Implement mounted AI runtime and retire legacy Feishu brains

AGENTS.md

@@ -76,6 +76,8 @@ frontend/                  # React 前端
 
 根据用户的具体需求，生成高质量、可直接运行的代码。
 
+- 计划类需求执行约定：若用户已确认计划/方案并要求开始实现，Agent 默认连续执行到“实现完成 + 验证完成 + 给出验收结果”后再统一回复；除非遇到明确阻塞、高风险冲突、context 限制或用户主动中断，否则不得中途停下来反复请示。
+
 ---
 
 ## 8. 服务边界约束与 AI 系统宪章（禁止越界）

CLAUDE.md

@@ -5,6 +5,7 @@
 - `CLAUDE.md`：面向 Claude 风格的项目说明文档（补充背景）
 - 建议：规则放 `AGENTS.md`，架构/目录说明放 `CLAUDE.md`
 - 当前 AI 行为的系统宪章已在 `AGENTS.md` 中升级为“事实层 / 评审层 / 生成层 / 执行层”四层分治；所有新增 AI 业务默认按该宪章接入，`CLAUDE.md` 仅保留结构与背景说明
+- 新增执行约定：计划类需求一旦确认进入实现阶段，默认连续执行到“实现完成 + 验证完成 + 统一回复验收结果”；仅在明确阻塞、高风险冲突、context 限制或用户主动中断时才中途打断。
 
 ## 2. 项目定位
 - 项目名：`PMFlow / Neural Sync`
@@ -117,7 +118,7 @@ backend/
 - 新增：`backend/app/core/startup/model_probe.py`（启动期模型连通性探针：优先走 DB 里的实际引擎配置，失败不污染 provider circuit）
 - 新增：`backend/app/services/sstc/charter_payload.py`（轻量宪章上下文提取模块：供 Signal Planner / SSTC 共享，避免导入环）
 - 新增：`backend/app/services/task_breakdown.py`（任务拆解业务对象归一、技能标签回填、自动指派阈值判断）
-- 新增：`backend/app/services/ai_pmo/consultant_dispatch.py`（轻量顾问式统一调度合同与注册表：保留现有业务 Prompt 拆分，统一生成 `domain/mode/risk_level/missing_fields/allow_execute/next_action`）
+- 新增：`backend/app/services/ai_pmo/consultant_dispatch.py`（轻量顾问式统一调度合同与注册表：当前仅保留给 `/rag/ask` 等顾问式问答合同使用；已退出 Feishu mention 主回复与 AI PMO ingress capture 主链）
 - 新增：`backend/app/config/ai_prompt_registry.py`（prompt registry / manifest，统一登记 prompt 场景、调用入口与契约）
 - 更新：`backend/app/config/ai_prompts.py`（默认角色 Prompt 调整为“产品经理（立项评审）”，强调查重、MVP、前景与需求清单）
 - 更新：`backend/app/config/ai_prompts_catalog/part_02.py`（`PROMPT_DRAFT_GENERATOR` 增加 active_projects / similar_projects / MVP / prospect / requirements 评审字段）
@@ -195,13 +196,18 @@ backend/
 - 新增：`backend/app/services/ai_pmo/workflow_semantic_planner/`（统一业务语义规划层：唯一允许承载 workflow bundle 语义关键词/正则/兼容 helper 的目录）
 - 新增：`backend/app/models/reply_iteration.py`（回复质量迭代模型：`QualityAsset/QualityAssetPair/JudgeRun/OptimizerCandidate/PromptRelease`）
 - 新增：`backend/app/models/training_pipeline.py`（训练/微调管线模型：`TrainingDataset/TrainingJob/TrainingEvaluation/TrainedModelArtifact/TrainedModelDeployment`）
+- 新增：`backend/app/models/mounted_ai_runtime.py`（Mounted Runtime 运行时模型：`MountedTopicWorkspace/MountedDraftWorkspace/MountedConversationRun/MountedAgentTrace/MountedToolCapabilityPolicy`）
 - 新增：`backend/app/services/reply_iteration/service.py`（反馈中心提升为学习资产、裁判分桶、候选改动、prompt release trial 服务）
 - 新增：`backend/app/services/training_pipeline/service.py`（训练数据集、训练作业、评测、artifact promote、deployment 决策服务）
 - 新增：`backend/app/services/prompt_release/service.py`（只读 QA prompt release 版本管理与 serving 渲染优先级）
+- 新增：`backend/app/services/mounted_ai_runtime_service.py`、`backend/app/services/mounted_ai_runtime_write_service.py`（Mounted Runtime 读模型与写路径：来源/主题/run 查询 + ingress topic/run/trace 落库）
 - 新增：`backend/app/api/ai_steward_api/routes_reply_iteration.py`（AI Steward 回复质量闭环接口：反馈提升、judge-runs、candidates、prompt-releases、trials）
 - 新增：`backend/app/api/ai_steward_api/routes_training_pipeline.py`（AI Steward 训练管线接口：datasets、jobs、evaluations、artifacts、deployments）
+- 新增：`backend/app/api/ai_steward_api/routes_mounted_ai.py`、`backend/app/schemas/mounted_ai_runtime.py`（Mounted Runtime 控制台接口：sources/topics/runs/run-detail/cutover-status/eval-suites）
 - 新增：`backend/alembic/versions/20260327_add_reply_iteration_training_tables.py`（回复质量迭代/训练管线主迁移：学习资产表、prompt release、训练数据集、训练作业与 experiment 扩展字段）
+- 新增：`backend/alembic/versions/20260405_add_mounted_ai_runtime_tables.py`（Mounted Runtime 建表迁移：topic workspace、draft workspace、conversation run、agent trace、tool capability policy）
 - 新增：`backend/tests/test_reply_iteration_service.py`、`test_training_pipeline_service.py`、`test_reply_iteration_api.py`、`test_reply_iteration_training_migration.py`（回复质量闭环与训练管线基础回归）
+- 新增：`backend/tests/test_mounted_ai_runtime_models.py`、`test_mounted_ai_runtime_migration.py`、`test_mounted_ai_runtime_api.py`、`test_mounted_ai_runtime_write_service.py`（Mounted Runtime 模型/迁移/API/写路径回归）
 - 新增：`backend/app/services/ai_pmo/ai_pmo_kernel/ai_pmo_workflow_runtime.py`（AI PMO `ContextAssembler` + `WorkflowRegistry/WorkflowHandler` 运行时骨架）
 - 新增：`backend/app/services/ai_pmo_decision_runtime.py`（AI PMO 决策脑层：`Strategic/Tactical/Simulation/ExecutionPolicy/DecisionEngine`）
 - 新增：`backend/app/services/ai_pmo/ai_pmo_kernel/ai_pmo_reply_renderer.py`（AI PMO 回执渲染出口 `ReplyRenderer`）
@@ -217,7 +223,7 @@ backend/
 - 新增：`backend/app/services/ai_pmo/ai_pmo_kernel/mixins/ingress/ai_pmo_kernel_mixins_ingress_storage.py`（Kernel 入站事件/工作项写入）
 - 新增：`backend/app/services/ai_pmo/ai_pmo_kernel/mixins/ingress/ai_pmo_kernel_mixins_ingress_render.py`（Kernel 入站结果组装与输出辅助）
 - 更新：`backend/app/services/ai_pmo/ai_pmo_kernel_ingress_flow/mixin_auto.py`（写操作改为“先起草后确认”，状态进入 `draft_pending_confirm`）
-- 更新：`backend/app/services/ai_pmo/ai_pmo_kernel_ingress_flow/mixin_entry.py`（会话命中 `draft_pending_confirm` 时优先路由草稿确认处理）
+- 更新：`backend/app/services/ai_pmo/ai_pmo_kernel_ingress_flow/mixin_entry.py`（会话命中 `draft_pending_confirm` 时优先路由草稿确认处理，并把 ingress 主链同步写入 Mounted Runtime 的 topic/run/trace）
 - 更新：`backend/app/models/work_item.py`（新增 `draft_content/draft_expires_at/draft_version/confirm_message_id`）
 - 新增：`backend/app/services/ai_pmo/ai_pmo_kernel/compositions/world_model.py`（Kernel world-model 组合入口）
 - 新增：`backend/app/services/ai_pmo/ai_pmo_kernel_world_model_core/`（Kernel world-model 核心构建与缓存）
@@ -371,6 +377,7 @@ frontend/
 - 新增：`frontend/src/api/taskBriefs.ts`（任务包中心 API 封装）
 - 新增：`frontend/src/api/modelRouting.ts`（模型路由 API 封装）
 - 新增：`frontend/src/api/knowledgeGraph.ts`（KG 能力汇总与 low/high 消费验证 API）
+- 新增：`frontend/src/api/mountedAi.ts`（Mounted Runtime 控制台 API：sources/topics/runs/run-detail/cutover-status）
 - 更新：`frontend/src/api/aiSteward.ts`（移除 OpenClaw dispatch 封装，新增 LLM 引擎 CRUD / settings 封装）
 - 更新：`frontend/src/api/aiSteward.ts`（新增 AI 管家 QA 封装：`/rag/ask`、QA sessions、feedback、会话改名；浮窗不再直接 `fetch`）
 - 更新：`frontend/src/api/aiPmo.ts`（新增 `processText` 封装，供 AI 管家浮窗把“允许执行”显式交接到 AI PMO Kernel）
@@ -383,6 +390,8 @@ frontend/
 - 更新：`frontend/src/pages/CharterDetailPage.tsx`、`ContractSyncBoardPage.tsx`、`DashboardPage.tsx`、`KnowledgeGraphContextBoardPage.tsx`、`LoginPage.tsx`、`ModelRoutingConfigPage.tsx`、`SchedulerCenterPage.tsx`、`SkillTasksPage.tsx`、`SkillsPage.tsx`、`TaskBriefCenterPage.tsx`、`TaskBriefRunsPage.tsx`、`TaskBriefTemplatePage.tsx`、`UserProfilePage.tsx`、`UserSkillsPage.tsx`、`WorkloadDetailPage.tsx`、`WorkloadsPage.tsx`（统一降级为轻量入口，转发到 `pages/*/` 子目录）
 - 新增：`frontend/src/pages/charter-detail/CharterDetailPage.tsx`、`contract-sync-board/ContractSyncBoardPage.tsx`、`dashboard/DashboardPage.tsx`、`knowledge-graph-context-board/KnowledgeGraphContextBoardPage.tsx`、`login/LoginPage.tsx`、`model-routing-config/ModelRoutingConfigPage.tsx`（剩余核心页面目录化收敛）
 - 新增：`frontend/src/pages/scheduler-center/SchedulerCenterPage.tsx`、`skill-tasks/SkillTasksPage.tsx`、`skills/SkillsPage.tsx`、`task-brief-center/TaskBriefCenterPage.tsx`、`task-brief-runs/TaskBriefRunsPage.tsx`、`task-brief-template/TaskBriefTemplatePage.tsx`（任务/调度/技能页面目录化收敛）
+- 新增：`frontend/src/pages/scheduler-center/components/MountedAISourceStatusPanel.tsx`、`MountedAITopicsPanel.tsx`、`MountedAIRunsPanel.tsx`、`MountedAIRunReplayPanel.tsx`、`MountedAIAcceptancePanel.tsx`（Mounted AI Console 来源状态、topic workspace、run 列表、replay 详情与验收门槛面板）
+- 更新：`frontend/src/pages/scheduler-center/SchedulerCenterPage.tsx`（Scheduler Center 升级为 Mounted AI Console：接入 mounted runtime 查询、run replay 与 PLAN3 验收门槛展示）
 - 新增：`frontend/src/pages/user-profile/UserProfilePage.tsx`、`user-skills/UserSkillsPage.tsx`、`workload-detail/WorkloadDetailPage.tsx`、`workloads/WorkloadsPage.tsx`（用户与负载页面目录化收敛）
 - 更新：`frontend/src/pages/InboxPage.tsx`（降级为轻量入口，转发到 `pages/inbox/InboxPage`）
 - 新增：`frontend/src/pages/inbox/helpers.ts`（Inbox 页面 tab 过滤、排序、project-scope 判断 helper）
@@ -578,6 +587,7 @@ uvicorn app.main:app --reload --host 0.0.0.0 --port 8000
 - `model-routing` 只暴露逻辑层字段：`function_name/tier_alias/source/rule_name/rule_priority/last_execution`；`PUT` 仅接受 `tier_alias=low|high`
 - 回复质量闭环接口：参考 `backend/app/api/ai_steward_api/routes_reply_iteration.py`（`POST /ai-steward/feedback-center/items/{id}/promote`、`/reply-iteration/judge-runs/candidates/prompt-releases/trials`）
 - 训练/微调管线接口：参考 `backend/app/api/ai_steward_api/routes_training_pipeline.py`（`GET/POST /ai-steward/training/datasets/jobs`、`GET /training/jobs/{id}/evaluations`、`POST /training/artifacts/{id}/promote`、`POST /training/deployments/{id}/decision`）
+- Mounted Runtime 控制台接口：参考 `backend/app/api/ai_steward_api/routes_mounted_ai.py`（`GET /ai-steward/mounted-ai/sources|topics|runs|runs/{id}|cutover-status|eval-suites|eval-suites/{id}`，统一 `{code,message,data}` 包络）；其中 `cutover-status` 必须返回模块级旧脑退役视图：`legacy_cutover_state / retired_item_count / active_item_count / legacy_cutover_items[]`
 - 任务包中心接口：参考 `backend/app/api/task_briefs.py`（`/task-briefs/center/templates/runs/generate/contract-sync`）
 - KG 上下文闭环接口：参考 `backend/app/api/knowledge_graph.py`（`GET /ai-steward/knowledge-graph/capability-summary`、`POST /ai-steward/knowledge-graph/verify-context-consumption`、`POST /ai-steward/knowledge-graph/sync/work-items`；旧 `POST /ai-steward/knowledge-graph/sync/feishu` 仅保留兼容别名）
 - 需求点落信号源：参考 `backend/app/services/signal/signal_source_orchestration/mixin_registry_execute.py`（`materialize_requirement_sources` 负责归一、去重、`raw_context` 持久化，并为每条 registry 同步绑定/物化正式 `Signal`）
@@ -603,7 +613,7 @@ uvicorn app.main:app --reload --host 0.0.0.0 --port 8000
 - RAG 问答主入口契约（`POST /rag/ask`）：HTTP 入口已收敛到 `AIPMOKernelService.process_ingress` 主链；请求侧统一透传 `asked_via/action_mode/message_intent/mixed_signal_part/consultant_mode_preference`，由 `ingress_options.rag_request_context` 下传到 query executor 并映射到 `AdaptiveRAGService.answer_question`，返回侧从 ingress `rag_result`/`direct_reply_text` 组装兼容响应
 - RAG 空回答保护：`POST /rag/ask` 若 `answer` 与 `reply_text` 同时为空，必须返回失败包络，禁止再用“当前暂无可用回答。”之类兜底文案冒充成功结果
 - RAG Grounding Runtime 合同补充：`POST /rag/ask` 的成功返回统一透传 `answer_mode/grounding_status/policy_version/citations/sources_used/evidence_pack/tool_trace/retrieval_metrics`；`AdaptiveRAG` 内部允许通过兼容层把异构 `sources` 映射成稳定 citation 结构，但业务语义判断仍只允许放在 `workflow_semantic_planner`
-- RAG / AI 管家顾问合同：`POST /rag/ask` 返回体会透传 `consultant_dispatch={domain,mode,risk_level,missing_fields,allow_execute,next_action}`；该合同由 `backend/app/services/ai_pmo/consultant_dispatch.py` 统一生成，飞书 mention 回复与 AI 管家问答共享同一份模式判断
+- RAG / AI 管家顾问合同：`POST /rag/ask` 返回体会透传 `consultant_dispatch={domain,mode,risk_level,missing_fields,allow_execute,next_action}`；该合同由 `backend/app/services/ai_pmo/consultant_dispatch.py` 统一生成，当前仅供 RAG / AI Steward 顾问式问答透传，已不再驱动 Feishu mention 回复或 AI PMO ingress capture 主链
 - RAG 顾问模式偏好约束：`consultant_mode_preference` 只允许取 `clarify/advise/challenge/brainstorm/execute`，且仅影响回答风格与 prompt wrapper；它不能替代 `workflow_semantic_planner` 去决定 `query vs execute`
 - RAG / AI Steward 路由包（`backend/app/api/rag_api/__init__.py`、`backend/app/api/ai_steward_api/__init__.py`）不再重复挂包级 prefix；路由 canonical 由 `shared.py` 内的共享 router 统一承担，避免运行中出现 `/rag/rag/*`、`/ai-steward/ai-steward/*`
 - AI PMO 消息台列表接口（`GET /ai-pmo/work-items`）：返回 `latest_policy_gate_code` 与规范化 `approval_requirement{required,reason,message}`（来源于最新 `decision_records` 投影）

