✨ feat(follow-up): extend follow-up chip suggestions to general chat (#15101)

* ✨ feat(follow-up): add foundation types for chat follow-up chips - FollowUpExtractInput.threadId for portal thread isolation - UserSystemAgentConfig.followUpAction (global enable + model) - LobeAgentChatConfig.enableFollowUpChips (per-agent opt-in) - ConversationHooks.onAssistantTurnSettled first-class member - Remove dead onGenerationStart/Complete/Cancelled hooks - DEFAULT_SYSTEM_AGENT_CONFIG.followUpAction off by default - DEFAULT_AGENT_CHAT_CONFIG.enableFollowUpChips false default * ♻️ refactor(follow-up): key follow-up store by conversation for concurrency - Convert useFollowUpActionStore from single-slot to slots map - conversationKey = messageMapKey(agentId, topicId, threadId?) for parity with chat store - contextSelectors.conversationKey exposes the key from ConversationProvider - FollowUpChips and ChatItem consume conversationKey - Onboarding hook adopts the new keyed API - Pass threadId through to extract (server filter lands in T3) * 🐛 fix(follow-up): address T2 code review feedback - Restore design-intent comments for 20s timeout and race guard - Remove dead pendingMessageId field from FollowUpActionSlot - Remove unused slotFor selector - Trim chipsFor / FollowUpActionSlot JSDoc to design intent only - Gate useOnboardingFollowUp against missing onboardingAgentId - removeSlot uses destructure; slotStatus uses ?? for falsy safety * ✨ feat(follow-up): filter extract by threadId for portal thread isolation - FollowUpActionService.extract honours optional threadId - threadId provided → eq(messages.threadId, threadId) - threadId absent → isNull(messages.threadId) so main topic never surfaces thread replies - Tests cover both branches * ✨ feat(conversation): emit onAssistantTurnSettled hook from provider - AssistantTurnSettledWatcher fires hooks.onAssistantTurnSettled(messageId, { reason }) once per turn - Reason derived from the most recent terminal Operation for the message id - Reason mapping: cancelled → stopped, type=regenerate → regenerated, type=continue → continued, else → completed - Settlement gated on idle + no pending tool intervention (mirrors Onboarding's logic) - Tests cover all four reason branches + intervention gating + no double-fire + fallback log - Onboarding bespoke prop untouched (migrates in T6) * 🐛 fix(conversation): scope settlement reason to turn-level operations - TURN_LEVEL_TYPES filter excludes child sub-ops (callLLM, executeToolCall, etc.) before sorting by endTime - Prevents successful regenerate/continue being misreported as 'completed' when a child finishes after the parent - Tests cover parent/child ordering for all reason branches * ✨ feat(follow-up): add useChatFollowUp hook and wire chat mount sites - New mergeConversationHooks composes multiple hooks with boolean short-circuit - useChatFollowUp computes effective enable (global × per-agent × valid model) - Registers onBeforeSendMessage/Continue/Regenerate to clear slot and onAssistantTurnSettled to extract - Mount sites: agent route ConversationArea, FloatingChatPanel, Portal Thread Chat (last in chain per §4.6) - Skips on reason='stopped'; skips when effective is false - Group chat intentionally not mounted * ♻️ refactor(onboarding): migrate settlement to ConversationHooks first-class - Drop bespoke onAssistantTurnSettled prop and duplicate useEffect from AgentOnboardingConversation - useOnboardingFollowUp returns ConversationHooks { onBeforeSendMessage, onAssistantTurnSettled } - Split settlement work: context-sync + builtin refresh runs first, chip extract runs after - Phase snapshot captured at memoize time preserves original prevPhase semantics - Settlement detection now lives solely in AssistantTurnSettledWatcher * ✨ feat(settings): add Follow-up suggestions controls (global + per-agent) - Global System Agent page: new Follow-up Suggestions panel (model picker + enable toggle) - Per-agent chat controls: enableFollowUpChips toggle with hint when global not configured - i18n keys: setting.systemAgent.followUpAction.*, setting.settingChat.enableFollowUpChips.* - Hint surfaces when user toggles per-agent ON but global is disabled/unmodeled * 🔧 chore(follow-up): T8 — scoped lint cleanup and comment discipline pass * 🐛 fix(follow-up): align conversationKey selector with callsite + wrap single hook - contextSelectors.conversationKey forwards full context (scope/isNew/groupId/subAgentId) so portal-thread NEW state matches callsite-computed keys - ConversationArea wraps chat-follow-up via mergeConversationHooks for spec §4.6 ordering robustness - Both per final-review Important concerns * ✅ test(settings): update follow-up defaults snapshots * ✨ feat(follow-up): surface model in service-model page + default to mini - Add followUpAction to /service-model OPTIONAL_FEATURE_ITEMS so model/provider and enable Switch render alongside inputCompletion and promptRewrite - Seed DEFAULT_FOLLOW_UP_ACTION_SYSTEM_AGENT_ITEM with DEFAULT_MINI model/provider so out-of-box config has a valid model; users only need to flip enabled - Sync settings selector snapshot
2026-06-13 19:20:04 +00:00 · 2026-05-23 00:31:15 +08:00
parent 6770d8f321
commit de9f7e092a
48 changed files with 2067 additions and 851 deletions
@@ -25,7 +25,7 @@ The two reference tools to read end-to-end:
 1. **先保证折叠态可读。** 每个 API 都必须有 Inspector；用户不展开也应该能看懂 “正在做什么 / 对什么做 / 当前结果是什么”。Inspector 不应该只展示函数名和原始参数。
 2. **Inspector 是一句话，不是详情页。** 优先表达动作、关键对象、数量、状态，例如 “分析图片 3 张”“搜索 12 个结果”“读取 config.json”。长文本、列表和结构化结果放到 Render 或 Portal。
 3. **Inspector 要覆盖执行生命周期。** `args` 还在 streaming、工具执行中、执行完成、执行失败时都应该有稳定展示；必要时同时读取 `args`、`partialArgs` 和 `pluginState`，避免出现空白、跳变或只显示半截参数。
-4. **文案要随状态切换时态。** 同一个动作在 loading 与 completed 两个阶段必须用不同的措辞：执行中用现在进行时（“正在创建任务 / Creating task / 正在搜索”），执行完成后切到完成态（“已创建任务 / Task created / 已找到 N 条”）。Inspector chip 会一直留在聊天记录里——如果一直挂着 “正在 xxx”，几小时后回看历史时会读起来像还在跑。约定的 i18n 形式是 `<api>.loading` / `<api>.completed` 一对键（见 `lobe-agent.apiName.callSubAgent.{loading,completed}` 与 `lobe-claude-code.task.{create,list,update,get}.{loading,completed}`），渲染时按 `isArgumentsStreaming || isLoading` 决定取哪一个。只读 / 查询类（“查看任务”这种本来就是名词性的）可以共用一个键。
+4. **文案要随状态切换时态。** 同一个动作在 loading 与 completed 两个阶段必须用不同的措辞：执行中用现在进行时（“正在创建任务 / Creating task / 正在搜索”），执行完成后切到完成态（“已创建任务 / Task created / 已找到 N 条”）。Inspector chip 会一直留在聊天记录里 —— 如果一直挂着 “正在 xxx”，几小时后回看历史时会读起来像还在跑。约定的 i18n 形式是 `<api>.loading` / `<api>.completed` 一对键（见 `lobe-agent.apiName.callSubAgent.{loading,completed}` 与 `lobe-claude-code.task.{create,list,update,get}.{loading,completed}`），渲染时按 `isArgumentsStreaming || isLoading` 决定取哪一个。只读 / 查询类（“查看任务” 这种本来就是名词性的）可以共用一个键。
 5. **只有结构化结果才需要 Render。** 如果工具结果只是自然语言总结，通常不需要 Render；如果结果包含列表、媒体、文件、表格、代码、diff、地图、时间线、权限请求等结构，就应该提供 Render。
 6. **Render 要帮助用户检查结果，而不是复述参数。** Render 的主体应该围绕工具产物组织：可预览、可比较、可筛选、可定位。参数只作为上下文辅助出现，不要把 Render 做成一块更大的 args dump。
 7. **参数和结果要一起参与渲染。** 好的 Tool UI 通常同时用 `args` 解释意图，用 `pluginState` 展示真实执行结果；但 `pluginState` 只放结果域数据，不要反向塞入可以从 `args` 推导出的内容。
@@ -346,25 +346,31 @@ describe('hetero exec command', () => {
    it('retries without --resume when the error stream event indicates the session is gone', async () => {
      // First spawn: exits non-zero, emits a resume-not-found error event
      const resumeNotFoundEvent = {
-        data: { error: 'No conversation found with session ID cc-stale', message: 'No conversation found with session ID cc-stale' },
+        data: {
+          error: 'No conversation found with session ID cc-stale',
+          message: 'No conversation found with session ID cc-stale',
+        },
        operationId: 'op-r1',
        stepIndex: 0,
        timestamp: 1,
        type: 'error',
      };
      mockSpawnAgent
-        .mockReturnValueOnce(
-          createFakeHandle({ events: [resumeNotFoundEvent], exitCode: 1 }),
-        )
+        .mockReturnValueOnce(createFakeHandle({ events: [resumeNotFoundEvent], exitCode: 1 }))
        // Second spawn: succeeds
        .mockReturnValueOnce(createFakeHandle({ exitCode: 0 }));

      await runCmd([
-        'hetero', 'exec',
-        '--type', 'claude-code',
-        '--prompt', 'do the thing',
-        '--resume', 'cc-stale',
-        '--operation-id', 'op-r1',
+        'hetero',
+        'exec',
+        '--type',
+        'claude-code',
+        '--prompt',
+        'do the thing',
+        '--resume',
+        'cc-stale',
+        '--operation-id',
+        'op-r1',
      ]);

      // Two spawns: first with --resume, retry without
@@ -389,11 +395,16 @@ describe('hetero exec command', () => {
        .mockReturnValueOnce(createFakeHandle({ exitCode: 0 }));

      await runCmd([
-        'hetero', 'exec',
-        '--type', 'claude-code',
-        '--prompt', 'continue',
-        '--resume', 'xyz',
-        '--operation-id', 'op-r2',
+        'hetero',
+        'exec',
+        '--type',
+        'claude-code',
+        '--prompt',
+        'continue',
+        '--resume',
+        'xyz',
+        '--operation-id',
+        'op-r2',
      ]);

      expect(mockSpawnAgent).toHaveBeenCalledTimes(2);
@@ -403,7 +414,10 @@ describe('hetero exec command', () => {

    it('retries without --resume when the error indicates context overflow', async () => {
      const contextOverflowEvent = {
-        data: { error: 'prompt is too long: 215168 tokens > 200000 maximum', message: 'prompt is too long: 215168 tokens > 200000 maximum' },
+        data: {
+          error: 'prompt is too long: 215168 tokens > 200000 maximum',
+          message: 'prompt is too long: 215168 tokens > 200000 maximum',
+        },
        operationId: 'op-ctx',
        stepIndex: 0,
        timestamp: 1,
@@ -414,11 +428,16 @@ describe('hetero exec command', () => {
        .mockReturnValueOnce(createFakeHandle({ exitCode: 0 }));

      await runCmd([
-        'hetero', 'exec',
-        '--type', 'claude-code',
-        '--prompt', 'next question',
-        '--resume', 'cc-longctx',
-        '--operation-id', 'op-ctx',
+        'hetero',
+        'exec',
+        '--type',
+        'claude-code',
+        '--prompt',
+        'next question',
+        '--resume',
+        'cc-longctx',
+        '--operation-id',
+        'op-ctx',
      ]);

      expect(mockSpawnAgent).toHaveBeenCalledTimes(2);
@@ -434,10 +453,14 @@ describe('hetero exec command', () => {
      );

      await runCmd([
-        'hetero', 'exec',
-        '--type', 'claude-code',
-        '--prompt', 'hi',
-        '--resume', 'cc-valid',
+        'hetero',
+        'exec',
+        '--type',
+        'claude-code',
+        '--prompt',
+        'hi',
+        '--resume',
+        'cc-valid',
      ]);

      expect(mockSpawnAgent).toHaveBeenCalledTimes(1);
@@ -455,10 +478,14 @@ describe('hetero exec command', () => {
      mockSpawnAgent.mockReturnValueOnce(createFakeHandle({ events: [errorEvent], exitCode: 1 }));

      await runCmd([
-        'hetero', 'exec',
-        '--type', 'claude-code',
-        '--prompt', 'fresh run',
-        '--operation-id', 'op-nr',
+        'hetero',
+        'exec',
+        '--type',
+        'claude-code',
+        '--prompt',
+        'fresh run',
+        '--operation-id',
+        'op-nr',
      ]);

      // No --resume → no interception → no retry
@@ -468,7 +495,10 @@ describe('hetero exec command', () => {

    it('does NOT suppress the resume-error event from JSONL output', async () => {
      const resumeNotFoundEvent = {
-        data: { error: 'No conversation found with session ID old', message: 'No conversation found with session ID old' },
+        data: {
+          error: 'No conversation found with session ID old',
+          message: 'No conversation found with session ID old',
+        },
        operationId: 'op-jsonl',
        stepIndex: 0,
        timestamp: 1,
@@ -479,11 +509,16 @@ describe('hetero exec command', () => {
        .mockReturnValueOnce(createFakeHandle({ exitCode: 0 }));

      await runCmd([
-        'hetero', 'exec',
-        '--type', 'claude-code',
-        '--prompt', 'do thing',
-        '--resume', 'old',
-        '--render', 'jsonl',
+        'hetero',
+        'exec',
+        '--type',
+        'claude-code',
+        '--prompt',
+        'do thing',
+        '--resume',
+        'old',
+        '--render',
+        'jsonl',
      ]);

      // The error event is still emitted to JSONL (for observability) even
@@ -492,7 +527,11 @@ describe('hetero exec command', () => {
        .map((c) => c[0])
        .filter((s): s is string => typeof s === 'string');
      const errorLine = lines.find((l) => {
-        try { return JSON.parse(l).type === 'error'; } catch { return false; }
+        try {
+          return JSON.parse(l).type === 'error';
+        } catch {
+          return false;
+        }
      });
      expect(errorLine).toBeDefined();
    });
@@ -341,7 +341,7 @@ const exec = async (options: ExecOptions): Promise<void> => {
    // into the ingester.  When intercepting resume errors, a matching
    // `error` event is withheld from the ingester and flags a retry instead.
    let resumeNotFound = false;
-    let ingestError = false;
+    const ingestError = false;
    try {
      for await (const event of handle.events) {
        if (interceptResumeErrors && event.type === 'error') {
@@ -393,7 +393,14 @@ const exec = async (options: ExecOptions): Promise<void> => {
      resumeNotFound = true;
    }

-    return { code, ingestError, resumeNotFound, sessionId: handle.sessionId, signal, stderrContent };
+    return {
+      code,
+      ingestError,
+      resumeNotFound,
+      sessionId: handle.sessionId,
+      signal,
+      stderrContent,
+    };
  };

  // ─── First run (with --resume if provided) ───────────────────────────────
@@ -254,7 +254,12 @@ class AgentManagementExecutor extends BaseExecutor<typeof AgentManagementApiName

        try {
          await dispatchNonHeteroSubAgent(
-            { kind: 'callAgent', targetAgentId: agentId, instruction, parentMessageId: ctx.messageId },
+            {
+              kind: 'callAgent',
+              targetAgentId: agentId,
+              instruction,
+              parentMessageId: ctx.messageId,
+            },
            {
              conversationContext,
              heterogeneousProvider: parentAgentConfig?.agencyConfig?.heterogeneousProvider,
@@ -242,7 +242,8 @@ export const MemoryManifest: BuiltinToolManifest = {
            type: 'string',
          },
          sourceIds: {
-            description: 'Stable source message ids that support this memory. Use [] when unavailable.',
+            description:
+              'Stable source message ids that support this memory. Use [] when unavailable.',
            items: { type: 'string' },
            type: ['array', 'null'],
          },
@@ -398,7 +399,8 @@ export const MemoryManifest: BuiltinToolManifest = {
            type: 'string',
          },
          sourceIds: {
-            description: 'Stable source message ids that support this memory. Use [] when unavailable.',
+            description:
+              'Stable source message ids that support this memory. Use [] when unavailable.',
            items: { type: 'string' },
            type: ['array', 'null'],
          },
@@ -561,7 +563,8 @@ export const MemoryManifest: BuiltinToolManifest = {
            type: 'string',
          },
          sourceIds: {
-            description: 'Stable source message ids that support this memory. Use [] when unavailable.',
+            description:
+              'Stable source message ids that support this memory. Use [] when unavailable.',
            items: { type: 'string' },
            type: ['array', 'null'],
          },
@@ -770,7 +773,8 @@ export const MemoryManifest: BuiltinToolManifest = {
            type: 'string',
          },
          sourceIds: {
-            description: 'Stable source message ids that support this memory. Use [] when unavailable.',
+            description:
+              'Stable source message ids that support this memory. Use [] when unavailable.',
            items: { type: 'string' },
            type: ['array', 'null'],
          },
@@ -27,6 +27,7 @@ export const DEFAULT_AGENT_CHAT_CONFIG: LobeAgentChatConfig = {
  enableAgentMode: true,
  enableCompressHistory: true,
  enableContextCompression: true,
+  enableFollowUpChips: false,
  enableHistoryCount: false,
  enableStreaming: true,
  historyCount: 20,
@@ -29,8 +29,15 @@ export const DEFAULT_INPUT_COMPLETION_SYSTEM_AGENT_ITEM: SystemAgentItem = {
  provider: DEFAULT_MINI_SYSTEM_AGENT_ITEM.provider,
 };

+export const DEFAULT_FOLLOW_UP_ACTION_SYSTEM_AGENT_ITEM: SystemAgentItem = {
+  enabled: false,
+  model: DEFAULT_MINI_SYSTEM_AGENT_ITEM.model,
+  provider: DEFAULT_MINI_SYSTEM_AGENT_ITEM.provider,
+};
+
 export const DEFAULT_SYSTEM_AGENT_CONFIG: UserSystemAgentConfig = {
  agentMeta: DEFAULT_SYSTEM_AGENT_ITEM,
+  followUpAction: DEFAULT_FOLLOW_UP_ACTION_SYSTEM_AGENT_ITEM,
  generationTopic: DEFAULT_MINI_SYSTEM_AGENT_ITEM,
  historyCompress: DEFAULT_SYSTEM_AGENT_ITEM,
  inputCompletion: DEFAULT_INPUT_COMPLETION_SYSTEM_AGENT_ITEM,
@@ -326,4 +326,3 @@ export interface AgentProcessConfig {
  /** Environment variables */
  env?: Record<string, string>;
 }
-
@@ -5,7 +5,6 @@ import OpenAI from 'openai';

 import { responsesAPIModels } from '../../const/models';
 import { buildDefaultAnthropicPayload } from '../../core/anthropicCompatibleFactory';
-import { assertToolLimits } from '../../utils/validateToolLimits';
 import { type LobeRuntimeAI } from '../../core/BaseAI';
 import {
  convertOpenAIMessages,
@@ -20,6 +19,7 @@ import { AgentRuntimeError } from '../../utils/createError';
 import { debugResponse, debugStream } from '../../utils/debugStream';
 import { getModelPricing } from '../../utils/getModelPricing';
 import { StreamingResponse } from '../../utils/response';
+import { assertToolLimits } from '../../utils/validateToolLimits';

 const COPILOT_BASE_URL = 'https://api.githubcopilot.com';
 const TOKEN_EXCHANGE_URL = 'https://api.github.com/copilot_internal/v2/token';
@@ -63,6 +63,7 @@ export interface LobeAgentChatConfig extends AgentMemoryChatConfig, AgentSelfIte
   * When enabled, old messages will be compressed into summaries when token threshold is reached
   */
  enableContextCompression?: boolean;
+  enableFollowUpChips?: boolean;
  /**
   * Enable historical message count
   */
@@ -206,6 +207,7 @@ export const AgentChatConfigSchema = z
    enableAutoScrollOnStreaming: z.boolean().optional(),
    enableCompressHistory: z.boolean().optional(),
    enableContextCompression: z.boolean().optional(),
+    enableFollowUpChips: z.boolean().optional(),
    enableHistoryCount: z.boolean().optional(),
    enableMaxTokens: z.boolean().optional(),
    enableReasoning: z.boolean().optional(),
@@ -19,6 +19,7 @@ export interface FollowUpModelConfig {
 export interface FollowUpExtractInput {
  hint?: FollowUpHint;
  modelConfig: FollowUpModelConfig;
+  threadId?: string;
  topicId: string;
 }

