own onEscape.
+ */
+export function BatchEvaluationPicker({
+  ctx,
+  core,
+  breadcrumb,
+  description,
+  onSelect,
+  onEscape,
+}: BatchEvaluationPickerProps) {
+  const opts = coreOptsFromCtx(ctx);
+  const navigate = useNavigate();
+  const goBack = onEscape ?? (() => navigate("/" + breadcrumb.slice(0, -1).join("/")));
+
+  return (
+    <PaginatedTablePicker
+      breadcrumb={breadcrumb}
+      description={description}
+      queryKey={["batch-evaluations", opts.region]}
+      loadPage={async (token, pageSize) => {
+        const response = await core.eval.listBatchEvaluations(token, pageSize, opts);
+        return {
+          items: response.batchEvaluations ?? [],
+          nextToken: response.nextToken,
+        };
+      }}
+      toRow={toRow}
+      columns={batchEvaluationColumns}
+      getValue={(row) => row.batchEvaluationId}
+      onSelect={onSelect}
+      onBack={goBack}
+      loadingMessage="Loading batch evaluations…"
+      errorMessage={(error) => `Error: ${error.message}`}
+      emptyMessage="No batch evaluations found in this Region."
+      emptyPageMessage="No batch evaluations on this page."
+    />
+  );
+}

src/components/JsonDetail.tsx

@@ -17,6 +17,10 @@ export interface JsonDetailProps {
   // loadingLabel names what's loading (e.g. "Loading endpoint…").
   loadingLabel: string;
   onRetry?: () => void;
+  // warning, when set, renders a persistent advisory above the JSON — the TUI's
+  // equivalent of the CLI's stderr warning (e.g. a CloudWatch results read that
+  // failed while the job metadata is still intact).
+  warning?: string;
 }
 
 // JsonDetail is the shared "show me the raw resource" screen body: a scrollable
@@ -30,6 +34,7 @@ export function JsonDetail({
   data,
   loadingLabel,
   onRetry,
+  warning,
 }: JsonDetailProps) {
   const navigate = useNavigate();
   const scrollRef = useRef<ScrollViewRef>(null);
@@ -67,6 +72,7 @@ export function JsonDetail({
         <Text color="red">Error: {error.message}</Text>
       ) : (
         <ScrollView ref={scrollRef}>
+          {warning ? <Text color="yellow">Warning: {warning}</Text> : null}
           <CodeBlock
             language="json"
             showLineNumbers={false}

src/components/Root.tsx

@@ -45,6 +45,9 @@ import {
   OnlineEvalGetScreen,
   OnlineEvalGetJsonScreen,
 } from "../handlers/eval/online-eval/get/screen.tsx";
+import { BatchEvaluationScreen } from "../handlers/eval/batch-evaluation/screen.tsx";
+import { BatchEvaluationListScreen } from "../handlers/eval/batch-evaluation/list/screen.tsx";
+import { BatchEvaluationGetJsonScreen } from "../handlers/eval/batch-evaluation/get/screen.tsx";
 import { MemoryEventScreen } from "../handlers/memory/event/screen.tsx";
 import { MemoryEventGetScreen } from "../handlers/memory/event/get/screen.tsx";
 import { MemoryEventListScreen } from "../handlers/memory/event/list/screen.tsx";
@@ -359,6 +362,24 @@ export function Root({ path, ctx, core, queryClient }: RootProps) {
             path="agentcore/eval/online-eval/get/:configId/json"
             element={<OnlineEvalGetJsonScreen ctx={ctx} core={core} />}
           />
+          <Route
+            path="agentcore/eval/batch-evaluation"
+            element={<BatchEvaluationScreen ctx={ctx} core={core} />}
+          />
+          <Route
+            path="agentcore/eval/batch-evaluation/list"
+            element={<BatchEvaluationListScreen ctx={ctx} core={core} />}
+          />
+          {/* Bare `get` (no id) has nothing to show — send the user to the list. */}
+          <Route
+            path="agentcore/eval/batch-evaluation/get"
+            element={<Navigate to="/agentcore/eval/batch-evaluation/list" replace />}
+          />
+          {/* get is raw JSON only — no metadata hub, so :id is the JSON view. */}
+          <Route
+            path="agentcore/eval/batch-evaluation/get/:batchEvaluationId"
+            element={<BatchEvaluationGetJsonScreen ctx={ctx} core={core} />}
+          />
           <Route
             path="agentcore/memory/event"
             element={<MemoryEventScreen ctx={ctx} core={core} />}