backend/alembic/versions/20260405_add_mounted_ai_runtime_tables.py

@@ -0,0 +1,261 @@
+"""add mounted ai runtime tables
+
+Revision ID: 20260405_add_mounted_ai_runtime_tables
+Revises: 20260331_add_qa_record_runtime_fields
+Create Date: 2026-04-05
+"""
+
+from __future__ import annotations
+
+from alembic import op
+import sqlalchemy as sa
+
+try:
+    from sqlalchemy.dialects.postgresql import JSONB
+except Exception:  # pragma: no cover
+    JSONB = sa.JSON
+
+
+revision = "20260405_add_mounted_ai_runtime_tables"
+down_revision = "20260331_add_qa_record_runtime_fields"
+branch_labels = None
+depends_on = None
+
+
+def _json_type(bind):
+    return JSONB() if bind.dialect.name == "postgresql" else sa.JSON()
+
+
+def _table_exists(bind, table_name: str) -> bool:
+    return sa.inspect(bind).has_table(table_name)
+
+
+def _index_exists(bind, table_name: str, index_name: str) -> bool:
+    inspector = sa.inspect(bind)
+    try:
+        indexes = inspector.get_indexes(table_name)
+    except Exception:
+        return False
+    return any(str(index.get("name")) == index_name for index in indexes)
+
+
+def _create_index_if_missing(bind, table_name: str, index_name: str, columns: list[str], *, unique: bool = False) -> None:
+    if not _index_exists(bind, table_name, index_name):
+        op.create_index(index_name, table_name, columns, unique=unique)
+
+
+def _base_columns() -> list[sa.Column]:
+    return [
+        sa.Column("tenant_id", sa.BigInteger(), nullable=False),
+        sa.Column("created_at", sa.TIMESTAMP(), nullable=False, server_default=sa.func.now()),
+        sa.Column("updated_at", sa.TIMESTAMP(), nullable=False, server_default=sa.func.now()),
+    ]
+
+
+def _create_mounted_topic_workspaces(bind) -> None:
+    if _table_exists(bind, "mounted_topic_workspaces"):
+        return
+    op.create_table(
+        "mounted_topic_workspaces",
+        sa.Column("id", sa.Integer(), primary_key=True, autoincrement=True),
+        sa.Column("topic_key", sa.VARCHAR(length=128), nullable=False),
+        sa.Column("source_system", sa.VARCHAR(length=32), nullable=False),
+        sa.Column("channel_ref", sa.VARCHAR(length=128), nullable=True),
+        sa.Column("thread_key", sa.VARCHAR(length=255), nullable=True),
+        sa.Column("root_message_id", sa.VARCHAR(length=128), nullable=True),
+        sa.Column("mode", sa.VARCHAR(length=24), nullable=False, server_default="chat"),
+        sa.Column("status", sa.VARCHAR(length=24), nullable=False, server_default="active"),
+        sa.Column("facts_json", _json_type(bind), nullable=True),
+        sa.Column("goals_json", _json_type(bind), nullable=True),
+        sa.Column("constraints_json", _json_type(bind), nullable=True),
+        sa.Column("options_json", _json_type(bind), nullable=True),
+        sa.Column("open_questions", _json_type(bind), nullable=True),
+        sa.Column("pending_actions", _json_type(bind), nullable=True),
+        sa.Column("bound_entities", _json_type(bind), nullable=True),
+        sa.Column("readiness_score", sa.Float(), nullable=True, server_default="0"),
+        sa.Column("last_message_id", sa.VARCHAR(length=128), nullable=True),
+        sa.Column("last_triggered_at", sa.TIMESTAMP(), nullable=True),
+        sa.Column("metadata_json", _json_type(bind), nullable=True),
+        *_base_columns(),
+        sa.UniqueConstraint(
+            "tenant_id",
+            "source_system",
+            "topic_key",
+            name="uq_mounted_topic_workspace_source_topic",
+        ),
+    )
+    for index_name, columns in (
+        ("ix_mounted_topic_workspaces_tenant_id", ["tenant_id"]),
+        ("ix_mounted_topic_workspaces_topic_key", ["topic_key"]),
+        ("ix_mounted_topic_workspaces_source_system", ["source_system"]),
+        ("ix_mounted_topic_workspaces_channel_ref", ["channel_ref"]),
+        ("ix_mounted_topic_workspaces_thread_key", ["thread_key"]),
+        ("ix_mounted_topic_workspaces_root_message_id", ["root_message_id"]),
+        ("ix_mounted_topic_workspaces_mode", ["mode"]),
+        ("ix_mounted_topic_workspaces_status", ["status"]),
+        ("ix_mounted_topic_workspaces_last_message_id", ["last_message_id"]),
+    ):
+        _create_index_if_missing(bind, "mounted_topic_workspaces", index_name, columns)
+
+
+def _create_mounted_draft_workspaces(bind) -> None:
+    if _table_exists(bind, "mounted_draft_workspaces"):
+        return
+    op.create_table(
+        "mounted_draft_workspaces",
+        sa.Column("id", sa.Integer(), primary_key=True, autoincrement=True),
+        sa.Column(
+            "topic_workspace_id",
+            sa.Integer(),
+            sa.ForeignKey("mounted_topic_workspaces.id", ondelete="CASCADE"),
+            nullable=False,
+        ),
+        sa.Column("state", sa.VARCHAR(length=24), nullable=False, server_default="collecting"),
+        sa.Column("draft_version", sa.Integer(), nullable=False, server_default="0"),
+        sa.Column("requirement_card", _json_type(bind), nullable=True),
+        sa.Column("draft_payload", _json_type(bind), nullable=True),
+        sa.Column("required_confirmations", _json_type(bind), nullable=True),
+        sa.Column("last_ready_at", sa.TIMESTAMP(), nullable=True),
+        sa.Column("confirm_message_id", sa.VARCHAR(length=128), nullable=True),
+        *_base_columns(),
+        sa.UniqueConstraint(
+            "tenant_id",
+            "topic_workspace_id",
+            name="uq_mounted_draft_workspace_topic",
+        ),
+    )
+    for index_name, columns in (
+        ("ix_mounted_draft_workspaces_tenant_id", ["tenant_id"]),
+        ("ix_mounted_draft_workspaces_topic_workspace_id", ["topic_workspace_id"]),
+        ("ix_mounted_draft_workspaces_state", ["state"]),
+        ("ix_mounted_draft_workspaces_confirm_message_id", ["confirm_message_id"]),
+    ):
+        _create_index_if_missing(bind, "mounted_draft_workspaces", index_name, columns)
+
+
+def _create_mounted_conversation_runs(bind) -> None:
+    if _table_exists(bind, "mounted_conversation_runs"):
+        return
+    op.create_table(
+        "mounted_conversation_runs",
+        sa.Column("id", sa.Integer(), primary_key=True, autoincrement=True),
+        sa.Column("trace_id", sa.VARCHAR(length=64), nullable=False),
+        sa.Column(
+            "topic_workspace_id",
+            sa.Integer(),
+            sa.ForeignKey("mounted_topic_workspaces.id", ondelete="SET NULL"),
+            nullable=True,
+        ),
+        sa.Column("source_system", sa.VARCHAR(length=32), nullable=False),
+        sa.Column("source_message_id", sa.VARCHAR(length=128), nullable=True),
+        sa.Column("mode_before", sa.VARCHAR(length=24), nullable=True),
+        sa.Column("mode_after", sa.VARCHAR(length=24), nullable=True),
+        sa.Column("active_agent", sa.VARCHAR(length=64), nullable=True),
+        sa.Column("status", sa.VARCHAR(length=24), nullable=False, server_default="queued"),
+        sa.Column("trigger_reason", sa.VARCHAR(length=128), nullable=True),
+        sa.Column("context_layers_hit", _json_type(bind), nullable=True),
+        sa.Column("tool_calls", _json_type(bind), nullable=True),
+        sa.Column("final_reply", sa.Text(), nullable=True),
+        sa.Column("execution_outcome", _json_type(bind), nullable=True),
+        sa.Column("error_text", sa.Text(), nullable=True),
+        sa.Column("started_at", sa.TIMESTAMP(), nullable=True),
+        sa.Column("completed_at", sa.TIMESTAMP(), nullable=True),
+        *_base_columns(),
+        sa.UniqueConstraint("tenant_id", "trace_id", name="uq_mounted_conversation_run_trace"),
+    )
+    for index_name, columns in (
+        ("ix_mounted_conversation_runs_tenant_id", ["tenant_id"]),
+        ("ix_mounted_conversation_runs_trace_id", ["trace_id"]),
+        ("ix_mounted_conversation_runs_topic_workspace_id", ["topic_workspace_id"]),
+        ("ix_mounted_conversation_runs_source_system", ["source_system"]),
+        ("ix_mounted_conversation_runs_source_message_id", ["source_message_id"]),
+        ("ix_mounted_conversation_runs_mode_before", ["mode_before"]),
+        ("ix_mounted_conversation_runs_mode_after", ["mode_after"]),
+        ("ix_mounted_conversation_runs_active_agent", ["active_agent"]),
+        ("ix_mounted_conversation_runs_status", ["status"]),
+        ("ix_mounted_conversation_runs_trigger_reason", ["trigger_reason"]),
+    ):
+        _create_index_if_missing(bind, "mounted_conversation_runs", index_name, columns)
+
+
+def _create_mounted_agent_traces(bind) -> None:
+    if _table_exists(bind, "mounted_agent_traces"):
+        return
+    op.create_table(
+        "mounted_agent_traces",
+        sa.Column("id", sa.Integer(), primary_key=True, autoincrement=True),
+        sa.Column(
+            "conversation_run_id",
+            sa.Integer(),
+            sa.ForeignKey("mounted_conversation_runs.id", ondelete="CASCADE"),
+            nullable=False,
+        ),
+        sa.Column("agent_name", sa.VARCHAR(length=64), nullable=False),
+        sa.Column("status", sa.VARCHAR(length=24), nullable=False, server_default="completed"),
+        sa.Column("latency_ms", sa.Integer(), nullable=True),
+        sa.Column("input_summary", _json_type(bind), nullable=True),
+        sa.Column("output_summary", _json_type(bind), nullable=True),
+        sa.Column("tool_trace", _json_type(bind), nullable=True),
+        sa.Column("failure_reason", sa.Text(), nullable=True),
+        *_base_columns(),
+    )
+    for index_name, columns in (
+        ("ix_mounted_agent_traces_tenant_id", ["tenant_id"]),
+        ("ix_mounted_agent_traces_conversation_run_id", ["conversation_run_id"]),
+        ("ix_mounted_agent_traces_agent_name", ["agent_name"]),
+        ("ix_mounted_agent_traces_status", ["status"]),
+    ):
+        _create_index_if_missing(bind, "mounted_agent_traces", index_name, columns)
+
+
+def _create_mounted_tool_capability_policies(bind) -> None:
+    if _table_exists(bind, "mounted_tool_capability_policies"):
+        return
+    op.create_table(
+        "mounted_tool_capability_policies",
+        sa.Column("id", sa.Integer(), primary_key=True, autoincrement=True),
+        sa.Column("tool_name", sa.VARCHAR(length=128), nullable=False),
+        sa.Column("risk_level", sa.VARCHAR(length=24), nullable=False, server_default="medium"),
+        sa.Column("confirm_required", sa.Boolean(), nullable=False, server_default=sa.true()),
+        sa.Column("auto_execute_allowed", sa.Boolean(), nullable=False, server_default=sa.false()),
+        sa.Column("is_active", sa.Boolean(), nullable=False, server_default=sa.true()),
+        sa.Column("source_scope", _json_type(bind), nullable=True),
+        sa.Column("tenant_scope", _json_type(bind), nullable=True),
+        sa.Column("policy_metadata", _json_type(bind), nullable=True),
+        sa.Column("notes", sa.Text(), nullable=True),
+        *_base_columns(),
+        sa.UniqueConstraint(
+            "tenant_id",
+            "tool_name",
+            name="uq_mounted_tool_capability_policy_tool",
+        ),
+    )
+    for index_name, columns in (
+        ("ix_mounted_tool_capability_policies_tenant_id", ["tenant_id"]),
+        ("ix_mounted_tool_capability_policies_tool_name", ["tool_name"]),
+        ("ix_mounted_tool_capability_policies_risk_level", ["risk_level"]),
+        ("ix_mounted_tool_capability_policies_is_active", ["is_active"]),
+    ):
+        _create_index_if_missing(bind, "mounted_tool_capability_policies", index_name, columns)
+
+
+def upgrade() -> None:
+    bind = op.get_bind()
+    _create_mounted_topic_workspaces(bind)
+    _create_mounted_draft_workspaces(bind)
+    _create_mounted_conversation_runs(bind)
+    _create_mounted_agent_traces(bind)
+    _create_mounted_tool_capability_policies(bind)
+
+
+def downgrade() -> None:
+    bind = op.get_bind()
+    for table_name in (
+        "mounted_tool_capability_policies",
+        "mounted_agent_traces",
+        "mounted_conversation_runs",
+        "mounted_draft_workspaces",
+        "mounted_topic_workspaces",
+    ):
+        if _table_exists(bind, table_name):
+            op.drop_table(table_name)

backend/app/api/ai_steward_api/__init__.py

@@ -8,6 +8,7 @@ from . import (
     routes_conflicts,
     routes_executive_briefing,
     routes_feedback_center,
+    routes_mounted_ai,
     routes_reply_iteration,
     routes_rag_sources,
     routes_stats_config,
@@ -22,6 +23,7 @@ for _module in [
     routes_conflicts,
     routes_executive_briefing,
     routes_feedback_center,
+    routes_mounted_ai,
     routes_reply_iteration,
     routes_brief_blockers,
     routes_stats_config,

backend/app/api/ai_steward_api/routes_mounted_ai.py

@@ -0,0 +1,127 @@
+from __future__ import annotations
+
+from fastapi import Depends, HTTPException, Query
+from sqlalchemy.ext.asyncio import AsyncSession
+
+from ...core.response import ApiResponse, success_response
+from ...dependencies.permissions import require_manager_user
+from ...dependencies.tenant import get_tenant_id
+from ...models import User
+from ...schemas.mounted_ai_runtime import (
+    MountedAICutoverStatusPayload,
+    MountedAIEvalSuiteDetailPayload,
+    MountedAIEvalSuiteListPayload,
+    MountedAIRunDetailPayload,
+    MountedAIRunListPayload,
+    MountedAISourceListPayload,
+    MountedAITopicListPayload,
+)
+from ...services.mounted_ai_runtime_service import MountedAIRuntimeService
+from .shared import *
+
+
+@router.get("/mounted-ai/sources", response_model=ApiResponse[MountedAISourceListPayload])
+async def list_mounted_ai_sources(
+    current_user: User = Depends(require_manager_user),
+    db: AsyncSession = Depends(get_db),
+    tenant_id: int = Depends(get_tenant_id),
+):
+    effective_tenant_id = int(getattr(current_user, "tenant_id", tenant_id) or tenant_id)
+    data = await MountedAIRuntimeService(db).list_sources(tenant_id=effective_tenant_id)
+    return success_response(data)
+
+
+@router.get("/mounted-ai/topics", response_model=ApiResponse[MountedAITopicListPayload])
+async def list_mounted_ai_topics(
+    mode: str | None = Query(default=None),
+    limit: int = Query(default=20, ge=1, le=100),
+    offset: int = Query(default=0, ge=0),
+    current_user: User = Depends(require_manager_user),
+    db: AsyncSession = Depends(get_db),
+    tenant_id: int = Depends(get_tenant_id),
+):
+    effective_tenant_id = int(getattr(current_user, "tenant_id", tenant_id) or tenant_id)
+    data = await MountedAIRuntimeService(db).list_topics(
+        tenant_id=effective_tenant_id,
+        limit=limit,
+        offset=offset,
+        mode=mode,
+    )
+    return success_response(data)
+
+
+@router.get("/mounted-ai/runs", response_model=ApiResponse[MountedAIRunListPayload])
+async def list_mounted_ai_runs(
+    status: str | None = Query(default=None),
+    limit: int = Query(default=20, ge=1, le=100),
+    offset: int = Query(default=0, ge=0),
+    current_user: User = Depends(require_manager_user),
+    db: AsyncSession = Depends(get_db),
+    tenant_id: int = Depends(get_tenant_id),
+):
+    effective_tenant_id = int(getattr(current_user, "tenant_id", tenant_id) or tenant_id)
+    data = await MountedAIRuntimeService(db).list_runs(
+        tenant_id=effective_tenant_id,
+        limit=limit,
+        offset=offset,
+        status=status,
+    )
+    return success_response(data)
+
+
+@router.get("/mounted-ai/runs/{run_id}", response_model=ApiResponse[MountedAIRunDetailPayload])
+async def get_mounted_ai_run_detail(
+    run_id: int,
+    current_user: User = Depends(require_manager_user),
+    db: AsyncSession = Depends(get_db),
+    tenant_id: int = Depends(get_tenant_id),
+):
+    effective_tenant_id = int(getattr(current_user, "tenant_id", tenant_id) or tenant_id)
+    try:
+        data = await MountedAIRuntimeService(db).get_run_detail(
+            tenant_id=effective_tenant_id,
+            run_id=run_id,
+        )
+    except ValueError as exc:
+        raise HTTPException(status_code=404, detail=str(exc)) from exc
+    return success_response(data)
+
+
+@router.get("/mounted-ai/cutover-status", response_model=ApiResponse[MountedAICutoverStatusPayload])
+async def get_mounted_ai_cutover_status(
+    current_user: User = Depends(require_manager_user),
+    db: AsyncSession = Depends(get_db),
+    tenant_id: int = Depends(get_tenant_id),
+):
+    effective_tenant_id = int(getattr(current_user, "tenant_id", tenant_id) or tenant_id)
+    data = await MountedAIRuntimeService(db).get_cutover_status(tenant_id=effective_tenant_id)
+    return success_response(data)
+
+
+@router.get("/mounted-ai/eval-suites", response_model=ApiResponse[MountedAIEvalSuiteListPayload])
+async def list_mounted_ai_eval_suites(
+    current_user: User = Depends(require_manager_user),
+    db: AsyncSession = Depends(get_db),
+    tenant_id: int = Depends(get_tenant_id),
+):
+    effective_tenant_id = int(getattr(current_user, "tenant_id", tenant_id) or tenant_id)
+    data = await MountedAIRuntimeService(db).list_eval_suites(tenant_id=effective_tenant_id)
+    return success_response(data)
+
+
+@router.get("/mounted-ai/eval-suites/{dataset_id}", response_model=ApiResponse[MountedAIEvalSuiteDetailPayload])
+async def get_mounted_ai_eval_suite_detail(
+    dataset_id: int,
+    current_user: User = Depends(require_manager_user),
+    db: AsyncSession = Depends(get_db),
+    tenant_id: int = Depends(get_tenant_id),
+):
+    effective_tenant_id = int(getattr(current_user, "tenant_id", tenant_id) or tenant_id)
+    try:
+        data = await MountedAIRuntimeService(db).get_eval_suite_detail(
+            tenant_id=effective_tenant_id,
+            dataset_id=dataset_id,
+        )
+    except ValueError as exc:
+        raise HTTPException(status_code=404, detail=str(exc)) from exc
+    return success_response(data)

backend/app/models/__init__.py

@@ -73,6 +73,13 @@ from .plan_bundle import PlanBundle
 from .reply_receipt import ReplyReceipt
 from .portfolio_snapshot import PortfolioSnapshot
 from .conversation_session import ConversationSession
+from .mounted_ai_runtime import (
+    MountedAgentTrace,
+    MountedConversationRun,
+    MountedDraftWorkspace,
+    MountedToolCapabilityPolicy,
+    MountedTopicWorkspace,
+)
 
 __all__ = [
     "Base",
@@ -153,4 +160,9 @@ __all__ = [
     "ReplyReceipt",
     "PortfolioSnapshot",
     "ConversationSession",
+    "MountedAgentTrace",
+    "MountedConversationRun",
+    "MountedDraftWorkspace",
+    "MountedToolCapabilityPolicy",
+    "MountedTopicWorkspace",
 ]