@@ -46,5 +47,6 @@ export const FollowUpModelConfigSchema = z.object({
 export const FollowUpExtractInputSchema = z.object({
  hint: FollowUpHintSchema.optional(),
  modelConfig: FollowUpModelConfigSchema,
+  threadId: z.string().optional(),
  topicId: z.string().min(1),
 });
@@ -11,6 +11,7 @@ export interface PromptRewriteSystemAgent extends Omit<SystemAgentItem, 'enabled

 export interface UserSystemAgentConfig {
  agentMeta: SystemAgentItem;
+  followUpAction: SystemAgentItem;
  generationTopic: SystemAgentItem;
  historyCompress: SystemAgentItem;
  inputCompletion: SystemAgentItem;
@@ -16,6 +16,8 @@ import ControlsForm from '@/features/ModelSwitchPanel/components/ControlsForm';
 import { useAgentStore } from '@/store/agent';
 import { agentByIdSelectors, chatConfigByIdSelectors } from '@/store/agent/selectors';
 import { aiModelSelectors, useAiInfraStore } from '@/store/aiInfra';
+import { useUserStore } from '@/store/user';
+import { systemAgentSelectors } from '@/store/user/selectors';
 import type { LobeAgentConfig } from '@/types/agent';

 import { useAgentId } from '../../hooks/useAgentId';
@@ -114,6 +116,11 @@ const styles = createStaticStyles(({ css }) => ({
    height: 1px;
    background: ${cssVar.colorSplit};
  `,
+  hint: css`
+    font-size: 12px;
+    line-height: 18px;
+    color: ${cssVar.colorTextTertiary};
+  `,
  form: css`
    margin: 0;
  `,
@@ -533,6 +540,11 @@ const Controls = memo<ControlsProps>(({ setUpdating, updating, variant = 'popove
    'enableAutoScrollOnStreaming',
  ]);
  const enableStreaming = form.getFieldValue(['chatConfig', 'enableStreaming']);
+  const enableFollowUpChips = form.getFieldValue(['chatConfig', 'enableFollowUpChips']);
+  const globalFollowUp = useUserStore(systemAgentSelectors.followUpAction, isEqual);
+  const globalFollowUpReady =
+    globalFollowUp.enabled === true && !!globalFollowUp.model && !!globalFollowUp.provider;
+  const showFollowUpHint = !globalFollowUpReady && Boolean(enableFollowUpChips);
  const enableReasoningEffort = form.getFieldValue(['chatConfig', 'enableReasoningEffort']);
  const reasoningEffortValue = form.getFieldValue(['params', 'reasoning_effort']);
  const disabledParams = useAiInfraStore(
@@ -783,6 +795,26 @@ const Controls = memo<ControlsProps>(({ setUpdating, updating, variant = 'popove
                />
              }
            />
+            <ControlRow
+              tag="followUpChips"
+              title={t('settingChat.enableFollowUpChips.title')}
+              tooltip={t('settingChat.enableFollowUpChips.desc')}
+              action={
+                <Switch
+                  checked={Boolean(enableFollowUpChips)}
+                  size={'small'}
+                  onChange={(checked) => {
+                    handleFieldChange(['chatConfig', 'enableFollowUpChips'], checked);
+                  }}
+                />
+              }
+            >
+              {showFollowUpHint && (
+                <div className={styles.hint}>
+                  {t('settingChat.enableFollowUpChips.notConfiguredHint')}
+                </div>
+              )}
+            </ControlRow>
            <ControlRow
              tag="inputTemplate"
              title={t('settingChat.inputTemplate.title')}
@@ -0,0 +1,286 @@
+/**
+ * @vitest-environment happy-dom
+ */
+import { render, waitFor } from '@testing-library/react';
+import debug from 'debug';
+import { beforeEach, describe, expect, it, vi } from 'vitest';
+
+import AssistantTurnSettledWatcher from './AssistantTurnSettledWatcher';
+
+const { mockState, mockChatState } = vi.hoisted(() => ({
+  mockChatState: {
+    operations: {} as Record<string, any>,
+    operationsByMessage: {} as Record<string, string[]>,
+  },
+  mockState: {
+    displayMessages: [] as Array<{ id: string; role: string }>,
+    generatingIds: new Set<string>(),
+    hooks: {} as Record<string, ((...args: any[]) => any) | undefined>,
+    pendingInterventions: [] as Array<{ id: string }>,
+  },
+}));
+
+vi.mock('./store', () => ({
+  contextSelectors: {
+    hook: (name: string) => (state: typeof mockState) => state.hooks[name],
+  },
+  conversationSelectors: {
+    displayMessages: (state: typeof mockState) => state.displayMessages,
+  },
+  dataSelectors: {
+    pendingInterventions: (state: typeof mockState) => state.pendingInterventions,
+  },
+  messageStateSelectors: {
+    isAssistantGroupItemGenerating: (id: string) => (state: typeof mockState) =>
+      state.generatingIds.has(id),
+  },
+  useConversationStore: (selector: (state: typeof mockState) => unknown) => selector(mockState),
+}));
+
+vi.mock('@/store/chat', () => ({
+  useChatStore: {
+    getState: () => mockChatState,
+  },
+}));
+
+interface SeedOp {
+  cancelReason?: string;
+  endTime?: number;
+  parentOperationId?: string;
+  status: string;
+  type: string;
+}
+
+const seedOperations = (messageId: string, ops: SeedOp[]) => {
+  mockChatState.operations = {};
+  mockChatState.operationsByMessage = {};
+  const ids: string[] = [];
+  ops.forEach((op, index) => {
+    const id = `op-${messageId}-${index}`;
+    ids.push(id);
+    mockChatState.operations[id] = {
+      id,
+      parentOperationId: op.parentOperationId,
+      status: op.status,
+      type: op.type,
+      metadata: {
+        cancelReason: op.cancelReason,
+        endTime: op.endTime ?? index + 1,
+        startTime: 0,
+      },
+    };
+  });
+  mockChatState.operationsByMessage[messageId] = ids;
+};
+
+const armAndSettle = (
+  rerender: (ui: React.ReactElement) => void,
+  hook: ReturnType<typeof vi.fn>,
+) => {
+  mockState.hooks.onAssistantTurnSettled = hook;
+  mockState.displayMessages = [
+    { id: 'user-1', role: 'user' },
+    { id: 'assistant-1', role: 'assistant' },
+  ];
+  mockState.generatingIds = new Set(['assistant-1']);
+  rerender(<AssistantTurnSettledWatcher />);
+
+  mockState.generatingIds = new Set();
+  rerender(<AssistantTurnSettledWatcher />);
+};
+
+describe('AssistantTurnSettledWatcher', () => {
+  beforeEach(() => {
+    mockState.displayMessages = [];
+    mockState.generatingIds = new Set();
+    mockState.pendingInterventions = [];
+    mockState.hooks = {};
+    mockChatState.operations = {};
+    mockChatState.operationsByMessage = {};
+  });
+
+  it('fires with reason "completed" when latest terminal op is sendMessage/completed', async () => {
+    const hook = vi.fn();
+    seedOperations('assistant-1', [{ status: 'completed', type: 'sendMessage' }]);
+
+    const { rerender } = render(<AssistantTurnSettledWatcher />);
+    armAndSettle(rerender, hook);
+
+    await waitFor(() => {
+      expect(hook).toHaveBeenCalledWith('assistant-1', { reason: 'completed' });
+    });
+    expect(hook).toHaveBeenCalledTimes(1);
+  });
+
+  it('fires with reason "regenerated" when latest terminal op type is regenerate', async () => {
+    const hook = vi.fn();
+    seedOperations('assistant-1', [{ status: 'completed', type: 'regenerate' }]);
+
+    const { rerender } = render(<AssistantTurnSettledWatcher />);
+    armAndSettle(rerender, hook);
+
+    await waitFor(() => {
+      expect(hook).toHaveBeenCalledWith('assistant-1', { reason: 'regenerated' });
+    });
+  });
+
+  it('fires with reason "continued" when latest terminal op type is continue', async () => {
+    const hook = vi.fn();
+    seedOperations('assistant-1', [{ status: 'completed', type: 'continue' }]);
+
+    const { rerender } = render(<AssistantTurnSettledWatcher />);
+    armAndSettle(rerender, hook);
+
+    await waitFor(() => {
+      expect(hook).toHaveBeenCalledWith('assistant-1', { reason: 'continued' });
+    });
+  });
+
+  it('fires with reason "stopped" when latest terminal op is cancelled, even on regenerate', async () => {
+    const hook = vi.fn();
+    seedOperations('assistant-1', [{ status: 'cancelled', type: 'regenerate' }]);
+
+    const { rerender } = render(<AssistantTurnSettledWatcher />);
+    armAndSettle(rerender, hook);
+
+    await waitFor(() => {
+      expect(hook).toHaveBeenCalledWith('assistant-1', { reason: 'stopped' });
+    });
+  });
+
+  it('derives "regenerated" when parent regenerate completes after child callLLM also completes (parent/child ordering)', async () => {
+    const hook = vi.fn();
+    seedOperations('assistant-1', [
+      { endTime: 1000, status: 'completed', type: 'regenerate' },
+      {
+        endTime: 1500,
+        parentOperationId: 'op-assistant-1-0',
+        status: 'completed',
+        type: 'callLLM',
+      },
+    ]);
+
+    const { rerender } = render(<AssistantTurnSettledWatcher />);
+    armAndSettle(rerender, hook);
+
+    await waitFor(() => {
+      expect(hook).toHaveBeenCalledWith('assistant-1', { reason: 'regenerated' });
+    });
+  });
+
+  it('derives "continued" when parent continue completes after child callLLM also completes', async () => {
+    const hook = vi.fn();
+    seedOperations('assistant-1', [
+      { endTime: 1000, status: 'completed', type: 'continue' },
+      {
+        endTime: 1500,
+        parentOperationId: 'op-assistant-1-0',
+        status: 'completed',
+        type: 'callLLM',
+      },
+    ]);
+
+    const { rerender } = render(<AssistantTurnSettledWatcher />);
+    armAndSettle(rerender, hook);
+
+    await waitFor(() => {
+      expect(hook).toHaveBeenCalledWith('assistant-1', { reason: 'continued' });
+    });
+  });
+
+  it('derives "stopped" when parent regenerate is cancelled with cancelled child callLLM finishing later', async () => {
+    const hook = vi.fn();
+    seedOperations('assistant-1', [
+      {
+        cancelReason: 'User cancelled',
+        endTime: 1000,
+        status: 'cancelled',
+        type: 'regenerate',
+      },
+      {
+        endTime: 1500,
+        parentOperationId: 'op-assistant-1-0',
+        status: 'cancelled',
+        type: 'callLLM',
+      },
+    ]);
+
+    const { rerender } = render(<AssistantTurnSettledWatcher />);
+    armAndSettle(rerender, hook);
+
+    await waitFor(() => {
+      expect(hook).toHaveBeenCalledWith('assistant-1', { reason: 'stopped' });
+    });
+  });
+
+  it('defers settlement while pending intervention exists, fires once it clears', async () => {
+    const hook = vi.fn();
+    seedOperations('assistant-1', [{ status: 'completed', type: 'sendMessage' }]);
+
+    const { rerender } = render(<AssistantTurnSettledWatcher />);
+
+    mockState.hooks.onAssistantTurnSettled = hook;
+    mockState.displayMessages = [
+      { id: 'user-1', role: 'user' },
+      { id: 'assistant-1', role: 'assistant' },
+    ];
+    mockState.generatingIds = new Set(['assistant-1']);
+    rerender(<AssistantTurnSettledWatcher />);
+
+    mockState.generatingIds = new Set();
+    mockState.pendingInterventions = [{ id: 'tool-1' }];
+    rerender(<AssistantTurnSettledWatcher />);
+    expect(hook).not.toHaveBeenCalled();
+
+    mockState.pendingInterventions = [];
+    rerender(<AssistantTurnSettledWatcher />);
+    expect(hook).not.toHaveBeenCalled();
+
+    mockState.generatingIds = new Set(['assistant-1']);
+    rerender(<AssistantTurnSettledWatcher />);
+    expect(hook).not.toHaveBeenCalled();
+
+    mockState.generatingIds = new Set();
+    rerender(<AssistantTurnSettledWatcher />);
+
+    await waitFor(() => {
+      expect(hook).toHaveBeenCalledWith('assistant-1', { reason: 'completed' });
+    });
+    expect(hook).toHaveBeenCalledTimes(1);
+  });
+
+  it('does not double-fire for the same message id across rerenders', async () => {
+    const hook = vi.fn();
+    seedOperations('assistant-1', [{ status: 'completed', type: 'sendMessage' }]);
+
+    const { rerender } = render(<AssistantTurnSettledWatcher />);
+    armAndSettle(rerender, hook);
+
+    await waitFor(() => {
+      expect(hook).toHaveBeenCalledTimes(1);
+    });
+
+    rerender(<AssistantTurnSettledWatcher />);
+    rerender(<AssistantTurnSettledWatcher />);
+    expect(hook).toHaveBeenCalledTimes(1);
+  });
+
+  it('falls back to reason "completed" and logs when no terminal op exists', async () => {
+    const hook = vi.fn();
+    const logSpy = vi.spyOn(debug, 'log').mockImplementation(() => {});
+    debug.enable('lobe-render:features:Conversation');
+
+    const { rerender } = render(<AssistantTurnSettledWatcher />);
+    armAndSettle(rerender, hook);
+
+    await waitFor(() => {
+      expect(hook).toHaveBeenCalledWith('assistant-1', { reason: 'completed' });
+    });
+    expect(logSpy).toHaveBeenCalled();
+    const logged = logSpy.mock.calls.flat().map(String).join(' ');
+    expect(logged).toContain('settlement fired without terminal op');
+
+    logSpy.mockRestore();
+    debug.disable();
+  });
+});
@@ -0,0 +1,108 @@
+'use client';
+
+import debug from 'debug';
+import { useEffect, useMemo, useRef } from 'react';
+
+import { useChatStore } from '@/store/chat';
+import { operationSelectors } from '@/store/chat/slices/operation/selectors';
+import { type Operation, type OperationType } from '@/store/chat/slices/operation/types';
+
+import {
+  contextSelectors,
+  conversationSelectors,
+  dataSelectors,
+  messageStateSelectors,
+  useConversationStore,
+} from './store';
+
+const log = debug('lobe-render:features:Conversation');
+
+const assistantLikeRoles = new Set(['assistant', 'assistantGroup', 'supervisor']);
+
+type SettlementReason = 'completed' | 'stopped' | 'regenerated' | 'continued';
+
+// Restrict reason derivation to turn-level intents so child sub-ops (callLLM, executeToolCall, …) cannot shadow the parent type at settlement
+const TURN_LEVEL_TYPES = new Set<OperationType>(['sendMessage', 'regenerate', 'continue']);
+
+const deriveReason = (op: Operation): SettlementReason => {
+  // cancelled wins over type: a stopped regenerate is still 'stopped'
+  if (op.status === 'cancelled') return 'stopped';
+  if (op.type === 'regenerate') return 'regenerated';
+  if (op.type === 'continue') return 'continued';
+  return 'completed';
+};
+
+const resolveReason = (messageId: string): SettlementReason => {
+  const operations = operationSelectors.getOperationsByMessage(messageId)(useChatStore.getState());
+  const terminal = operations
+    .filter((op) => TURN_LEVEL_TYPES.has(op.type))
+    .filter(
+      (op) => op.status === 'completed' || op.status === 'cancelled' || op.status === 'failed',
+    )
+    .sort((a, b) => (b.metadata.endTime ?? 0) - (a.metadata.endTime ?? 0))[0];
+
+  if (!terminal) {
+    log('settlement fired without terminal op for messageId=%s', messageId);
+    return 'completed';
+  }
+
+  return deriveReason(terminal);
+};
+
+const AssistantTurnSettledWatcher = () => {
+  const displayMessages = useConversationStore(conversationSelectors.displayMessages);
+  const onAssistantTurnSettled = useConversationStore(
+    contextSelectors.hook('onAssistantTurnSettled'),
+  );
+
+  const latestAssistantMessageId = useMemo(() => {
+    const latest = displayMessages.at(-1);
+    if (!latest || !assistantLikeRoles.has(latest.role)) return undefined;
+    return latest.id;
+  }, [displayMessages]);
+
+  const isLatestAssistantGenerating = useConversationStore((s) =>
+    latestAssistantMessageId
+      ? messageStateSelectors.isAssistantGroupItemGenerating(latestAssistantMessageId)(s)
+      : false,
+  );
+
+  const pendingInterventionCount = useConversationStore(
+    (s) => dataSelectors.pendingInterventions(s).length,
+  );
+
+  const armedSettledMessageIdRef = useRef<string>(undefined);
+  const firedSettledMessageIdRef = useRef<string>(undefined);
+
+  useEffect(() => {
+    if (!onAssistantTurnSettled || !latestAssistantMessageId) return;
+
+    if (pendingInterventionCount > 0) {
+      armedSettledMessageIdRef.current = undefined;
+      return;
+    }
+
+    if (isLatestAssistantGenerating) {
+      armedSettledMessageIdRef.current = latestAssistantMessageId;
+      return;
+    }
+
+    if (armedSettledMessageIdRef.current !== latestAssistantMessageId) return;
+    if (firedSettledMessageIdRef.current === latestAssistantMessageId) return;
+
+    firedSettledMessageIdRef.current = latestAssistantMessageId;
+    armedSettledMessageIdRef.current = undefined;
+
+    const reason = resolveReason(latestAssistantMessageId);
+    void onAssistantTurnSettled(latestAssistantMessageId, { reason });
+  }, [
+    isLatestAssistantGenerating,
+    latestAssistantMessageId,
+    onAssistantTurnSettled,
+    pendingInterventionCount,
+  ]);
+
+  return null;
+};
+
+export default AssistantTurnSettledWatcher;
@@ -44,7 +44,7 @@ const ChatItem = memo<ChatItemProps>(
    ...rest
  }) => {
    const isUser = placement === 'right';
-    const topicId = useConversationStore(contextSelectors.topicId);
+    const conversationKey = useConversationStore(contextSelectors.conversationKey);
    const isEmptyMessage =
      !message || String(message).trim() === '' || message === placeholderMessage;
    const errorContent = error && (
@@ -118,7 +118,9 @@ const ChatItem = memo<ChatItemProps>(
          )}
          {belowMessage}
        </Flexbox>
-        {id && topicId && <FollowUpChips messageId={id} topicId={topicId} />}
+        {id && conversationKey && (
+          <FollowUpChips conversationKey={conversationKey} messageId={id} />
+        )}
        {actions && <Actions actions={actions} placement={placement} />}
      </Flexbox>
    );
@@ -48,321 +48,323 @@ interface VirtualizedListProps {
 */
 const VirtualizedList = memo<VirtualizedListProps>(
  ({ dataSource, footerSlot, headerSlot, itemContent }) => {
-  const virtuaRef = useRef<VListHandle>(null);
-  const scrollEndTimerRef = useRef<ReturnType<typeof setTimeout> | null>(null);
-  const lastUserScrollIntentAtRef = useRef(0);
+    const virtuaRef = useRef<VListHandle>(null);
+    const scrollEndTimerRef = useRef<ReturnType<typeof setTimeout> | null>(null);
+    const lastUserScrollIntentAtRef = useRef(0);

-  // Per-topic scroll restoration. Provider does not remount on topic switch,
-  // so we key the scroll snapshot by the message-map key derived from
-  // ConversationStore's `context`.
-  const contextKey = useConversationStore((s) => messageMapKey(s.context));
-  const { recordScroll } = useTopicScrollPersist({
-    contextKey,
-    dataSourceLength: dataSource.length,
-    virtuaRef,
-  });
+    // Per-topic scroll restoration. Provider does not remount on topic switch,
+    // so we key the scroll snapshot by the message-map key derived from
+    // ConversationStore's `context`.
+    const contextKey = useConversationStore((s) => messageMapKey(s.context));
+    const { recordScroll } = useTopicScrollPersist({
+      contextKey,
+      dataSourceLength: dataSource.length,
+      virtuaRef,
+    });