src/handlers/eval/batch-evaluation/batch-evaluation.screen.test.tsx

@@ -0,0 +1,200 @@
+import { afterEach, describe, expect, test } from "bun:test";
+import type {
+  BatchEvaluationSummary,
+  GetBatchEvaluationResponse,
+} from "@aws-sdk/client-bedrock-agentcore";
+import {
+  cleanupScreens,
+  renderScreen,
+  TestCoreClient,
+  waitFor,
+  waitForText,
+} from "../../../testing";
+
+afterEach(cleanupScreens);
+
+const evalEndpointUrl = "https://eval.test";
+
+function summary(overrides: Partial<BatchEvaluationSummary> = {}): BatchEvaluationSummary {
+  return {
+    batchEvaluationArn: "arn:aws:bedrock-agentcore:us-east-1:123456789012:batch-evaluate/be-1",
+    batchEvaluationId: "be-1",
+    batchEvaluationName: "nightly_regression",
+    status: "COMPLETED",
+    createdAt: new Date("2026-07-19T01:02:03.000Z"),
+    updatedAt: new Date("2026-07-20T12:34:56.000Z"),
+    ...overrides,
+  };
+}
+
+function getResponse(
+  overrides: Partial<GetBatchEvaluationResponse> = {},
+): GetBatchEvaluationResponse {
+  return {
+    batchEvaluationArn: "arn:aws:bedrock-agentcore:us-east-1:123456789012:batch-evaluate/be-1",
+    batchEvaluationId: "be-1",
+    batchEvaluationName: "nightly_regression",
+    status: "COMPLETED",
+    createdAt: new Date("2026-07-19T01:02:03.000Z"),
+    updatedAt: new Date("2026-07-20T12:34:56.000Z"),
+    evaluators: [{ evaluatorId: "Builtin.Correctness" }],
+    outputConfig: {
+      cloudWatchConfig: {
+        logGroupName: "/aws/bedrock-agentcore/evaluations/batch-evaluations/results/default",
+        logStreamName: "run-be-1",
+      },
+    },
+    evaluationResults: {
+      numberOfSessionsCompleted: 3,
+      totalNumberOfSessions: 3,
+      numberOfSessionsFailed: 0,
+      evaluatorSummaries: [
+        {
+          evaluatorId: "Builtin.Correctness",
+          statistics: { averageScore: 1 },
+          totalEvaluated: 3,
+          totalFailed: 0,
+        },
+      ],
+    },
+    ...overrides,
+  };
+}
+
+function coreWithBatchEvals(items: BatchEvaluationSummary[]): TestCoreClient {
+  const core = new TestCoreClient();
+  core.eval.setBatchEvalListResponse({ batchEvaluations: items });
+  return core;
+}
+
+describe("batch-evaluation menu", () => {
+  test("offers get and list", async () => {
+    const screen = renderScreen("/agentcore/eval/batch-evaluation");
+
+    await waitForText(screen.lastFrame, "list batch evaluations");
+    const frame = screen.lastFrame()!;
+    expect(frame).toContain("list");
+    expect(frame).toContain("get");
+  });
+});
+
+describe("batch-evaluation picker", () => {
+  test("renders name, status, and updated time", async () => {
+    const core = coreWithBatchEvals([
+      summary({
+        batchEvaluationName: "staging_eval",
+        status: "FAILED",
+        updatedAt: new Date("2026-07-21T02:03:04.000Z"),
+      }),
+    ]);
+    const screen = renderScreen("/agentcore/eval/batch-evaluation/list", { core });
+
+    await waitForText(screen.lastFrame, "staging_eval");
+    const frame = screen.lastFrame()!;
+    expect(frame).toContain("FAILED");
+    expect(frame).toContain("2026-07-21 02:03");
+  });
+
+  test("calls listBatchEvaluations with exact Core options", async () => {
+    const core = coreWithBatchEvals([summary()]);
+    renderScreen("/agentcore/eval/batch-evaluation/list", { core, endpointUrl: evalEndpointUrl });
+
+    await waitFor(() => core.eval.calls.some((call) => call.method === "listBatchEvaluations"));
+    expect(core.eval.calls.filter((call) => call.method === "listBatchEvaluations")).toEqual([
+      {
+        method: "listBatchEvaluations",
+        args: [
+          undefined,
+          expect.any(Number),
+          { region: "us-east-1", endpointUrl: evalEndpointUrl },
+        ],
+      },
+    ]);
+  });
+
+  test("bare get redirects to the picker", async () => {
+    const core = coreWithBatchEvals([
+      summary({ batchEvaluationId: "redirected-be", batchEvaluationName: "redirected_eval" }),