backend/app/models/mounted_ai_runtime.py

@@ -0,0 +1,149 @@
+"""Mounted AI Runtime 核心运行时对象。"""
+
+from __future__ import annotations
+
+from typing import Any, Optional
+
+from sqlalchemy import Boolean, Float, ForeignKey, Integer, Text, TIMESTAMP, UniqueConstraint, VARCHAR
+from sqlalchemy.orm import Mapped, mapped_column
+
+from .base import Base
+from .types import JSONB
+
+
+class MountedTopicWorkspace(Base):
+    __tablename__ = "mounted_topic_workspaces"
+    __table_args__ = (
+        UniqueConstraint(
+            "tenant_id",
+            "source_system",
+            "topic_key",
+            name="uq_mounted_topic_workspace_source_topic",
+        ),
+    )
+
+    id: Mapped[int] = mapped_column(Integer, primary_key=True, autoincrement=True)
+    topic_key: Mapped[str] = mapped_column(VARCHAR(128), nullable=False, index=True)
+    source_system: Mapped[str] = mapped_column(VARCHAR(32), nullable=False, index=True)
+    channel_ref: Mapped[Optional[str]] = mapped_column(VARCHAR(128), nullable=True, index=True)
+    thread_key: Mapped[Optional[str]] = mapped_column(VARCHAR(255), nullable=True, index=True)
+    root_message_id: Mapped[Optional[str]] = mapped_column(VARCHAR(128), nullable=True, index=True)
+    mode: Mapped[str] = mapped_column(VARCHAR(24), nullable=False, default="chat", index=True)
+    status: Mapped[str] = mapped_column(VARCHAR(24), nullable=False, default="active", index=True)
+    facts_json: Mapped[Optional[Any]] = mapped_column(JSONB, nullable=True)
+    goals_json: Mapped[Optional[Any]] = mapped_column(JSONB, nullable=True)
+    constraints_json: Mapped[Optional[Any]] = mapped_column(JSONB, nullable=True)
+    options_json: Mapped[Optional[Any]] = mapped_column(JSONB, nullable=True)
+    open_questions: Mapped[Optional[Any]] = mapped_column(JSONB, nullable=True)
+    pending_actions: Mapped[Optional[Any]] = mapped_column(JSONB, nullable=True)
+    bound_entities: Mapped[Optional[Any]] = mapped_column(JSONB, nullable=True)
+    readiness_score: Mapped[Optional[float]] = mapped_column(Float, nullable=True, default=0.0)
+    last_message_id: Mapped[Optional[str]] = mapped_column(VARCHAR(128), nullable=True, index=True)
+    last_triggered_at: Mapped[Optional[TIMESTAMP]] = mapped_column(TIMESTAMP, nullable=True)
+    metadata_json: Mapped[Optional[Any]] = mapped_column(JSONB, nullable=True)
+
+
+class MountedDraftWorkspace(Base):
+    __tablename__ = "mounted_draft_workspaces"
+    __table_args__ = (
+        UniqueConstraint(
+            "tenant_id",
+            "topic_workspace_id",
+            name="uq_mounted_draft_workspace_topic",
+        ),
+    )
+
+    id: Mapped[int] = mapped_column(Integer, primary_key=True, autoincrement=True)
+    topic_workspace_id: Mapped[int] = mapped_column(
+        Integer,
+        ForeignKey("mounted_topic_workspaces.id", ondelete="CASCADE"),
+        nullable=False,
+        index=True,
+    )
+    state: Mapped[str] = mapped_column(VARCHAR(24), nullable=False, default="collecting", index=True)
+    draft_version: Mapped[int] = mapped_column(Integer, nullable=False, default=0)
+    requirement_card: Mapped[Optional[Any]] = mapped_column(JSONB, nullable=True)
+    draft_payload: Mapped[Optional[Any]] = mapped_column(JSONB, nullable=True)
+    required_confirmations: Mapped[Optional[Any]] = mapped_column(JSONB, nullable=True)
+    last_ready_at: Mapped[Optional[TIMESTAMP]] = mapped_column(TIMESTAMP, nullable=True)
+    confirm_message_id: Mapped[Optional[str]] = mapped_column(VARCHAR(128), nullable=True, index=True)
+
+
+class MountedConversationRun(Base):
+    __tablename__ = "mounted_conversation_runs"
+    __table_args__ = (
+        UniqueConstraint("tenant_id", "trace_id", name="uq_mounted_conversation_run_trace"),
+    )
+
+    id: Mapped[int] = mapped_column(Integer, primary_key=True, autoincrement=True)
+    trace_id: Mapped[str] = mapped_column(VARCHAR(64), nullable=False, index=True)
+    topic_workspace_id: Mapped[Optional[int]] = mapped_column(
+        Integer,
+        ForeignKey("mounted_topic_workspaces.id", ondelete="SET NULL"),
+        nullable=True,
+        index=True,
+    )
+    source_system: Mapped[str] = mapped_column(VARCHAR(32), nullable=False, index=True)
+    source_message_id: Mapped[Optional[str]] = mapped_column(VARCHAR(128), nullable=True, index=True)
+    mode_before: Mapped[Optional[str]] = mapped_column(VARCHAR(24), nullable=True, index=True)
+    mode_after: Mapped[Optional[str]] = mapped_column(VARCHAR(24), nullable=True, index=True)
+    active_agent: Mapped[Optional[str]] = mapped_column(VARCHAR(64), nullable=True, index=True)
+    status: Mapped[str] = mapped_column(VARCHAR(24), nullable=False, default="queued", index=True)
+    trigger_reason: Mapped[Optional[str]] = mapped_column(VARCHAR(128), nullable=True, index=True)
+    context_layers_hit: Mapped[Optional[Any]] = mapped_column(JSONB, nullable=True)
+    tool_calls: Mapped[Optional[Any]] = mapped_column(JSONB, nullable=True)
+    final_reply: Mapped[Optional[str]] = mapped_column(Text, nullable=True)
+    execution_outcome: Mapped[Optional[Any]] = mapped_column(JSONB, nullable=True)
+    error_text: Mapped[Optional[str]] = mapped_column(Text, nullable=True)
+    started_at: Mapped[Optional[TIMESTAMP]] = mapped_column(TIMESTAMP, nullable=True)
+    completed_at: Mapped[Optional[TIMESTAMP]] = mapped_column(TIMESTAMP, nullable=True)
+
+
+class MountedAgentTrace(Base):
+    __tablename__ = "mounted_agent_traces"
+
+    id: Mapped[int] = mapped_column(Integer, primary_key=True, autoincrement=True)
+    conversation_run_id: Mapped[int] = mapped_column(
+        Integer,
+        ForeignKey("mounted_conversation_runs.id", ondelete="CASCADE"),
+        nullable=False,
+        index=True,
+    )
+    agent_name: Mapped[str] = mapped_column(VARCHAR(64), nullable=False, index=True)
+    status: Mapped[str] = mapped_column(VARCHAR(24), nullable=False, default="completed", index=True)
+    latency_ms: Mapped[Optional[int]] = mapped_column(Integer, nullable=True)
+    input_summary: Mapped[Optional[Any]] = mapped_column(JSONB, nullable=True)
+    output_summary: Mapped[Optional[Any]] = mapped_column(JSONB, nullable=True)
+    tool_trace: Mapped[Optional[Any]] = mapped_column(JSONB, nullable=True)
+    failure_reason: Mapped[Optional[str]] = mapped_column(Text, nullable=True)
+
+
+class MountedToolCapabilityPolicy(Base):
+    __tablename__ = "mounted_tool_capability_policies"
+    __table_args__ = (
+        UniqueConstraint(
+            "tenant_id",
+            "tool_name",
+            name="uq_mounted_tool_capability_policy_tool",
+        ),
+    )
+
+    id: Mapped[int] = mapped_column(Integer, primary_key=True, autoincrement=True)
+    tool_name: Mapped[str] = mapped_column(VARCHAR(128), nullable=False, index=True)
+    risk_level: Mapped[str] = mapped_column(VARCHAR(24), nullable=False, default="medium", index=True)
+    confirm_required: Mapped[bool] = mapped_column(Boolean, nullable=False, default=True)
+    auto_execute_allowed: Mapped[bool] = mapped_column(Boolean, nullable=False, default=False)
+    is_active: Mapped[bool] = mapped_column(Boolean, nullable=False, default=True, index=True)
+    source_scope: Mapped[Optional[Any]] = mapped_column(JSONB, nullable=True)
+    tenant_scope: Mapped[Optional[Any]] = mapped_column(JSONB, nullable=True)
+    policy_metadata: Mapped[Optional[Any]] = mapped_column(JSONB, nullable=True)
+    notes: Mapped[Optional[str]] = mapped_column(Text, nullable=True)
+
+
+__all__ = [
+    "MountedAgentTrace",
+    "MountedConversationRun",
+    "MountedDraftWorkspace",
+    "MountedToolCapabilityPolicy",
+    "MountedTopicWorkspace",
+]

backend/app/schemas/mounted_ai_runtime.py

@@ -0,0 +1,150 @@
+from __future__ import annotations
+
+from typing import Any, Optional
+
+from pydantic import BaseModel, ConfigDict, Field
+
+
+class MountedAISourceItem(BaseModel):
+    model_config = ConfigDict(extra="ignore")
+
+    source_system: str
+    run_count: int = 0
+    topic_count: int = 0
+    active_topic_count: int = 0
+    last_seen_at: Optional[str] = None
+
+
+class MountedAISourceListPayload(BaseModel):
+    model_config = ConfigDict(extra="ignore")
+
+    items: list[MountedAISourceItem] = Field(default_factory=list)
+
+
+class MountedAITopicItem(BaseModel):
+    model_config = ConfigDict(extra="ignore")
+
+    id: int
+    topic_key: str
+    source_system: str
+    channel_ref: Optional[str] = None
+    thread_key: Optional[str] = None
+    mode: str
+    status: str
+    readiness_score: Optional[float] = None
+    draft_state: Optional[str] = None
+    draft_version: Optional[int] = None
+    updated_at: Optional[str] = None
+
+
+class MountedAITopicListPayload(BaseModel):
+    model_config = ConfigDict(extra="ignore")
+
+    items: list[MountedAITopicItem] = Field(default_factory=list)
+    total: int = 0
+
+
+class MountedAIRunItem(BaseModel):
+    model_config = ConfigDict(extra="ignore")
+
+    id: int
+    trace_id: str
+    source_system: str
+    source_message_id: Optional[str] = None
+    mode_before: Optional[str] = None
+    mode_after: Optional[str] = None
+    active_agent: Optional[str] = None
+    status: str
+    trigger_reason: Optional[str] = None
+    final_reply: Optional[str] = None
+    trace_count: int = 0
+    updated_at: Optional[str] = None
+
+
+class MountedAIRunListPayload(BaseModel):
+    model_config = ConfigDict(extra="ignore")
+
+    items: list[MountedAIRunItem] = Field(default_factory=list)
+    total: int = 0
+
+
+class MountedAIAgentTraceItem(BaseModel):
+    model_config = ConfigDict(extra="ignore")
+
+    id: int
+    agent_name: str
+    status: str
+    latency_ms: Optional[int] = None
+    failure_reason: Optional[str] = None
+    created_at: Optional[str] = None
+
+
+class MountedAIRunDetailPayload(BaseModel):
+    model_config = ConfigDict(extra="ignore")
+
+    item: MountedAIRunItem
+    agent_traces: list[MountedAIAgentTraceItem] = Field(default_factory=list)
+    context_layers_hit: Optional[Any] = None
+    tool_calls: Optional[Any] = None
+    execution_outcome: Optional[Any] = None
+
+
+class MountedAICutoverStatusPayload(BaseModel):
+    model_config = ConfigDict(extra="ignore")
+
+    mounted_runtime_ready: bool
+    topic_workspace_count: int = 0
+    conversation_run_count: int = 0
+    tool_policy_count: int = 0
+    observed_source_systems: list[str] = Field(default_factory=list)
+    legacy_cutover_state: str
+    retired_item_count: int = 0
+    active_item_count: int = 0
+    legacy_cutover_items: list["MountedAILegacyCutoverItem"] = Field(default_factory=list)
+
+
+class MountedAILegacyCutoverItem(BaseModel):
+    model_config = ConfigDict(extra="ignore")
+
+    key: str
+    title: str
+    status: str
+    summary: str
+    evidence: str
+
+
+class MountedAIEvalSuiteItem(BaseModel):
+    model_config = ConfigDict(extra="ignore")
+
+    id: int
+    dataset_type: str
+    version: str
+    status: str
+    suite_kind: str
+    sample_count: int = 0
+    latest_score: Optional[float] = None
+    last_evaluated_at: Optional[str] = None
+
+
+class MountedAIEvalSuiteListPayload(BaseModel):
+    model_config = ConfigDict(extra="ignore")
+
+    items: list[MountedAIEvalSuiteItem] = Field(default_factory=list)
+    total: int = 0
+
+
+class MountedAIEvalResultItem(BaseModel):
+    model_config = ConfigDict(extra="ignore")
+
+    evaluation_type: str
+    status: str
+    score: Optional[float] = None
+
+
+class MountedAIEvalSuiteDetailPayload(BaseModel):
+    model_config = ConfigDict(extra="ignore")
+
+    item: MountedAIEvalSuiteItem
+    selection_rules: Optional[Any] = None
+    notes: Optional[str] = None
+    evaluations: list[MountedAIEvalResultItem] = Field(default_factory=list)

backend/app/services/ai_pmo/ai_pmo_kernel/mixins/ingress/ai_pmo_kernel_mixins_ingress_render.py

@@ -3,7 +3,6 @@
 from __future__ import annotations
 
 from .... import ai_pmo_kernel_common_core as _common
-from ....consultant_dispatch import build_consultant_dispatch_contract
 from ....workflow_semantic_planner import WorkflowSemanticPlanner
 
 for _name, _value in vars(_common).items():
@@ -25,17 +24,7 @@ class KernelIngressRenderMixin:
     ) -> dict[str, Any]:
         linked_project_ids = [int(work_item.project_id)] if work_item.project_id is not None else []
         direct_reply_text = str(render.get("text") or "").strip()
-        primary_operation_code = self._primary_operation_code(str(intent.name or ""))
-        risk_level = self._risk_level_for_intent(str(intent.name or ""))
-        missing_fields = (
-            execution_result.get("missing_fields")
-            if isinstance(execution_result.get("missing_fields"), list)
-            else (
-                (intent.metadata or {}).get("missing_fields")
-                if isinstance(intent.metadata, dict) and isinstance((intent.metadata or {}).get("missing_fields"), list)
-                else []
-            )
-        )
+        workflow_route = str(work_item.intent or execution_result.get("workflow_route") or "").strip() or None
         mixed_sidecar_reply = (
             str((intent.metadata or {}).get("mixed_sidecar_reply") or "").strip()
             if isinstance(intent.metadata, dict)
@@ -62,17 +51,6 @@ class KernelIngressRenderMixin:
             signal_ids = [primary_signal_id, *signal_ids]
         if signal_ids and primary_signal_id is None:
             primary_signal_id = signal_ids[0]
-        consultant_dispatch = build_consultant_dispatch_contract(
-            intent_name=str(intent.name or ""),
-            message_text=str(getattr(work_item, "raw_message", "") or direct_reply_text or intent.signal_summary or ""),
-            source_system=str(getattr(work_item, "source_system", "") or "pmflow"),
-            status=status,
-            classification=classification,
-            primary_operation_code=primary_operation_code,
-            risk_level=risk_level,
-            missing_fields=missing_fields,
-            direct_reply_text=direct_reply_text,
-        )
         result = {
             "handled": True,
             "status": status,
@@ -97,10 +75,9 @@ class KernelIngressRenderMixin:
                 }
             ],
             "intent": intent.name,
-            "workflow_route": str(work_item.intent or execution_result.get("workflow_route") or "").strip() or None,
+            "workflow_route": workflow_route,
             "direct_reply_required": bool(direct_reply_required),
             "direct_reply_text": direct_reply_text,
-            "consultant_dispatch": consultant_dispatch,
             "reply_template_key": str(render.get("template_key") or "").strip(),
             "work_item_id": int(work_item.id),
             "decision_level": work_item.decision_level,
@@ -121,7 +98,7 @@ class KernelIngressRenderMixin:
                 intent_name=str(intent.name or ""),
                 message_text=str(getattr(work_item, "raw_message", "") or direct_reply_text or intent.signal_summary or ""),
                 source_system=str(getattr(work_item, "source_system", "") or "unknown"),
-                workflow_route=str(work_item.intent or execution_result.get("workflow_route") or "").strip(),
+                workflow_route=str(workflow_route or ""),
                 project_id=_safe_int(work_item.project_id),
                 metadata=intent.metadata if isinstance(intent.metadata, dict) else {},
             ).to_dict()
@@ -135,11 +112,8 @@ class KernelIngressRenderMixin:
             draft_expires_at = payload.get("draft_expires_at")
             if isinstance(draft_expires_at, str) and draft_expires_at.strip():
                 result["draft_expires_at"] = draft_expires_at.strip()
-        if intent.name == INTENT_QUERY and isinstance(execution_result.get("rag_result"), dict):
-            rag_result = dict(execution_result["rag_result"])
-            if not isinstance(rag_result.get("consultant_dispatch"), dict):
-                rag_result["consultant_dispatch"] = consultant_dispatch
-            result["rag_result"] = rag_result
+        if workflow_route == ROUTE_QUERY_ANSWER and isinstance(execution_result.get("rag_result"), dict):
+            result["rag_result"] = dict(execution_result["rag_result"])
         if intent.name == INTENT_PROJECT_CREATE:
             raw_payload = execution_result.get("raw_payload") if isinstance(execution_result.get("raw_payload"), dict) else {}
             result["project"] = raw_payload.get("project")

backend/app/services/ai_pmo/ai_pmo_kernel_ingress_flow/mixin_entry.py

@@ -7,6 +7,7 @@ from app.services.ai_pmo.workflow_semantic_planner.constants import (
     QUERY_SCENE_BUSINESS_REVIEW,
     QUERY_SCENE_CORRECTION_ESCALATION,
 )
+from app.services.mounted_ai_runtime_write_service import MountedAIRuntimeWriteService
 
 class KernelIngressFlowEntryMixin:
     _MIXED_HINT_CONNECTORS = ("顺便", "另外", "同时", "并且", "还有", "再问")
@@ -291,6 +292,8 @@ class KernelIngressFlowEntryMixin:
         perf_work_item_id: Optional[int] = None
         perf_intent_name = ""
         thread_context = self._extract_thread_context_from_ingress_options(ingress_options)
+        mounted_runtime = MountedAIRuntimeWriteService(self.db)
+        mounted_tracking = None
         def _log_perf(stage_name: str, stage_started_at: float) -> None:
             elapsed_seconds = time.perf_counter() - stage_started_at
             perf_line = (
@@ -300,6 +303,86 @@ class KernelIngressFlowEntryMixin:
             )
             logger.info(perf_line)
             print(perf_line, flush=True)
+
+        async def _safe_begin_mounted_runtime(*, runtime_message_id: str, runtime_message_text: str) -> None:
+            nonlocal mounted_tracking
+            if mounted_tracking is not None:
+                return
+            try:
+                mounted_tracking = await mounted_runtime.begin_ingress_run(
+                    tenant_id=int(self.tenant_id),
+                    source_system=source_system,
+                    source_message_id=runtime_message_id,
+                    message_text=runtime_message_text,
+                    sender=sender,
+                    channel_ref=channel_ref,
+                    thread_context=thread_context,
+                )
+            except Exception as runtime_exc:
+                logger.warning(
+                    "mounted_ai_runtime_begin_failed",
+                    tenant_id=self.tenant_id,
+                    source_system=source_system,
+                    source_message_id=runtime_message_id,
+                    error=str(runtime_exc),
+                )
+
+        async def _safe_record_mounted_decision(
+            *,
+            workflow_route: str,
+            intent_name: str,
+            intent_confidence: float,
+            trigger_reason: str,
+            auto_execute: bool,
+            work_item_id: Optional[int],
+            route_metadata: Optional[dict[str, Any]],
+            mode_override: Optional[str] = None,
+        ) -> None:
+            if mounted_tracking is None:
+                return
+            try:
+                await mounted_runtime.record_orchestrator_decision(
+                    tracking=mounted_tracking,
+                    workflow_route=workflow_route,
+                    intent_name=intent_name,
+                    intent_confidence=float(intent_confidence or 0.0),
+                    trigger_reason=trigger_reason,
+                    auto_execute=auto_execute,
+                    work_item_id=work_item_id,
+                    route_metadata=route_metadata,
+                    mode_override=mode_override,
+                )
+            except Exception as runtime_exc:
+                logger.warning(
+                    "mounted_ai_runtime_record_failed",
+                    tenant_id=self.tenant_id,
+                    source_system=source_system,
+                    source_message_id=str(getattr(getattr(mounted_tracking, "conversation_run", None), "source_message_id", "") or ""),
+                    error=str(runtime_exc),
+                )
+
+        async def _safe_finalize_mounted_runtime(
+            *,
+            result: Optional[dict[str, Any]] = None,
+            error_text: Optional[str] = None,
+        ) -> None:
+            if mounted_tracking is None:
+                return
+            try:
+                await mounted_runtime.finalize_run(
+                    tracking=mounted_tracking,
+                    result=result if isinstance(result, dict) else None,
+                    error_text=error_text,
+                )
+            except Exception as runtime_exc:
+                logger.warning(
+                    "mounted_ai_runtime_finalize_failed",
+                    tenant_id=self.tenant_id,
+                    source_system=source_system,
+                    source_message_id=str(getattr(getattr(mounted_tracking, "conversation_run", None), "source_message_id", "") or ""),
+                    error=str(runtime_exc),
+                )
+
         ingress_prepare = await self._prepare_ingress_envelope(
             source_system=source_system,
             source_message_id=source_message_id,
@@ -321,6 +404,10 @@ class KernelIngressFlowEntryMixin:
         operator = ingress_prepare["operator"]
         work_item = ingress_prepare["work_item"]
         session = ingress_prepare["session"]
+        await _safe_begin_mounted_runtime(
+            runtime_message_id=synthetic_message_id,
+            runtime_message_text=clean_text,
+        )
         self._merge_ingress_thread_context(
             ingest_event=ingest_event,
             ingress_options=ingress_options,
@@ -344,6 +431,25 @@ class KernelIngressFlowEntryMixin:
                 ingress_options=ingress_options,
             )
             if draft_capture is not None:
+                draft_mode_override = (
+                    "draft_ready"
+                    if str(draft_capture.get("reply_template_key") or "").strip().lower() == "draft_pending_confirm"
+                    else "plan"
+                )
+                await _safe_record_mounted_decision(
+                    workflow_route=str(draft_capture.get("workflow_route") or ROUTE_FOLLOWUP_CONFIRM).strip() or ROUTE_FOLLOWUP_CONFIRM,
+                    intent_name=INTENT_FOLLOWUP_CONFIRM,
+                    intent_confidence=0.95,
+                    trigger_reason="draft_confirmation",
+                    auto_execute=False,
+                    work_item_id=_safe_int(draft_capture.get("work_item_id")) or perf_work_item_id,
+                    route_metadata={
+                        "draft_confirmation": True,
+                        "reply_template_key": str(draft_capture.get("reply_template_key") or "").strip() or None,
+                    },
+                    mode_override=draft_mode_override,
+                )
+                await _safe_finalize_mounted_runtime(result=draft_capture)
                 return draft_capture
         mixed_meta: dict[str, Any] = {}
         if not self._is_short_confirmation_message(
@@ -454,7 +560,7 @@ class KernelIngressFlowEntryMixin:
             resolved_project_for_session = _safe_int(workflow_route.project_id) or _safe_int(world_model.get("resolved_project_id"))
             fallback_intent = intent
             if intent.name == INTENT_IGNORE:
-                return await self._finalize_ignored_ingress(
+                capture_result = await self._finalize_ignored_ingress(
                     work_item=work_item,
                     ingest_event=ingest_event,
                     session=session,
@@ -465,6 +571,18 @@ class KernelIngressFlowEntryMixin:
                     sender=sender,
                     resolved_project_for_session=resolved_project_for_session,
                 )
+                await _safe_record_mounted_decision(
+                    workflow_route=workflow_route.key,
+                    intent_name=intent.name,
+                    intent_confidence=float(intent.confidence or 0.0),
+                    trigger_reason="ignored",
+                    auto_execute=False,
+                    work_item_id=int(work_item.id),
+                    route_metadata=route_metadata,
+                    mode_override="chat",
+                )
+                await _safe_finalize_mounted_runtime(result=capture_result)
+                return capture_result
 
             work_item.intent = workflow_route.key
             work_item.project_id = workflow_route.project_id
@@ -562,9 +680,24 @@ class KernelIngressFlowEntryMixin:
                     started=started,
                 )
                 _log_perf("_build_capture_result", stage_started_at)
+                await _safe_record_mounted_decision(
+                    workflow_route=workflow_route.key,
+                    intent_name=intent.name,
+                    intent_confidence=float(intent.confidence or 0.0),
+                    trigger_reason=manual_review_code if requires_manual_review else str(gate.gate_code or "manual_review"),
+                    auto_execute=False,
+                    work_item_id=int(work_item.id),
+                    route_metadata=route_metadata,
+                    mode_override=(
+                        "draft_ready"
+                        if str(capture_result.get("reply_template_key") or "").strip().lower() == "draft_pending_confirm"
+                        else None
+                    ),
+                )
+                await _safe_finalize_mounted_runtime(result=capture_result)
                 return capture_result
 