-  // Second-to-last message is the user turn when sending (user + assistant pair)
-  const isSecondLastMessageFromUser = useConversationStore(
-    dataSelectors.isSecondLastMessageFromUser,
-  );
+    // Second-to-last message is the user turn when sending (user + assistant pair)
+    const isSecondLastMessageFromUser = useConversationStore(
+      dataSelectors.isSecondLastMessageFromUser,
+    );

-  const {
-    isScrollShrinking,
-    isSpacerMessage,
-    listData,
-    onScrollOffset,
-    registerSpacerNode,
-    spacerActive,
-    spacerHeight,
-  } = useConversationScroll({
-    dataSource,
-    isSecondLastMessageFromUser,
-    virtuaRef,
-  });
+    const {
+      isScrollShrinking,
+      isSpacerMessage,
+      listData,
+      onScrollOffset,
+      registerSpacerNode,
+      spacerActive,
+      spacerHeight,
+    } = useConversationScroll({
+      dataSource,
+      isSecondLastMessageFromUser,
+      virtuaRef,
+    });

-  const isAutoScrollEnabled = useAutoScrollEnabled();
+    const isAutoScrollEnabled = useAutoScrollEnabled();

-  // Store actions
-  const registerVirtuaScrollMethods = useConversationStore((s) => s.registerVirtuaScrollMethods);
-  const setScrollState = useConversationStore((s) => s.setScrollState);
-  const resetVisibleItems = useConversationStore((s) => s.resetVisibleItems);
-  const setActiveIndex = useConversationStore((s) => s.setActiveIndex);
-  const activeIndex = useConversationStore(virtuaListSelectors.activeIndex);
+    // Store actions
+    const registerVirtuaScrollMethods = useConversationStore((s) => s.registerVirtuaScrollMethods);
+    const setScrollState = useConversationStore((s) => s.setScrollState);
+    const resetVisibleItems = useConversationStore((s) => s.resetVisibleItems);
+    const setActiveIndex = useConversationStore((s) => s.setActiveIndex);
+    const activeIndex = useConversationStore(virtuaListSelectors.activeIndex);

-  const markUserScrollIntent = useCallback(() => {
-    lastUserScrollIntentAtRef.current = Date.now();
-  }, []);
+    const markUserScrollIntent = useCallback(() => {
+      lastUserScrollIntentAtRef.current = Date.now();
+    }, []);

-  const handlePointerMove = useCallback(
-    (event: PointerEvent<HTMLDivElement>) => {
-      if (event.buttons > 0) {
-        markUserScrollIntent();
+    const handlePointerMove = useCallback(
+      (event: PointerEvent<HTMLDivElement>) => {
+        if (event.buttons > 0) {
+          markUserScrollIntent();
+        }
+      },
+      [markUserScrollIntent],
+    );
+
+    const handleKeyDown = useCallback(
+      (event: KeyboardEvent<HTMLDivElement>) => {
+        if (SCROLL_KEYS.has(event.key)) {
+          markUserScrollIntent();
+        }
+      },
+      [markUserScrollIntent],
+    );
+
+    // Check if at bottom based on scroll position
+    const checkAtBottom = useCallback(() => {
+      const ref = virtuaRef.current;
+      if (!ref) return false;
+
+      const scrollOffset = ref.scrollOffset;
+      const scrollSize = ref.scrollSize;
+      const viewportSize = ref.viewportSize;
+
+      return scrollSize - scrollOffset - viewportSize <= AT_BOTTOM_THRESHOLD;
+    }, []);
+
+    // Handle scroll events
+    const handleScroll = useCallback(() => {
+      const refForActive = virtuaRef.current;
+      const activeFromFindRaw =
+        refForActive && typeof refForActive.findItemIndex === 'function'
+          ? refForActive.findItemIndex(refForActive.scrollOffset + refForActive.viewportSize * 0.25)
+          : null;
+      const activeFromFind =
+        typeof activeFromFindRaw === 'number' && activeFromFindRaw >= 0 ? activeFromFindRaw : null;
+
+      if (activeFromFind !== activeIndex) setActiveIndex(activeFromFind);
+
+      setScrollState({ isScrolling: true });
+
+      // Shrink spacer on scroll up when not streaming
+      const ref = virtuaRef.current;
+      if (ref) {
+        const hasUserScrollIntent =
+          Date.now() - lastUserScrollIntentAtRef.current <= USER_SCROLL_INTENT_TTL_MS;
+        onScrollOffset(ref.scrollOffset, hasUserScrollIntent);
      }
-    },
-    [markUserScrollIntent],
-  );

-  const handleKeyDown = useCallback(
-    (event: KeyboardEvent<HTMLDivElement>) => {
-      if (SCROLL_KEYS.has(event.key)) {
-        markUserScrollIntent();
+      // Check if at bottom
+      const isAtBottom = checkAtBottom();
+      setScrollState({ atBottom: isAtBottom });
+
+      if (ref) {
+        recordScroll(ref.scrollOffset, isAtBottom);
      }
-    },
-    [markUserScrollIntent],
-  );

-  // Check if at bottom based on scroll position
-  const checkAtBottom = useCallback(() => {
-    const ref = virtuaRef.current;
-    if (!ref) return false;
-
-    const scrollOffset = ref.scrollOffset;
-    const scrollSize = ref.scrollSize;
-    const viewportSize = ref.viewportSize;
-
-    return scrollSize - scrollOffset - viewportSize <= AT_BOTTOM_THRESHOLD;
-  }, []);
-
-  // Handle scroll events
-  const handleScroll = useCallback(() => {
-    const refForActive = virtuaRef.current;
-    const activeFromFindRaw =
-      refForActive && typeof refForActive.findItemIndex === 'function'
-        ? refForActive.findItemIndex(refForActive.scrollOffset + refForActive.viewportSize * 0.25)
-        : null;
-    const activeFromFind =
-      typeof activeFromFindRaw === 'number' && activeFromFindRaw >= 0 ? activeFromFindRaw : null;
-
-    if (activeFromFind !== activeIndex) setActiveIndex(activeFromFind);
-
-    setScrollState({ isScrolling: true });
-
-    // Shrink spacer on scroll up when not streaming
-    const ref = virtuaRef.current;
-    if (ref) {
-      const hasUserScrollIntent =
-        Date.now() - lastUserScrollIntentAtRef.current <= USER_SCROLL_INTENT_TTL_MS;
-      onScrollOffset(ref.scrollOffset, hasUserScrollIntent);
-    }
-
-    // Check if at bottom
-    const isAtBottom = checkAtBottom();
-    setScrollState({ atBottom: isAtBottom });
-
-    if (ref) {
-      recordScroll(ref.scrollOffset, isAtBottom);
-    }
-
-    // Clear existing timer
-    if (scrollEndTimerRef.current) {
-      clearTimeout(scrollEndTimerRef.current);
-    }
-
-    // Set new timer for scroll end
-    scrollEndTimerRef.current = setTimeout(() => {
-      setScrollState({ isScrolling: false });
-    }, 150);
-  }, [activeIndex, checkAtBottom, onScrollOffset, recordScroll, setActiveIndex, setScrollState]);
-
-  const handleScrollEnd = useCallback(() => {
-    setScrollState({ isScrolling: false });
-  }, [setScrollState]);
-
-  // Register scroll methods to store on mount
-  useEffect(() => {
-    const ref = virtuaRef.current;
-    if (ref) {
-      registerVirtuaScrollMethods({
-        getItemOffset: (index) => ref.getItemOffset(index),
-        getItemSize: (index) => ref.getItemSize(index),
-        getScrollOffset: () => ref.scrollOffset,
-        getScrollSize: () => ref.scrollSize,
-        getTotalCount: () => totalCountRef.current,
-        getViewportSize: () => ref.viewportSize,
-        scrollTo: (offset) => ref.scrollTo(offset),
-        scrollToIndex: (index, options) => ref.scrollToIndex(index, options),
-      });
-
-      // Seed active index once on mount (avoid requiring user scroll)
-      const initialActiveRaw = ref.findItemIndex(ref.scrollOffset + ref.viewportSize * 0.25);
-      const initialActive =
-        typeof initialActiveRaw === 'number' && initialActiveRaw >= 0 ? initialActiveRaw : null;
-      setActiveIndex(initialActive);
-    }
-
-    return () => {
-      registerVirtuaScrollMethods(null);
-    };
-  }, [registerVirtuaScrollMethods, setActiveIndex]);
-
-  // Cleanup on unmount
-  useEffect(() => {
-    return () => {
-      resetVisibleItems();
+      // Clear existing timer
      if (scrollEndTimerRef.current) {
        clearTimeout(scrollEndTimerRef.current);
      }
-    };
-  }, [resetVisibleItems]);

-  // Keep currently-streaming items mounted so vlist recycling never triggers
-  // Markdown animation replay when the user scrolls them back into view.
-  const streamingIndices = useConversationStore(
-    useShallow((s) => {
-      const indices: number[] = [];
+      // Set new timer for scroll end
+      scrollEndTimerRef.current = setTimeout(() => {
+        setScrollState({ isScrolling: false });
+      }, 150);
+    }, [activeIndex, checkAtBottom, onScrollOffset, recordScroll, setActiveIndex, setScrollState]);
+
+    const handleScrollEnd = useCallback(() => {
+      setScrollState({ isScrolling: false });
+    }, [setScrollState]);
+
+    // Register scroll methods to store on mount
+    useEffect(() => {
+      const ref = virtuaRef.current;
+      if (ref) {
+        registerVirtuaScrollMethods({
+          getItemOffset: (index) => ref.getItemOffset(index),
+          getItemSize: (index) => ref.getItemSize(index),
+          getScrollOffset: () => ref.scrollOffset,
+          getScrollSize: () => ref.scrollSize,
+          getTotalCount: () => totalCountRef.current,
+          getViewportSize: () => ref.viewportSize,
+          scrollTo: (offset) => ref.scrollTo(offset),
+          scrollToIndex: (index, options) => ref.scrollToIndex(index, options),
+        });
+
+        // Seed active index once on mount (avoid requiring user scroll)
+        const initialActiveRaw = ref.findItemIndex(ref.scrollOffset + ref.viewportSize * 0.25);
+        const initialActive =
+          typeof initialActiveRaw === 'number' && initialActiveRaw >= 0 ? initialActiveRaw : null;
+        setActiveIndex(initialActive);
+      }
+
+      return () => {
+        registerVirtuaScrollMethods(null);
+      };
+    }, [registerVirtuaScrollMethods, setActiveIndex]);
+
+    // Cleanup on unmount
+    useEffect(() => {
+      return () => {
+        resetVisibleItems();
+        if (scrollEndTimerRef.current) {
+          clearTimeout(scrollEndTimerRef.current);
+        }
+      };
+    }, [resetVisibleItems]);
+
+    // Keep currently-streaming items mounted so vlist recycling never triggers
+    // Markdown animation replay when the user scrolls them back into view.
+    const streamingIndices = useConversationStore(
+      useShallow((s) => {
+        const indices: number[] = [];
+        for (let i = 0; i < dataSource.length; i++) {
+          const id = dataSource[i];
+          if (!id) continue;
+          if (messageStateSelectors.isMessageGenerating(id)(s)) indices.push(i);
+        }
+        return indices;
+      }),
+    );
+
+    // Also keep items that host the active text selection — unmounting a node
+    // containing a Selection endpoint would silently drop the user's highlight.
+    const selectionMessageIds = useSelectionMessageIds();
+
+    const keepMountedIndices = useMemo(() => {
+      if (selectionMessageIds.size === 0) return streamingIndices;
+      const merged = new Set<number>(streamingIndices);
      for (let i = 0; i < dataSource.length; i++) {
        const id = dataSource[i];
-        if (!id) continue;
-        if (messageStateSelectors.isMessageGenerating(id)(s)) indices.push(i);
+        if (id && selectionMessageIds.has(id)) merged.add(i);
      }
-      return indices;
-    }),
-  );
+      if (merged.size === streamingIndices.length) return streamingIndices;
+      return [...merged].sort((a, b) => a - b);
+    }, [dataSource, streamingIndices, selectionMessageIds]);

-  // Also keep items that host the active text selection — unmounting a node
-  // containing a Selection endpoint would silently drop the user's highlight.
-  const selectionMessageIds = useSelectionMessageIds();
+    const atBottom = useConversationStore(virtuaListSelectors.atBottom);
+    const scrollToBottom = useConversationStore((s) => s.scrollToBottom);

-  const keepMountedIndices = useMemo(() => {
-    if (selectionMessageIds.size === 0) return streamingIndices;
-    const merged = new Set<number>(streamingIndices);
-    for (let i = 0; i < dataSource.length; i++) {
-      const id = dataSource[i];
-      if (id && selectionMessageIds.has(id)) merged.add(i);
-    }
-    if (merged.size === streamingIndices.length) return streamingIndices;
-    return [...merged].sort((a, b) => a - b);
-  }, [dataSource, streamingIndices, selectionMessageIds]);
+    // The ChatInput's floating overlay (TodoProgress + QueueTray) covers the
+    // bottom of this scroll viewport like a layer. Extend VList's internal
+    // padding-bottom by the overlay height so the last message can still be
+    // scrolled into view *above* the overlay; the +12 compensates for the
+    // ChatInput's `marginTop: -12` (skipScrollMarginWithList) so the last
+    // message lands exactly on the overlay's top edge.
+    const overlayHeight = useConversationStore(inputSelectors.chatInputOverlayHeight);
+    const paddingBottom = Math.max(24, overlayHeight + 12);

-  const atBottom = useConversationStore(virtuaListSelectors.atBottom);
-  const scrollToBottom = useConversationStore((s) => s.scrollToBottom);
+    const dataWithSlots = useMemo(
+      () => [
+        ...(headerSlot ? [CONVERSATION_HEADER_ID] : []),
+        ...listData,
+        ...(footerSlot ? [CONVERSATION_FOOTER_ID] : []),
+      ],
+      [footerSlot, headerSlot, listData],
+    );

-  // The ChatInput's floating overlay (TodoProgress + QueueTray) covers the
-  // bottom of this scroll viewport like a layer. Extend VList's internal
-  // padding-bottom by the overlay height so the last message can still be
-  // scrolled into view *above* the overlay; the +12 compensates for the
-  // ChatInput's `marginTop: -12` (skipScrollMarginWithList) so the last
-  // message lands exactly on the overlay's top edge.
-  const overlayHeight = useConversationStore(inputSelectors.chatInputOverlayHeight);
-  const paddingBottom = Math.max(24, overlayHeight + 12);
+    const keepMountedIndicesWithSlots = useMemo(
+      () => (headerSlot ? keepMountedIndices.map((index) => index + 1) : keepMountedIndices),
+      [headerSlot, keepMountedIndices],
+    );

-  const dataWithSlots = useMemo(
-    () => [
-      ...(headerSlot ? [CONVERSATION_HEADER_ID] : []),
-      ...listData,
-      ...(footerSlot ? [CONVERSATION_FOOTER_ID] : []),
-    ],
-    [footerSlot, headerSlot, listData],
-  );
+    // Mirror the latest data length into a ref so the scroll-methods registered
+    // once on mount can read the current total count (including spacer/footer)
+    // without re-registering on every render.
+    const totalCountRef = useRef(dataWithSlots.length);
+    totalCountRef.current = dataWithSlots.length;

-  const keepMountedIndicesWithSlots = useMemo(
-    () => (headerSlot ? keepMountedIndices.map((index) => index + 1) : keepMountedIndices),
-    [headerSlot, keepMountedIndices],
-  );
-
-  // Mirror the latest data length into a ref so the scroll-methods registered
-  // once on mount can read the current total count (including spacer/footer)
-  // without re-registering on every render.
-  const totalCountRef = useRef(dataWithSlots.length);
-  totalCountRef.current = dataWithSlots.length;
-
-  return (
-    <div
-      style={{ height: '100%', position: 'relative' }}
-      onKeyDownCapture={handleKeyDown}
-      onPointerDownCapture={markUserScrollIntent}
-      onPointerMoveCapture={handlePointerMove}
-      onTouchMoveCapture={markUserScrollIntent}
-      onWheelCapture={markUserScrollIntent}
-    >
-      {/* Debug Inspector - placed outside VList so it won't be recycled by the virtual list */}
-      {OPEN_DEV_INSPECTOR && <DebugInspector />}
-      <VList
-        bufferSize={typeof window !== 'undefined' ? window.innerHeight : 0}
-        data={dataWithSlots}
-        keepMounted={keepMountedIndicesWithSlots}
-        ref={virtuaRef}
-        style={{ height: '100%', overflowAnchor: 'none', paddingBottom }}
-        onScroll={handleScroll}
-        onScrollEnd={handleScrollEnd}
+    return (
+      <div
+        style={{ height: '100%', position: 'relative' }}
+        onKeyDownCapture={handleKeyDown}
+        onPointerDownCapture={markUserScrollIntent}
+        onPointerMoveCapture={handlePointerMove}
+        onTouchMoveCapture={markUserScrollIntent}
+        onWheelCapture={markUserScrollIntent}
      >
-        {(messageId, index): ReactElement => {
-          if (messageId === CONVERSATION_HEADER_ID) {
-            return (
-              <WideScreenContainer key={messageId} style={{ position: 'relative' }}>
-                {headerSlot}
-              </WideScreenContainer>
-            );
-          }
-          if (messageId === CONVERSATION_FOOTER_ID) {
-            return (
-              <WideScreenContainer key={messageId} style={{ position: 'relative' }}>
-                {footerSlot}
-              </WideScreenContainer>
-            );
-          }
-          if (isSpacerMessage(messageId)) {
-            // Only animate the collapse-to-zero (unmount). Any non-zero height
-            // change (initial mount, shrink as assistant grows) is applied
-            // instantly so virtua's scrollSize updates in a single frame and
-            // scrollToIndex can reach the user message without trailing behind
-            // a 200ms transition.
-            const shouldAnimate = !isScrollShrinking && spacerHeight === 0;
-            return (
-              <WideScreenContainer key={messageId} style={{ position: 'relative' }}>
-                <div
-                  aria-hidden
-                  ref={registerSpacerNode}
-                  style={{
-                    height: spacerHeight,
-                    pointerEvents: 'none',
-                    transition: shouldAnimate
-                      ? `height ${CONVERSATION_SPACER_TRANSITION_MS}ms ease`
-                      : 'none',
-                    width: '100%',
-                  }}
-                />
-              </WideScreenContainer>
-            );
-          }
+        {/* Debug Inspector - placed outside VList so it won't be recycled by the virtual list */}
+        {OPEN_DEV_INSPECTOR && <DebugInspector />}
+        <VList
+          bufferSize={typeof window !== 'undefined' ? window.innerHeight : 0}
+          data={dataWithSlots}
+          keepMounted={keepMountedIndicesWithSlots}
+          ref={virtuaRef}
+          style={{ height: '100%', overflowAnchor: 'none', paddingBottom }}
+          onScroll={handleScroll}
+          onScrollEnd={handleScrollEnd}
+        >
+          {(messageId, index): ReactElement => {
+            if (messageId === CONVERSATION_HEADER_ID) {
+              return (
+                <WideScreenContainer key={messageId} style={{ position: 'relative' }}>
+                  {headerSlot}
+                </WideScreenContainer>
+              );
+            }
+            if (messageId === CONVERSATION_FOOTER_ID) {
+              return (
+                <WideScreenContainer key={messageId} style={{ position: 'relative' }}>
+                  {footerSlot}
+                </WideScreenContainer>
+              );
+            }
+            if (isSpacerMessage(messageId)) {
+              // Only animate the collapse-to-zero (unmount). Any non-zero height
+              // change (initial mount, shrink as assistant grows) is applied
+              // instantly so virtua's scrollSize updates in a single frame and
+              // scrollToIndex can reach the user message without trailing behind
+              // a 200ms transition.
+              const shouldAnimate = !isScrollShrinking && spacerHeight === 0;
+              return (
+                <WideScreenContainer key={messageId} style={{ position: 'relative' }}>
+                  <div
+                    aria-hidden
+                    ref={registerSpacerNode}
+                    style={{
+                      height: spacerHeight,
+                      pointerEvents: 'none',
+                      transition: shouldAnimate
+                        ? `height ${CONVERSATION_SPACER_TRANSITION_MS}ms ease`
+                        : 'none',
+                      width: '100%',
+                    }}
+                  />
+                </WideScreenContainer>
+              );
+            }

