Fix flaky knowledge and quota e2e assertions

packages/commands/tests/e2e/knowledge/journeys/j1-cold-start.e2e.test.ts

@@ -159,7 +159,10 @@ describe.skipIf(!isKbAdminE2EReady())("journey J1: 冷启动首答 (live, 自清
         ],
       );
       expect(chatBadVersionRun.exitCode, "未发布版本号应被服务端拒绝").not.toBe(0);
-      expect(chatBadVersionRun.stderr).toMatch(/HTTP 400/);
+      // JSON 错误通道写 HTTP_STATUS/400 或 http_status:400，不一定有 "HTTP 400" 字面量
+      expect(chatBadVersionRun.stderr).toMatch(
+        /HTTP_STATUS\/400|"http_status"\s*:\s*400|ParamField\.Error/i,
+      );
     } finally {
       for (const agentId of serviceAgentIds) {
         const deleteRun = await reporter.runStep(

packages/commands/tests/e2e/knowledge/journeys/journey-helpers.ts

@@ -325,7 +325,6 @@ export async function createKbWithDocs(
   const createRun = await reporter.runStep("kb create --wait", routes, [
     "knowledge",
     "create",
-    "--yes",
     "--name",
     kbName,
     "--description",

packages/commands/tests/e2e/knowledge/knowledge-chunk-category-file.e2e.test.ts

@@ -1126,7 +1126,6 @@ describe.skipIf(!isKbAdminE2EReady())(
         const kbCreateRun = await runCommandE2e(KNOWLEDGE_KB_DELETE_ROUTES, [
           "knowledge",
           "create",
-          "--yes",
           "--name",
           `e2e-cate-kb-${Date.now() % 100000000}`.slice(0, 20),
           "--description",
@@ -1247,7 +1246,6 @@ describe.skipIf(!isKbAdminE2EReady())(
       const createRun = await runCommandE2e(KNOWLEDGE_KB_DELETE_ROUTES, [
         "knowledge",
         "create",
-        "--yes",
         "--name",
         `e2e-ck-${Date.now() % 100000000}`,
         "--description",
@@ -1803,7 +1801,6 @@ describe.skipIf(!isKbAdminE2EReady())("e2e: chunk/category/file 参数补全 (li
       const kbCreateRun = await runCommandE2e(KNOWLEDGE_KB_DELETE_ROUTES, [
         "knowledge",
         "create",
-        "--yes",
         "--name",
         `e2e-pk-${Date.now() % 100000000}`.slice(0, 20),
         "--description",
@@ -1914,11 +1911,8 @@ describe.skipIf(!isKbAdminE2EReady())("e2e: chunk/category/file 参数补全 (li
         expect(chunkListEmptyPageRun.exitCode, chunkListEmptyPageRun.stderr).toBe(0);
         expect(chunkListEmptyPageRun.stdout).toMatch(/No chunks found\./);
 
-        // ── contract: updating a nonexistent chunk id — the server currently
-        // reports success (silent upsert-like semantics, verified live). The CLI
-        // passes the server result through without an existence pre-check; this
-        // assertion pins the known server behavior so a future server-side change
-        // (rejecting unknown ids) surfaces here immediately ──
+        // ── contract: updating a nonexistent chunk id — server rejects invalid
+        // chunkId (Index.InvalidParameter). CLI 不做存在性预检，原样透传服务端错误 ──
         const chunkUpdateMissingRun = await runCommandE2e(KNOWLEDGE_CHUNK_CATEGORY_FILE_ROUTES, [
           "knowledge",
           "chunk",
@@ -1934,8 +1928,8 @@ describe.skipIf(!isKbAdminE2EReady())("e2e: chunk/category/file 参数补全 (li
           "--workspace-id",
           workspaceId,
         ]);
-        expect(chunkUpdateMissingRun.exitCode, chunkUpdateMissingRun.stderr).toBe(0);
-        expect(chunkUpdateMissingRun.stdout).toMatch(/updated: chunk_nonexistent_e2e_0000/);
+        expect(chunkUpdateMissingRun.exitCode).not.toBe(0);
+        expect(chunkUpdateMissingRun.stderr).toMatch(/chunkId|InvalidParameter/i);
 