-            return await self._execute_auto_ingress_flow(
+            capture_result = await self._execute_auto_ingress_flow(
                 work_item=work_item,
                 intent=intent,
                 world_model=world_model,
@@ -586,7 +719,19 @@ class KernelIngressFlowEntryMixin:
                     session=session,
                     log_perf_fn=_log_perf,
                 )
+            await _safe_record_mounted_decision(
+                workflow_route=workflow_route.key,
+                intent_name=intent.name,
+                intent_confidence=float(intent.confidence or 0.0),
+                trigger_reason="auto_execute",
+                auto_execute=True,
+                work_item_id=int(work_item.id),
+                route_metadata=route_metadata,
+            )
+            await _safe_finalize_mounted_runtime(result=capture_result)
+            return capture_result
         except Exception as exc:
+            await _safe_finalize_mounted_runtime(error_text=str(exc))
             await self._persist_kernel_failure_with_fresh_session(
                 work_item_id=_safe_int(getattr(work_item, "id", None)),
                 ingest_event_id=_safe_int(getattr(ingest_event, "id", None)),

backend/app/services/integrations/feishu_integration_service/mixin_reply_llm_core.py

@@ -1,7 +1,6 @@
 from __future__ import annotations
 
 from .shared import *
-from ...ai_pmo.consultant_dispatch import build_channel_consultant_reply
 from ...ai_pmo.workflow_semantic_planner import WorkflowSemanticPlanner
 
 class FeishuReplyLLMCoreMixin:
@@ -80,19 +79,6 @@ class FeishuReplyLLMCoreMixin:
             )
             return
 
-        consultant_reply = self._build_consultant_dispatch_reply(capture_result)
-        if consultant_reply:
-            await self._safe_reply_text(
-                chat_id=chat_id,
-                mention_open_id=sender_open_id,
-                mention_name=sender_name,
-                text=consultant_reply,
-                trace_id=trace_id,
-                source_message_id=source_message_id,
-                perf_t0=perf_t0,
-            )
-            return
-
         try:
             status = str(capture_result.get("status") or "unknown")
             classification = str(capture_result.get("classification") or "unknown")
@@ -246,14 +232,8 @@ class FeishuReplyLLMCoreMixin:
 
     @staticmethod
     def _build_consultant_dispatch_reply(capture_result: Dict[str, Any]) -> Optional[str]:
-        if not isinstance(capture_result, dict):
+        _ = capture_result
         return None
-        reply = build_channel_consultant_reply(
-            consultant_dispatch=capture_result.get("consultant_dispatch"),
-            direct_reply_text=str(capture_result.get("direct_reply_text") or "").strip(),
-            channel="feishu",
-        )
-        return str(reply or "").strip() or None
 
     @classmethod
     def _build_action_request_ack_reply(

backend/app/services/integrations/feishu_signal_pipeline/mixin_followup_a.py

@@ -2,8 +2,6 @@
 from __future__ import annotations
 
 from .shared import *
-from ...ai_pmo.workflow_semantic_planner import WorkflowSemanticPlanner
-from app.utils.thread_context import extract_thread_anchor
 
 
 class FeishuPipelineFollowupAMixin:
@@ -18,229 +16,18 @@ class FeishuPipelineFollowupAMixin:
         process_started_at: float,
         thread_context: Optional[dict[str, Any]] = None,
     ) -> Optional[dict[str, Any]]:
-        if not source_chat_id:
+        _ = (
+            message,
+            sender,
+            source_chat_id,
+            source_sender_open_id,
+            feishu_message_id,
+            process_started_at,
+            thread_context,
+        )
+        # Legacy feishu_items queue follow-up is retired. Runtime callers should route
+        # status/continue requests through mounted work-item/topic flows instead.
         return None
-        wants_continue = self._looks_like_continue_followup(message)
-        wants_status = WorkflowSemanticPlanner.wants_followup_status(message)
-        if not wants_continue and not wants_status:
-            return None
-
-        recent_items = await self._load_recent_chat_items(
-            source_chat_id=source_chat_id,
-            source_sender_open_id=source_sender_open_id,
-            sender=sender,
-        )
-        thread_anchor = extract_thread_anchor(thread_context)
-        anchor_item = (
-            next(
-                (
-                    item
-                    for item in recent_items
-                    if thread_anchor and str(getattr(item, "feishu_message_id", "") or "").strip() == thread_anchor
-                ),
-                None,
-            )
-            if thread_anchor
-            else None
-        )
-        if thread_anchor and anchor_item is None:
-            return None
-        if not recent_items:
-            reply_text = "当前群里还没有可追踪的飞书处理项。"
-            decision = ROUTING_QUEUE_FOLLOWUP_BLOCKED
-            reason = "群上下文中不存在可追踪的最近处理项。"
-            classification = "question"
-            status = "answered"
-            mutations: list[dict[str, Any]] = []
-            results: list[dict[str, Any]] = []
-        elif wants_continue:
-            if anchor_item is not None and self._normalize_route_status(anchor_item.route_status) == ROUTE_STATUS_QUEUED:
-                pending_items = [anchor_item]
-            else:
-                pending_items = [item for item in recent_items if self._normalize_route_status(item.route_status) == ROUTE_STATUS_QUEUED]
-            if len(pending_items) != 1:
-                reply_text = self._build_followup_pending_clarification(pending_items or recent_items[:3])
-                decision = ROUTING_QUEUE_FOLLOWUP_BLOCKED
-                reason = "群上下文中存在多个待处理项，无法自动确定继续处理目标。"
-                classification = "command"
-                status = "captured"
-                mutations = []
-                results = []
-            else:
-                operator_user = await self._resolve_command_operator_user(sender=sender)
-                if operator_user is None:
-                    reply_text = "未执行：当前租户下没有可用操作者，无法继续处理该项。"
-                    decision = ROUTING_QUEUE_FOLLOWUP_BLOCKED
-                    reason = "继续处理被阻断：当前租户下没有可用操作者。"
-                    classification = "command"
-                    status = "captured"
-                    mutations = []
-                    results = []
-                else:
-                    target_item = pending_items[0]
-                    if (target_item.level or "").lower() == "project":
-                        suggestion = await self._get_project_suggestion(int(target_item.id))
-                        if suggestion is None:
-                            reply_text = "未执行：最近待处理项目候选缺少建议记录，无法继续处理。"
-                            decision = ROUTING_QUEUE_FOLLOWUP_BLOCKED
-                            reason = "继续处理被阻断：最近待处理项目候选缺少建议记录。"
-                            classification = "command"
-                            status = "captured"
-                            mutations = []
-                            results = []
-                        else:
-                            candidate_fields = self._resolve_project_candidate_fields(item=target_item, suggestion=suggestion)
-                            if not candidate_fields["ready"]:
-                                reply_text = (
-                                    str(target_item.blocking_reason or "").strip()
-                                    or self._build_project_clarification_message(candidate_fields["missing_fields"])
-                                )
-                                decision = ROUTING_QUEUE_FOLLOWUP_BLOCKED
-                                reason = "继续处理被阻断：最近待处理建项目项仍缺少必填字段。"
-                                classification = "command"
-                                status = "captured"
-                                mutations = []
-                                results = []
-                            else:
-                                confirm_result = await self.confirm_item(item_id=int(target_item.id), user_id=int(operator_user.id))
-                                reply_text = self._build_followup_confirmation_reply(target_item=target_item, confirm_result=confirm_result)
-                                decision = ROUTING_QUEUE_FOLLOWUP_CONFIRM
-                                reason = "命中继续处理指令，已按最近唯一待处理项继续执行。"
-                                classification = "command"
-                                status = "captured"
-                                mutations = self._summarize_db_mutations(
-                                    {
-                                        "results": [
-                                            {
-                                                "item_id": int(target_item.id),
-                                                "level": target_item.level,
-                                                "route_status": ROUTE_STATUS_AUTO_DONE,
-                                                "project_id": confirm_result.get("project_id"),
-                                                "signal_id": confirm_result.get("signal_id"),
-                                            }
-                                        ]
-                                    }
-                                )
-                                results = [
-                                    {
-                                        "item_id": int(target_item.id),
-                                        "level": target_item.level,
-                                        "route_status": ROUTE_STATUS_AUTO_DONE,
-                                        "project_id": _safe_int(confirm_result.get("project_id")),
-                                        "signal_id": _safe_int(confirm_result.get("signal_id")),
-                                    }
-                                ]
-                    else:
-                        signal = await self._get_signal_by_item(int(target_item.id))
-                        if (target_item.level or "").lower() == "signal" and not (signal and (signal.matched_project_id or target_item.matched_project_id)):
-                            reply_text = "未执行：最近待处理信号仍缺少明确项目归属，请先补充项目名或项目 ID。"
-                            decision = ROUTING_QUEUE_FOLLOWUP_BLOCKED
-                            reason = "继续处理被阻断：最近待处理信号缺少项目归属。"
-                            classification = "command"
-                            status = "captured"
-                            mutations = []
-                            results = []
-                        else:
-                            confirm_result = await self.confirm_item(item_id=int(target_item.id), user_id=int(operator_user.id))
-                            reply_text = self._build_followup_confirmation_reply(target_item=target_item, confirm_result=confirm_result)
-                            decision = ROUTING_QUEUE_FOLLOWUP_CONFIRM
-                            reason = "命中继续处理指令，已按最近唯一待处理项继续执行。"
-                            classification = "command"
-                            status = "captured"
-                            mutations = self._summarize_db_mutations(
-                                {
-                                    "results": [
-                                        {
-                                            "item_id": int(target_item.id),
-                                            "level": target_item.level,
-                                            "route_status": ROUTE_STATUS_AUTO_DONE,
-                                            "project_id": confirm_result.get("project_id"),
-                                            "signal_id": confirm_result.get("signal_id"),
-                                        }
-                                    ]
-                                }
-                            )
-                            results = [
-                                {
-                                    "item_id": int(target_item.id),
-                                    "level": target_item.level,
-                                    "route_status": ROUTE_STATUS_AUTO_DONE,
-                                    "project_id": _safe_int(confirm_result.get("project_id")),
-                                    "signal_id": _safe_int(confirm_result.get("signal_id")),
-                                }
-                            ]
-        else:
-            target_item = anchor_item or self._select_followup_status_target(
-                message=message,
-                recent_items=recent_items,
-                thread_context=thread_context,
-            )
-            if target_item is None:
-                reply_text = self._build_followup_pending_clarification(recent_items[:3])
-                decision = ROUTING_QUEUE_FOLLOWUP_BLOCKED
-                reason = "状态追问缺少唯一锚点，系统拒绝猜测最近一条处理项。"
-                classification = "question"
-                status = "answered"
-                mutations = []
-                results = []
-            else:
-                reply_text = await self._build_followup_status_reply(target_item, message=message)
-                decision = ROUTING_QUEUE_FOLLOWUP_QUERY
-                reason = "命中进度/状态追问，已直接基于真实 DB 返回唯一处理项状态。"
-                classification = "question"
-                status = "answered"
-                mutations = []
-                results = []
-
-        await self._write_processing_log(
-            feishu_message_id=feishu_message_id,
-            feishu_item_id=None,
-            raw_message=message,
-            ai_raw_output=json.dumps(
-                {
-                    "followup_reply": reply_text,
-                    "recent_item_ids": [int(item.id) for item in recent_items[:3]],
-                    "wants_continue": wants_continue,
-                    "wants_status": wants_status,
-                },
-                ensure_ascii=False,
-            )[:4000],
-            project_score=None,
-            signal_confidence=None,
-            level="followup",
-            routing_decision=decision,
-            routing_reason=reason,
-            reasoning_bullets=[
-                "follow-up 查询/指令命中飞书线程锚点或唯一目标。",
-                "回复内容只基于真实 DB 状态，不使用自由生成。",
-                f"recent_items={len(recent_items)}",
-            ],
-            processing_status="success",
-            error_message=None,
-            duration_ms=int((time.perf_counter() - process_started_at) * 1000),
-        )
-
-        latest = recent_items[0] if recent_items else None
-        linked_project_ids: list[int] = []
-        if results and results[0].get("project_id") is not None:
-            linked_project_ids.append(int(results[0]["project_id"]))
-        if latest and latest.matched_project_id is not None:
-            linked_project_ids.append(int(latest.matched_project_id))
-        linked_project_ids = sorted(set(linked_project_ids))
-        return {
-            "status": status,
-            "classification": classification,
-            "project_scope": classification,
-            "linked_project_ids": linked_project_ids,
-            "signal_id": str(results[0].get("signal_id")) if results else None,
-            "created_project_id": str(results[0].get("project_id")) if results else None,
-            "auto_tasks": [],
-            "ai_trace_id": f"feishu-followup-{(feishu_message_id or 'noid')[:32]}",
-            "results": results,
-            "direct_reply_required": True,
-            "direct_reply_text": reply_text,
-            "db_mutations": mutations,
-        }
 