-          const isAgentCouncil = messageId.includes('agentCouncil');
-          const messageIndex = headerSlot ? index - 1 : index;
-          const isLastItem = messageIndex === dataSource.length - 1;
-          const content = itemContent(messageIndex, messageId);
+            const isAgentCouncil = messageId.includes('agentCouncil');
+            const messageIndex = headerSlot ? index - 1 : index;
+            const isLastItem = messageIndex === dataSource.length - 1;
+            const content = itemContent(messageIndex, messageId);
+
+            if (isAgentCouncil) {
+              // AgentCouncil needs full width for horizontal scroll
+              return (
+                <div key={messageId} style={{ position: 'relative', width: '100%' }}>
+                  {content}
+                  {/* AutoScroll is placed inside the last Item so it only triggers when the last Item is visible */}
+                  {isLastItem && isAutoScrollEnabled && !spacerActive && <AutoScroll />}
+                </div>
+              );
+            }

-          if (isAgentCouncil) {
-            // AgentCouncil needs full width for horizontal scroll
            return (
-              <div key={messageId} style={{ position: 'relative', width: '100%' }}>
+              <WideScreenContainer key={messageId} style={{ position: 'relative' }}>
                {content}
-                {/* AutoScroll is placed inside the last Item so it only triggers when the last Item is visible */}
                {isLastItem && isAutoScrollEnabled && !spacerActive && <AutoScroll />}
-              </div>
+              </WideScreenContainer>
            );
-          }
-
-          return (
-            <WideScreenContainer key={messageId} style={{ position: 'relative' }}>
-              {content}
-              {isLastItem && isAutoScrollEnabled && !spacerActive && <AutoScroll />}
-            </WideScreenContainer>
-          );
-        }}
-      </VList>
-      {/* BackBottom is placed outside VList so it remains visible regardless of scroll position */}
-      <WideScreenContainer style={{ position: 'relative' }}>
-        <BackBottom
-          atBottom={atBottom}
-          bottomOffset={overlayHeight}
-          visible={!atBottom}
-          onScrollToBottom={() => scrollToBottom(true)}
-        />
-      </WideScreenContainer>
-    </div>
-  );
-}, isEqual);
+          }}
+        </VList>
+        {/* BackBottom is placed outside VList so it remains visible regardless of scroll position */}
+        <WideScreenContainer style={{ position: 'relative' }}>
+          <BackBottom
+            atBottom={atBottom}
+            bottomOffset={overlayHeight}
+            visible={!atBottom}
+            onScrollToBottom={() => scrollToBottom(true)}
+          />
+        </WideScreenContainer>
+      </div>
+    );
+  },
+  isEqual,
+);

 VirtualizedList.displayName = 'ConversationVirtualizedList';

@@ -8,6 +8,7 @@ import { memo, useMemo } from 'react';

 import { messageMapKey } from '@/store/chat/utils/messageMapKey';

+import AssistantTurnSettledWatcher from './AssistantTurnSettledWatcher';
 import { createStore, Provider } from './store';
 import StoreUpdater from './StoreUpdater';
 import {
@@ -103,6 +104,7 @@ export const ConversationProvider = memo<ConversationProviderProps>(
          skipFetch={skipFetch}
          onMessagesChange={onMessagesChange}
        />
+        <AssistantTurnSettledWatcher />
        {children}
      </Provider>
    );
@@ -21,8 +21,8 @@ vi.hoisted(() => {
 const MSG = 'msg-1';
 const OTHER_MSG = 'msg-2';
 const CHILD_MSG = 'msg-1-child-answer';
-const TOPIC = 'topic-1';
-const OTHER_TOPIC = 'topic-2';
+const KEY = 'main_agent-a_topic-1';
+const OTHER_KEY = 'main_agent-a_topic-2';

 const updateInputMessageMock = vi.fn();
 const editorSetDocumentMock = vi.fn();
@@ -58,89 +58,109 @@ describe('<FollowUpChips />', () => {

  it('renders nothing when status is not ready', () => {
    useFollowUpActionStore.setState({
-      chips: [{ label: 'x', message: 'x' }],
-      messageId: MSG,
-      status: 'loading',
-      topicId: TOPIC,
+      slots: {
+        [KEY]: {
+          chips: [{ label: 'x', message: 'x' }],
+          messageId: MSG,
+          status: 'loading',
+        },
+      },
    });
-    const { container } = render(<FollowUpChips messageId={MSG} topicId={TOPIC} />);
+    const { container } = render(<FollowUpChips conversationKey={KEY} messageId={MSG} />);
    expect(container).toBeEmptyDOMElement();
  });

  it('renders nothing when messageId mismatches and is not a child id', () => {
    useFollowUpActionStore.setState({
-      chips: [{ label: 'x', message: 'x' }],
-      messageId: OTHER_MSG,
-      status: 'ready',
-      topicId: TOPIC,
+      slots: {
+        [KEY]: {
+          chips: [{ label: 'x', message: 'x' }],
+          messageId: OTHER_MSG,
+          status: 'ready',
+        },
+      },
    });
-    const { container } = render(<FollowUpChips messageId={MSG} topicId={TOPIC} />);
+    const { container } = render(<FollowUpChips conversationKey={KEY} messageId={MSG} />);
    expect(container).toBeEmptyDOMElement();
  });

-  it('renders nothing when topicId mismatches', () => {
+  it('renders nothing when conversationKey mismatches', () => {
    useFollowUpActionStore.setState({
-      chips: [{ label: 'a', message: 'a' }],
-      messageId: MSG,
-      status: 'ready',
-      topicId: TOPIC,
+      slots: {
+        [KEY]: {
+          chips: [{ label: 'a', message: 'a' }],
+          messageId: MSG,
+          status: 'ready',
+        },
+      },
    });
-    const { container } = render(<FollowUpChips messageId={MSG} topicId={OTHER_TOPIC} />);
+    const { container } = render(<FollowUpChips conversationKey={OTHER_KEY} messageId={MSG} />);
    expect(container).toBeEmptyDOMElement();
  });

  it('renders nothing while the bound message is generating', () => {
    isGeneratingMock = true;
    useFollowUpActionStore.setState({
-      chips: [{ label: 'a', message: 'a' }],
-      messageId: MSG,
-      status: 'ready',
-      topicId: TOPIC,
+      slots: {
+        [KEY]: {
+          chips: [{ label: 'a', message: 'a' }],
+          messageId: MSG,
+          status: 'ready',
+        },
+      },
    });
-    const { container } = render(<FollowUpChips messageId={MSG} topicId={TOPIC} />);
+    const { container } = render(<FollowUpChips conversationKey={KEY} messageId={MSG} />);
    expect(container).toBeEmptyDOMElement();
  });

  it('renders one button per chip when both ids match and not generating', () => {
    useFollowUpActionStore.setState({
-      chips: [
-        { label: 'a', message: 'a' },
-        { label: 'b', message: 'b' },
-        { label: 'c', message: 'c' },
-      ],
-      messageId: MSG,
-      status: 'ready',
-      topicId: TOPIC,
+      slots: {
+        [KEY]: {
+          chips: [
+            { label: 'a', message: 'a' },
+            { label: 'b', message: 'b' },
+            { label: 'c', message: 'c' },
+          ],
+          messageId: MSG,
+          status: 'ready',
+        },
+      },
    });
-    render(<FollowUpChips messageId={MSG} topicId={TOPIC} />);
+    render(<FollowUpChips conversationKey={KEY} messageId={MSG} />);
    expect(screen.getAllByRole('button')).toHaveLength(3);
  });

  it('renders chips when the stored messageId matches a child block id of the bound group', () => {
    displayMessagesMock = [{ children: [{ id: CHILD_MSG }], id: MSG }];
    useFollowUpActionStore.setState({
-      chips: [{ label: 'a', message: 'a' }],
-      messageId: CHILD_MSG,
-      status: 'ready',
-      topicId: TOPIC,
+      slots: {
+        [KEY]: {
+          chips: [{ label: 'a', message: 'a' }],
+          messageId: CHILD_MSG,
+          status: 'ready',
+        },
+      },
    });
-    render(<FollowUpChips messageId={MSG} topicId={TOPIC} />);
+    render(<FollowUpChips conversationKey={KEY} messageId={MSG} />);
    expect(screen.getAllByRole('button')).toHaveLength(1);
  });

  it('fills the input and consumes on click instead of sending', () => {
    useFollowUpActionStore.setState({
-      chips: [{ label: 'go', message: 'go ahead' }],
-      messageId: MSG,
-      status: 'ready',
-      topicId: TOPIC,
+      slots: {
+        [KEY]: {
+          chips: [{ label: 'go', message: 'go ahead' }],
+          messageId: MSG,
+          status: 'ready',
+        },
+      },
    });
-    render(<FollowUpChips messageId={MSG} topicId={TOPIC} />);
+    render(<FollowUpChips conversationKey={KEY} messageId={MSG} />);
    fireEvent.click(screen.getByRole('button', { name: 'go' }));
    expect(updateInputMessageMock).toHaveBeenCalledWith('go ahead');
    expect(editorSetDocumentMock).toHaveBeenCalledWith('text', 'go ahead');
    expect(editorFocusMock).toHaveBeenCalled();
-    // The chip is not consumed on click — it stays ready until the user sends.
-    expect(useFollowUpActionStore.getState().status).toBe('ready');
+    expect(useFollowUpActionStore.getState().slots[KEY]?.status).toBe('ready');
  });
 });
@@ -11,27 +11,22 @@ import { messageStateSelectors } from '../store';
 import { styles } from './style';

 interface FollowUpChipsProps {
+  conversationKey: string;
  messageId: string;
-  topicId: string;
 }

-const FollowUpChips = memo<FollowUpChipsProps>(({ messageId, topicId }) => {
-  // For assistantGroup, the server resolves the latest answer message id which
-  // lives inside `children`, not as the top-level group id. Collect children ids
-  // as a stable primitive so the followUpAction selector can match either.
+const FollowUpChips = memo<FollowUpChipsProps>(({ conversationKey, messageId }) => {
  const childIdsKey = useConversationStore((s) => {
    const m = s.displayMessages.find((x) => x.id === messageId);
    return m?.children?.map((c) => c.id).join('|') ?? '';
  });
  const selector = useMemo(
-    () => followUpActionSelectors.chipsFor({ childIdsKey, messageId, topicId }),
-    [childIdsKey, messageId, topicId],
+    () => followUpActionSelectors.chipsFor({ childIdsKey, conversationKey, messageId }),
+    [childIdsKey, conversationKey, messageId],
  );
  const chips = useFollowUpActionStore(selector);
  const updateInputMessage = useConversationStore((s) => s.updateInputMessage);
  const editor = useConversationStore((s) => s.editor);
-  // Hide chips while the bound group/message is still being generated — chips
-  // are only valid for a fully settled assistant turn.
  const isGenerating = useConversationStore(
    messageStateSelectors.isAssistantGroupItemGenerating(messageId),
  );
@@ -0,0 +1,233 @@
+import { type LobeAgentChatConfig } from '@lobechat/types';
+import { renderHook } from '@testing-library/react';
+import { afterEach, beforeEach, describe, expect, it, vi } from 'vitest';
+
+import { useFollowUpActionStore } from '@/store/followUpAction';
+import { useUserStore } from '@/store/user';
+
+import { useChatFollowUp } from './useChatFollowUp';
+
+const VALID_GLOBAL = {
+  enabled: true,
+  model: 'global-model',
+  provider: 'global-provider',
+};
+
+const VALID_AGENT: LobeAgentChatConfig = {
+  enableFollowUpChips: true,
+} as LobeAgentChatConfig;
+
+const CONVERSATION_KEY = 'main_agent-1_topic-1';
+const TOPIC_ID = 'topic-1';
+
+type Mock = ReturnType<typeof vi.fn>;
+
+vi.mock('@/store/user', () => ({
+  useUserStore: vi.fn(),
+}));
+
+describe('useChatFollowUp', () => {
+  let fetchFor: ReturnType<typeof vi.fn>;
+  let clear: ReturnType<typeof vi.fn>;
+
+  beforeEach(() => {
+    fetchFor = vi.fn().mockResolvedValue(undefined);
+    clear = vi.fn();
+
+    vi.spyOn(useFollowUpActionStore, 'getState').mockReturnValue({
+      fetchFor,
+      clear,
+    } as any);
+
+    (useUserStore as unknown as Mock).mockImplementation((selector: any) =>
+      selector({
+        settings: { systemAgent: { followUpAction: VALID_GLOBAL } },
+      }),
+    );
+  });
+
+  afterEach(() => {
+    vi.restoreAllMocks();
+    (useUserStore as unknown as Mock).mockReset();
+  });
+
+  describe('disabled effective state — returns empty hooks', () => {
+    const renderWith = (config: {
+      agentChatConfig?: LobeAgentChatConfig;
+      conversationKey?: string;
+      topicId?: string;
+    }) =>
+      renderHook(() =>
+        useChatFollowUp({
+          agentChatConfig: config.agentChatConfig,
+          conversationKey: config.conversationKey,
+          topicId: config.topicId,
+        }),
+      );
+
+    const assertEmpty = (hooks: ReturnType<typeof useChatFollowUp>) => {
+      expect(hooks.onBeforeSendMessage).toBeUndefined();
+      expect(hooks.onBeforeContinue).toBeUndefined();
+      expect(hooks.onBeforeRegenerate).toBeUndefined();
+      expect(hooks.onAssistantTurnSettled).toBeUndefined();
+    };
+
+    it('when global enabled = false', () => {
+      (useUserStore as unknown as Mock).mockImplementation((selector: any) =>
+        selector({
+          settings: {
+            systemAgent: { followUpAction: { ...VALID_GLOBAL, enabled: false } },
+          },
+        }),
+      );
+      const { result } = renderWith({
+        agentChatConfig: VALID_AGENT,
+        conversationKey: CONVERSATION_KEY,
+        topicId: TOPIC_ID,
+      });
+      assertEmpty(result.current);
+    });
+
+    it('when global model is empty', () => {
+      (useUserStore as unknown as Mock).mockImplementation((selector: any) =>
+        selector({
+          settings: {
+            systemAgent: { followUpAction: { ...VALID_GLOBAL, model: '' } },
+          },
+        }),
+      );
+      const { result } = renderWith({
+        agentChatConfig: VALID_AGENT,
+        conversationKey: CONVERSATION_KEY,
+        topicId: TOPIC_ID,
+      });
+      assertEmpty(result.current);
+    });
+
+    it('when global provider is empty', () => {
+      (useUserStore as unknown as Mock).mockImplementation((selector: any) =>
+        selector({
+          settings: {
+            systemAgent: { followUpAction: { ...VALID_GLOBAL, provider: '' } },
+          },
+        }),
+      );
+      const { result } = renderWith({
+        agentChatConfig: VALID_AGENT,
+        conversationKey: CONVERSATION_KEY,
+        topicId: TOPIC_ID,
+      });
+      assertEmpty(result.current);
+    });
+
+    it('when per-agent enableFollowUpChips is false', () => {
+      const { result } = renderWith({
+        agentChatConfig: { enableFollowUpChips: false } as LobeAgentChatConfig,
+        conversationKey: CONVERSATION_KEY,
+        topicId: TOPIC_ID,
+      });
+      assertEmpty(result.current);
+    });
+
+    it('when conversationKey is missing', () => {
+      const { result } = renderWith({
+        agentChatConfig: VALID_AGENT,
+        conversationKey: undefined,
+        topicId: TOPIC_ID,
+      });
+      assertEmpty(result.current);
+    });
+
+    it('when topicId is missing', () => {
+      const { result } = renderWith({
+        agentChatConfig: VALID_AGENT,
+        conversationKey: CONVERSATION_KEY,
+        topicId: undefined,
+      });
+      assertEmpty(result.current);
+    });
+
+    it('does not call clear or fetchFor when invoked through empty hooks', async () => {
+      const { result } = renderWith({
+        agentChatConfig: { enableFollowUpChips: false } as LobeAgentChatConfig,
+        conversationKey: CONVERSATION_KEY,
+        topicId: TOPIC_ID,
+      });
+      await result.current.onBeforeSendMessage?.({} as any);
+      await result.current.onAssistantTurnSettled?.('m', { reason: 'completed' });
+      expect(clear).not.toHaveBeenCalled();
+      expect(fetchFor).not.toHaveBeenCalled();
+    });
+  });
+
+  describe('enabled effective state', () => {
+    const renderEnabled = (overrides: Partial<Parameters<typeof useChatFollowUp>[0]> = {}) =>
+      renderHook(() =>
+        useChatFollowUp({
+          agentChatConfig: VALID_AGENT,
+          conversationKey: CONVERSATION_KEY,
+          topicId: TOPIC_ID,
+          ...overrides,
+        }),
+      );
+
+    it('onBeforeSendMessage clears the slot', async () => {
+      const { result } = renderEnabled();
+      await result.current.onBeforeSendMessage?.({} as any);
+      expect(clear).toHaveBeenCalledWith(CONVERSATION_KEY);
+    });
+
+    it('onBeforeContinue clears the slot', async () => {
+      const { result } = renderEnabled();
+      await result.current.onBeforeContinue?.('m');
+      expect(clear).toHaveBeenCalledWith(CONVERSATION_KEY);
+    });
+
+    it('onBeforeRegenerate clears the slot', async () => {
+      const { result } = renderEnabled();
+      await result.current.onBeforeRegenerate?.('m');
+      expect(clear).toHaveBeenCalledWith(CONVERSATION_KEY);
+    });
+
+    it('onAssistantTurnSettled with reason=stopped skips fetch', async () => {
+      const { result } = renderEnabled();
+      await result.current.onAssistantTurnSettled?.('m', { reason: 'stopped' });
+      expect(fetchFor).not.toHaveBeenCalled();
+    });
+
+    it('onAssistantTurnSettled with reason=completed fires fetchFor with full params', async () => {
+      const { result } = renderEnabled({ threadId: 'thread-1' });
+      await result.current.onAssistantTurnSettled?.('m', { reason: 'completed' });
+      expect(fetchFor).toHaveBeenCalledWith(CONVERSATION_KEY, {
+        hint: { kind: 'chat' },
+        modelConfig: { model: VALID_GLOBAL.model, provider: VALID_GLOBAL.provider },
+        threadId: 'thread-1',
+        topicId: TOPIC_ID,
+      });
+    });
+
+    it('onAssistantTurnSettled with reason=regenerated fires fetchFor', async () => {
+      const { result } = renderEnabled();
+      await result.current.onAssistantTurnSettled?.('m', { reason: 'regenerated' });
+      expect(fetchFor).toHaveBeenCalledTimes(1);
+    });
+
+    it('onAssistantTurnSettled with reason=continued fires fetchFor', async () => {
+      const { result } = renderEnabled();
+      await result.current.onAssistantTurnSettled?.('m', { reason: 'continued' });
+      expect(fetchFor).toHaveBeenCalledTimes(1);
+    });
+
+    it('clear is scoped to the passed conversationKey — different keys do not collide', async () => {
+      const { result: a } = renderEnabled({ conversationKey: 'key-a' });
+      const { result: b } = renderEnabled({ conversationKey: 'key-b' });
+
+      await a.current.onBeforeSendMessage?.({} as any);
+      expect(clear).toHaveBeenCalledWith('key-a');
+      expect(clear).not.toHaveBeenCalledWith('key-b');
+
+      await b.current.onBeforeSendMessage?.({} as any);
+      expect(clear).toHaveBeenCalledWith('key-b');
+    });
+  });
+});
@@ -0,0 +1,77 @@
+import { type LobeAgentChatConfig } from '@lobechat/types';
+import { useMemo } from 'react';
+
+import { useFollowUpActionStore } from '@/store/followUpAction';
+import { useUserStore } from '@/store/user';
+import { systemAgentSelectors } from '@/store/user/slices/settings/selectors/systemAgent';
+
+import { type ConversationHooks } from '../types';
+
+interface UseChatFollowUpParams {
+  agentChatConfig: LobeAgentChatConfig | undefined;
+  conversationKey: string | undefined;
+  threadId?: string;
+  topicId: string | undefined;
+}
+
+/**
+ * Wire the chat-side Follow-up Chips lifecycle.
+ *
+ * Effective enable = `systemAgent.followUpAction.enabled` AND a valid global
+ * model/provider AND per-agent `chatConfig.enableFollowUpChips` — otherwise
+ * returns an empty `ConversationHooks` object so the merge chain treats it as
+ * identity.
+ *
+ * Registration ordering note: callers MUST compose this hook LAST in a
+ * `mergeConversationHooks(...)` chain. The hook's
+ * `onBeforeSendMessage`/`onBeforeContinue`/`onBeforeRegenerate` clear the chip
+ * slot; if a preceding validator returns `false`, the chain short-circuits
+ * before the clear runs and chips persist for the blocked send.
+ */
+export const useChatFollowUp = ({
+  agentChatConfig,
+  conversationKey,
+  threadId,
+  topicId,
+}: UseChatFollowUpParams): ConversationHooks => {
+  const globalConfig = useUserStore(systemAgentSelectors.followUpAction);
+
+  const effective = useMemo(() => {
+    const globalEnabled = globalConfig.enabled === true;
+    const hasValidModel = !!globalConfig.model && !!globalConfig.provider;
+    const perAgentEnabled = agentChatConfig?.enableFollowUpChips === true;
+    return globalEnabled && hasValidModel && perAgentEnabled;
+  }, [
+    globalConfig.enabled,
+    globalConfig.model,
+    globalConfig.provider,
+    agentChatConfig?.enableFollowUpChips,
+  ]);
+
+  return useMemo<ConversationHooks>(() => {
+    if (!effective || !conversationKey || !topicId) return {};
+
+    const clearSlot = () => useFollowUpActionStore.getState().clear(conversationKey);
+
+    return {
+      onAssistantTurnSettled: async (_messageId, { reason }) => {
+        if (reason === 'stopped') return;
+        await useFollowUpActionStore.getState().fetchFor(conversationKey, {
+          hint: { kind: 'chat' },
+          modelConfig: { model: globalConfig.model, provider: globalConfig.provider },
+          threadId,
+          topicId,
+        });
+      },
+      onBeforeContinue: async () => {
+        clearSlot();
+      },
+      onBeforeRegenerate: async () => {
+        clearSlot();
+      },
+      onBeforeSendMessage: async () => {
+        clearSlot();
+      },
+    };
+  }, [effective, conversationKey, globalConfig.model, globalConfig.provider, threadId, topicId]);
+};
@@ -1,3 +1,5 @@
+import { messageMapKey } from '@/store/chat/utils/messageMapKey';
+
 import { type State } from '../../initialState';

 const context = (s: State) => s.context;
@@ -7,12 +9,26 @@ const threadId = (s: State) => s.context.threadId;
 const isThread = (s: State) => !!s.context.threadId;
 const isTopic = (s: State) => !!s.context.topicId;