         // ── P4: chunk update --content-file ──
         if (addedChunkId) {

packages/commands/tests/e2e/knowledge/knowledge-doc-status.e2e.test.ts

@@ -150,7 +150,6 @@ describe.skipIf(!isKbAdminE2EReady())("e2e: knowledge doc status (live, 自清
       const createRun = await runCommandE2e(KNOWLEDGE_DOC_STATUS_ROUTES, [
         "knowledge",
         "create",
-        "--yes",
         "--name",
         `e2e-st-${Date.now() % 100000000}`,
         "--description",

packages/commands/tests/e2e/knowledge/knowledge-kb-delete.e2e.test.ts

@@ -111,7 +111,6 @@ describe.skipIf(!isKbAdminE2EReady())("e2e: knowledge kb 写链路 (live, 自清
       const createRun = await runCommandE2e(KNOWLEDGE_KB_DELETE_ROUTES, [
         "knowledge",
         "create",
-        "--yes",
         "--name",
         kbName,
         "--description",

packages/commands/tests/e2e/memory/journeys/helpers.ts

@@ -199,35 +199,53 @@ export class MemoryJourney {
     return result.events!;
   }
 
-  async add(text: string, options: { direct?: boolean; schemaId?: string } = {}): Promise<string> {
+  async add(
+    text: string,
+    options: { direct?: boolean; schemaId?: string; requireExtraction?: boolean } = {},
+  ): Promise<string> {
     const payload = options.direct
       ? ["--content", text]
       : ["--messages", JSON.stringify([{ role: "user", content: text }])];
-    this.uncertainSubmission = true;
-    this.flush();
-    const receipt = await this.command<MemoryAddBody>("add", [
-      "add",
-      "--user-id",
-      this.userId,
-      ...payload,
-      ...(options.schemaId ? ["--profile-schema", options.schemaId] : []),
-      "--wait",
-      "0",
-    ]);
-    expect(receipt.event_id).toBeTruthy();
-    this.events.add(receipt.event_id!);
-    this.pending.add(receipt.event_id!);
-    this.uncertainSubmission = false;
-    this.flush();
-    const events = await this.settle(receipt.event_id!);
-    expect(
-      events.some((event) => event.status === "FAILED"),
-      JSON.stringify(events),
-    ).toBe(false);
-    expect(events.some((event) => ["SUCCEEDED", "SUCCESS"].includes(event.status ?? ""))).toBe(
-      true,
+    // Server-side observation extraction is nondeterministic: a terminal
+    // SUCCEEDED event may still carry an empty result. An empty terminal
+    // result never materializes later, so retry instead of waiting it out.
+    const maxAttempts = options.requireExtraction ? 3 : 1;
+    let eventId = "";
+    for (let attempt = 1; attempt <= maxAttempts; attempt += 1) {
+      this.uncertainSubmission = true;
+      this.flush();
+      const receipt = await this.command<MemoryAddBody>(
+        attempt === 1 ? "add" : `add retry ${attempt - 1}`,
+        [
+          "add",
+          "--user-id",
+          this.userId,
+          ...payload,
+          ...(options.schemaId ? ["--profile-schema", options.schemaId] : []),
+          "--wait",
+          "0",
+        ],
+      );
+      expect(receipt.event_id).toBeTruthy();
+      this.events.add(receipt.event_id!);
+      this.pending.add(receipt.event_id!);
+      this.uncertainSubmission = false;
+      this.flush();
+      eventId = receipt.event_id!;
+      const events = await this.settle(eventId);
+      expect(
+        events.some((event) => event.status === "FAILED"),
+        JSON.stringify(events),
+      ).toBe(false);
+      expect(events.some((event) => ["SUCCEEDED", "SUCCESS"].includes(event.status ?? ""))).toBe(
+        true,
+      );
+      if (!options.requireExtraction) return eventId;
+      if (events.some((event) => (event.result ?? []).length > 0)) return eventId;
+    }
+    throw new Error(
+      `add extracted nothing after ${maxAttempts} attempts; evidence ${this.outputDir}`,
     );
-    return receipt.event_id!;
   }
 