     async def _load_recent_chat_items(
         self,

backend/app/services/mounted_ai_runtime_service.py

@@ -0,0 +1,528 @@
+from __future__ import annotations
+
+from pathlib import Path
+from typing import Any, Optional
+
+from sqlalchemy import case, func, select
+from sqlalchemy.ext.asyncio import AsyncSession
+
+from ..models import (
+    MountedAgentTrace,
+    MountedConversationRun,
+    MountedDraftWorkspace,
+    MountedToolCapabilityPolicy,
+    MountedTopicWorkspace,
+    TrainingDataset,
+    TrainingEvaluation,
+    TrainingJob,
+)
+
+
+class MountedAIRuntimeService:
+    def __init__(self, db: AsyncSession):
+        self.db = db
+
+    @staticmethod
+    def _dt(value: Any) -> Optional[str]:
+        return value.isoformat() if value is not None else None
+
+    @staticmethod
+    def _backend_root() -> Path:
+        return Path(__file__).resolve().parents[2]
+
+    @classmethod
+    def _read_backend_source(cls, relative_path: str) -> str:
+        try:
+            return (cls._backend_root() / relative_path).read_text(encoding="utf-8")
+        except OSError:
+            return ""
+
+    @staticmethod
+    def _legacy_cutover_item(
+        *,
+        key: str,
+        title: str,
+        retired: bool,
+        summary: str,
+        evidence: str,
+    ) -> dict[str, Any]:
+        return {
+            "key": key,
+            "title": title,
+            "status": "retired" if retired else "active",
+            "summary": summary,
+            "evidence": evidence,
+        }
+
+    @classmethod
+    def _build_legacy_cutover_items(cls) -> list[dict[str, Any]]:
+        ingress_render = cls._read_backend_source("app/services/ai_pmo/ai_pmo_kernel/mixins/ingress/ai_pmo_kernel_mixins_ingress_render.py")
+        feishu_reply = cls._read_backend_source("app/services/integrations/feishu_integration_service/mixin_reply_llm_core.py")
+        signal_capture_service = cls._read_backend_source("app/services/signal/signal_capture_service/service.py")
+        feishu_pipeline_service = cls._read_backend_source("app/services/integrations/feishu_signal_pipeline/service.py")
+        feishu_process_core = cls._read_backend_source("app/services/integrations/feishu_signal_pipeline/mixin_process_message_core.py")
+        feishu_followup_a = cls._read_backend_source("app/services/integrations/feishu_signal_pipeline/mixin_followup_a.py")
+
+        return [
+            cls._legacy_cutover_item(
+                key="ai_pmo_ingress_consultant_capture",
+                title="AI PMO ingress consultant capture",
+                retired="build_consultant_dispatch_contract" not in ingress_render and '"consultant_dispatch"' not in ingress_render,
+                summary="capture result 不再生成或透传 consultant_dispatch 到 AI PMO ingress 主链。",
+                evidence="ai_pmo_kernel_mixins_ingress_render.py 不再包含 consultant dispatch contract/import 与 rag_result 注入。",
+            ),
+            cls._legacy_cutover_item(
+                key="feishu_mention_consultant_reply",
+                title="Feishu mention consultant reply",
+                retired="build_channel_consultant_reply" not in feishu_reply and "_build_consultant_dispatch_reply(capture_result)" not in feishu_reply,
+                summary="飞书 mention 主回复不再走 consultant reply 分支。",
+                evidence="mixin_reply_llm_core.py 已移除 consultant reply import 和 active reply branch；helper 仅保留 no-op。",
+            ),
+            cls._legacy_cutover_item(
+                key="signal_capture_legacy_entry",
+                title="Signal capture legacy entry",
+                retired="SignalCaptureLegacyMixin" not in signal_capture_service,
+                summary="主动信号捕获服务不再继承 legacy capture mixin。",
+                evidence="signal_capture_service/service.py 的 ProactiveSignalCaptureService 继承链已去掉 SignalCaptureLegacyMixin。",
+            ),
+            cls._legacy_cutover_item(
+                key="legacy_feishu_queue_followup",
+                title="Legacy Feishu queue follow-up",
+                retired=(
+                    "FeishuPipelineFollowupAMixin" not in feishu_pipeline_service
+                    and "FeishuPipelineFollowupBMixin" not in feishu_pipeline_service
+                    and "_maybe_handle_queue_followup(" not in feishu_process_core
+                    and "return None" in feishu_followup_a
+                    and "Legacy feishu_items queue follow-up is retired." in feishu_followup_a
+                ),
+                summary="旧 feishu_items 队列 continue/status follow-up 已从真实入口退役。",
+                evidence="service.py 不再混入 FollowupA/B，process_message_core 不再调用 queue followup，mixin_followup_a 已降为 retired stub。",
+            ),
+        ]
+
+    def _topic_snapshot(
+        self,
+        row: MountedTopicWorkspace,
+        *,
+        draft: Optional[MountedDraftWorkspace] = None,
+    ) -> dict[str, Any]:
+        return {
+            "id": int(row.id),
+            "topic_key": row.topic_key,
+            "source_system": row.source_system,
+            "channel_ref": row.channel_ref,
+            "thread_key": row.thread_key,
+            "mode": row.mode,
+            "status": row.status,
+            "readiness_score": float(row.readiness_score or 0.0) if row.readiness_score is not None else None,
+            "draft_state": draft.state if draft is not None else None,
+            "draft_version": int(draft.draft_version) if draft is not None else None,
+            "updated_at": self._dt(row.updated_at),
+        }
+
+    def _run_snapshot(self, row: MountedConversationRun, *, trace_count: int = 0) -> dict[str, Any]:
+        return {
+            "id": int(row.id),
+            "trace_id": row.trace_id,
+            "source_system": row.source_system,
+            "source_message_id": row.source_message_id,
+            "mode_before": row.mode_before,
+            "mode_after": row.mode_after,
+            "active_agent": row.active_agent,
+            "status": row.status,
+            "trigger_reason": row.trigger_reason,
+            "final_reply": row.final_reply,
+            "trace_count": int(trace_count),
+            "updated_at": self._dt(row.updated_at),
+        }
+
+    @staticmethod
+    def _trace_snapshot(row: MountedAgentTrace) -> dict[str, Any]:
+        return {
+            "id": int(row.id),
+            "agent_name": row.agent_name,
+            "status": row.status,
+            "latency_ms": int(row.latency_ms) if row.latency_ms is not None else None,
+            "failure_reason": row.failure_reason,
+            "created_at": row.created_at.isoformat() if row.created_at else None,
+        }
+
+    @staticmethod
+    def _evaluation_snapshot(row: TrainingEvaluation) -> dict[str, Any]:
+        return {
+            "evaluation_type": row.evaluation_type,
+            "status": row.status,
+            "score": float(row.score or 0.0) if row.score is not None else None,
+        }
+
+    @staticmethod
+    def _eval_suite_snapshot(
+        row: TrainingDataset,
+        *,
+        latest_score: Optional[float],
+        last_evaluated_at: Optional[Any],
+    ) -> dict[str, Any]:
+        selection_rules = row.selection_rules if isinstance(row.selection_rules, dict) else {}
+        suite_kind = str(selection_rules.get("suite_kind") or "custom").strip() or "custom"
+        return {
+            "id": int(row.id),
+            "dataset_type": row.dataset_type,
+            "version": row.version,
+            "status": row.status,
+            "suite_kind": suite_kind,
+            "sample_count": int(row.sample_count or 0),
+            "latest_score": latest_score,
+            "last_evaluated_at": MountedAIRuntimeService._dt(last_evaluated_at),
+        }
+
+    async def list_sources(self, *, tenant_id: int) -> dict[str, Any]:
+        source_map: dict[str, dict[str, Any]] = {}
+
+        run_rows = (
+            await self.db.execute(
+                select(
+                    MountedConversationRun.source_system,
+                    func.count(MountedConversationRun.id).label("run_count"),
+                    func.max(MountedConversationRun.updated_at).label("last_seen_at"),
+                )
+                .where(MountedConversationRun.tenant_id == int(tenant_id))
+                .group_by(MountedConversationRun.source_system)
+            )
+        ).all()
+        for row in run_rows:
+            source_system = str(row.source_system or "").strip()
+            source_map[source_system] = {
+                "source_system": source_system,
+                "run_count": int(row.run_count or 0),
+                "topic_count": 0,
+                "active_topic_count": 0,
+                "last_seen_at": self._dt(row.last_seen_at),
+            }
+
+        topic_rows = (
+            await self.db.execute(
+                select(
+                    MountedTopicWorkspace.source_system,
+                    func.count(MountedTopicWorkspace.id).label("topic_count"),
+                    func.sum(
+                        case((MountedTopicWorkspace.status == "active", 1), else_=0)
+                    ).label("active_topic_count"),
+                    func.max(MountedTopicWorkspace.updated_at).label("topic_last_seen_at"),
+                )
+                .where(MountedTopicWorkspace.tenant_id == int(tenant_id))
+                .group_by(MountedTopicWorkspace.source_system)
+            )
+        ).all()
+        for row in topic_rows:
+            source_system = str(row.source_system or "").strip()
+            payload = source_map.setdefault(
+                source_system,
+                {
+                    "source_system": source_system,
+                    "run_count": 0,
+                    "topic_count": 0,
+                    "active_topic_count": 0,
+                    "last_seen_at": None,
+                },
+            )
+            payload["topic_count"] = int(row.topic_count or 0)
+            payload["active_topic_count"] = int(row.active_topic_count or 0)
+            topic_last_seen_at = self._dt(row.topic_last_seen_at)
+            if topic_last_seen_at and not payload.get("last_seen_at"):
+                payload["last_seen_at"] = topic_last_seen_at
+
+        items = sorted(source_map.values(), key=lambda item: (item["source_system"] or ""))
+        return {"items": items}
+
+    async def list_topics(
+        self,
+        *,
+        tenant_id: int,
+        limit: int,
+        offset: int,
+        mode: Optional[str] = None,
+    ) -> dict[str, Any]:
+        base_stmt = select(MountedTopicWorkspace).where(MountedTopicWorkspace.tenant_id == int(tenant_id))
+        if mode:
+            base_stmt = base_stmt.where(MountedTopicWorkspace.mode == str(mode).strip())
+
+        total = (
+            await self.db.execute(
+                select(func.count()).select_from(base_stmt.order_by(None).subquery())
+            )
+        ).scalar_one()
+
+        rows = (
+            await self.db.execute(
+                base_stmt.order_by(MountedTopicWorkspace.updated_at.desc(), MountedTopicWorkspace.id.desc())
+                .offset(int(offset))
+                .limit(int(limit))
+            )
+        ).scalars().all()
+
+        topic_ids = [int(row.id) for row in rows]
+        draft_map: dict[int, MountedDraftWorkspace] = {}
+        if topic_ids:
+            draft_rows = (
+                await self.db.execute(
+                    select(MountedDraftWorkspace).where(
+                        MountedDraftWorkspace.tenant_id == int(tenant_id),
+                        MountedDraftWorkspace.topic_workspace_id.in_(topic_ids),
+                    )
+                )
+            ).scalars().all()
+            draft_map = {int(row.topic_workspace_id): row for row in draft_rows}
+
+        return {
+            "items": [self._topic_snapshot(row, draft=draft_map.get(int(row.id))) for row in rows],
+            "total": int(total or 0),
+        }
+
+    async def list_runs(
+        self,
+        *,
+        tenant_id: int,
+        limit: int,
+        offset: int,
+        status: Optional[str] = None,
+    ) -> dict[str, Any]:
+        base_stmt = select(MountedConversationRun).where(MountedConversationRun.tenant_id == int(tenant_id))
+        if status:
+            base_stmt = base_stmt.where(MountedConversationRun.status == str(status).strip())
+
+        total = (
+            await self.db.execute(
+                select(func.count()).select_from(base_stmt.order_by(None).subquery())
+            )
+        ).scalar_one()
+
+        rows = (
+            await self.db.execute(
+                base_stmt.order_by(MountedConversationRun.updated_at.desc(), MountedConversationRun.id.desc())
+                .offset(int(offset))
+                .limit(int(limit))
+            )
+        ).scalars().all()
+
+        run_ids = [int(row.id) for row in rows]
+        trace_counts: dict[int, int] = {}
+        if run_ids:
+            trace_rows = (
+                await self.db.execute(
+                    select(
+                        MountedAgentTrace.conversation_run_id,
+                        func.count(MountedAgentTrace.id).label("trace_count"),
+                    )
+                    .where(
+                        MountedAgentTrace.tenant_id == int(tenant_id),
+                        MountedAgentTrace.conversation_run_id.in_(run_ids),
+                    )
+                    .group_by(MountedAgentTrace.conversation_run_id)
+                )
+            ).all()
+            trace_counts = {int(row.conversation_run_id): int(row.trace_count or 0) for row in trace_rows}
+
+        return {
+            "items": [self._run_snapshot(row, trace_count=trace_counts.get(int(row.id), 0)) for row in rows],
+            "total": int(total or 0),
+        }
+
+    async def get_run_detail(self, *, tenant_id: int, run_id: int) -> dict[str, Any]:
+        row = (
+            await self.db.execute(
+                select(MountedConversationRun).where(
+                    MountedConversationRun.tenant_id == int(tenant_id),
+                    MountedConversationRun.id == int(run_id),
+                )
+            )
+        ).scalar_one_or_none()
+        if row is None:
+            raise ValueError("mounted ai run not found")
+
+        trace_rows = (
+            await self.db.execute(
+                select(MountedAgentTrace)
+                .where(
+                    MountedAgentTrace.tenant_id == int(tenant_id),
+                    MountedAgentTrace.conversation_run_id == int(row.id),
+                )
+                .order_by(MountedAgentTrace.id.asc())
+            )
+        ).scalars().all()
+
+        return {
+            "item": self._run_snapshot(row, trace_count=len(trace_rows)),
+            "agent_traces": [self._trace_snapshot(trace) for trace in trace_rows],
+            "context_layers_hit": row.context_layers_hit,
+            "tool_calls": row.tool_calls,
+            "execution_outcome": row.execution_outcome,
+        }
+
+    async def get_cutover_status(self, *, tenant_id: int) -> dict[str, Any]:
+        topic_workspace_count = (
+            await self.db.execute(
+                select(func.count(MountedTopicWorkspace.id)).where(
+                    MountedTopicWorkspace.tenant_id == int(tenant_id)
+                )
+            )
+        ).scalar_one()
+        conversation_run_count = (
+            await self.db.execute(
+                select(func.count(MountedConversationRun.id)).where(
+                    MountedConversationRun.tenant_id == int(tenant_id)
+                )
+            )
+        ).scalar_one()
+        tool_policy_count = (
+            await self.db.execute(
+                select(func.count(MountedToolCapabilityPolicy.id)).where(
+                    MountedToolCapabilityPolicy.tenant_id == int(tenant_id)
+                )
+            )
+        ).scalar_one()
+        observed_sources = (
+            await self.db.execute(
+                select(MountedConversationRun.source_system)
+                .where(MountedConversationRun.tenant_id == int(tenant_id))
+                .distinct()
+                .order_by(MountedConversationRun.source_system.asc())
+            )
+        ).scalars().all()
+        legacy_cutover_items = self._build_legacy_cutover_items()
+        active_item_count = sum(1 for item in legacy_cutover_items if item["status"] == "active")
+        retired_item_count = sum(1 for item in legacy_cutover_items if item["status"] == "retired")
+        legacy_cutover_state = "retired" if active_item_count == 0 else "active_legacy_paths_remaining"
+        return {
+            "mounted_runtime_ready": True,
+            "topic_workspace_count": int(topic_workspace_count or 0),
+            "conversation_run_count": int(conversation_run_count or 0),
+            "tool_policy_count": int(tool_policy_count or 0),
+            "observed_source_systems": [str(item) for item in observed_sources if str(item or "").strip()],
+            "legacy_cutover_state": legacy_cutover_state,
+            "retired_item_count": int(retired_item_count),
+            "active_item_count": int(active_item_count),
+            "legacy_cutover_items": legacy_cutover_items,
+        }
+
+    async def list_eval_suites(self, *, tenant_id: int) -> dict[str, Any]:
+        datasets = (
+            await self.db.execute(
+                select(TrainingDataset)
+                .where(
+                    TrainingDataset.tenant_id == int(tenant_id),
+                    TrainingDataset.dataset_type == "eval",
+                )
+                .order_by(TrainingDataset.id.desc())
+            )
+        ).scalars().all()
+        if not datasets:
+            return {"items": [], "total": 0}
+
+        dataset_ids = [int(row.id) for row in datasets]
+        jobs = (
+            await self.db.execute(
+                select(TrainingJob)
+                .where(
+                    TrainingJob.tenant_id == int(tenant_id),
+                    TrainingJob.dataset_id.in_(dataset_ids),
+                )
+                .order_by(TrainingJob.id.desc())
+            )
+        ).scalars().all()
+        latest_job_by_dataset: dict[int, TrainingJob] = {}
+        for job in jobs:
+            latest_job_by_dataset.setdefault(int(job.dataset_id), job)
+
+        job_ids = [int(job.id) for job in latest_job_by_dataset.values()]
+        evaluations = []
+        if job_ids:
+            evaluations = (
+                await self.db.execute(
+                    select(TrainingEvaluation)
+                    .where(
+                        TrainingEvaluation.tenant_id == int(tenant_id),
+                        TrainingEvaluation.training_job_id.in_(job_ids),
+                    )
+                    .order_by(TrainingEvaluation.id.desc())
+                )
+            ).scalars().all()
+        evaluations_by_job: dict[int, list[TrainingEvaluation]] = {}
+        for item in evaluations:
+            evaluations_by_job.setdefault(int(item.training_job_id), []).append(item)
+
+        items: list[dict[str, Any]] = []
+        for dataset in datasets:
+            latest_job = latest_job_by_dataset.get(int(dataset.id))
+            latest_evaluations = evaluations_by_job.get(int(latest_job.id), []) if latest_job is not None else []
+            completed_scores = [
+                float(row.score or 0.0)
+                for row in latest_evaluations
+                if row.score is not None and str(row.status or "").strip().lower() == "completed"
+            ]
+            latest_score = round(sum(completed_scores) / len(completed_scores), 2) if completed_scores else None
+            last_evaluated_at = None
+            if latest_evaluations:
+                last_evaluated_at = latest_evaluations[0].updated_at
+            items.append(
+                self._eval_suite_snapshot(
+                    dataset,
+                    latest_score=latest_score,
+                    last_evaluated_at=last_evaluated_at,
+                )
+            )
+        return {"items": items, "total": len(items)}
+
+    async def get_eval_suite_detail(self, *, tenant_id: int, dataset_id: int) -> dict[str, Any]:
+        dataset = (
+            await self.db.execute(
+                select(TrainingDataset).where(
+                    TrainingDataset.tenant_id == int(tenant_id),
+                    TrainingDataset.id == int(dataset_id),
+                    TrainingDataset.dataset_type == "eval",
+                )
+            )
+        ).scalar_one_or_none()
+        if dataset is None:
+            raise ValueError("mounted ai eval suite not found")
+
+        latest_job = (
+            await self.db.execute(
+                select(TrainingJob)
+                .where(
+                    TrainingJob.tenant_id == int(tenant_id),
+                    TrainingJob.dataset_id == int(dataset.id),
+                )
+                .order_by(TrainingJob.id.desc())
+                .limit(1)
+            )
+        ).scalar_one_or_none()
+        evaluations: list[TrainingEvaluation] = []
+        if latest_job is not None:
+            evaluations = (
+                await self.db.execute(
+                    select(TrainingEvaluation)
+                    .where(
+                        TrainingEvaluation.tenant_id == int(tenant_id),
+                        TrainingEvaluation.training_job_id == int(latest_job.id),
+                    )
+                    .order_by(TrainingEvaluation.id.asc())
+                )
+            ).scalars().all()
+
+        completed_scores = [
+            float(row.score or 0.0)
+            for row in evaluations
+            if row.score is not None and str(row.status or "").strip().lower() == "completed"
+        ]
+        latest_score = round(sum(completed_scores) / len(completed_scores), 2) if completed_scores else None
+        last_evaluated_at = evaluations[-1].updated_at if evaluations else None
+        return {
+            "item": self._eval_suite_snapshot(
+                dataset,
+                latest_score=latest_score,
+                last_evaluated_at=last_evaluated_at,
+            ),
+            "selection_rules": dataset.selection_rules if isinstance(dataset.selection_rules, dict) else {},
+            "notes": dataset.notes,
+            "evaluations": [self._evaluation_snapshot(row) for row in evaluations],
+        }

backend/app/services/mounted_ai_runtime_write_service.py

@@ -0,0 +1,292 @@
+from __future__ import annotations
+
+import hashlib
+from dataclasses import dataclass
+from typing import Any, Optional
+
+from sqlalchemy import select
+from sqlalchemy.ext.asyncio import AsyncSession
+
+from ..models import MountedAgentTrace, MountedConversationRun, MountedTopicWorkspace
+from ..utils.timezone import now_cn_naive
+
+
+@dataclass
+class MountedIngressTracking:
+    topic_workspace: MountedTopicWorkspace
+    conversation_run: MountedConversationRun
+    context_layers_hit: dict[str, bool]
+
+
+class MountedAIRuntimeWriteService:
+    def __init__(self, db: AsyncSession):
+        self.db = db
+
+    @staticmethod
+    def _compact(value: Any, *, max_len: int = 128) -> Optional[str]:
+        text = str(value or "").strip()
+        if not text:
+            return None
+        return text[:max_len]
+
+    @classmethod
+    def build_topic_key(
+        cls,
+        *,
+        source_system: str,
+        channel_ref: Optional[str],
+        source_message_id: Optional[str],
+        sender: str,
+        message_text: str,
+        thread_context: Optional[dict[str, Any]],
+    ) -> str:
+        normalized_source = cls._compact(source_system, max_len=32) or "unknown"
+        normalized_channel = cls._compact(channel_ref, max_len=128) or "direct"
+        root_message_id = cls._compact((thread_context or {}).get("root_message_id"), max_len=128)
+        if root_message_id:
+            return f"{normalized_source}:{normalized_channel}:root:{root_message_id}"
+        thread_id = cls._compact((thread_context or {}).get("thread_id"), max_len=128)
+        if thread_id:
+            return f"{normalized_source}:{normalized_channel}:thread:{thread_id}"
+        message_id = cls._compact(source_message_id, max_len=128)
+        if message_id:
+            return f"{normalized_source}:{normalized_channel}:message:{message_id}"
+        digest = hashlib.sha1(
+            f"{normalized_source}|{normalized_channel}|{sender}|{message_text}".encode("utf-8"),
+            usedforsecurity=False,
+        ).hexdigest()
+        return f"{normalized_source}:{normalized_channel}:synthetic:{digest[:20]}"
+
+    @classmethod
+    def build_run_trace_id(
+        cls,
+        *,
+        source_system: str,
+        source_message_id: Optional[str],
+        topic_key: str,
+    ) -> str:
+        normalized_source = cls._compact(source_system, max_len=32) or "unknown"
+        normalized_message_id = cls._compact(source_message_id, max_len=128)
+        if normalized_message_id:
+            return f"mounted:{normalized_source}:{normalized_message_id}"[:64]
+        digest = hashlib.sha1(topic_key.encode("utf-8"), usedforsecurity=False).hexdigest()
+        return f"mounted:{normalized_source}:{digest[:48]}"[:64]
+
+    @staticmethod
+    def _mode_for_route(
+        *,
+        workflow_route: str,
+        auto_execute: bool,
+        mode_override: Optional[str] = None,
+    ) -> str:
+        if str(mode_override or "").strip():
+            return str(mode_override).strip()
+        normalized_route = str(workflow_route or "").strip().lower()
+        if auto_execute:
+            return "execute"
+        if normalized_route.startswith("query."):
+            return "chat"
+        if normalized_route.startswith("followup."):
+            return "plan"
+        return "plan"
+
+    async def begin_ingress_run(
+        self,
+        *,
+        tenant_id: int,
+        source_system: str,
+        source_message_id: Optional[str],
+        message_text: str,
+        sender: str,
+        channel_ref: Optional[str],
+        thread_context: Optional[dict[str, Any]],
+    ) -> MountedIngressTracking:
+        topic_key = self.build_topic_key(
+            source_system=source_system,
+            channel_ref=channel_ref,
+            source_message_id=source_message_id,
+            sender=sender,
+            message_text=message_text,
+            thread_context=thread_context,
+        )
+        topic_row = (
+            await self.db.execute(
+                select(MountedTopicWorkspace).where(
+                    MountedTopicWorkspace.tenant_id == int(tenant_id),
+                    MountedTopicWorkspace.source_system == (self._compact(source_system, max_len=32) or "unknown"),
+                    MountedTopicWorkspace.topic_key == topic_key,
+                )
+            )
+        ).scalar_one_or_none()
+        topic_exists = topic_row is not None
+        now_at = now_cn_naive()
+        if topic_row is None:
+            topic_row = MountedTopicWorkspace(
+                tenant_id=int(tenant_id),
+                topic_key=topic_key,
+                source_system=self._compact(source_system, max_len=32) or "unknown",
+                channel_ref=self._compact(channel_ref, max_len=128),
+                thread_key=self._compact((thread_context or {}).get("thread_id"), max_len=255),
+                root_message_id=self._compact((thread_context or {}).get("root_message_id"), max_len=128),
+                mode="chat",
+                status="active",
+                metadata_json={},
+            )
+            self.db.add(topic_row)
+            await self.db.flush()
+        topic_row.channel_ref = self._compact(channel_ref, max_len=128)
+        topic_row.thread_key = self._compact((thread_context or {}).get("thread_id"), max_len=255)
+        topic_row.root_message_id = self._compact((thread_context or {}).get("root_message_id"), max_len=128)
+        topic_row.last_message_id = self._compact(source_message_id, max_len=128)
+        topic_row.last_triggered_at = now_at
+        topic_row.metadata_json = {
+            **(topic_row.metadata_json if isinstance(topic_row.metadata_json, dict) else {}),
+            "last_sender": self._compact(sender, max_len=128),
+            "last_message_preview": self._compact(message_text, max_len=200),
+            "last_reply_to_message_id": self._compact((thread_context or {}).get("reply_to_message_id"), max_len=128),
+        }
+
+        context_layers_hit = {
+            "current_message": True,
+            "thread": bool((thread_context or {}).get("root_message_id") or (thread_context or {}).get("thread_id")),
+            "topic": bool(topic_exists),
+        }
+        trace_id = self.build_run_trace_id(
+            source_system=source_system,
+            source_message_id=source_message_id,
+            topic_key=topic_key,
+        )
+        run_row = (
+            await self.db.execute(
+                select(MountedConversationRun).where(
+                    MountedConversationRun.tenant_id == int(tenant_id),
+                    MountedConversationRun.trace_id == trace_id,
+                )
+            )
+        ).scalar_one_or_none()
+        if run_row is None:
+            run_row = MountedConversationRun(
+                tenant_id=int(tenant_id),
+                trace_id=trace_id,
+                topic_workspace_id=int(topic_row.id),
+                source_system=self._compact(source_system, max_len=32) or "unknown",
+                source_message_id=self._compact(source_message_id, max_len=128),
+                mode_before=self._compact(topic_row.mode, max_len=24) if topic_exists else None,
+                active_agent="ConversationOrchestrator",
+                status="running",
+                trigger_reason=f"ingress:{self._compact(source_system, max_len=32) or 'unknown'}",
+                context_layers_hit=context_layers_hit,
+                started_at=now_at,
+            )
+            self.db.add(run_row)
+            await self.db.flush()
+        else:
+            run_row.topic_workspace_id = int(topic_row.id)
+            run_row.status = "running"
+            run_row.active_agent = "ConversationOrchestrator"
+            run_row.context_layers_hit = context_layers_hit
+            if run_row.started_at is None:
+                run_row.started_at = now_at
+
+        return MountedIngressTracking(
+            topic_workspace=topic_row,
+            conversation_run=run_row,
+            context_layers_hit=context_layers_hit,
+        )
+
+    async def record_orchestrator_decision(
+        self,
+        *,
+        tracking: MountedIngressTracking,
+        workflow_route: str,
+        intent_name: str,
+        intent_confidence: float,
+        trigger_reason: str,
+        auto_execute: bool,
+        work_item_id: Optional[int],
+        route_metadata: Optional[dict[str, Any]],
+        mode_override: Optional[str] = None,
+    ) -> str:
+        mode_after = self._mode_for_route(
+            workflow_route=workflow_route,
+            auto_execute=auto_execute,
+            mode_override=mode_override,
+        )
+        topic_row = tracking.topic_workspace
+        run_row = tracking.conversation_run
+        topic_row.mode = mode_after
+        topic_row.metadata_json = {
+            **(topic_row.metadata_json if isinstance(topic_row.metadata_json, dict) else {}),
+            "last_workflow_route": str(workflow_route or "").strip() or None,
+            "last_intent_name": str(intent_name or "").strip() or None,
+            "last_intent_confidence": float(intent_confidence or 0.0),
+            "last_work_item_id": int(work_item_id) if work_item_id is not None else None,
+        }
+        run_row.mode_after = mode_after
+        run_row.active_agent = "ConversationOrchestrator"
+        run_row.trigger_reason = self._compact(trigger_reason, max_len=128)
+        self.db.add(
+            MountedAgentTrace(
+                tenant_id=int(topic_row.tenant_id),
+                conversation_run_id=int(run_row.id),
+                agent_name="ConversationOrchestrator",
+                status="completed",
+                input_summary={
+                    "mode_before": run_row.mode_before,
+                    "context_layers_hit": tracking.context_layers_hit,
+                },
+                output_summary={
+                    "mode_after": mode_after,
+                    "workflow_route": str(workflow_route or "").strip() or None,
+                    "intent_name": str(intent_name or "").strip() or None,
+                    "auto_execute": bool(auto_execute),
+                    "route_metadata": dict(route_metadata or {}),
+                },
+            )
+        )
+        await self.db.flush()
+        return mode_after
+
+    async def finalize_run(
+        self,
+        *,
+        tracking: MountedIngressTracking,
+        result: Optional[dict[str, Any]] = None,
+        error_text: Optional[str] = None,
+    ) -> None:
+        topic_row = tracking.topic_workspace
+        run_row = tracking.conversation_run
+        now_at = now_cn_naive()
+        normalized_error = str(error_text or "").strip()
+        final_reply = str((result or {}).get("direct_reply_text") or "").strip() if isinstance(result, dict) else ""
+        if normalized_error:
+            run_row.status = "failed"
+            run_row.active_agent = "ConversationOrchestrator"
+            run_row.error_text = normalized_error
+        else:
+            run_row.status = "completed"
+            run_row.active_agent = "ReplyComposer"
+            run_row.final_reply = final_reply or None
+            run_row.execution_outcome = {
+                "handled": bool((result or {}).get("handled")) if isinstance(result, dict) else False,
+                "status": str((result or {}).get("status") or "").strip() if isinstance(result, dict) else "",
+                "classification": str((result or {}).get("classification") or "").strip() if isinstance(result, dict) else "",
+                "workflow_route": str((result or {}).get("workflow_route") or "").strip() if isinstance(result, dict) else "",
+                "work_item_id": (result or {}).get("work_item_id") if isinstance(result, dict) else None,
+            }
+        run_row.completed_at = now_at
+        self.db.add(
+            MountedAgentTrace(
+                tenant_id=int(topic_row.tenant_id),
+                conversation_run_id=int(run_row.id),
+                agent_name="ReplyComposer" if not normalized_error else "ConversationOrchestrator",
+                status="failed" if normalized_error else "completed",
+                output_summary=(
+                    {"final_reply": final_reply or None, "status": str((result or {}).get("status") or "").strip() or None}
+                    if not normalized_error
+                    else None
+                ),
+                failure_reason=normalized_error or None,
+            )
+        )
+        await self.db.flush()