+const conversationKey = (s: State): string => {
+  const { agentId, topicId, threadId, scope, isNew, groupId, subAgentId } = s.context;
+  return messageMapKey({
+    agentId,
+    groupId,
+    isNew,
+    scope,
+    subAgentId,
+    threadId: threadId ?? undefined,
+    topicId: topicId ?? undefined,
+  });
+};
+
 const hooks = (s: State) => s.hooks;
 const hook = (hookName: keyof State['hooks']) => (s: State) => s.hooks[hookName];

 export const contextSelectors = {
  agentId,
  context,
+  conversationKey,
  hook,
  hooks,
  isThread,
@@ -69,6 +69,26 @@ export interface ConversationHooks {
   */
  onAfterSendMessage?: () => Promise<void>;

+  /**
+   * Fires once per assistant turn after streaming ends and any pending tool
+   * intervention has cleared. Use for post-turn side effects (e.g. extracting
+   * follow-up chips, syncing onboarding phase).
+   *
+   * Overlap: `onGenerationStop` also fires when the user stops generation;
+   * `onAssistantTurnSettled` will additionally fire in that case with
+   * `reason: 'stopped'`. Consumers tracking stop events should pick one and
+   * avoid double-counting.
+   *
+   * Registration order: consumers should be registered LAST in any hook-merge
+   * chain — the settlement handler may clear caches that depend on prior
+   * validators short-circuiting (e.g. a `false`-returning `onBeforeSendMessage`
+   * upstream must short-circuit before the chip slot is cleared).
+   */
+  onAssistantTurnSettled?: (
+    messageId: string,
+    meta: { reason: 'completed' | 'stopped' | 'regenerated' | 'continued' },
+  ) => Promise<unknown> | void;
+
  /**
   * Called before continuing generation
   *
@@ -103,6 +123,10 @@ export interface ConversationHooks {
   */
  onBeforeSendMessage?: (params: SendMessageParams) => Promise<boolean | void>;

+  // ========================================
+  // Generation State Change Hooks
+  // ========================================
+
  /**
   * Called after continue generation completes
   *
@@ -110,31 +134,6 @@ export interface ConversationHooks {
   */
  onContinueComplete?: (messageId: string) => void;

-  // ========================================
-  // Generation State Change Hooks
-  // ========================================
-
-  /**
-   * Called when AI generation is cancelled
-   *
-   * @param operationId - The operation ID for this generation
-   */
-  onGenerationCancelled?: (operationId: string) => void;
-
-  /**
-   * Called when AI generation completes successfully
-   *
-   * @param operationId - The operation ID for this generation
-   */
-  onGenerationComplete?: (operationId: string) => void;
-
-  /**
-   * Called when AI generation starts
-   *
-   * @param operationId - The operation ID for this generation
-   */
-  onGenerationStart?: (operationId: string) => void;
-
  /**
   * Called when generation is stopped by user
   */
@@ -0,0 +1,103 @@
+import { describe, expect, it, vi } from 'vitest';
+
+import { type ConversationHooks } from '../types';
+import { mergeConversationHooks } from './mergeConversationHooks';
+
+describe('mergeConversationHooks', () => {
+  it('returns empty object when no hooks supplied', () => {
+    expect(mergeConversationHooks()).toEqual({});
+    expect(mergeConversationHooks(undefined)).toEqual({});
+    expect(mergeConversationHooks(undefined, undefined)).toEqual({});
+  });
+
+  it('returns the single hook object when only one is defined', () => {
+    const onMessageCopied = vi.fn();
+    const merged = mergeConversationHooks({ onMessageCopied });
+    expect(merged.onMessageCopied).toBe(onMessageCopied);
+  });
+
+  it('invokes hooks of disjoint members from both inputs', async () => {
+    const onMessageCopied = vi.fn();
+    const onAfterSendMessage = vi.fn();
+
+    const merged = mergeConversationHooks({ onMessageCopied }, { onAfterSendMessage });
+
+    merged.onMessageCopied?.('m-1');
+    await merged.onAfterSendMessage?.();
+
+    expect(onMessageCopied).toHaveBeenCalledWith('m-1');
+    expect(onAfterSendMessage).toHaveBeenCalledTimes(1);
+  });
+
+  it('invokes both onAssistantTurnSettled handlers in order with same args', async () => {
+    const order: string[] = [];
+    const a = vi.fn(async () => {
+      order.push('a');
+    });
+    const b = vi.fn(async () => {
+      order.push('b');
+    });
+
+    const merged = mergeConversationHooks(
+      { onAssistantTurnSettled: a },
+      { onAssistantTurnSettled: b },
+    );
+
+    await merged.onAssistantTurnSettled?.('msg-1', { reason: 'completed' });
+
+    expect(a).toHaveBeenCalledWith('msg-1', { reason: 'completed' });
+    expect(b).toHaveBeenCalledWith('msg-1', { reason: 'completed' });
+    expect(order).toEqual(['a', 'b']);
+  });
+
+  it('short-circuits onBeforeSendMessage chain on first false', async () => {
+    const first = vi.fn(async () => undefined);
+    const second = vi.fn(async () => false);
+    const third = vi.fn(async () => undefined);
+
+    const merged = mergeConversationHooks(
+      { onBeforeSendMessage: first as ConversationHooks['onBeforeSendMessage'] },
+      { onBeforeSendMessage: second as ConversationHooks['onBeforeSendMessage'] },
+      { onBeforeSendMessage: third as ConversationHooks['onBeforeSendMessage'] },
+    );
+
+    const result = await merged.onBeforeSendMessage?.({ message: 'hi' } as any);
+
+    expect(result).toBe(false);
+    expect(first).toHaveBeenCalledTimes(1);
+    expect(second).toHaveBeenCalledTimes(1);
+    expect(third).not.toHaveBeenCalled();
+  });
+
+  it('continues the chain when before-hooks return undefined / true', async () => {
+    const first = vi.fn(async () => undefined);
+    const second = vi.fn(async () => true);
+    const third = vi.fn(async () => undefined);
+
+    const merged = mergeConversationHooks(
+      { onBeforeRegenerate: first as ConversationHooks['onBeforeRegenerate'] },
+      { onBeforeRegenerate: second as ConversationHooks['onBeforeRegenerate'] },
+      { onBeforeRegenerate: third as ConversationHooks['onBeforeRegenerate'] },
+    );
+
+    const result = await merged.onBeforeRegenerate?.('msg-1');
+
+    expect(result).toBeUndefined();
+    expect(first).toHaveBeenCalled();
+    expect(second).toHaveBeenCalled();
+    expect(third).toHaveBeenCalled();
+  });
+
+  it('skips undefined entries in the input array', async () => {
+    const onAfterSendMessage = vi.fn();
+    const merged = mergeConversationHooks(undefined, { onAfterSendMessage }, undefined);
+    await merged.onAfterSendMessage?.();
+    expect(onAfterSendMessage).toHaveBeenCalledTimes(1);
+  });
+
+  it('treats absent fields as no-ops without throwing', async () => {
+    const merged = mergeConversationHooks({}, {});
+    expect(merged.onAssistantTurnSettled).toBeUndefined();
+    expect(merged.onBeforeSendMessage).toBeUndefined();
+  });
+});
@@ -0,0 +1,52 @@
+import { type ConversationHooks } from '../types';
+
+const BLOCKING_HOOK_KEYS = new Set([
+  'onBeforeContinue',
+  'onBeforeRegenerate',
+  'onBeforeSendMessage',
+]);
+
+const collectHookNames = (hooks: ConversationHooks[]): Set<string> => {
+  const names = new Set<string>();
+  for (const h of hooks) {
+    for (const key of Object.keys(h)) {
+      if (typeof h[key] === 'function') names.add(key);
+    }
+  }
+  return names;
+};
+
+export const mergeConversationHooks = (
+  ...hooks: (ConversationHooks | undefined)[]
+): ConversationHooks => {
+  const defined = hooks.filter((h): h is ConversationHooks => !!h);
+  if (defined.length === 0) return {};
+  if (defined.length === 1) return defined[0];
+
+  const merged: ConversationHooks = {};
+  const names = collectHookNames(defined);
+
+  for (const name of names) {
+    if (BLOCKING_HOOK_KEYS.has(name)) {
+      merged[name] = async (...args: unknown[]) => {
+        for (const h of defined) {
+          const fn = h[name];
+          if (typeof fn !== 'function') continue;
+          const result = await fn(...args);
+          if (result === false) return false;
+        }
+        return undefined;
+      };
+    } else {
+      merged[name] = async (...args: unknown[]) => {
+        for (const h of defined) {
+          const fn = h[name];
+          if (typeof fn !== 'function') continue;
+          await fn(...args);
+        }
+      };
+    }
+  }
+
+  return merged;
+};
@@ -11,9 +11,13 @@ import {
  type ConversationHooks,
  ConversationProvider,
 } from '@/features/Conversation';
+import { useChatFollowUp } from '@/features/Conversation/hooks/useChatFollowUp';
 import { type ConversationContext } from '@/features/Conversation/types';
+import { mergeConversationHooks } from '@/features/Conversation/utils/mergeConversationHooks';
 import { useOperationState } from '@/hooks/useOperationState';
 import { useActionsBarConfig } from '@/routes/(main)/agent/features/Conversation/useActionsBarConfig';
+import { useAgentStore } from '@/store/agent';
+import { chatConfigByIdSelectors } from '@/store/agent/selectors';
 import { useChatStore } from '@/store/chat';
 import { messageMapKey } from '@/store/chat/utils/messageMapKey';

@@ -158,17 +162,28 @@ const FloatingChatPanel = memo<FloatingChatPanelProps>(
    const [open, setOpen] = useState(true);
    const [activeSnapPoint, setActiveSnapPoint] = useState<number>(REST_SNAP_POINT);

+    const agentChatConfig = useAgentStore(chatConfigByIdSelectors.getChatConfigById(agentId));
+    const chatFollowUpHooks = useChatFollowUp({
+      agentChatConfig,
+      conversationKey: chatKey,
+      threadId: threadId ?? undefined,
+      topicId: topicId ?? undefined,
+    });
+
    const mergedHooks = useMemo<ConversationHooks>(
-      () => ({
-        ...hooks,
-        // Expand the sheet the moment the user presses Send, so the chat grows
-        // into view before the AI response streams in — not after it finishes.
-        onBeforeSendMessage: async (params) => {
-          setActiveSnapPoint(MAX_SNAP_POINT);
-          return hooks?.onBeforeSendMessage?.(params);
-        },
-      }),
-      [hooks],
+      () =>
+        mergeConversationHooks(
+          hooks,
+          {
+            // Expand the sheet the moment the user presses Send, so the chat grows
+            // into view before the AI response streams in — not after it finishes.
+            onBeforeSendMessage: async () => {
+              setActiveSnapPoint(MAX_SNAP_POINT);
+            },
+          },
+          chatFollowUpHooks,
+        ),
+      [hooks, chatFollowUpHooks],
    );

    const sheetProps: FloatingSheetProps = {
@@ -1,4 +1,4 @@
-import { render, screen, waitFor } from '@testing-library/react';
+import { render, screen } from '@testing-library/react';
 import type { ReactNode } from 'react';
 import { beforeEach, describe, expect, it, vi } from 'vitest';

@@ -17,7 +17,6 @@ const { chatInputSpy, messageItemSpy, mockState } = vi.hoisted(() => ({
  messageItemSpy: vi.fn(),
  mockState: {
    displayMessages: [] as Array<{ content?: string; id: string; role: string }>,
-    generatingIds: new Set<string>(),
    pendingInterventions: [] as Array<{ id: string }>,
  },
 }));
@@ -91,7 +90,6 @@ describe('AgentOnboardingConversation', () => {
    chatInputSpy.mockClear();
    messageItemSpy.mockClear();
    mockState.displayMessages = [];
-    mockState.generatingIds = new Set();
    mockState.pendingInterventions = [];
  });

@@ -195,101 +193,6 @@ describe('AgentOnboardingConversation', () => {
    );
  });

-  it('fires the assistant-settled callback after the latest assistant stops generating', async () => {
-    const onAssistantTurnSettled = vi.fn();
-    mockState.displayMessages = [
-      { id: 'user-1', role: 'user' },
-      { id: 'assistant-1', role: 'assistant' },
-    ];
-    mockState.generatingIds = new Set(['assistant-1']);
-
-    const { rerender } = render(
-      <AgentOnboardingConversation
-        discoveryUserMessageCount={0}
-        topicId="topic-1"
-        onAssistantTurnSettled={onAssistantTurnSettled}
-      />,
-    );
-
-    expect(onAssistantTurnSettled).not.toHaveBeenCalled();
-
-    mockState.generatingIds = new Set();
-    rerender(
-      <AgentOnboardingConversation
-        discoveryUserMessageCount={1}
-        topicId="topic-1"
-        onAssistantTurnSettled={onAssistantTurnSettled}
-      />,
-    );
-
-    await waitFor(() => {
-      expect(onAssistantTurnSettled).toHaveBeenCalledWith('assistant-1');
-    });
-    expect(onAssistantTurnSettled).toHaveBeenCalledTimes(1);
-  });
-
-  it('waits for resumed generation after a pending intervention clears', async () => {
-    const onAssistantTurnSettled = vi.fn();
-    mockState.displayMessages = [
-      { id: 'user-1', role: 'user' },
-      { id: 'assistant-1', role: 'assistant' },
-    ];
-    mockState.generatingIds = new Set(['assistant-1']);
-
-    const { rerender } = render(
-      <AgentOnboardingConversation
-        discoveryUserMessageCount={0}
-        topicId="topic-1"
-        onAssistantTurnSettled={onAssistantTurnSettled}
-      />,
-    );
-
-    mockState.generatingIds = new Set();
-    mockState.pendingInterventions = [{ id: 'tool-1' }];
-    rerender(
-      <AgentOnboardingConversation
-        discoveryUserMessageCount={1}
-        topicId="topic-1"
-        onAssistantTurnSettled={onAssistantTurnSettled}
-      />,
-    );
-    expect(onAssistantTurnSettled).not.toHaveBeenCalled();
-
-    mockState.pendingInterventions = [];
-    rerender(
-      <AgentOnboardingConversation
-        discoveryUserMessageCount={2}
-        topicId="topic-1"
-        onAssistantTurnSettled={onAssistantTurnSettled}
-      />,
-    );
-    expect(onAssistantTurnSettled).not.toHaveBeenCalled();
-
-    mockState.generatingIds = new Set(['assistant-1']);
-    rerender(
-      <AgentOnboardingConversation
-        discoveryUserMessageCount={3}
-        topicId="topic-1"
-        onAssistantTurnSettled={onAssistantTurnSettled}
-      />,
-    );
-    expect(onAssistantTurnSettled).not.toHaveBeenCalled();
-
-    mockState.generatingIds = new Set();
-    rerender(
-      <AgentOnboardingConversation
-        discoveryUserMessageCount={4}
-        topicId="topic-1"
-        onAssistantTurnSettled={onAssistantTurnSettled}
-      />,
-    );
-
-    await waitFor(() => {
-      expect(onAssistantTurnSettled).toHaveBeenCalledWith('assistant-1');
-    });
-    expect(onAssistantTurnSettled).toHaveBeenCalledTimes(1);
-  });
-
  it('renders normal message items outside the greeting state', () => {
    mockState.displayMessages = [
      { id: 'assistant-1', role: 'assistant' },
@@ -323,8 +226,4 @@ vi.mock('@/features/Conversation/store', () => ({
  dataSelectors: {
    pendingInterventions: (state: typeof mockState) => state.pendingInterventions,
  },
-  messageStateSelectors: {
-    isAssistantGroupItemGenerating: (id: string) => (state: typeof mockState) =>
-      state.generatingIds.has(id),
-  },
 }));
@@ -16,7 +16,7 @@ import {
  MessageItem,
  useConversationStore,
 } from '@/features/Conversation';
-import { dataSelectors, messageStateSelectors } from '@/features/Conversation/store';
+import { dataSelectors } from '@/features/Conversation/store';
 import WideScreenContainer from '@/features/WideScreenContainer';
 import { useIsMobile } from '@/hooks/useIsMobile';
 import type { OnboardingPhase } from '@/types/user';
@@ -29,8 +29,6 @@ import WelcomeMobile from './Welcome.mobile';
 import WelcomeMessage from './WelcomeMessage';
 import WrapUpHint from './WrapUpHint';

-const assistantLikeRoles = new Set(['assistant', 'assistantGroup', 'supervisor']);
-
 interface AgentOnboardingConversationProps {
  discoveryUserMessageCount?: number;
  feedbackSubmitted?: boolean;
@@ -46,7 +44,6 @@ interface AgentOnboardingConversationProps {
  // is ready to handle the first send.
  isInputReady?: boolean;
  onAfterWrapUp?: () => Promise<unknown> | void;
-  onAssistantTurnSettled?: (messageId: string) => Promise<unknown> | void;
  onboardingFinished?: boolean;
  phase?: OnboardingPhase;
  readOnly?: boolean;
@@ -70,7 +67,6 @@ const AgentOnboardingConversation = memo<AgentOnboardingConversationProps>(
    hasMessages,
    isInputReady = true,
    onAfterWrapUp,
-    onAssistantTurnSettled,
    onboardingFinished,
    phase,
    readOnly,
@@ -79,9 +75,6 @@ const AgentOnboardingConversation = memo<AgentOnboardingConversationProps>(
  }) => {
    const isMobile = useIsMobile();
    const displayMessages = useConversationStore(conversationSelectors.displayMessages);
-    const pendingInterventionCount = useConversationStore(
-      (s) => dataSelectors.pendingInterventions(s).length,
-    );
    // The agent-marketplace intervention renders as an absolute overlay anchored
    // to the chat input area, which would otherwise occlude the last message.
    // Reserve matching scroll headroom inside ChatList so the latest message can
@@ -105,23 +98,8 @@ const AgentOnboardingConversation = memo<AgentOnboardingConversationProps>(
      [hasMessages, displayMessages],
    );