   async remove(nodeId: string, allowMissing = false) {

packages/commands/tests/e2e/memory/journeys/main.live.test.ts

@@ -47,14 +47,15 @@ describe.skipIf(!liveEnabled())("Memory J1–J5 (real cloud, isolated CLI subpro
     "J3 a later correction becomes a searchable current preference",
     async () =>
       journey("J3", async (fixture) => {
-        await fixture.add("请记住，我当前最喜欢的运动是徒步。");
+        await fixture.add("请记住，我当前最喜欢的运动是徒步。", { requireExtraction: true });
         await eventually(
           () => fixture.search("当前最喜欢的运动是什么？"),
           (nodes) => nodes.some((node) => /徒步/.test(node.content)),
           "initial preference",
         );
         await fixture.add(
           "更新我的长期偏好：我以前最喜欢徒步，但现在已经改成游泳。游泳是我当前最喜欢的运动，徒步只是过去的偏好。",
+          { requireExtraction: true },
         );
         await eventually(
           () => fixture.list(),

packages/commands/tests/e2e/memory/live-helpers.ts

@@ -24,6 +24,11 @@ export function assertMemoryServiceRejection(
   );
 }
 
+/** True only for the flaky server-side project-existence rejection (see add). */
+function isTransientProjectRejection(error: unknown): boolean {
+  return error instanceof Error && /project_ids got a project does not exist/i.test(error.message);
+}
+
 /** Each journey owns only a fresh user; cleanup never touches existing user data. */
 export function memoryJourney(
   libraryId: string | undefined = process.env.BAILIAN_E2E_MEMORY_LIBRARY_ID,
@@ -71,7 +76,17 @@ export function memoryJourney(
     for (const [index, argument] of args.entries()) {
       if (argument === "--project-id") projects.add(args[index + 1]);
     }
-    return invoke(["add", "--user-id", userId, ...args]);
+    // Project-existence validation is transiently inconsistent: a valid
+    // project can be reported as missing, then accepted minutes later.
+    // Only that exact rejection is retryable; anything else fails at once.
+    for (let attempt = 1; ; attempt += 1) {
+      try {
+        return await invoke(["add", "--user-id", userId, ...args]);
+      } catch (error) {
+        if (attempt >= 3 || !isTransientProjectRejection(error)) throw error;
+        await delay(attempt * 5_000);
+      }
+    }
   }
   async function search(args: string[]): Promise<Node[]> {
     const result = await invoke<MemoryNodeListBody>(["search", "--user-id", userId, ...args]);

packages/commands/tests/e2e/quota.e2e.test.ts

@@ -311,19 +311,35 @@ describe.skipIf(!isDashScopeE2EReady())("e2e: quota（DashScope）", () => {
   });
 
   test("quota list --model 精确查询返回模型限流", async () => {
+    // 账号限流目录不一定含固定模型名（如 qwen3-max）；先模糊取一条再测精确查询。
+    const listRun = await runCommandE2e(QUOTA_ROUTES, [
+      "quota",
+      "list",
+      "--name",
+      "qwen",
+      "--page-size",
+      "1",
+      "--output",
+      "json",
+    ]);
+    expect(listRun.exitCode, listRun.stderr).toBe(0);
+    const listed = parseStdoutJson<{ items?: { model?: string }[] }>(listRun.stdout);
+    const model = listed.items?.[0]?.model;
+    expect(model, "quota list --name qwen 应至少返回一条模型限流").toBeTruthy();
+
     const { stdout, stderr, exitCode } = await runCommandE2e(QUOTA_ROUTES, [
       "quota",
       "list",
       "--model",
-      "qwen3-max",
+      model!,
       "--output",
       "json",
     ]);
     expect(exitCode, stderr).toBe(0);
     const data = parseStdoutJson<{
       items?: { model?: string; model_limit?: { request_limit?: number | null } | null }[];
     }>(stdout);
-    expect(data.items?.[0]?.model).toBe("qwen3-max");
+    expect(data.items?.[0]?.model).toBe(model);
     expect(data.items?.[0]).toHaveProperty("model_limit");
   });
 

packages/commands/tests/output/local-command-output.test.ts

@@ -43,6 +43,7 @@ const retained = { future: { values: [false, 0, null, []] } };
 beforeEach(() => {
   vi.stubEnv("NO_COLOR", "1");
   vi.stubEnv("BAILIAN_INSTALL_METHOD", "brew");
+  vi.stubEnv("DASHSCOPE_API_KEY", undefined as unknown as string);
   vi.spyOn(httpTransport, "fetch").mockImplementation(async (input, init) => {
     const url = typeof input === "string" ? input : input instanceof URL ? input.href : input.url;
     if (url.endsWith("/index.json")) return Response.json({ version: 1, skills: registrySkills });