backend/app/services/signal/signal_capture_service/service.py

@@ -2,7 +2,6 @@ from __future__ import annotations
 
 from .mixin_analysis_normalize import SignalCaptureAnalysisNormalizeMixin
 from .mixin_capture_entry import SignalCaptureEntryMixin
-from .mixin_capture_legacy import SignalCaptureLegacyMixin
 from .mixin_external_sources import SignalCaptureExternalSourcesMixin
 from .mixin_new_project import SignalCaptureNewProjectMixin
 from .mixin_pipeline_cache import SignalCapturePipelineCacheMixin
@@ -14,7 +13,6 @@ from .mixin_task_orchestration_team import SignalCaptureTaskOrchestrationTeamMix
 
 class ProactiveSignalCaptureService(
     SignalCaptureEntryMixin,
-    SignalCaptureLegacyMixin,
     SignalCapturePipelineCacheMixin,
     SignalCaptureExternalSourcesMixin,
     SignalCaptureTaskOrchestrationMainMixin,

backend/tests/test_ai_pmo_ingress_mainline.py

@@ -22,6 +22,7 @@ from app.services.ai_pmo.ai_pmo_kernel.ai_agent_kernel import gate_execution
 from app.services.ai_pmo.ai_pmo_kernel.ai_pmo_reply_renderer import ReplyRenderer
 from app.services.ai_pmo.ai_pmo_kernel import ai_pmo_typed_executor_mixins_parse
 from app.services.ai_pmo.ai_pmo_kernel_ingress_flow import mixin_entry
+from app.services.ai_pmo.ai_pmo_kernel.mixins.ingress import ai_pmo_kernel_mixins_ingress_render
 from app.services.ai_pmo.ai_pmo_kernel_ingress_flow import mixin_prepare_manual
 from app.services.ai_pmo.ai_pmo_kernel_world_model_core import mixin_project
 from app.services.ai_pmo.ai_pmo_typed_executor_execute import mixin_execute_query_project
@@ -218,6 +219,92 @@ async def test_process_ingress_prioritizes_draft_pending_confirm(monkeypatch) ->
     assert "草稿" in result["direct_reply_text"]
 
 
+@pytest.mark.asyncio
+async def test_process_ingress_records_mounted_runtime_for_draft_confirmation(monkeypatch) -> None:
+    captured: dict[str, object] = {"calls": []}
+
+    class FakeMountedAIRuntimeWriteService:
+        def __init__(self, db) -> None:
+            captured["db"] = db
+
+        async def begin_ingress_run(self, **kwargs):
+            captured["calls"].append(("begin", kwargs))
+            return SimpleNamespace(
+                topic_workspace=SimpleNamespace(id=91, tenant_id=1),
+                conversation_run=SimpleNamespace(id=92),
+                context_layers_hit={"current_message": True, "thread": True, "topic": False},
+            )
+
+        async def record_orchestrator_decision(self, **kwargs):
+            captured["calls"].append(("record", kwargs))
+            return "plan"
+
+        async def finalize_run(self, **kwargs):
+            captured["calls"].append(("finalize", kwargs))
+
+    class FakeKernel(mixin_entry.KernelIngressFlowEntryMixin):
+        def __init__(self) -> None:
+            self.db = object()
+            self.tenant_id = 1
+
+        async def _prepare_ingress_envelope(self, **kwargs):
+            _ = kwargs
+            return {
+                "early_result": None,
+                "clean_text": "确认",
+                "synthetic_message_id": "s-msg-mounted-1",
+                "closure_payload": {},
+                "ingest_event": SimpleNamespace(id=100),
+                "operator": SimpleNamespace(id=8),
+                "work_item": SimpleNamespace(id=55),
+                "session": SimpleNamespace(last_intent=mixin_entry.STATE_DRAFT_PENDING_CONFIRM),
+            }
+
+        async def _maybe_handle_draft_confirmation_ingress(self, **kwargs):
+            _ = kwargs
+            return {
+                "handled": True,
+                "status": "answered",
+                "classification": "command",
+                "workflow_route": "followup.confirm",
+                "reply_template_key": "action_succeeded",
+                "work_item_id": 55,
+                "direct_reply_required": True,
+                "direct_reply_text": "已确认草稿并继续执行。",
+            }
+
+        def _is_short_confirmation_message(self, **kwargs):
+            _ = kwargs
+            return False
+
+        async def _prepare_mixed_sidecar_context(self, **kwargs):
+            _ = kwargs
+            raise AssertionError("draft confirmation flow should bypass mixed sidecar precheck")
+
+    monkeypatch.setattr(mixin_entry, "MountedAIRuntimeWriteService", FakeMountedAIRuntimeWriteService)
+
+    kernel = FakeKernel()
+    result = await kernel.process_ingress(
+        source_system="feishu",
+        source_message_id="feishu-mounted-1",
+        message_text="确认",
+        sender="tester",
+        channel_ref="chat-1",
+        source_sender_open_id="ou_xxx",
+        ingress_options={"thread_context": {"reply_to_message_id": "msg-root-1"}},
+    )
+
+    assert result["status"] == "answered"
+    assert [name for name, _ in captured["calls"]] == ["begin", "record", "finalize"]
+    begin_kwargs = captured["calls"][0][1]
+    record_kwargs = captured["calls"][1][1]
+    finalize_kwargs = captured["calls"][2][1]
+    assert begin_kwargs["source_system"] == "feishu"
+    assert begin_kwargs["source_message_id"] == "s-msg-mounted-1"
+    assert record_kwargs["workflow_route"] == "followup.confirm"
+    assert finalize_kwargs["result"]["direct_reply_text"] == "已确认草稿并继续执行。"
+
+
 def test_should_route_to_draft_confirmation_requires_anchor_or_explicit_id() -> None:
     session = SimpleNamespace(last_intent=mixin_entry.STATE_DRAFT_PENDING_CONFIRM)
 
@@ -1113,6 +1200,51 @@ def test_consultant_dispatch_uses_short_next_action_for_query_answers() -> None:
     assert "/rag/ask" not in dispatch["next_action"]
 
 
+def test_build_capture_result_no_longer_exposes_consultant_dispatch() -> None:
+    class _FakeKernel(ai_pmo_kernel_mixins_ingress_render.KernelIngressRenderMixin):
+        tenant_id = 1
+
+    work_item = SimpleNamespace(
+        id=18,
+        project_id=7,
+        signal_id=None,
+        state="succeeded",
+        intent="query.answer",
+        reply_template_key="action_succeeded",
+        decision_level="team",
+        review_summary={"verified": True},
+        raw_message="现在进展怎么样？",
+        source_system="pmflow_manual",
+        draft_version=0,
+    )
+    intent = CanonicalIntent(
+        name="query.answer",
+        confidence=0.93,
+        project_id=7,
+        signal_title="",
+        signal_summary="当前主链正常。",
+        metadata={},
+    )
+    result = _FakeKernel()._build_capture_result(
+        work_item=work_item,
+        intent=intent,
+        execution_result={
+            "workflow_route": "query.answer",
+            "rag_result": {"answer": "当前主链正常。"},
+        },
+        render={
+            "template_key": "action_succeeded",
+            "text": "当前主链正常。",
+        },
+        direct_reply_required=True,
+        status="answered",
+        classification="question",
+    )
+
+    assert "consultant_dispatch" not in result
+    assert "consultant_dispatch" not in result["rag_result"]
+
+
 @pytest.mark.asyncio
 async def test_low_tier_rag_failure_retries_high_tier(monkeypatch) -> None:
     captured: dict[str, object] = {"logs": []}

backend/tests/test_draft_confirm_handler.py

@@ -203,7 +203,7 @@ def test_consultant_dispatch_contract_marks_scope_expansion_as_challenge() -> No
     assert contract["next_action"]
 
 
-def test_feishu_reply_llm_core_builds_contract_reply_for_clarify_and_challenge() -> None:
+def test_feishu_reply_llm_core_no_longer_uses_consultant_dispatch_reply() -> None:
     clarify = FeishuReplyLLMCoreMixin._build_consultant_dispatch_reply(
         {
             "direct_reply_text": "",
@@ -217,22 +217,8 @@ def test_feishu_reply_llm_core_builds_contract_reply_for_clarify_and_challenge()
             },
         }
     )
-    challenge = FeishuReplyLLMCoreMixin._build_consultant_dispatch_reply(
-        {
-            "direct_reply_text": "",
-            "consultant_dispatch": {
-                "domain": "project",
-                "mode": "challenge",
-                "risk_level": "high",
-                "missing_fields": [],
-                "allow_execute": False,
-                "next_action": "建议先拆分范围，再确认是否继续推进。",
-            },
-        }
-    )
 
-    assert "补充" in str(clarify or "")
-    assert "建议" in str(challenge or "")
+    assert clarify is None
 
 
 @pytest.mark.asyncio
@@ -986,9 +972,28 @@ def test_adaptive_rag_process_entry_and_mixed_format() -> None:
     assert mixed.endswith("，已入库处理")
 
 
-def test_adaptive_rag_select_scene_prompt_prefers_mixed() -> None:
+class _FakePromptResult:
+    @staticmethod
+    def scalar_one_or_none():
+        return None
+
+
+class _FakePromptDB:
+    async def execute(self, query):  # noqa: ARG002
+        return _FakePromptResult()
+
+
+def _make_adaptive_rag_mixin() -> AdaptiveRAGDeepGuardAnswerMixin:
     mixin = AdaptiveRAGDeepGuardAnswerMixin()
-    scene_key, prompt = mixin._select_scene_prompt(
+    mixin.db = _FakePromptDB()
+    mixin.tenant_id = 1
+    return mixin
+
+
+@pytest.mark.asyncio
+async def test_adaptive_rag_select_scene_prompt_prefers_mixed() -> None:
+    mixin = _make_adaptive_rag_mixin()
+    scene_key, prompt = await mixin._select_scene_prompt(
         question="现在进展怎么样？",
         project_id=7,
         asked_via="pmflow_manual",
@@ -1007,9 +1012,10 @@ def test_adaptive_rag_select_scene_prompt_prefers_mixed() -> None:
     assert "新增任务确认流" in prompt
 
 
-def test_adaptive_rag_select_scene_prompt_ai_pmo_manual_entry() -> None:
-    mixin = AdaptiveRAGDeepGuardAnswerMixin()
-    scene_key, prompt = mixin._select_scene_prompt(
+@pytest.mark.asyncio
+async def test_adaptive_rag_select_scene_prompt_ai_pmo_manual_entry() -> None:
+    mixin = _make_adaptive_rag_mixin()
+    scene_key, prompt = await mixin._select_scene_prompt(
         question="帮我总结这个项目",
         project_id=9,
         asked_via="pmflow_manual",
@@ -1027,9 +1033,10 @@ def test_adaptive_rag_select_scene_prompt_ai_pmo_manual_entry() -> None:
     assert "帮我总结这个项目" in prompt
 
 
-def test_adaptive_rag_select_scene_prompt_prefers_decision_choice() -> None:
-    mixin = AdaptiveRAGDeepGuardAnswerMixin()
-    scene_key, prompt = mixin._select_scene_prompt(
+@pytest.mark.asyncio
+async def test_adaptive_rag_select_scene_prompt_prefers_decision_choice() -> None:
+    mixin = _make_adaptive_rag_mixin()
+    scene_key, prompt = await mixin._select_scene_prompt(
         question="如果 AI 助手第一次上线只允许一种提醒，你会把它做成桌面通知还是飞书消息？为什么？",
         project_id=None,
         asked_via="feishu",
@@ -1048,9 +1055,10 @@ def test_adaptive_rag_select_scene_prompt_prefers_decision_choice() -> None:
     assert "飞书消息" in prompt
 
 
-def test_adaptive_rag_select_scene_prompt_prefers_industry_trend_without_project_context() -> None:
-    mixin = AdaptiveRAGDeepGuardAnswerMixin()
-    scene_key, prompt = mixin._select_scene_prompt(
+@pytest.mark.asyncio
+async def test_adaptive_rag_select_scene_prompt_prefers_industry_trend_without_project_context() -> None:
+    mixin = _make_adaptive_rag_mixin()
+    scene_key, prompt = await mixin._select_scene_prompt(
         question="帮我看看宠物行业最新动态",
         project_id=None,
         asked_via="feishu",
@@ -1069,9 +1077,10 @@ def test_adaptive_rag_select_scene_prompt_prefers_industry_trend_without_project
     assert "project_context" not in prompt
 
 
-def test_adaptive_rag_select_scene_prompt_filters_member_context_noise() -> None:
-    mixin = AdaptiveRAGDeepGuardAnswerMixin()
-    scene_key, prompt = mixin._select_scene_prompt(
+@pytest.mark.asyncio
+async def test_adaptive_rag_select_scene_prompt_filters_member_context_noise() -> None:
+    mixin = _make_adaptive_rag_mixin()
+    scene_key, prompt = await mixin._select_scene_prompt(
         question="杨宇现在负载情况如何？",
         project_id=None,
         asked_via="feishu",
@@ -1094,9 +1103,10 @@ def test_adaptive_rag_select_scene_prompt_filters_member_context_noise() -> None
     assert '"projects"' not in prompt
 
 
-def test_adaptive_rag_select_scene_prompt_ignores_negated_project_context_for_trend() -> None:
-    mixin = AdaptiveRAGDeepGuardAnswerMixin()
-    scene_key, prompt = mixin._select_scene_prompt(
+@pytest.mark.asyncio
+async def test_adaptive_rag_select_scene_prompt_ignores_negated_project_context_for_trend() -> None:
+    mixin = _make_adaptive_rag_mixin()
+    scene_key, prompt = await mixin._select_scene_prompt(
         question="帮我看看宠物行业最新动态，重点只说消费侧和产品侧的变化，不要提项目。",
         project_id=7,
         asked_via="feishu",
@@ -1115,9 +1125,10 @@ def test_adaptive_rag_select_scene_prompt_ignores_negated_project_context_for_tr
     assert "project_context" not in prompt
 
 
-def test_adaptive_rag_select_scene_prompt_uses_project_trend_when_project_context_explicit() -> None:
-    mixin = AdaptiveRAGDeepGuardAnswerMixin()
-    scene_key, prompt = mixin._select_scene_prompt(
+@pytest.mark.asyncio
+async def test_adaptive_rag_select_scene_prompt_uses_project_trend_when_project_context_explicit() -> None:
+    mixin = _make_adaptive_rag_mixin()
+    scene_key, prompt = await mixin._select_scene_prompt(
         question="这个项目所在行业最新动态怎么样？",
         project_id=7,
         asked_via="feishu",
@@ -1136,9 +1147,10 @@ def test_adaptive_rag_select_scene_prompt_uses_project_trend_when_project_contex
     assert "外部搜索结果" in prompt
 