-    const latestAssistantMessageId = useMemo(() => {
-      const latest = displayMessages.at(-1);
-      if (!latest || !assistantLikeRoles.has(latest.role)) return undefined;
-
-      return latest.id;
-    }, [displayMessages]);
-
-    const isLatestAssistantGenerating = useConversationStore((s) =>
-      latestAssistantMessageId
-        ? messageStateSelectors.isAssistantGroupItemGenerating(latestAssistantMessageId)(s)
-        : false,
-    );
-
    const [showGreeting, setShowGreeting] = useState(isGreetingState);
    const prevGreetingRef = useRef(isGreetingState);
-    const armedSettledMessageIdRef = useRef<string>(undefined);
-    const firedSettledMessageIdRef = useRef<string>(undefined);

    useEffect(() => {
      if (prevGreetingRef.current && !isGreetingState) {
@@ -140,37 +118,9 @@ const AgentOnboardingConversation = memo<AgentOnboardingConversationProps>(
      prevGreetingRef.current = isGreetingState;
    }, [isGreetingState]);

-    useEffect(() => {
-      if (!onAssistantTurnSettled || !latestAssistantMessageId) return;
-
-      if (pendingInterventionCount > 0) {
-        armedSettledMessageIdRef.current = undefined;
-        return;
-      }
-
-      if (isLatestAssistantGenerating) {
-        armedSettledMessageIdRef.current = latestAssistantMessageId;
-        return;
-      }
-
-      if (armedSettledMessageIdRef.current !== latestAssistantMessageId) return;
-      if (firedSettledMessageIdRef.current === latestAssistantMessageId) return;
-
-      firedSettledMessageIdRef.current = latestAssistantMessageId;
-      armedSettledMessageIdRef.current = undefined;
-      void onAssistantTurnSettled(latestAssistantMessageId);
-    }, [
-      isLatestAssistantGenerating,
-      latestAssistantMessageId,
-      onAssistantTurnSettled,
-      pendingInterventionCount,
-    ]);
-
    const hasPersistedAssistantOpener = displayMessages.at(0)?.role === 'assistant';
    const shouldShowSyntheticWelcome =
-      !onboardingFinished &&
-      !hasPersistedAssistantOpener &&
-      displayMessages.length > 0;
+      !onboardingFinished && !hasPersistedAssistantOpener && displayMessages.length > 0;
    const shouldShowGreetingWelcome = showGreeting && !onboardingFinished;
    const shouldShowGreetingActions = showGreeting && !onboardingFinished;

@@ -14,6 +14,8 @@ import { useNavigate } from 'react-router-dom';

 import Loading from '@/components/Loading/BrandTextLoading';
 import { ONBOARDING_PRODUCTION_DEFAULT_MODEL } from '@/const/onboarding';
+import { type ConversationHooks } from '@/features/Conversation/types';
+import { mergeConversationHooks } from '@/features/Conversation/utils/mergeConversationHooks';
 import ModeSwitch from '@/features/Onboarding/components/ModeSwitch';
 import { useOnboardingAgentTemplates } from '@/hooks/useOnboardingAgentTemplates';
 import { useClientDataSWR, useOnlyFetchOnceSWR } from '@/libs/swr';
@@ -141,12 +143,14 @@ const AgentOnboardingPage = memo(() => {
    [onboardingAgentConfig?.model, onboardingAgentConfig?.provider],
  );

-  const onboardingFollowUp = useOnboardingFollowUp({
+  const onboardingFollowUpHooks = useOnboardingFollowUp({
    enabled: !onboardingFinished && !viewingHistoricalTopic,
    isGreeting,
    modelConfig: onboardingFollowUpModelConfig,
+    onboardingAgentId,
+    phase: data?.context?.phase,
+    topicId: effectiveTopicId,
  });
-  const { onBeforeSendMessage, triggerExtract } = onboardingFollowUp;

  // Re-entry latch for the fresh-state first-send orchestration. The combination
  // of advisory lock + this ref ensures rapid double-submit cannot create two
@@ -157,7 +161,6 @@ const AgentOnboardingPage = memo(() => {
  const composedOnBeforeSendMessage = useCallback(
    async (params: SendMessageParams): Promise<boolean> => {
      params.metadata = { ...params.metadata, trigger: RequestTrigger.Onboarding };
-      await onBeforeSendMessage();

      if (!onboardingAgentId) {
        // ChatInput is gated by `isInputReady`; this branch should be unreachable.
@@ -212,7 +215,7 @@ const AgentOnboardingPage = memo(() => {
      await orchestration;
      return false;
    },
-    [effectiveTopicId, mutate, onBeforeSendMessage, onboardingAgentId],
+    [effectiveTopicId, mutate, onboardingAgentId],
  );

  const syncOnboardingContext = useCallback(async () => {
@@ -223,43 +226,52 @@ const AgentOnboardingPage = memo(() => {
    return nextContext;
  }, [mutate, mutateHistoryTopics, onboardingAgentId]);

-  const handleAssistantTurnSettled = useCallback(async () => {
-    if (!effectiveTopicId) return;
+  const onboardingTurnSettledHook = useMemo<ConversationHooks>(() => {
+    if (onboardingFinished || viewingHistoricalTopic) return {};

-    const prevPhase = data?.context?.phase;
-    const prevFinishedAt = agentOnboarding?.finishedAt;
+    return {
+      onAssistantTurnSettled: async () => {
+        if (!effectiveTopicId) return;

-    const extractPromise = triggerExtract(effectiveTopicId, prevPhase);
+        const prevPhase = data?.context?.phase;
+        const prevFinishedAt = agentOnboarding?.finishedAt;

-    // Sync first to learn the next phase/finishedAt; only then decide whether
-    // the heavier user-store / builtin-agent refreshes are needed this turn.
-    const [nextContext] = await Promise.all([syncOnboardingContext(), extractPromise]);
+        const nextContext = await syncOnboardingContext();
+        const newPhase = nextContext?.context?.phase;
+        const newFinishedAt = nextContext?.agentOnboarding?.finishedAt;

-    const newPhase = nextContext?.context?.phase;
-    const newFinishedAt = nextContext?.agentOnboarding?.finishedAt;
-
-    const refreshes: Promise<unknown>[] = [];
-    if (newFinishedAt !== prevFinishedAt) refreshes.push(refreshUserState());
-    if (newPhase !== prevPhase) {
-      refreshes.push(refreshBuiltinAgent(BUILTIN_AGENT_SLUGS.webOnboarding));
-    }
-    if (refreshes.length > 0) await Promise.all(refreshes);
+        const refreshes: Promise<unknown>[] = [];
+        if (newFinishedAt !== prevFinishedAt) refreshes.push(refreshUserState());
+        if (newPhase !== prevPhase) {
+          refreshes.push(refreshBuiltinAgent(BUILTIN_AGENT_SLUGS.webOnboarding));
+        }
+        if (refreshes.length > 0) await Promise.all(refreshes);
+      },
+    };
  }, [
-    agentOnboarding?.finishedAt,
-    data?.context?.phase,
+    onboardingFinished,
+    viewingHistoricalTopic,
    effectiveTopicId,
+    data?.context?.phase,
+    agentOnboarding?.finishedAt,
    refreshBuiltinAgent,
    refreshUserState,
    syncOnboardingContext,
-    triggerExtract,
  ]);
-  const assistantTurnSettledHandler =
-    onboardingFinished || viewingHistoricalTopic ? undefined : handleAssistantTurnSettled;

-  const conversationHooks = useMemo(
-    () => (onboardingFinished ? undefined : { onBeforeSendMessage: composedOnBeforeSendMessage }),
-    [onboardingFinished, composedOnBeforeSendMessage],
-  );
+  const conversationHooks = useMemo(() => {
+    if (onboardingFinished) return undefined;
+    return mergeConversationHooks(
+      { onBeforeSendMessage: composedOnBeforeSendMessage },
+      onboardingTurnSettledHook,
+      onboardingFollowUpHooks,
+    );
+  }, [
+    onboardingFinished,
+    composedOnBeforeSendMessage,
+    onboardingTurnSettledHook,
+    onboardingFollowUpHooks,
+  ]);

  if (error) {
    return (
@@ -316,7 +328,6 @@ const AgentOnboardingPage = memo(() => {
              showFeedback={!viewingHistoricalTopic}
              topicId={effectiveTopicId}
              onAfterWrapUp={syncOnboardingContext}
-              onAssistantTurnSettled={assistantTurnSettledHandler}
            />
          </ErrorBoundary>
        </OnboardingConversationProvider>
@@ -1,6 +1,7 @@
 import { renderHook } from '@testing-library/react';
 import { afterEach, beforeEach, describe, expect, it, vi } from 'vitest';

+import { messageMapKey } from '@/store/chat/utils/messageMapKey';
 import { useFollowUpActionStore } from '@/store/followUpAction';

 import { useOnboardingFollowUp } from './useOnboardingFollowUp';
@@ -9,6 +10,9 @@ const MODEL_CONFIG = {
  model: 'scene-model',
  provider: 'scene-provider',
 };
+const AGENT_ID = 'agent-onboarding';
+const TOPIC_ID = 'topic-1';
+const CONVERSATION_KEY = messageMapKey({ agentId: AGENT_ID, topicId: TOPIC_ID });

 describe('useOnboardingFollowUp', () => {
  let fetchFor: ReturnType<typeof vi.fn>;
@@ -27,65 +31,186 @@ describe('useOnboardingFollowUp', () => {
    vi.restoreAllMocks();
  });

-  it('triggerExtract skips when disabled', async () => {
+  it('returns no hooks when disabled', async () => {
    const { result } = renderHook(() =>
-      useOnboardingFollowUp({ enabled: false, isGreeting: false, modelConfig: MODEL_CONFIG }),
+      useOnboardingFollowUp({
+        enabled: false,
+        isGreeting: false,
+        modelConfig: MODEL_CONFIG,
+        onboardingAgentId: AGENT_ID,
+        phase: 'discovery',
+        topicId: TOPIC_ID,
+      }),
    );
-    await result.current.triggerExtract('topic-1', 'discovery');
+    expect(result.current.onAssistantTurnSettled).toBeUndefined();
+    expect(result.current.onBeforeSendMessage).toBeUndefined();
+  });
+
+  it('returns no hooks when onboardingAgentId is missing', async () => {
+    const { result } = renderHook(() =>
+      useOnboardingFollowUp({
+        enabled: true,
+        isGreeting: false,
+        modelConfig: MODEL_CONFIG,
+        onboardingAgentId: undefined,
+        phase: 'discovery',
+        topicId: TOPIC_ID,
+      }),
+    );
+    expect(result.current.onAssistantTurnSettled).toBeUndefined();
+    expect(result.current.onBeforeSendMessage).toBeUndefined();
+  });
+
+  it('returns no hooks when topicId is missing', async () => {
+    const { result } = renderHook(() =>
+      useOnboardingFollowUp({
+        enabled: true,
+        isGreeting: false,
+        modelConfig: MODEL_CONFIG,
+        onboardingAgentId: AGENT_ID,
+        phase: 'discovery',
+        topicId: undefined,
+      }),
+    );
+    expect(result.current.onAssistantTurnSettled).toBeUndefined();
+    expect(result.current.onBeforeSendMessage).toBeUndefined();
+  });
+
+  it('onAssistantTurnSettled skips when phase is undefined', async () => {
+    const { result } = renderHook(() =>
+      useOnboardingFollowUp({
+        enabled: true,
+        isGreeting: false,
+        modelConfig: MODEL_CONFIG,
+        onboardingAgentId: AGENT_ID,
+        phase: undefined,
+        topicId: TOPIC_ID,
+      }),
+    );
+    await result.current.onAssistantTurnSettled?.('msg-1', { reason: 'completed' });
    expect(fetchFor).not.toHaveBeenCalled();
  });

-  it('triggerExtract skips when phase is undefined', async () => {
+  it('onAssistantTurnSettled skips when phase is summary', async () => {
    const { result } = renderHook(() =>
-      useOnboardingFollowUp({ enabled: true, isGreeting: false, modelConfig: MODEL_CONFIG }),
+      useOnboardingFollowUp({
+        enabled: true,
+        isGreeting: false,
+        modelConfig: MODEL_CONFIG,
+        onboardingAgentId: AGENT_ID,
+        phase: 'summary',
+        topicId: TOPIC_ID,
+      }),
    );
-    await result.current.triggerExtract('topic-1', undefined);
+    await result.current.onAssistantTurnSettled?.('msg-1', { reason: 'completed' });
    expect(fetchFor).not.toHaveBeenCalled();
  });

-  it('triggerExtract skips when phase is summary', async () => {
+  it('onAssistantTurnSettled skips when isGreeting is true', async () => {
    const { result } = renderHook(() =>
-      useOnboardingFollowUp({ enabled: true, isGreeting: false, modelConfig: MODEL_CONFIG }),
+      useOnboardingFollowUp({
+        enabled: true,
+        isGreeting: true,
+        modelConfig: MODEL_CONFIG,
+        onboardingAgentId: AGENT_ID,
+        phase: 'agent_identity',
+        topicId: TOPIC_ID,
+      }),
    );
-    await result.current.triggerExtract('topic-1', 'summary');
+    await result.current.onAssistantTurnSettled?.('msg-1', { reason: 'completed' });
    expect(fetchFor).not.toHaveBeenCalled();
  });

-  it('triggerExtract skips when isGreeting is true', async () => {
+  it('onAssistantTurnSettled skips when reason is stopped', async () => {
    const { result } = renderHook(() =>
-      useOnboardingFollowUp({ enabled: true, isGreeting: true, modelConfig: MODEL_CONFIG }),
+      useOnboardingFollowUp({
+        enabled: true,
+        isGreeting: false,
+        modelConfig: MODEL_CONFIG,
+        onboardingAgentId: AGENT_ID,
+        phase: 'discovery',
+        topicId: TOPIC_ID,
+      }),
    );
-    await result.current.triggerExtract('topic-1', 'agent_identity');
+    await result.current.onAssistantTurnSettled?.('msg-1', { reason: 'stopped' });
    expect(fetchFor).not.toHaveBeenCalled();
  });

-  it('triggerExtract fires fetchFor with onboarding hint on a normal turn', async () => {
+  it('onAssistantTurnSettled fires fetchFor with onboarding hint on a normal turn', async () => {
    const { result } = renderHook(() =>
-      useOnboardingFollowUp({ enabled: true, isGreeting: false, modelConfig: MODEL_CONFIG }),
+      useOnboardingFollowUp({
+        enabled: true,
+        isGreeting: false,
+        modelConfig: MODEL_CONFIG,
+        onboardingAgentId: AGENT_ID,
+        phase: 'discovery',
+        topicId: TOPIC_ID,
+      }),
    );
-    await result.current.triggerExtract('topic-1', 'discovery');
-    expect(fetchFor).toHaveBeenCalledWith('topic-1', {
+    await result.current.onAssistantTurnSettled?.('msg-1', { reason: 'completed' });
+    expect(fetchFor).toHaveBeenCalledWith(CONVERSATION_KEY, {
      hint: {
        kind: 'onboarding',
        phase: 'discovery',
      },
      modelConfig: MODEL_CONFIG,
+      topicId: TOPIC_ID,
+    });
+  });
+
+  it('onAssistantTurnSettled uses the phase snapshot captured at memoize time', async () => {
+    const { result, rerender } = renderHook(
+      (props: { phase: 'discovery' | 'agent_identity' }) =>
+        useOnboardingFollowUp({
+          enabled: true,
+          isGreeting: false,
+          modelConfig: MODEL_CONFIG,
+          onboardingAgentId: AGENT_ID,
+          phase: props.phase,
+          topicId: TOPIC_ID,
+        }),
+      { initialProps: { phase: 'discovery' } },
+    );
+    const fired = result.current.onAssistantTurnSettled;
+    rerender({ phase: 'agent_identity' });
+    await fired?.('msg-1', { reason: 'completed' });
+    expect(fetchFor).toHaveBeenCalledWith(CONVERSATION_KEY, {
+      hint: {
+        kind: 'onboarding',
+        phase: 'discovery',
+      },
+      modelConfig: MODEL_CONFIG,
+      topicId: TOPIC_ID,
    });
  });

  it('onBeforeSendMessage clears when enabled', async () => {
    const { result } = renderHook(() =>
-      useOnboardingFollowUp({ enabled: true, isGreeting: false, modelConfig: MODEL_CONFIG }),
+      useOnboardingFollowUp({
+        enabled: true,
+        isGreeting: false,
+        modelConfig: MODEL_CONFIG,
+        onboardingAgentId: AGENT_ID,
+        phase: 'discovery',
+        topicId: TOPIC_ID,
+      }),
    );
-    await result.current.onBeforeSendMessage();
-    expect(clear).toHaveBeenCalledTimes(1);
+    await result.current.onBeforeSendMessage?.({} as any);
+    expect(clear).toHaveBeenCalledWith(CONVERSATION_KEY);
  });

-  it('onBeforeSendMessage does nothing when disabled', async () => {
+  it('onBeforeSendMessage is absent when disabled', async () => {
    const { result } = renderHook(() =>
-      useOnboardingFollowUp({ enabled: false, isGreeting: false, modelConfig: MODEL_CONFIG }),
+      useOnboardingFollowUp({
+        enabled: false,
+        isGreeting: false,
+        modelConfig: MODEL_CONFIG,
+        onboardingAgentId: AGENT_ID,
+        phase: 'discovery',
+        topicId: TOPIC_ID,
+      }),
    );
-    await result.current.onBeforeSendMessage();
+    expect(result.current.onBeforeSendMessage).toBeUndefined();
    expect(clear).not.toHaveBeenCalled();
  });
 });
@@ -1,6 +1,8 @@
 import type { FollowUpModelConfig } from '@lobechat/types';
-import { useCallback } from 'react';
+import { useMemo } from 'react';

+import { type ConversationHooks } from '@/features/Conversation/types';
+import { messageMapKey } from '@/store/chat/utils/messageMapKey';
 import { useFollowUpActionStore } from '@/store/followUpAction';
 import type { OnboardingPhase } from '@/types/user';

@@ -8,37 +10,48 @@ interface UseOnboardingFollowUpParams {
  enabled: boolean;
  isGreeting: boolean;
  modelConfig: FollowUpModelConfig;
-}
-
-interface OnboardingFollowUpHandlers {
-  onBeforeSendMessage: () => Promise<void>;
-  triggerExtract: (topicId: string, phase: OnboardingPhase | undefined) => Promise<void>;
+  onboardingAgentId: string | undefined;
+  phase: OnboardingPhase | undefined;
+  topicId: string | undefined;
 }

 export const useOnboardingFollowUp = ({
  enabled,
  isGreeting,
  modelConfig,
-}: UseOnboardingFollowUpParams): OnboardingFollowUpHandlers => {
-  const triggerExtract = useCallback(
-    async (topicId: string, phase: OnboardingPhase | undefined) => {
-      if (!enabled) return;
-      if (!phase) return;
-      if (phase === 'summary') return;
-      if (isGreeting) return;
+  onboardingAgentId,
+  phase,
+  topicId,
+}: UseOnboardingFollowUpParams): ConversationHooks => {
+  return useMemo<ConversationHooks>(() => {
+    if (!enabled || !onboardingAgentId || !topicId) return {};

-      await useFollowUpActionStore.getState().fetchFor(topicId, {
-        hint: { kind: 'onboarding', phase },
-        modelConfig,
-      });
-    },
-    [enabled, isGreeting, modelConfig],
-  );
+    const conversationKey = messageMapKey({ agentId: onboardingAgentId, topicId });
+    const phaseSnapshot = phase;