+    ]);
+    const screen = renderScreen("/agentcore/eval/batch-evaluation/get", { core });
+
+    await waitForText(screen.lastFrame, "redirected_eval");
+    expect(core.eval.calls[0]?.method).toBe("listBatchEvaluations");
+  });
+
+  test("selection opens the matching batch evaluation JSON", async () => {
+    const core = coreWithBatchEvals([summary({ batchEvaluationId: "be-1" })]);
+    core.eval.setBatchEvalGetResponse(getResponse({ batchEvaluationId: "be-1" }));
+    const screen = renderScreen("/agentcore/eval/batch-evaluation/list", { core });
+
+    await waitForText(screen.lastFrame, "nightly_regression");
+    await screen.press("return");
+    await waitForText(screen.lastFrame, "agentcore → eval → batch-evaluation → get → be-1");
+    await waitFor(() =>
+      core.eval.calls.some(
+        (call) => call.method === "getBatchEvaluation" && call.args[0] === "be-1",
+      ),
+    );
+  });
+
+  test("shows the empty state", async () => {
+    const empty = renderScreen("/agentcore/eval/batch-evaluation/list");
+    await waitForText(empty.lastFrame, "No batch evaluations found in this Region.");
+  });
+});
+
+describe("batch-evaluation detail (raw JSON)", () => {
+  test("renders the full response, including merged results", async () => {
+    const core = new TestCoreClient();
+    core.eval.setBatchEvalGetResponse(getResponse());
+    core.eval.setBatchEvalResults([
+      { evaluatorId: "Builtin.Correctness", level: "Trace", sessionId: "s1", score: 1 },
+    ]);
+    const screen = renderScreen("/agentcore/eval/batch-evaluation/get/be-1", {
+      core,
+      endpointUrl: evalEndpointUrl,
+    });
+
+    await waitForText(screen.lastFrame, "nightly_regression");
+    const frame = screen.lastFrame()!;
+    expect(frame).toContain('"status"');
+    expect(frame).toContain("COMPLETED");
+    expect(frame).toContain('"evaluationResults"');
+    expect(frame).toContain('"results"');
+    // Screen requests results by default (no --disable-cw-results in the TUI): it
+    // calls getBatchEvaluation(id, opts) with no includeResults override, so the
+    // options object defaults to {} (includeResults defaults to true in Core).
+    expect(core.eval.calls.find((call) => call.method === "getBatchEvaluation")).toEqual({
+      method: "getBatchEvaluation",
+      args: ["be-1", { region: "us-east-1", endpointUrl: evalEndpointUrl }, {}],
+    });
+  });
+
+  test("a CloudWatch results failure still renders the metadata", async () => {
+    const core = new TestCoreClient();
+    core.eval.setBatchEvalGetResponse(getResponse());
+    core.eval.setBatchEvalResultsError(new Error("AccessDenied"));
+    const screen = renderScreen("/agentcore/eval/batch-evaluation/get/be-1", { core });
+
+    await waitForText(screen.lastFrame, "nightly_regression");
+    const frame = screen.lastFrame()!;
+    expect(frame).toContain("COMPLETED"); // status intact
+    expect(frame).not.toContain('"results"'); // results omitted, screen didn't crash
+    expect(frame).toContain("could not retrieve CloudWatch results"); // warned, not silent
+    expect(frame).toContain("AccessDenied");
+  });
+
+  test("retries a failed detail query", async () => {
+    const core = new TestCoreClient();
+    core.eval.setError(new Error("job unavailable"));
+    const screen = renderScreen("/agentcore/eval/batch-evaluation/get/be-1", { core });
+
+    await waitForText(screen.lastFrame, "job unavailable");
+    expect(screen.lastFrame()).toContain("[r] retry");
+
+    core.eval.setError(undefined);
+    core.eval.setBatchEvalGetResponse(getResponse());
+    await screen.write("r");
+    await waitForText(screen.lastFrame, "nightly_regression");
+  });
+});

src/handlers/eval/batch-evaluation/get/index.tsx

@@ -58,3 +58,5 @@ function warnCloudWatchFailure(
       `Job status is unaffected.${link}`,
   );
 }
+
+export { BatchEvaluationGetJsonScreen } from "./screen.tsx";