 
-def test_adaptive_rag_select_scene_prompt_prefers_project_status_for_progress_questions() -> None:
-    mixin = AdaptiveRAGDeepGuardAnswerMixin()
-    scene_key, prompt = mixin._select_scene_prompt(
+@pytest.mark.asyncio
+async def test_adaptive_rag_select_scene_prompt_prefers_project_status_for_progress_questions() -> None:
+    mixin = _make_adaptive_rag_mixin()
+    scene_key, prompt = await mixin._select_scene_prompt(
         question="这个项目最新进展怎么样？",
         project_id=7,
         asked_via="feishu",

backend/tests/test_feishu_pipeline_queue_contract.py

@@ -214,6 +214,12 @@ def test_feishu_signal_pipeline_service_runtime_excludes_legacy_followup_mixins(
     assert "FeishuPipelineFollowupBMixin" not in source
 
 
+def test_proactive_signal_capture_service_runtime_excludes_legacy_capture_mixin() -> None:
+    source = inspect.getsource(ProactiveSignalCaptureService)
+
+    assert "SignalCaptureLegacyMixin" not in source
+
+
 def test_decorate_capture_result_adds_queue_action_card_metadata() -> None:
     result = FeishuPipelineProcessMessageHelpersBMixin._decorate_capture_result(
         {

backend/tests/test_mounted_ai_runtime_api.py

@@ -0,0 +1,253 @@
+from __future__ import annotations
+
+import os
+from types import SimpleNamespace
+
+import pytest
+import pytest_asyncio
+from fastapi import FastAPI
+from httpx import ASGITransport, AsyncClient
+from sqlalchemy import JSON
+from sqlalchemy.ext.asyncio import async_sessionmaker, create_async_engine
+
+os.environ["DATABASE_URL"] = "postgresql+asyncpg://tester:tester@localhost:5432/pmflow_test"
+
+from app.api.ai_steward_api import routes_mounted_ai
+from app.core.database import get_db
+from app.dependencies.auth import get_current_user
+from app.dependencies.tenant import get_tenant_id
+from app.models.base import Base
+from app.models.mounted_ai_runtime import (
+    MountedAgentTrace,
+    MountedConversationRun,
+    MountedDraftWorkspace,
+    MountedToolCapabilityPolicy,
+    MountedTopicWorkspace,
+)
+from app.models.training_pipeline import TrainingDataset, TrainingEvaluation, TrainingJob
+
+
+def _use_sqlite_json_types() -> None:
+    for table in (
+        MountedTopicWorkspace.__table__,
+        MountedDraftWorkspace.__table__,
+        MountedConversationRun.__table__,
+        MountedAgentTrace.__table__,
+        MountedToolCapabilityPolicy.__table__,
+        TrainingDataset.__table__,
+        TrainingJob.__table__,
+        TrainingEvaluation.__table__,
+    ):
+        for column in table.columns:
+            if column.name in {
+                "facts_json",
+                "goals_json",
+                "constraints_json",
+                "options_json",
+                "open_questions",
+                "pending_actions",
+                "bound_entities",
+                "metadata_json",
+                "requirement_card",
+                "draft_payload",
+                "required_confirmations",
+                "context_layers_hit",
+                "tool_calls",
+                "execution_outcome",
+                "input_summary",
+                "output_summary",
+                "tool_trace",
+                "source_scope",
+                "tenant_scope",
+                "policy_metadata",
+                "selection_rules",
+                "channel_scope",
+                "metrics_json",
+                "evaluation_summary",
+            }:
+                column.type = JSON()
+
+
+@pytest_asyncio.fixture
+async def db_session():
+    engine = create_async_engine("sqlite+aiosqlite:///:memory:")
+    _use_sqlite_json_types()
+    async with engine.begin() as conn:
+        await conn.run_sync(
+            lambda sync_conn: Base.metadata.create_all(
+                sync_conn,
+                tables=[
+                    MountedTopicWorkspace.__table__,
+                    MountedDraftWorkspace.__table__,
+                    MountedConversationRun.__table__,
+                    MountedAgentTrace.__table__,
+                    MountedToolCapabilityPolicy.__table__,
+                    TrainingDataset.__table__,
+                    TrainingJob.__table__,
+                    TrainingEvaluation.__table__,
+                ],
+            )
+        )
+    maker = async_sessionmaker(engine, expire_on_commit=False)
+    async with maker() as session:
+        yield session
+    await engine.dispose()
+
+
+def _build_app(db_session):
+    app = FastAPI()
+    app.include_router(routes_mounted_ai.router)
+
+    async def _fake_get_db():
+        yield db_session
+
+    app.dependency_overrides[get_db] = _fake_get_db
+    app.dependency_overrides[get_tenant_id] = lambda: 1
+    app.dependency_overrides[get_current_user] = lambda: SimpleNamespace(
+        id=9001,
+        tenant_id=1,
+        role="ADMIN",
+        name="管理员",
+        username="admin",
+    )
+    return app
+
+
+@pytest.mark.asyncio
+async def test_mounted_ai_admin_routes_return_runtime_views(db_session) -> None:
+    topic = MountedTopicWorkspace(
+        tenant_id=1,
+        topic_key="feishu:chat_1:topic_1",
+        source_system="feishu",
+        channel_ref="chat_1",
+        thread_key="thread_1",
+        mode="plan",
+        status="active",
+        open_questions=["负责人是谁"],
+        readiness_score=0.4,
+    )
+    db_session.add(topic)
+    await db_session.flush()
+
+    draft = MountedDraftWorkspace(
+        tenant_id=1,
+        topic_workspace_id=topic.id,
+        state="collecting",
+        draft_version=2,
+        requirement_card={"goals": ["做挂载 AI"]},
+    )
+    run = MountedConversationRun(
+        tenant_id=1,
+        trace_id="mounted-run-1",
+        topic_workspace_id=topic.id,
+        source_system="feishu",
+        source_message_id="om_123",
+        mode_before="plan",
+        mode_after="clarify",
+        active_agent="Clarifier",
+        status="completed",
+        trigger_reason="missing_owner",
+        context_layers_hit={"thread": True, "topic": True},
+        final_reply="还差负责人信息。",
+    )
+    db_session.add_all([draft, run])
+    await db_session.flush()
+
+    trace = MountedAgentTrace(
+        tenant_id=1,
+        conversation_run_id=run.id,
+        agent_name="Clarifier",
+        status="completed",
+        latency_ms=128,
+        input_summary={"mode": "plan"},
+        output_summary={"question": "这件事打算谁来负责？"},
+    )
+    policy = MountedToolCapabilityPolicy(
+        tenant_id=1,
+        tool_name="project.create",
+        risk_level="high",
+        confirm_required=True,
+        auto_execute_allowed=False,
+    )
+    eval_dataset = TrainingDataset(
+        tenant_id=1,
+        dataset_type="eval",
+        version="teacher-frozen-v1",
+        status="ready",
+        sample_count=18,
+        selection_rules={"suite_kind": "teacher_eval_frozen_set"},
+    )
+    db_session.add_all([trace, policy, eval_dataset])
+    await db_session.flush()
+
+    eval_job = TrainingJob(
+        tenant_id=1,
+        dataset_id=eval_dataset.id,
+        job_type="sft",
+        base_model_family="gpt-5.4",
+        target_slot="high",
+        status="queued",
+        provider_backend="openai",
+    )
+    db_session.add(eval_job)
+    await db_session.flush()
+
+    db_session.add(
+        TrainingEvaluation(
+            tenant_id=1,
+            training_job_id=eval_job.id,
+            evaluation_type="offline_judge",
+            status="completed",
+            score=91.5,
+            evaluation_summary={"pass": True},
+        )
+    )
+    await db_session.commit()
+
+    app = _build_app(db_session)
+    transport = ASGITransport(app=app)
+    async with AsyncClient(transport=transport, base_url="http://test") as client:
+        sources_resp = await client.get("/ai-steward/mounted-ai/sources")
+        assert sources_resp.status_code == 200
+        assert sources_resp.json()["data"]["items"][0]["source_system"] == "feishu"
+        assert sources_resp.json()["data"]["items"][0]["topic_count"] == 1
+
+        topics_resp = await client.get("/ai-steward/mounted-ai/topics")
+        assert topics_resp.status_code == 200
+        assert topics_resp.json()["data"]["items"][0]["topic_key"] == "feishu:chat_1:topic_1"
+        assert topics_resp.json()["data"]["items"][0]["draft_state"] == "collecting"
+
+        runs_resp = await client.get("/ai-steward/mounted-ai/runs")
+        assert runs_resp.status_code == 200
+        assert runs_resp.json()["data"]["items"][0]["active_agent"] == "Clarifier"
+
+        detail_resp = await client.get(f"/ai-steward/mounted-ai/runs/{run.id}")
+        assert detail_resp.status_code == 200
+        assert detail_resp.json()["data"]["agent_traces"][0]["agent_name"] == "Clarifier"
+
+        cutover_resp = await client.get("/ai-steward/mounted-ai/cutover-status")
+        assert cutover_resp.status_code == 200
+        assert cutover_resp.json()["data"]["mounted_runtime_ready"] is True
+        assert cutover_resp.json()["data"]["tool_policy_count"] == 1
+        assert cutover_resp.json()["data"]["legacy_cutover_state"] == "retired"
+        assert cutover_resp.json()["data"]["retired_item_count"] >= 4
+        assert cutover_resp.json()["data"]["active_item_count"] == 0
+        assert {
+            item["key"]: item["status"]
+            for item in cutover_resp.json()["data"]["legacy_cutover_items"]
+        } == {
+            "ai_pmo_ingress_consultant_capture": "retired",
+            "feishu_mention_consultant_reply": "retired",
+            "signal_capture_legacy_entry": "retired",
+            "legacy_feishu_queue_followup": "retired",
+        }
+
+        eval_suites_resp = await client.get("/ai-steward/mounted-ai/eval-suites")
+        assert eval_suites_resp.status_code == 200
+        assert eval_suites_resp.json()["data"]["items"][0]["suite_kind"] == "teacher_eval_frozen_set"
+        assert eval_suites_resp.json()["data"]["items"][0]["latest_score"] == 91.5
+
+        eval_suite_detail_resp = await client.get(f"/ai-steward/mounted-ai/eval-suites/{eval_dataset.id}")
+        assert eval_suite_detail_resp.status_code == 200
+        assert eval_suite_detail_resp.json()["data"]["item"]["sample_count"] == 18
+        assert eval_suite_detail_resp.json()["data"]["evaluations"][0]["evaluation_type"] == "offline_judge"

backend/tests/test_mounted_ai_runtime_migration.py

@@ -0,0 +1,54 @@
+from __future__ import annotations
+
+import importlib.util
+from pathlib import Path
+
+import sqlalchemy as sa
+from alembic.migration import MigrationContext
+from alembic.operations import Operations
+
+
+def _load_migration_module():
+    path = (
+        Path(__file__).resolve().parents[1]
+        / "alembic"
+        / "versions"
+        / "20260405_add_mounted_ai_runtime_tables.py"
+    )
+    spec = importlib.util.spec_from_file_location("mounted_ai_runtime_tables_migration", path)
+    module = importlib.util.module_from_spec(spec)
+    assert spec is not None
+    assert spec.loader is not None
+    spec.loader.exec_module(module)
+    return module
+
+
+def test_mounted_ai_runtime_migration_adds_tables() -> None:
+    migration = _load_migration_module()
+    engine = sa.create_engine("sqlite:///:memory:")
+
+    with engine.begin() as conn:
+        context = MigrationContext.configure(conn)
+        migration.op = Operations(context)
+        migration.upgrade()
+
+        inspector = sa.inspect(conn)
+        tables = set(inspector.get_table_names())
+
+        assert "mounted_topic_workspaces" in tables
+        assert "mounted_draft_workspaces" in tables
+        assert "mounted_conversation_runs" in tables
+        assert "mounted_agent_traces" in tables
+        assert "mounted_tool_capability_policies" in tables
+
+        migration.downgrade()
+
+        inspector = sa.inspect(conn)
+        tables = set(inspector.get_table_names())
+
+        assert "mounted_topic_workspaces" not in tables
+        assert "mounted_draft_workspaces" not in tables
+        assert "mounted_conversation_runs" not in tables
+        assert "mounted_agent_traces" not in tables
+        assert "mounted_tool_capability_policies" not in tables
+

backend/tests/test_mounted_ai_runtime_models.py

@@ -0,0 +1,185 @@
+from __future__ import annotations
+
+import os
+
+import pytest
+from sqlalchemy import JSON, create_engine
+from sqlalchemy.exc import IntegrityError
+from sqlalchemy.orm import Session
+
+os.environ["DATABASE_URL"] = "sqlite:///:memory:"
+
+from app.models.base import Base
+from app.models.mounted_ai_runtime import (
+    MountedAgentTrace,
+    MountedConversationRun,
+    MountedDraftWorkspace,
+    MountedToolCapabilityPolicy,
+    MountedTopicWorkspace,
+)
+
+
+def _use_sqlite_json_types() -> None:
+    for table in (
+        MountedTopicWorkspace.__table__,
+        MountedDraftWorkspace.__table__,
+        MountedConversationRun.__table__,
+        MountedAgentTrace.__table__,
+        MountedToolCapabilityPolicy.__table__,
+    ):
+        for column in table.columns:
+            if column.name in {
+                "facts_json",
+                "goals_json",
+                "constraints_json",
+                "options_json",
+                "open_questions",
+                "pending_actions",
+                "bound_entities",
+                "metadata_json",
+                "requirement_card",
+                "draft_payload",
+                "required_confirmations",
+                "context_layers_hit",
+                "tool_calls",
+                "execution_outcome",
+                "input_summary",
+                "output_summary",
+                "tool_trace",
+                "source_scope",
+                "tenant_scope",
+                "policy_metadata",
+            }:
+                column.type = JSON()
+
+
+def _create_session() -> Session:
+    engine = create_engine("sqlite:///:memory:")
+    _use_sqlite_json_types()
+    Base.metadata.create_all(
+        engine,
+        tables=[
+            MountedTopicWorkspace.__table__,
+            MountedDraftWorkspace.__table__,
+            MountedConversationRun.__table__,
+            MountedAgentTrace.__table__,
+            MountedToolCapabilityPolicy.__table__,
+        ],
+    )
+    return Session(engine)
+
+
+def test_mounted_ai_runtime_models_roundtrip_and_defaults() -> None:
+    session = _create_session()
+
+    topic = MountedTopicWorkspace(
+        tenant_id=1,
+        topic_key="feishu:chat_1:topic_1",
+        source_system="feishu",
+        channel_ref="chat_1",
+        thread_key="thread_1",
+        root_message_id="om_root_1",
+        mode="plan",
+        facts_json={"problem": "直播复盘"},
+        open_questions=["负责人是谁"],
+        readiness_score=0.35,
+    )
+    session.add(topic)
+    session.flush()
+
+    draft = MountedDraftWorkspace(
+        tenant_id=1,
+        topic_workspace_id=topic.id,
+        draft_version=1,
+        requirement_card={"goals": ["整理复盘需求"]},
+        required_confirmations=[{"type": "user_confirm"}],
+    )
+    run = MountedConversationRun(
+        tenant_id=1,
+        trace_id="trace-mounted-1",
+        topic_workspace_id=topic.id,
+        source_system="feishu",
+        source_message_id="om_123",
+        mode_before="plan",
+        mode_after="clarify",
+        active_agent="Clarifier",
+        trigger_reason="missing_owner",
+        context_layers_hit={"thread": True, "topic": True},
+        final_reply="还差负责人信息。",
+    )
+    session.add_all([draft, run])
+    session.flush()
+
+    trace = MountedAgentTrace(
+        tenant_id=1,
+        conversation_run_id=run.id,
+        agent_name="Clarifier",
+        status="completed",
+        latency_ms=128,
+        input_summary={"mode": "plan"},
+        output_summary={"question": "这件事打算谁来负责？"},
+    )
+    policy = MountedToolCapabilityPolicy(
+        tenant_id=1,
+        tool_name="project.create",
+        risk_level="high",
+        source_scope={"systems": ["feishu"]},
+    )
+    session.add_all([trace, policy])
+    session.commit()
+
+    saved_topic = session.query(MountedTopicWorkspace).one()
+    saved_draft = session.query(MountedDraftWorkspace).one()
+    saved_run = session.query(MountedConversationRun).one()
+    saved_trace = session.query(MountedAgentTrace).one()
+    saved_policy = session.query(MountedToolCapabilityPolicy).one()
+
+    assert saved_topic.status == "active"
+    assert saved_draft.state == "collecting"
+    assert saved_run.status == "queued"
+    assert saved_trace.tool_trace is None
+    assert saved_policy.confirm_required is True
+    assert saved_policy.auto_execute_allowed is False
+    assert saved_policy.is_active is True
+
+
+def test_mounted_ai_runtime_models_enforce_unique_scopes() -> None:
+    session = _create_session()
+
+    session.add(
+        MountedTopicWorkspace(
+            tenant_id=9,
+            topic_key="feishu:chat_9:topic_9",
+            source_system="feishu",
+        )
+    )
+    session.add(
+        MountedToolCapabilityPolicy(
+            tenant_id=9,
+            tool_name="project.create",
+            risk_level="high",
+        )
+    )
+    session.commit()
+
+    session.add(
+        MountedTopicWorkspace(
+            tenant_id=9,
+            topic_key="feishu:chat_9:topic_9",
+            source_system="feishu",
+        )
+    )
+    with pytest.raises(IntegrityError):
+        session.commit()
+    session.rollback()
+
+    session.add(
+        MountedToolCapabilityPolicy(
+            tenant_id=9,
+            tool_name="project.create",
+            risk_level="high",
+        )
+    )
+    with pytest.raises(IntegrityError):
+        session.commit()
+

backend/tests/test_mounted_ai_runtime_write_service.py

@@ -0,0 +1,163 @@
+from __future__ import annotations
+
+import os
+
+import pytest
+import pytest_asyncio
+from sqlalchemy import JSON, select
+from sqlalchemy.ext.asyncio import async_sessionmaker, create_async_engine
+
+os.environ["DATABASE_URL"] = "postgresql+asyncpg://tester:tester@localhost:5432/pmflow_test"
+
+from app.models.base import Base
+from app.models.mounted_ai_runtime import (
+    MountedAgentTrace,
+    MountedConversationRun,
+    MountedTopicWorkspace,
+)
+from app.services.mounted_ai_runtime_write_service import MountedAIRuntimeWriteService
+
+
+def _use_sqlite_json_types() -> None:
+    for table in (
+        MountedTopicWorkspace.__table__,
+        MountedConversationRun.__table__,
+        MountedAgentTrace.__table__,
+    ):
+        for column in table.columns:
+            if column.name in {
+                "facts_json",
+                "goals_json",
+                "constraints_json",
+                "options_json",
+                "open_questions",
+                "pending_actions",
+                "bound_entities",
+                "metadata_json",
+                "context_layers_hit",
+                "tool_calls",
+                "execution_outcome",
+                "input_summary",
+                "output_summary",
+                "tool_trace",
+            }:
+                column.type = JSON()
+
+
+@pytest_asyncio.fixture
+async def db_session():
+    engine = create_async_engine("sqlite+aiosqlite:///:memory:")
+    _use_sqlite_json_types()
+    async with engine.begin() as conn:
+        await conn.run_sync(
+            lambda sync_conn: Base.metadata.create_all(
+                sync_conn,
+                tables=[
+                    MountedTopicWorkspace.__table__,
+                    MountedConversationRun.__table__,
+                    MountedAgentTrace.__table__,
+                ],
+            )
+        )
+    maker = async_sessionmaker(engine, expire_on_commit=False)
+    async with maker() as session:
+        yield session
+    await engine.dispose()
+
+
+@pytest.mark.asyncio
+async def test_begin_ingress_run_uses_thread_anchor_and_sets_runtime_context(db_session) -> None:
+    service = MountedAIRuntimeWriteService(db_session)
+
+    tracking = await service.begin_ingress_run(
+        tenant_id=1,
+        source_system="feishu",
+        source_message_id="om_message_1",
+        message_text="帮我复盘一下这个直播需求。",
+        sender="林宇航",
+        channel_ref="chat_1",
+        thread_context={
+            "thread_id": "thread_1",
+            "root_message_id": "om_root_1",
+            "reply_to_message_id": "om_reply_1",
+        },
+    )
+    await db_session.commit()
+
+    saved_topic = (
+        await db_session.execute(select(MountedTopicWorkspace))
+    ).scalar_one()
+    saved_run = (
+        await db_session.execute(select(MountedConversationRun))
+    ).scalar_one()
+
+    assert tracking.topic_workspace.id == saved_topic.id
+    assert saved_topic.topic_key == "feishu:chat_1:root:om_root_1"
+    assert saved_topic.thread_key == "thread_1"
+    assert saved_topic.last_message_id == "om_message_1"
+    assert saved_run.status == "running"
+    assert saved_run.mode_before is None
+    assert saved_run.context_layers_hit == {
+        "current_message": True,
+        "thread": True,
+        "topic": False,
+    }
+
+
+@pytest.mark.asyncio
+async def test_record_decision_and_finalize_run_persist_topic_mode_and_traces(db_session) -> None:
+    service = MountedAIRuntimeWriteService(db_session)
+
+    tracking = await service.begin_ingress_run(
+        tenant_id=1,
+        source_system="feishu",
+        source_message_id="om_message_2",
+        message_text="分配给谁做比较合适？",
+        sender="林宇航",
+        channel_ref="chat_2",
+        thread_context={"root_message_id": "om_root_2"},
+    )
+
+    await service.record_orchestrator_decision(
+        tracking=tracking,
+        workflow_route="project_signal.create",
+        intent_name="signal_create",
+        intent_confidence=0.82,
+        trigger_reason="manual_review",
+        auto_execute=False,
+        work_item_id=321,
+        route_metadata={"workflow_family": "project_signal", "workflow_action": "create"},
+    )
+    await service.finalize_run(
+        tracking=tracking,
+        result={
+            "handled": True,
+            "status": "captured",
+            "classification": "command",
+            "work_item_id": 321,
+            "workflow_route": "project_signal.create",
+            "direct_reply_text": "我先帮你收敛需求，再给负责人建议。",
+        },
+    )
+    await db_session.commit()
+
+    saved_topic = (
+        await db_session.execute(select(MountedTopicWorkspace))
+    ).scalar_one()
+    saved_run = (
+        await db_session.execute(select(MountedConversationRun))
+    ).scalar_one()
+    traces = (
+        await db_session.execute(
+            select(MountedAgentTrace).order_by(MountedAgentTrace.id.asc())
+        )
+    ).scalars().all()
+
+    assert saved_topic.mode == "plan"
+    assert saved_topic.metadata_json["last_workflow_route"] == "project_signal.create"
+    assert saved_run.mode_after == "plan"
+    assert saved_run.status == "completed"
+    assert saved_run.final_reply == "我先帮你收敛需求，再给负责人建议。"
+    assert len(traces) == 2
+    assert traces[0].agent_name == "ConversationOrchestrator"
+    assert traces[1].agent_name == "ReplyComposer"