-  const onBeforeSendMessage = useCallback(async () => {
-    if (!enabled) return;
-    useFollowUpActionStore.getState().clear();
-  }, [enabled]);
-
-  return { onBeforeSendMessage, triggerExtract };
+    return {
+      onAssistantTurnSettled: async (_messageId, { reason }) => {
+        if (reason === 'stopped') return;
+        if (isGreeting) return;
+        if (!phaseSnapshot) return;
+        if (phaseSnapshot === 'summary') return;
+        await useFollowUpActionStore.getState().fetchFor(conversationKey, {
+          hint: { kind: 'onboarding', phase: phaseSnapshot },
+          modelConfig,
+          topicId,
+        });
+      },
+      onBeforeSendMessage: async () => {
+        useFollowUpActionStore.getState().clear(conversationKey);
+      },
+    };
+  }, [
+    enabled,
+    isGreeting,
+    modelConfig.model,
+    modelConfig.provider,
+    onboardingAgentId,
+    phase,
+    topicId,
+  ]);
 };
@@ -13,7 +13,11 @@ import {
  useConversationStore,
 } from '@/features/Conversation';
 import SkeletonList from '@/features/Conversation/components/SkeletonList';
+import { useChatFollowUp } from '@/features/Conversation/hooks/useChatFollowUp';
+import { mergeConversationHooks } from '@/features/Conversation/utils/mergeConversationHooks';
 import { useOperationState } from '@/hooks/useOperationState';
+import { useAgentStore } from '@/store/agent';
+import { chatConfigByIdSelectors } from '@/store/agent/selectors';
 import { useChatStore } from '@/store/chat';
 import { portalThreadSelectors, threadSelectors } from '@/store/chat/selectors';
 import { type MessageMapKeyInput } from '@/store/chat/utils/messageMapKey';
@@ -187,30 +191,44 @@ const ThreadChat = memo(() => {
  // Get operation state for reactive updates
  const operationState = useOperationState(context);

+  const agentChatConfig = useAgentStore(
+    chatConfigByIdSelectors.getChatConfigById(activeAgentId || ''),
+  );
+  const chatFollowUpHooks = useChatFollowUp({
+    agentChatConfig,
+    conversationKey: chatKey,
+    threadId: portalThreadId ?? undefined,
+    topicId: activeTopicId ?? undefined,
+  });
+
  // Hooks to handle post-message-creation tasks for new thread
  const hooks: ConversationHooks = useMemo(
-    () => ({
-      onAfterMessageCreate: async ({ createdThreadId }) => {
-        if (!createdThreadId) return;
+    () =>
+      mergeConversationHooks(
+        {
+          onAfterMessageCreate: async ({ createdThreadId }) => {
+            if (!createdThreadId) return;

-        const state = useChatStore.getState();
+            const state = useChatStore.getState();

-        // Refresh threads list
-        await state.refreshThreads();
-        // Refresh messages to include new thread messages
-        await state.refreshMessages();
-        // Open the newly created thread in portal
-        state.openThreadInPortal(createdThreadId, threadStartMessageId);
+            // Refresh threads list
+            await state.refreshThreads();
+            // Refresh messages to include new thread messages
+            await state.refreshMessages();
+            // Open the newly created thread in portal
+            state.openThreadInPortal(createdThreadId, threadStartMessageId);

-        // Summarize thread title for new thread
-        const portalThread = threadSelectors.currentPortalThread(useChatStore.getState());
-        if (portalThread) {
-          const chats = threadSelectors.portalAIChats(useChatStore.getState());
-          await useChatStore.getState().summaryThreadTitle(portalThread.id, chats);
-        }
-      },
-    }),
-    [threadStartMessageId],
+            // Summarize thread title for new thread
+            const portalThread = threadSelectors.currentPortalThread(useChatStore.getState());
+            if (portalThread) {
+              const chats = threadSelectors.portalAIChats(useChatStore.getState());
+              await useChatStore.getState().summaryThreadTitle(portalThread.id, chats);
+            }
+          },
+        },
+        chatFollowUpHooks,
+      ),
+    [chatFollowUpHooks, threadStartMessageId],
  );

  return (
@@ -29,6 +29,7 @@ const SYSTEM_AGENT_MODEL_ITEMS: SystemAgentModelItem[] = [
 ];

 const OPTIONAL_FEATURE_ITEMS: SystemAgentModelItem[] = [
+  { key: 'followUpAction' },
  { key: 'inputCompletion' },
  { key: 'promptRewrite' },
 ];
@@ -163,7 +164,9 @@ const ModelAssignmentsForm = memo(() => {
  });

  const isOptionalFeatureLoading =
-    loadingKey === 'inputCompletion' || loadingKey === 'promptRewrite';
+    loadingKey === 'followUpAction' ||
+    loadingKey === 'inputCompletion' ||
+    loadingKey === 'promptRewrite';
  const isModelAssignmentLoading = loadingKey && !isOptionalFeatureLoading;

  const modelAssignments: FormGroupItemType = {
@@ -618,6 +618,11 @@ export default {
  'settingChat.enableAutoScrollOnStreaming.desc': 'Override global setting for this assistant',
  'settingChat.enableAutoScrollOnStreaming.title': 'Auto-scroll During AI Response',
  'settingChat.enableCompressHistory.title': 'Enable Automatic Summary of Chat History',
+  'settingChat.enableFollowUpChips.desc':
+    'After each reply, show one-click follow-up reply chips below the message. Requires the global Follow-up model to be configured.',
+  'settingChat.enableFollowUpChips.notConfiguredHint':
+    'Configure the global Follow-up model first to enable this.',
+  'settingChat.enableFollowUpChips.title': 'Follow-up Suggestions',
  'settingChat.enableHistoryCount.alias': 'Unlimited',
  'settingChat.enableHistoryCount.limited': 'Include only {{number}} conversation messages',
  'settingChat.enableHistoryCount.setlimited': 'Set limited history messages',
@@ -969,6 +974,10 @@ When I am ___, I need ___
    'Once filled out, the system agent will use the custom prompt when generating content',
  'systemAgent.customPrompt.placeholder': 'Please enter custom prompt',
  'systemAgent.customPrompt.title': 'Custom Prompt',
+  'systemAgent.followUpAction.label': 'Follow-up Suggestions Model',
+  'systemAgent.followUpAction.modelDesc':
+    'Model used to suggest one-click follow-up replies under each assistant message',
+  'systemAgent.followUpAction.title': 'Follow-up Suggestions',
  'systemAgent.generationTopic.label': 'Model',
  'systemAgent.generationTopic.modelDesc': 'Model used to name AI image topics',
  'systemAgent.generationTopic.title': 'AI Image Topic Naming',
@@ -9,11 +9,13 @@ import { useTranslation } from 'react-i18next';
 import AgentHome from '@/features/AgentHome';
 import ChatMiniMap from '@/features/ChatMiniMap';
 import { ChatList, ConversationProvider } from '@/features/Conversation';
+import { useChatFollowUp } from '@/features/Conversation/hooks/useChatFollowUp';
+import { mergeConversationHooks } from '@/features/Conversation/utils/mergeConversationHooks';
 import ZenModeToast from '@/features/ZenModeToast';
 import { useGatewayReconnect } from '@/hooks/useGatewayReconnect';
 import { useOperationState } from '@/hooks/useOperationState';
 import { useAgentStore } from '@/store/agent';
-import { agentSelectors } from '@/store/agent/selectors';
+import { agentChatConfigSelectors, agentSelectors } from '@/store/agent/selectors';
 import { useChatStore } from '@/store/chat';
 import { threadSelectors, topicSelectors } from '@/store/chat/selectors';
 import { messageMapKey } from '@/store/chat/utils/messageMapKey';
@@ -71,11 +73,22 @@ const Conversation = memo(() => {
  );
  useGatewayReconnect(context.topicId, runningOperation);

+  const agentChatConfig = useAgentStore(agentChatConfigSelectors.currentChatConfig);
+  const chatFollowUpHooks = useChatFollowUp({
+    agentChatConfig,
+    conversationKey: chatKey,
+    threadId: context.threadId ?? undefined,
+    topicId: context.topicId ?? undefined,
+  });
+
+  const hooks = useMemo(() => mergeConversationHooks(chatFollowUpHooks), [chatFollowUpHooks]);
+
  return (
    <ConversationProvider
      actionsBar={actionsBarConfig}
      context={context}
      hasInitMessages={!!messages}
+      hooks={hooks}
      messages={messages}
      operationState={operationState}
      onMessagesChange={(messages, ctx) => {
@@ -16,6 +16,7 @@ const Page = () => {
      <SystemAgentForm systemAgentKey="translation" />
      <SystemAgentForm systemAgentKey="historyCompress" />
      <SystemAgentForm systemAgentKey="agentMeta" />
+      <SystemAgentForm allowDisable systemAgentKey="followUpAction" />
      <SystemAgentForm allowDisable systemAgentKey="inputCompletion" />
      <SystemAgentForm allowDisable systemAgentKey="promptRewrite" />
    </>
@@ -143,6 +143,48 @@ describe('FollowUpActionService.extract', () => {
    expect(result).toEqual({ chips: [], messageId: FOUND_MSG });
  });

+  const captureWhereOps = () => {
+    const arg = queryFindFirstSpy.mock.calls[0][0];
+    const fakeTable = {
+      content: { col: 'content' },
+      createdAt: { col: 'createdAt' },
+      id: { col: 'id' },
+      role: { col: 'role' },
+      threadId: { col: 'threadId' },
+      topicId: { col: 'topicId' },
+      userId: { col: 'userId' },
+    };
+    const ops = {
+      and: (...parts: any[]) => ({ op: 'and', parts }),
+      eq: (col: any, value: any) => ({ col, op: 'eq', value }),
+      isNotNull: (col: any) => ({ col, op: 'isNotNull' }),
+      isNull: (col: any) => ({ col, op: 'isNull' }),
+      ne: (col: any, value: any) => ({ col, op: 'ne', value }),
+    };
+    const result = arg.where(fakeTable, ops);
+    return { parts: result.parts as any[], table: fakeTable };
+  };
+
+  it('filters by threadId when provided (thread isolation)', async () => {
+    queryFindFirstSpy.mockResolvedValue(undefined);
+    await svc.extract({
+      modelConfig: MODEL_CONFIG,
+      threadId: 'thread-A',
+      topicId: TEST_TOPIC,
+    });
+    const { parts, table } = captureWhereOps();
+    expect(parts).toContainEqual({ col: table.threadId, op: 'eq', value: 'thread-A' });
+    expect(parts.some((p) => p.op === 'isNull' && p.col === table.threadId)).toBe(false);
+  });
+
+  it('filters by isNull(threadId) when no threadId provided (main topic only)', async () => {
+    queryFindFirstSpy.mockResolvedValue(undefined);
+    await svc.extract({ modelConfig: MODEL_CONFIG, topicId: TEST_TOPIC });
+    const { parts, table } = captureWhereOps();
+    expect(parts).toContainEqual({ col: table.threadId, op: 'isNull' });
+    expect(parts.some((p) => p.op === 'eq' && p.col === table.threadId)).toBe(false);
+  });
+
  it('appends onboarding addendum to system prompt when hint is onboarding', async () => {
    queryFindFirstSpy.mockResolvedValue({ id: FOUND_MSG, content: 'q?' });
    runtimeMock.generateObject.mockResolvedValue({ chips: [] });
@@ -22,6 +22,7 @@ export class FollowUpActionService {

  async extract({
    topicId,
+    threadId,
    hint,
    modelConfig,
  }: FollowUpExtractInput): Promise<FollowUpExtractResult> {
@@ -30,10 +31,13 @@ export class FollowUpActionService {
    const row = await this.db.query.messages.findFirst({
      columns: { content: true, id: true },
      orderBy: (m, { desc }) => desc(m.createdAt),
-      where: (m, { and, eq, isNotNull, ne }) =>
+      where: (m, { and, eq, isNotNull, isNull, ne }) =>
        and(
          eq(m.userId, this.userId),
          eq(m.topicId, topicId),
+          // Discriminate thread vs main topic: an absent threadId must NOT
+          // surface a thread reply that lives under the same topicId.
+          threadId ? eq(m.threadId, threadId) : isNull(m.threadId),
          eq(m.role, 'assistant'),
          isNotNull(m.content),
          ne(m.content, ''),
@@ -158,9 +158,7 @@ export function buildCloudHeteroContext(params: {
          : entry.content;
      return `<${entry.role}>\n${body}\n</${entry.role}>`;
    });
-    parts.push(
-      `<previous_conversation>\n${entries.join('\n')}\n</previous_conversation>`,
-    );
+    parts.push(`<previous_conversation>\n${entries.join('\n')}\n</previous_conversation>`);
  }

  return parts.join('\n\n');
@@ -266,7 +266,9 @@ describe('OnboardingService', () => {
    expect(result.success).toBe(true);
    expect(result.savedFields).toEqual(['fullName']);
    expect(result.ignoredFields).toEqual(['agentName', 'agentEmoji']);
-    expect(result.content).toContain('Skipped agent identity because agentName matches the user identity');
+    expect(result.content).toContain(
+      'Skipped agent identity because agentName matches the user identity',
+    );
    expect(persistedUserState.fullName).toBe('anbex');
    expect(mockAgentModel.update).not.toHaveBeenCalled();
  });
@@ -3,19 +3,55 @@ import type { FollowUpChip, FollowUpHint, FollowUpModelConfig } from '@lobechat/
 import { followUpActionService } from '@/services/followUpAction';
 import { type StoreSetter } from '@/store/types';

+import { type FollowUpActionSlot } from './initialState';
 import { type FollowUpActionStore } from './store';

 // LLM `generateObject` for chip extraction routinely takes 8-12s end-to-end.
 // Anything below ~20s aborts before the model can respond.
 const TIMEOUT_MS = 20_000;

+const IDLE_SLOT: FollowUpActionSlot = { chips: [], status: 'idle' };
+
 type Setter = StoreSetter<FollowUpActionStore>;

 interface FetchForParams {
  hint?: FollowUpHint;
  modelConfig: FollowUpModelConfig;
+  threadId?: string;
+  topicId: string;
 }

+const writeSlot = (
+  set: Setter,
+  conversationKey: string,
+  slot: FollowUpActionSlot,
+  action: string,
+): void => {
+  set(
+    (state) => ({
+      slots: {
+        ...state.slots,
+        [conversationKey]: slot,
+      },
+    }),
+    false,
+    action,
+  );
+};
+
+const removeSlot = (set: Setter, conversationKey: string, action: string): void => {
+  set(
+    (state) => {
+      if (!state.slots[conversationKey]) return state;
+
+      const { [conversationKey]: _, ...rest } = state.slots;
+      return { slots: rest };
+    },
+    false,
+    action,
+  );
+};
+
 export const createFollowUpActionSlice = (
  set: Setter,
  get: () => FollowUpActionStore,
@@ -32,94 +68,76 @@ export class FollowUpActionImpl {
    this.#get = get;
  }

-  fetchFor = async (topicId: string, params: FetchForParams): Promise<void> => {
-    const cur = this.#get();
-    // Dedupe: skip if already loading/ready for the same topic
-    if (cur.pendingTopicId === topicId && cur.status !== 'idle') return;
+  fetchFor = async (conversationKey: string, params: FetchForParams): Promise<void> => {
+    const existing = this.#get().slots[conversationKey];
+    if (existing?.status === 'loading') return;

-    cur.abortController?.abort();
+    existing?.abortController?.abort();

    const controller = new AbortController();
    const timeoutId = setTimeout(() => controller.abort(), TIMEOUT_MS);

-    this.#set(
+    writeSlot(
+      this.#set,
+      conversationKey,
      {
        abortController: controller,
        chips: [],
-        messageId: undefined,
-        pendingTopicId: topicId,
        status: 'loading',
-        topicId: undefined,
      },
-      false,
      'fetchFor:start',
    );

-    const result = await followUpActionService.extract({ ...params, topicId }, controller.signal);
+    const result = await followUpActionService.extract(
+      {
+        hint: params.hint,
+        modelConfig: params.modelConfig,
+        threadId: params.threadId,
+        topicId: params.topicId,
+      },
+      controller.signal,
+    );
    clearTimeout(timeoutId);

-    // Discard stale results: if the active controller in state is no longer
-    // this one, our call has been superseded — either by clear()/abort()
-    // (e.g., user sent a new message) or by a newer fetchFor for the same
-    // topic (next turn). Identity beats topicId here because a same-topic
-    // follow-up turn would otherwise let an in-flight prior result overwrite
-    // the new turn's chips when the network abort race is lost.
-    if (this.#get().abortController !== controller) return;
+    // Identity guard: a same-key follow-up turn (next assistant settle) would
+    // otherwise let an in-flight prior result overwrite the new turn's chips
+    // when the network abort race is lost.
+    if (this.#get().slots[conversationKey]?.abortController !== controller) return;

    if (!result || !result.messageId || result.chips.length === 0) {
-      this.#set(
-        {
-          abortController: undefined,
-          chips: [],
-          messageId: undefined,
-          pendingTopicId: undefined,
-          status: 'idle',
-          topicId: undefined,
-        },
-        false,
-        'fetchFor:fail',
-      );
+      writeSlot(this.#set, conversationKey, { ...IDLE_SLOT }, 'fetchFor:fail');
      return;
    }

-    this.#set(
+    writeSlot(
+      this.#set,
+      conversationKey,
      {
-        abortController: undefined,
        chips: result.chips,
        messageId: result.messageId,
-        pendingTopicId: undefined,
        status: 'ready',
-        topicId,
      },
-      false,
      'fetchFor:ready',
    );
  };

-  abort = (): void => {
-    const cur = this.#get();
-    cur.abortController?.abort();
-    this.#set(
-      {
-        abortController: undefined,
-        chips: [],
-        messageId: undefined,
-        pendingTopicId: undefined,
-        status: 'idle',
-        topicId: undefined,
-      },
-      false,
-      'abort',
-    );
+  abort = (conversationKey: string): void => {
+    const slot = this.#get().slots[conversationKey];
+    if (!slot) return;
+    slot.abortController?.abort();
+    writeSlot(this.#set, conversationKey, { ...IDLE_SLOT }, 'abort');
  };

-  clear = (): void => {
-    this.abort();
+  clear = (conversationKey: string): void => {
+    const slot = this.#get().slots[conversationKey];
+    if (!slot) return;
+    slot.abortController?.abort();
+    removeSlot(this.#set, conversationKey, 'clear');
  };

-  consume = (chip: FollowUpChip): void => {
+  consume = (conversationKey: string, chip: FollowUpChip): void => {
    void chip;
-    this.clear();
+    this.clear(conversationKey);
  };
 }

@@ -2,13 +2,20 @@ import { afterEach, beforeEach, describe, expect, it, vi } from 'vitest';

 import { followUpActionService } from '@/services/followUpAction';

+import { followUpActionSelectors } from './selectors';
 import { useFollowUpActionStore } from './store';

-const TOPIC = 'topic-1';
-const NEW_TOPIC = 'topic-2';
+const KEY_A = 'main_agent-a_topic-a';
+const KEY_B = 'main_agent-b_topic-b';
+const TOPIC_A = 'topic-a';
+const TOPIC_B = 'topic-b';
 const MSG = 'msg-real';
 const MODEL_CONFIG = { model: 'scene-model', provider: 'scene-provider' };
-const FETCH_PARAMS = { modelConfig: MODEL_CONFIG };
+const FETCH_PARAMS_A = { modelConfig: MODEL_CONFIG, topicId: TOPIC_A };
+const FETCH_PARAMS_B = { modelConfig: MODEL_CONFIG, topicId: TOPIC_B };
+
+const slotA = () => useFollowUpActionStore.getState().slots[KEY_A];
+const slotB = () => useFollowUpActionStore.getState().slots[KEY_B];

 describe('useFollowUpActionStore', () => {
  beforeEach(() => {
@@ -26,84 +33,106 @@ describe('useFollowUpActionStore', () => {
      chips: [{ label: 'a', message: 'a' }],
    });

