test(claude): split the thinking-restart regression into two tests

Each case now states its own setup instead of branching inside one test body.

packages/server/src/server/agent/providers/claude/agent.interrupt-restart-regression.test.ts

@@ -12,6 +12,7 @@ interface QueryMock {
   close: ReturnType<typeof vi.fn>;
   setPermissionMode: ReturnType<typeof vi.fn>;
   setModel: ReturnType<typeof vi.fn>;
+  applyFlagSettings: ReturnType<typeof vi.fn>;
   supportedModels: ReturnType<typeof vi.fn>;
   supportedCommands: ReturnType<typeof vi.fn>;
   rewindFiles: ReturnType<typeof vi.fn>;
@@ -148,6 +149,7 @@ function createScriptedQuery(params: {
     close: vi.fn(() => undefined),
     setPermissionMode: vi.fn(async () => undefined),
     setModel: vi.fn(async () => undefined),
+    applyFlagSettings: vi.fn(async () => undefined),
     supportedModels: vi.fn(async () => [{ value: "opus", displayName: "Opus" }]),
     supportedCommands: vi.fn(async () => []),
     rewindFiles: vi.fn(async () => ({ canRewind: true })),
@@ -276,6 +278,64 @@ afterEach(() => {
   queryFactory.mockReset();
 });
 
+async function createThinkingRestartSession() {
+  const queries: ScriptedQuery[] = [];
+  queryFactory.mockImplementation(({ prompt }: { prompt: AsyncIterable<unknown> }) => {
+    const scriptedQuery = createScriptedQuery({
+      prompt,
+      sessionId: "thinking-restart-session",
+      handlePrompt: ({ query }) => query.emit(buildSuccessResult("thinking-restart-session")),
+    });
+    queries.push(scriptedQuery);
+    return scriptedQuery;
+  });
+  const session = await new ClaudeAgentClient({
+    logger: createTestLogger(),
+    queryFactory,
+    resolveBinary: async () => "/test/claude/bin",
+  }).createSession({ provider: "claude", cwd: process.cwd() });
+  return { session, queries };
+}
+
+test("submits after a thinking change before the first turn", async () => {
+  const { session, queries } = await createThinkingRestartSession();
+
+  try {
+    await session.setThinkingOption!("medium");
+    const events = await collectUntilTerminal(
+      streamSession(session, "continue", { clientMessageId: "client-continue" }),
+    );
+    expect(events.filter((event) => event.type === "turn_failed")).toEqual([]);
+    expect(events.at(-1)?.type).toBe("turn_completed");
+    expect(queryFactory).toHaveBeenCalledTimes(1);
+    expect(queries[0].prompts.map((prompt) => prompt.text)).toEqual(["continue"]);
+  } finally {
+    await session.close();
+  }
+});
+
+test("submits after a thinking change once the previous query ended", async () => {
+  const { session, queries } = await createThinkingRestartSession();
+
+  try {
+    await collectUntilTerminal(streamSession(session, "first prompt"));
+    queries[0].end();
+    // Let the pump retire the old query before changing thinking.
+    await new Promise((resolve) => setTimeout(resolve, 0));
+
+    await session.setThinkingOption!("medium");
+    const events = await collectUntilTerminal(
+      streamSession(session, "continue", { clientMessageId: "client-continue" }),
+    );
+    expect(events.filter((event) => event.type === "turn_failed")).toEqual([]);
+    expect(events.at(-1)?.type).toBe("turn_completed");
+    expect(queryFactory).toHaveBeenCalledTimes(2);
+    expect(queries[1].prompts.map((prompt) => prompt.text)).toEqual(["continue"]);
+  } finally {
+    await session.close();
+  }
+});
+
 test("interrupt only calls query.interrupt and leaves the query open", async () => {
   const logger = createTestLogger();
   const queries: ScriptedQuery[] = [];

packages/server/src/server/agent/providers/claude/agent.ts

@@ -3275,6 +3275,9 @@ class ClaudeAgentSession implements AgentSession {
       }
     }
 
+    // A fresh query already uses the current settings, even when there was no old query to retire.
+    this.queryRestartNeeded = false;
+
     // Preserve claudeSessionId across query recreation so buildOptions() passes
     // resume: sessionId and the new query continues the existing conversation.
     this.persistence = null;
@@ -3877,11 +3880,12 @@ class ClaudeAgentSession implements AgentSession {
   }
 
   private startQueryPump(): void {
-    if (this.closed || this.queryPumpPromise) {
+    if (this.closed || this.queryPumpPromise || !this.query) {
       return;
     }
 
-    const pump = this.runQueryPump().catch((error) => {
+    // Drain the initialized query; a pending settings change belongs to the next turn.
+    const pump = this.runQueryPump(this.query).catch((error) => {
       this.logger.trace(
         {
           agentId: this.agentId,
@@ -3902,25 +3906,7 @@ class ClaudeAgentSession implements AgentSession {
     });
   }
 
-  private async runQueryPump(): Promise<void> {
-    let activeQuery: Query;
-    try {
-      activeQuery = await this.ensureQuery();
-    } catch (error) {
-      this.logger.trace(
-        {
-          agentId: this.agentId,
-          provider: "claude",
-          sessionId: this.claudeSessionId,
-          turnId: this.activeForegroundTurnId ?? this.autonomousTurn?.id ?? undefined,
-          err: error,
-        },
-        "provider.claude.query_pump.init_failed",
-      );
-      this.failActiveTurns(error instanceof Error ? error.message : "Claude stream failed");
-      return;
-    }
-
+  private async runQueryPump(activeQuery: Query): Promise<void> {
     let consecutiveInterruptAbortRecoveries = 0;
     const logRawMessage = (message: SDKMessage): void => {
       this.logger.trace(

packages/server/src/server/daemon-e2e/claude-thinking-memory.real.e2e.test.ts

@@ -115,4 +115,43 @@ describe("daemon E2E (real claude) - thinking effort memory", () => {
       rmSync(cwd, { recursive: true, force: true });
     }
   }, 420_000);
+
+  test("a thinking change before the first turn does not fail that turn", async () => {
+    const logger = pino({ level: "silent" });
+    const cwd = tmpCwd();
+    const daemon = await createTestPaseoDaemon({
+      agentClients: createRealProviderClients(["claude"], logger),
+      logger,
+    });
+    const client = new DaemonClient({ url: `ws://127.0.0.1:${daemon.port}/ws` });
+
+    try {
+      await client.connect();
+      await client.fetchAgents({ subscribe: {} });
+
+      const modelList = await client.listProviderModels("claude", { cwd });
+      const model = modelList.models.find(modelHasLowThinkingOption);
+      if (!model) {
+        throw new Error("No Claude Sonnet model with low thinking effort returned");
+      }
+
+      const agent = await client.createAgent({
+        cwd,
+        title: "claude-thinking-before-first-turn-real",
+        ...getRealProviderConfig("claude"),
+        model: model.id,
+      });
+      await client.setAgentThinkingOption(agent.id, "low");
+
+      await client.sendMessage(agent.id, "Reply exactly: ACK_FIRST");
+      const finish = await client.waitForFinish(agent.id, 180_000);
+      expect(finish.final?.lastError).toBeUndefined();
+      expect(finish.status).toBe("idle");
+      expect(compactText(await getAssistantText(client, agent.id))).toContain("ack_first");
+    } finally {
+      await client.close().catch(() => undefined);
+      await daemon.close().catch(() => undefined);
+      rmSync(cwd, { recursive: true, force: true });
+    }
+  }, 300_000);
 });