-    const promise = useFollowUpActionStore.getState().fetchFor(TOPIC, FETCH_PARAMS);
-    expect(useFollowUpActionStore.getState().status).toBe('loading');
+    const promise = useFollowUpActionStore.getState().fetchFor(KEY_A, FETCH_PARAMS_A);
+    expect(slotA().status).toBe('loading');
    await promise;
    expect(spy).toHaveBeenCalledOnce();
-    expect(useFollowUpActionStore.getState().status).toBe('ready');
-    expect(useFollowUpActionStore.getState().chips).toHaveLength(1);
-    expect(useFollowUpActionStore.getState().messageId).toBe(MSG);
-    expect(useFollowUpActionStore.getState().topicId).toBe(TOPIC);
+    expect(slotA().status).toBe('ready');
+    expect(slotA().chips).toHaveLength(1);
+    expect(slotA().messageId).toBe(MSG);
  });

-  it('fetchFor forwards modelConfig to the service', async () => {
+  it('fetchFor forwards modelConfig, topicId, and threadId to the service', async () => {
    const spy = vi.spyOn(followUpActionService, 'extract').mockResolvedValue({
      messageId: MSG,
      chips: [{ label: 'a', message: 'a' }],
    });
-    await useFollowUpActionStore.getState().fetchFor(TOPIC, {
+    await useFollowUpActionStore.getState().fetchFor(KEY_A, {
      hint: { kind: 'onboarding', phase: 'discovery' },
      modelConfig: MODEL_CONFIG,
+      threadId: 'thd-1',
+      topicId: TOPIC_A,
    });

    expect(spy).toHaveBeenCalledWith(
      {
        hint: { kind: 'onboarding', phase: 'discovery' },
        modelConfig: MODEL_CONFIG,
-        topicId: TOPIC,
+        threadId: 'thd-1',
+        topicId: TOPIC_A,
      },
      expect.any(AbortSignal),
    );
  });

-  it('fetchFor returns idle when service returns null', async () => {
+  it('fetchFor leaves slot idle when service returns null', async () => {
    vi.spyOn(followUpActionService, 'extract').mockResolvedValue(null);
-    await useFollowUpActionStore.getState().fetchFor(TOPIC, FETCH_PARAMS);
-    expect(useFollowUpActionStore.getState().status).toBe('idle');
-    expect(useFollowUpActionStore.getState().chips).toHaveLength(0);
-    expect(useFollowUpActionStore.getState().messageId).toBeUndefined();
+    await useFollowUpActionStore.getState().fetchFor(KEY_A, FETCH_PARAMS_A);
+    expect(slotA().status).toBe('idle');
+    expect(slotA().chips).toHaveLength(0);
+    expect(slotA().messageId).toBeUndefined();
  });

-  it('fetchFor returns idle when service returns empty messageId', async () => {
+  it('fetchFor leaves slot idle when service returns empty messageId', async () => {
    vi.spyOn(followUpActionService, 'extract').mockResolvedValue({ chips: [], messageId: '' });
-    await useFollowUpActionStore.getState().fetchFor(TOPIC, FETCH_PARAMS);
-    expect(useFollowUpActionStore.getState().status).toBe('idle');
-    expect(useFollowUpActionStore.getState().messageId).toBeUndefined();
+    await useFollowUpActionStore.getState().fetchFor(KEY_A, FETCH_PARAMS_A);
+    expect(slotA().status).toBe('idle');
+    expect(slotA().messageId).toBeUndefined();
  });

-  it('fetchFor dedupes same topicId while still loading', async () => {
+  it('fetchFor dedupes while the slot is loading', async () => {
    const spy = vi
      .spyOn(followUpActionService, 'extract')
      .mockImplementation(() => new Promise(() => {}));
-    const p1 = useFollowUpActionStore.getState().fetchFor(TOPIC, FETCH_PARAMS);
-    const p2 = useFollowUpActionStore.getState().fetchFor(TOPIC, FETCH_PARAMS);
-    void p1;
-    void p2;
+    void useFollowUpActionStore.getState().fetchFor(KEY_A, FETCH_PARAMS_A);
+    void useFollowUpActionStore.getState().fetchFor(KEY_A, FETCH_PARAMS_A);
    expect(spy).toHaveBeenCalledTimes(1);
  });

-  it('fetchFor with new topicId aborts the old controller', async () => {
-    let firstSignal: AbortSignal | undefined;
-    vi.spyOn(followUpActionService, 'extract').mockImplementation(async (_, signal) => {
-      if (!firstSignal) firstSignal = signal;
+  it('fetchFor on a different key does not abort an in-flight fetch on another key', async () => {
+    let signalA: AbortSignal | undefined;
+    let signalB: AbortSignal | undefined;
+    vi.spyOn(followUpActionService, 'extract').mockImplementation(async (input, signal) => {
+      if (input.topicId === TOPIC_A) signalA = signal;
+      else signalB = signal;
      return new Promise(() => {});
    });
-    const p1 = useFollowUpActionStore.getState().fetchFor(TOPIC, FETCH_PARAMS);
-    void p1;
+    void useFollowUpActionStore.getState().fetchFor(KEY_A, FETCH_PARAMS_A);
    await Promise.resolve();
    await Promise.resolve();
-    void useFollowUpActionStore.getState().fetchFor(NEW_TOPIC, FETCH_PARAMS);
-    expect(firstSignal?.aborted).toBe(true);
+    void useFollowUpActionStore.getState().fetchFor(KEY_B, FETCH_PARAMS_B);
+    await Promise.resolve();
+    expect(signalA?.aborted).toBe(false);
+    expect(signalB?.aborted).toBe(false);
+    expect(slotA().status).toBe('loading');
+    expect(slotB().status).toBe('loading');
  });

-  it('clear() aborts and resets state', async () => {
+  it('clear(keyA) does not touch slots[keyB]', async () => {
    vi.spyOn(followUpActionService, 'extract').mockImplementation(() => new Promise(() => {}));
-    const p = useFollowUpActionStore.getState().fetchFor(TOPIC, FETCH_PARAMS);
-    void p;
-    useFollowUpActionStore.getState().clear();
-    expect(useFollowUpActionStore.getState().status).toBe('idle');
-    expect(useFollowUpActionStore.getState().messageId).toBeUndefined();
-    expect(useFollowUpActionStore.getState().pendingTopicId).toBeUndefined();
+    void useFollowUpActionStore.getState().fetchFor(KEY_A, FETCH_PARAMS_A);
+    void useFollowUpActionStore.getState().fetchFor(KEY_B, FETCH_PARAMS_B);
+    useFollowUpActionStore.getState().clear(KEY_A);
+    expect(slotA()).toBeUndefined();
+    expect(slotB()?.status).toBe('loading');
+  });
+
+  it('abort(keyA) does not affect keyB controller', async () => {
+    let signalA: AbortSignal | undefined;
+    let signalB: AbortSignal | undefined;
+    vi.spyOn(followUpActionService, 'extract').mockImplementation(async (input, signal) => {
+      if (input.topicId === TOPIC_A) signalA = signal;
+      else signalB = signal;
+      return new Promise(() => {});
+    });
+    void useFollowUpActionStore.getState().fetchFor(KEY_A, FETCH_PARAMS_A);
+    void useFollowUpActionStore.getState().fetchFor(KEY_B, FETCH_PARAMS_B);
+    await Promise.resolve();
+    useFollowUpActionStore.getState().abort(KEY_A);
+    expect(signalA?.aborted).toBe(true);
+    expect(signalB?.aborted).toBe(false);
+    expect(slotA().status).toBe('idle');
+    expect(slotB()?.status).toBe('loading');
  });

  it('20s timeout aborts the in-flight call', async () => {
@@ -112,23 +141,30 @@ describe('useFollowUpActionStore', () => {
      signal = s;
      return new Promise(() => {});
    });
-    const p = useFollowUpActionStore.getState().fetchFor(TOPIC, FETCH_PARAMS);
-    void p;
+    void useFollowUpActionStore.getState().fetchFor(KEY_A, FETCH_PARAMS_A);
    await Promise.resolve();
    vi.advanceTimersByTime(20_000);
    expect(signal?.aborted).toBe(true);
  });

-  it('consume(chip) clears state', () => {
+  it('consume(key, chip) clears the slot for that key only', () => {
    useFollowUpActionStore.setState({
-      chips: [{ label: 'x', message: 'hello' }],
-      messageId: MSG,
-      status: 'ready',
+      slots: {
+        [KEY_A]: {
+          chips: [{ label: 'x', message: 'hello' }],
+          messageId: MSG,
+          status: 'ready',
+        },
+        [KEY_B]: {
+          chips: [{ label: 'y', message: 'hello' }],
+          messageId: MSG,
+          status: 'ready',
+        },
+      },
    });
-    useFollowUpActionStore.getState().consume({ label: 'x', message: 'hello' });
-    expect(useFollowUpActionStore.getState().status).toBe('idle');
-    expect(useFollowUpActionStore.getState().messageId).toBeUndefined();
-    expect(useFollowUpActionStore.getState().chips).toHaveLength(0);
+    useFollowUpActionStore.getState().consume(KEY_A, { label: 'x', message: 'hello' });
+    expect(slotA()).toBeUndefined();
+    expect(slotB()?.status).toBe('ready');
  });

  it('discards stale results when controller is replaced (race protection)', async () => {
@@ -145,46 +181,121 @@ describe('useFollowUpActionStore', () => {
        messageId: 'msg-new',
      });

-    // First fetchFor is in flight (does not yet resolve).
-    const p1 = useFollowUpActionStore.getState().fetchFor(TOPIC, FETCH_PARAMS);
+    const p1 = useFollowUpActionStore.getState().fetchFor(KEY_A, FETCH_PARAMS_A);
    void p1;
    await Promise.resolve();

-    // User sends a new message → clear() aborts and resets.
-    useFollowUpActionStore.getState().clear();
-    expect(useFollowUpActionStore.getState().status).toBe('idle');
+    useFollowUpActionStore.getState().clear(KEY_A);
+    expect(slotA()).toBeUndefined();

-    // Next turn starts another fetchFor for the SAME topic.
-    const p2 = useFollowUpActionStore.getState().fetchFor(TOPIC, FETCH_PARAMS);
+    const p2 = useFollowUpActionStore.getState().fetchFor(KEY_A, FETCH_PARAMS_A);

-    // The first call now resolves with a stale result. It must be discarded
-    // because its controller is no longer the active one — even though the
-    // topicId still matches.
    resolveFirst!({ chips: [{ label: 'a', message: 'a' }], messageId: 'msg-old' });
    await p1;

-    expect(useFollowUpActionStore.getState().messageId).not.toBe('msg-old');
+    expect(slotA()?.messageId).not.toBe('msg-old');

-    // Second call still writes through normally.
    await p2;
    expect(spy).toHaveBeenCalledTimes(2);
-    expect(useFollowUpActionStore.getState().status).toBe('ready');
-    expect(useFollowUpActionStore.getState().messageId).toBe('msg-new');
+    expect(slotA()?.status).toBe('ready');
+    expect(slotA()?.messageId).toBe('msg-new');
  });

-  it('reset aborts in-flight request and resets state', async () => {
-    let signal: AbortSignal | undefined;
+  it('reset aborts all in-flight requests and clears every slot', async () => {
+    const signals: AbortSignal[] = [];
    vi.spyOn(followUpActionService, 'extract').mockImplementation(async (_, s) => {
-      signal = s;
+      if (s) signals.push(s);
      return new Promise(() => {});
    });
-    const p = useFollowUpActionStore.getState().fetchFor(TOPIC, FETCH_PARAMS);
-    void p;
+    void useFollowUpActionStore.getState().fetchFor(KEY_A, FETCH_PARAMS_A);
+    void useFollowUpActionStore.getState().fetchFor(KEY_B, FETCH_PARAMS_B);
    await Promise.resolve();
    useFollowUpActionStore.getState().reset();
-    expect(signal?.aborted).toBe(true);
-    expect(useFollowUpActionStore.getState().status).toBe('idle');
-    expect(useFollowUpActionStore.getState().messageId).toBeUndefined();
-    expect(useFollowUpActionStore.getState().pendingTopicId).toBeUndefined();
+    expect(signals.every((s) => s.aborted)).toBe(true);
+    expect(useFollowUpActionStore.getState().slots).toEqual({});
+  });
+});
+
+describe('followUpActionSelectors.chipsFor', () => {
+  beforeEach(() => {
+    useFollowUpActionStore.getState().reset?.();
+  });
+
+  it('returns chips when the slot matches messageId', () => {
+    useFollowUpActionStore.setState({
+      slots: {
+        [KEY_A]: {
+          chips: [{ label: 'a', message: 'a' }],
+          messageId: MSG,
+          status: 'ready',
+        },
+      },
+    });
+    const chips = followUpActionSelectors.chipsFor({
+      conversationKey: KEY_A,
+      messageId: MSG,
+    })(useFollowUpActionStore.getState());
+    expect(chips).toHaveLength(1);
+  });
+
+  it('returns empty when slot is missing', () => {
+    const chips = followUpActionSelectors.chipsFor({
+      conversationKey: KEY_A,
+      messageId: MSG,
+    })(useFollowUpActionStore.getState());
+    expect(chips).toHaveLength(0);
+  });
+
+  it('returns empty when slot is not ready', () => {
+    useFollowUpActionStore.setState({
+      slots: {
+        [KEY_A]: {
+          chips: [{ label: 'a', message: 'a' }],
+          messageId: MSG,
+          status: 'loading',
+        },
+      },
+    });
+    const chips = followUpActionSelectors.chipsFor({
+      conversationKey: KEY_A,
+      messageId: MSG,
+    })(useFollowUpActionStore.getState());
+    expect(chips).toHaveLength(0);
+  });
+
+  it('matches a child id via childIdsKey (assistantGroup case)', () => {
+    const CHILD = 'msg-child';
+    useFollowUpActionStore.setState({
+      slots: {
+        [KEY_A]: {
+          chips: [{ label: 'a', message: 'a' }],
+          messageId: CHILD,
+          status: 'ready',
+        },
+      },
+    });
+    const chips = followUpActionSelectors.chipsFor({
+      childIdsKey: `${CHILD}|other`,
+      conversationKey: KEY_A,
+      messageId: 'group-id',
+    })(useFollowUpActionStore.getState());
+    expect(chips).toHaveLength(1);
+  });
+
+  it('does not leak across conversation keys', () => {
+    useFollowUpActionStore.setState({
+      slots: {
+        [KEY_A]: {
+          chips: [{ label: 'a', message: 'a' }],
+          messageId: MSG,
+          status: 'ready',
+        },
+      },
+    });
+    const chips = followUpActionSelectors.chipsFor({
+      conversationKey: KEY_B,
+      messageId: MSG,
+    })(useFollowUpActionStore.getState());
+    expect(chips).toHaveLength(0);
  });
 });
@@ -2,16 +2,18 @@ import type { FollowUpChip } from '@lobechat/types';

 export type FollowUpActionStatus = 'idle' | 'loading' | 'ready';

-export interface FollowUpActionState {
+/** Per-conversation slot — concurrent surfaces (inbox, popup, thread) own their own slot. */
+export interface FollowUpActionSlot {
  abortController?: AbortController;
  chips: FollowUpChip[];
  messageId?: string;
-  pendingTopicId?: string;
  status: FollowUpActionStatus;
-  topicId?: string;
+}
+
+export interface FollowUpActionState {
+  slots: Record<string, FollowUpActionSlot>;
 }

 export const initialFollowUpActionState: FollowUpActionState = {
-  chips: [],
-  status: 'idle',
+  slots: {},
 };
@@ -1,43 +1,33 @@
 import type { FollowUpChip } from '@lobechat/types';

-import { type FollowUpActionState } from './initialState';
+import { type FollowUpActionState, type FollowUpActionStatus } from './initialState';

 const EMPTY_CHIPS: readonly FollowUpChip[] = [];

 interface ChipsForArgs {
-  /**
-   * Pipe-joined ids of the displayMessage's children blocks (for assistantGroup).
-   * Server-side resolves the latest answer message id, which inside an
-   * assistantGroup is a child block id rather than the group id, so we accept
-   * any child id as a valid match in addition to the top-level id.
-   */
+  /** Pipe-joined ids of the assistantGroup's child blocks — the server resolves the latest answer to a child block id, not the group id. */
  childIdsKey?: string;
+  conversationKey: string | undefined;
  messageId: string | undefined;
-  topicId: string | undefined;
 }

-/**
- * Chips render only when ALL hold:
- * - status === 'ready'
- * - the stored topicId matches
- * - the stored messageId matches the bound message id OR one of its child block ids
- *
- * Topic-only matching would let stale chips from a previous turn render under
- * a newly streaming assistant message in the same topic, so messageId membership
- * is required.
- */
 const chipsFor =
-  ({ childIdsKey, messageId, topicId }: ChipsForArgs) =>
+  ({ childIdsKey, conversationKey, messageId }: ChipsForArgs) =>
  (s: FollowUpActionState): readonly FollowUpChip[] => {
-    if (s.status !== 'ready') return EMPTY_CHIPS;
-    if (!messageId || !topicId) return EMPTY_CHIPS;
-    if (s.topicId !== topicId) return EMPTY_CHIPS;
-    if (!s.messageId) return EMPTY_CHIPS;
-    if (s.messageId === messageId) return s.chips;
-    if (childIdsKey && childIdsKey.split('|').includes(s.messageId)) return s.chips;
+    if (!conversationKey || !messageId) return EMPTY_CHIPS;
+    const slot = s.slots[conversationKey];
+    if (!slot || slot.status !== 'ready' || !slot.messageId) return EMPTY_CHIPS;
+    if (slot.messageId === messageId) return slot.chips;
+    if (childIdsKey && childIdsKey.split('|').includes(slot.messageId)) return slot.chips;
    return EMPTY_CHIPS;
  };

+const slotStatus =
+  (conversationKey: string | undefined) =>
+  (s: FollowUpActionState): FollowUpActionStatus =>
+    (conversationKey ? s.slots[conversationKey]?.status : undefined) ?? 'idle';
+
 export const followUpActionSelectors = {
  chipsFor,
+  slotStatus,
 };
@@ -28,22 +28,11 @@ class FollowUpActionStoreResetAction implements ResetableStore {
  }

  reset = () => {
-    // Cancel any in-flight LLM call before wiping state, otherwise the AbortController is leaked.
    const current = this.#api.getState();
-    current.abortController?.abort();
-    // Explicitly include undefined fields so zustand's merge-mode setState clears them.
-    this.#set(
-      {
-        abortController: undefined,
-        chips: [],
-        messageId: undefined,
-        pendingTopicId: undefined,
-        status: 'idle',
-        topicId: undefined,
-      },
-      false,
-      'resetFollowUpActionStore',
-    );
+    for (const slot of Object.values(current.slots)) {
+      slot.abortController?.abort();
+    }
+    this.#set({ slots: {} }, false, 'resetFollowUpActionStore');
  };
 }

@@ -55,6 +55,11 @@ exports[`settingsSelectors > currentSystemAgent > should merge DEFAULT_SYSTEM_AG
    "provider": "deepseek",
  },
  "enableAutoReply": true,
+  "followUpAction": {
+    "enabled": false,
+    "model": "gpt-5.4-mini",
+    "provider": "openai",
+  },
  "generationTopic": {
    "model": "gpt-5.4-mini",
    "provider": "openai",
@@ -107,6 +112,7 @@ exports[`settingsSelectors > defaultAgent > should merge DEFAULT_AGENT and s.set
      "enableAgentMode": true,
      "enableCompressHistory": true,
      "enableContextCompression": true,
+      "enableFollowUpChips": false,
      "enableHistoryCount": false,
      "enableStreaming": true,
      "historyCount": 20,
@@ -150,6 +156,7 @@ exports[`settingsSelectors > defaultAgentConfig > should merge DEFAULT_AGENT_CON
    "enableAgentMode": true,
    "enableCompressHistory": true,
    "enableContextCompression": true,
+    "enableFollowUpChips": false,
    "enableHistoryCount": false,
    "enableStreaming": true,
    "historyCount": 20,
@@ -15,9 +15,11 @@ const promptRewrite = (s: UserStore) => currentSystemAgent(s).promptRewrite;
 const historyCompress = (s: UserStore) => currentSystemAgent(s).historyCompress;
 const generationTopic = (s: UserStore) => currentSystemAgent(s).generationTopic;
 const inputCompletion = (s: UserStore) => currentSystemAgent(s).inputCompletion;
+const followUpAction = (s: UserStore) => currentSystemAgent(s).followUpAction;

 export const systemAgentSelectors = {
  agentMeta,
+  followUpAction,
  generationTopic,
  historyCompress,
  inputCompletion,