frontend/src/api/mountedAi.ts

@@ -0,0 +1,158 @@
+import { api } from "@/lib/api";
+
+type ApiEnvelope<T> = { code: number; message: string; data: T };
+
+function unwrap<T>(raw: ApiEnvelope<T> | T): T {
+  if (raw && typeof raw === "object" && "data" in (raw as Record<string, unknown>)) {
+    return (raw as ApiEnvelope<T>).data;
+  }
+  return raw as T;
+}
+
+export interface MountedAISourceItem {
+  source_system: string;
+  run_count: number;
+  topic_count: number;
+  active_topic_count: number;
+  last_seen_at?: string | null;
+}
+
+export interface MountedAISourceListPayload {
+  items: MountedAISourceItem[];
+}
+
+export interface MountedAITopicItem {
+  id: number;
+  topic_key: string;
+  source_system: string;
+  channel_ref?: string | null;
+  thread_key?: string | null;
+  mode: string;
+  status: string;
+  readiness_score?: number | null;
+  draft_state?: string | null;
+  draft_version?: number | null;
+  updated_at?: string | null;
+}
+
+export interface MountedAITopicListPayload {
+  items: MountedAITopicItem[];
+  total: number;
+}
+
+export interface MountedAIRunItem {
+  id: number;
+  trace_id: string;
+  source_system: string;
+  source_message_id?: string | null;
+  mode_before?: string | null;
+  mode_after?: string | null;
+  active_agent?: string | null;
+  status: string;
+  trigger_reason?: string | null;
+  final_reply?: string | null;
+  trace_count: number;
+  updated_at?: string | null;
+}
+
+export interface MountedAIRunListPayload {
+  items: MountedAIRunItem[];
+  total: number;
+}
+
+export interface MountedAIAgentTraceItem {
+  id: number;
+  agent_name: string;
+  status: string;
+  latency_ms?: number | null;
+  failure_reason?: string | null;
+  created_at?: string | null;
+}
+
+export interface MountedAIRunDetailPayload {
+  item: MountedAIRunItem;
+  agent_traces: MountedAIAgentTraceItem[];
+  context_layers_hit?: Record<string, unknown> | null;
+  tool_calls?: unknown;
+  execution_outcome?: unknown;
+}
+
+export interface MountedAICutoverStatusPayload {
+  mounted_runtime_ready: boolean;
+  topic_workspace_count: number;
+  conversation_run_count: number;
+  tool_policy_count: number;
+  observed_source_systems: string[];
+  legacy_cutover_state: string;
+  retired_item_count: number;
+  active_item_count: number;
+  legacy_cutover_items: MountedAILegacyCutoverItem[];
+}
+
+export interface MountedAILegacyCutoverItem {
+  key: string;
+  title: string;
+  status: string;
+  summary: string;
+  evidence: string;
+}
+
+export interface MountedAIEvalSuiteItem {
+  id: number;
+  dataset_type: string;
+  version: string;
+  status: string;
+  suite_kind: string;
+  sample_count: number;
+  latest_score?: number | null;
+  last_evaluated_at?: string | null;
+}
+
+export interface MountedAIEvalSuiteListPayload {
+  items: MountedAIEvalSuiteItem[];
+  total: number;
+}
+
+export const mountedAiApi = {
+  async getSources(): Promise<MountedAISourceListPayload> {
+    const { data } = await api.get<ApiEnvelope<MountedAISourceListPayload>>("/ai-steward/mounted-ai/sources");
+    return unwrap(data);
+  },
+
+  async getTopics(params?: { limit?: number; offset?: number; mode?: string }): Promise<MountedAITopicListPayload> {
+    const { data } = await api.get<ApiEnvelope<MountedAITopicListPayload>>("/ai-steward/mounted-ai/topics", {
+      params: {
+        limit: params?.limit ?? 20,
+        offset: params?.offset ?? 0,
+        mode: params?.mode,
+      },
+    });
+    return unwrap(data);
+  },
+
+  async getRuns(params?: { limit?: number; offset?: number; status?: string }): Promise<MountedAIRunListPayload> {
+    const { data } = await api.get<ApiEnvelope<MountedAIRunListPayload>>("/ai-steward/mounted-ai/runs", {
+      params: {
+        limit: params?.limit ?? 20,
+        offset: params?.offset ?? 0,
+        status: params?.status,
+      },
+    });
+    return unwrap(data);
+  },
+
+  async getRunDetail(runId: number): Promise<MountedAIRunDetailPayload> {
+    const { data } = await api.get<ApiEnvelope<MountedAIRunDetailPayload>>(`/ai-steward/mounted-ai/runs/${runId}`);
+    return unwrap(data);
+  },
+
+  async getCutoverStatus(): Promise<MountedAICutoverStatusPayload> {
+    const { data } = await api.get<ApiEnvelope<MountedAICutoverStatusPayload>>("/ai-steward/mounted-ai/cutover-status");
+    return unwrap(data);
+  },
+
+  async getEvalSuites(): Promise<MountedAIEvalSuiteListPayload> {
+    const { data } = await api.get<ApiEnvelope<MountedAIEvalSuiteListPayload>>("/ai-steward/mounted-ai/eval-suites");
+    return unwrap(data);
+  },
+};

frontend/src/pages/scheduler-center/SchedulerCenterPage.tsx

@@ -1,7 +1,168 @@
+import { useEffect, useMemo, useState } from "react";
+import { useQuery } from "@tanstack/react-query";
+import { Bot, RefreshCw, ShieldCheck, Waypoints } from "lucide-react";
+
+import { mountedAiApi } from "@/api/mountedAi";
+import { Button } from "@/components/ui/button";
+import { Card, CardContent, CardDescription, CardHeader, CardTitle } from "@/components/ui/card";
+import { MountedAIAcceptancePanel } from "./components/MountedAIAcceptancePanel";
+import { MountedAIRunReplayPanel } from "./components/MountedAIRunReplayPanel";
+import { MountedAIRunsPanel } from "./components/MountedAIRunsPanel";
+import { MountedAISourceStatusPanel } from "./components/MountedAISourceStatusPanel";
+import { MountedAITopicsPanel } from "./components/MountedAITopicsPanel";
+
 export default function SchedulerCenterPage() {
+  const [selectedRunId, setSelectedRunId] = useState<number | null>(null);
+
+  const sourcesQuery = useQuery({
+    queryKey: ["mounted-ai", "sources"],
+    queryFn: () => mountedAiApi.getSources(),
+  });
+  const topicsQuery = useQuery({
+    queryKey: ["mounted-ai", "topics"],
+    queryFn: () => mountedAiApi.getTopics({ limit: 10 }),
+  });
+  const runsQuery = useQuery({
+    queryKey: ["mounted-ai", "runs"],
+    queryFn: () => mountedAiApi.getRuns({ limit: 10 }),
+  });
+  const cutoverQuery = useQuery({
+    queryKey: ["mounted-ai", "cutover-status"],
+    queryFn: () => mountedAiApi.getCutoverStatus(),
+  });
+  const evalSuitesQuery = useQuery({
+    queryKey: ["mounted-ai", "eval-suites"],
+    queryFn: () => mountedAiApi.getEvalSuites(),
+  });
+  const selectedRunDetailQuery = useQuery({
+    queryKey: ["mounted-ai", "runs", "detail", selectedRunId],
+    queryFn: () => mountedAiApi.getRunDetail(selectedRunId as number),
+    enabled: selectedRunId != null,
+  });
+
+  useEffect(() => {
+    if (selectedRunId == null && (runsQuery.data?.items.length || 0) > 0) {
+      setSelectedRunId(runsQuery.data?.items[0]?.id ?? null);
+    }
+  }, [runsQuery.data?.items, selectedRunId]);
+
+  const loading =
+    sourcesQuery.isLoading || topicsQuery.isLoading || runsQuery.isLoading || cutoverQuery.isLoading || evalSuitesQuery.isLoading;
+  const runtimeReady = !!cutoverQuery.data?.mounted_runtime_ready;
+  const frozenEvalSuite =
+    evalSuitesQuery.data?.items.find((item) => item.suite_kind === "teacher_eval_frozen_set") ||
+    evalSuitesQuery.data?.items[0] ||
+    null;
+  const observedSourceText = useMemo(() => {
+    const items = cutoverQuery.data?.observed_source_systems || [];
+    return items.length > 0 ? items.join(" / ") : "暂无";
+  }, [cutoverQuery.data?.observed_source_systems]);
+
+  const refreshAll = async () => {
+    await Promise.all([
+      sourcesQuery.refetch(),
+      topicsQuery.refetch(),
+      runsQuery.refetch(),
+      cutoverQuery.refetch(),
+      evalSuitesQuery.refetch(),
+    ]);
+  };
+
   return (
-    <div className="rounded-2xl border border-dashed p-6 text-center text-sm text-muted-foreground">
-      调度中心功能建设中
+    <div className="space-y-6">
+      <div className="rounded-3xl border border-slate-200/80 bg-gradient-to-br from-slate-950 via-slate-900 to-cyan-950 px-6 py-6 text-white shadow-xl">
+        <div className="flex flex-col gap-4 lg:flex-row lg:items-start lg:justify-between">
+          <div className="space-y-3">
+            <div className="inline-flex items-center gap-2 rounded-full border border-white/15 bg-white/10 px-3 py-1 text-xs font-medium text-white/80">
+              <Bot className="h-3.5 w-3.5" />
+              Mounted AI Console
+            </div>
+            <div>
+              <h1 className="text-2xl font-semibold tracking-tight">挂载 Runtime 调度中心</h1>
+              <p className="mt-2 max-w-3xl text-sm leading-6 text-slate-200/80">
+                这里展示当前 Mounted Runtime 的来源观测、topic workspace、conversation runs 与 cutover 状态。
+              </p>
+            </div>
+          </div>
+          <div className="flex flex-wrap items-center gap-2">
+            <Button
+              variant="outline"
+              size="sm"
+              className="border-white/20 bg-white/10 text-white hover:bg-white/15"
+              onClick={() => void refreshAll()}
+              disabled={loading}
+            >
+              <RefreshCw className={`mr-1 h-4 w-4 ${loading ? "animate-spin" : ""}`} />
+              {loading ? "刷新中..." : "刷新控制台"}
+            </Button>
+          </div>
+        </div>
+      </div>
+
+      <div className="grid gap-4 md:grid-cols-2 xl:grid-cols-4">
+        <Card className="border-slate-200/80 bg-white/90 shadow-sm dark:border-slate-800 dark:bg-slate-900/85">
+          <CardHeader className="pb-2">
+            <CardDescription>Runtime 就绪度</CardDescription>
+            <CardTitle className="flex items-center gap-2 text-base">
+              <ShieldCheck className="h-4 w-4" />
+              {runtimeReady ? "已就绪" : "待接管"}
+            </CardTitle>
+          </CardHeader>
+          <CardContent className="text-xs text-slate-500 dark:text-slate-400">
+            legacy 状态：{cutoverQuery.data?.legacy_cutover_state || "-"}
+          </CardContent>
+        </Card>
+
+        <Card className="border-slate-200/80 bg-white/90 shadow-sm dark:border-slate-800 dark:bg-slate-900/85">
+          <CardHeader className="pb-2">
+            <CardDescription>已观测来源</CardDescription>
+            <CardTitle className="text-base">{observedSourceText}</CardTitle>
+          </CardHeader>
+          <CardContent className="text-xs text-slate-500 dark:text-slate-400">
+            来源数量：{cutoverQuery.data?.observed_source_systems.length || 0}
+          </CardContent>
+        </Card>
+
+        <Card className="border-slate-200/80 bg-white/90 shadow-sm dark:border-slate-800 dark:bg-slate-900/85">
+          <CardHeader className="pb-2">
+            <CardDescription>Topic Workspace</CardDescription>
+            <CardTitle className="text-base">{cutoverQuery.data?.topic_workspace_count ?? "-"}</CardTitle>
+          </CardHeader>
+          <CardContent className="text-xs text-slate-500 dark:text-slate-400">
+            当前活跃主题总数
+          </CardContent>
+        </Card>
+
+        <Card className="border-slate-200/80 bg-white/90 shadow-sm dark:border-slate-800 dark:bg-slate-900/85">
+          <CardHeader className="pb-2">
+            <CardDescription>Conversation Runs</CardDescription>
+            <CardTitle className="flex items-center gap-2 text-base">
+              <Waypoints className="h-4 w-4" />
+              {cutoverQuery.data?.conversation_run_count ?? "-"}
+            </CardTitle>
+          </CardHeader>
+          <CardContent className="text-xs text-slate-500 dark:text-slate-400">
+            Tool Policy：{cutoverQuery.data?.tool_policy_count ?? "-"}
+          </CardContent>
+        </Card>
+      </div>
+
+      <div className="grid gap-4 xl:grid-cols-2">
+        <MountedAISourceStatusPanel items={sourcesQuery.data?.items || []} loading={sourcesQuery.isLoading} />
+        <MountedAITopicsPanel items={topicsQuery.data?.items || []} loading={topicsQuery.isLoading} />
+      </div>
+
+      <div className="grid gap-4 xl:grid-cols-[1.2fr_0.8fr]">
+        <MountedAIRunsPanel
+          items={runsQuery.data?.items || []}
+          loading={runsQuery.isLoading}
+          selectedRunId={selectedRunId}
+          onSelectRun={setSelectedRunId}
+        />
+        <MountedAIRunReplayPanel detail={selectedRunDetailQuery.data} loading={selectedRunDetailQuery.isLoading} />
+      </div>
+
+      <MountedAIAcceptancePanel frozenSuite={frozenEvalSuite} cutoverStatus={cutoverQuery.data} />
     </div>
   );
 }

frontend/src/pages/scheduler-center/components/MountedAIAcceptancePanel.tsx

@@ -0,0 +1,121 @@
+import { AlertTriangle, ShieldCheck, Target } from "lucide-react";
+
+import type { MountedAICutoverStatusPayload, MountedAIEvalSuiteItem } from "@/api/mountedAi";
+import { Badge } from "@/components/ui/badge";
+import { Card, CardContent, CardDescription, CardHeader, CardTitle } from "@/components/ui/card";
+import { formatBeijingDateTime } from "@/lib/datetime";
+
+interface MountedAIAcceptancePanelProps {
+  frozenSuite?: MountedAIEvalSuiteItem | null;
+  cutoverStatus?: MountedAICutoverStatusPayload | null;
+}
+
+export function MountedAIAcceptancePanel({ frozenSuite, cutoverStatus }: MountedAIAcceptancePanelProps) {
+  const latestScore = frozenSuite?.latest_score ?? null;
+  const passed = latestScore != null && latestScore >= 90;
+  const legacyItems = cutoverStatus?.legacy_cutover_items || [];
+
+  return (
+    <Card className="border-slate-200/80 bg-white/90 shadow-sm dark:border-slate-800 dark:bg-slate-900/85">
+      <CardHeader>
+        <CardTitle className="text-base">Acceptance Gates</CardTitle>
+        <CardDescription>按 PLAN3 固定展示当前终局验收门槛，避免实现阶段和验收标准脱节。</CardDescription>
+      </CardHeader>
+      <CardContent className="space-y-4">
+        <div className="grid gap-3 md:grid-cols-3">
+          <div className="rounded-2xl border border-slate-200/80 p-4 dark:border-slate-700">
+            <div className="flex items-center gap-2 text-sm font-medium">
+              <Target className="h-4 w-4 text-cyan-600" />
+              综合分
+            </div>
+            <div className="mt-2 text-2xl font-semibold">{latestScore != null ? latestScore.toFixed(1) : "≥ 90"}</div>
+            <div className="mt-1 text-xs text-slate-500 dark:text-slate-400">
+              {latestScore != null ? "当前读取的是 frozen suite 最新均分" : "低于 90 不予验收"}
+            </div>
+          </div>
+          <div className="rounded-2xl border border-slate-200/80 p-4 dark:border-slate-700">
+            <div className="flex items-center gap-2 text-sm font-medium">
+              <ShieldCheck className="h-4 w-4 text-emerald-600" />
+              严重误执行
+            </div>
+            <div className="mt-2 text-2xl font-semibold">0</div>
+            <div className="mt-1 text-xs text-slate-500 dark:text-slate-400">任何高危误执行直接判失败</div>
+          </div>
+          <div className="rounded-2xl border border-slate-200/80 p-4 dark:border-slate-700">
+            <div className="flex items-center gap-2 text-sm font-medium">
+              <AlertTriangle className="h-4 w-4 text-amber-600" />
+              当前状态
+            </div>
+            <div className="mt-2">
+              <Badge variant={passed ? "secondary" : "outline"}>
+                {passed ? "已达到门槛" : frozenSuite ? "尚未达标" : "待教师冻结集评测"}
+              </Badge>
+            </div>
+            <div className="mt-1 text-xs text-slate-500 dark:text-slate-400">
+              {frozenSuite
+                ? `${frozenSuite.version} · ${frozenSuite.sample_count} 样本 · ${formatBeijingDateTime(frozenSuite.last_evaluated_at)}`
+                : "训练后置，但门槛已锁定"}
+            </div>
+          </div>
+        </div>
+
+        <div className="grid gap-2 md:grid-cols-2 xl:grid-cols-4">
+          <div className="rounded-xl border border-dashed border-slate-300 p-3 text-sm dark:border-slate-700">
+            `mode accuracy` ≥ 95
+          </div>
+          <div className="rounded-xl border border-dashed border-slate-300 p-3 text-sm dark:border-slate-700">
+            `topic binding` ≥ 93
+          </div>
+          <div className="rounded-xl border border-dashed border-slate-300 p-3 text-sm dark:border-slate-700">
+            `clarify correctness` ≥ 90
+          </div>
+          <div className="rounded-xl border border-dashed border-slate-300 p-3 text-sm dark:border-slate-700">
+            `draft readiness` ≥ 90
+          </div>
+          <div className="rounded-xl border border-dashed border-slate-300 p-3 text-sm dark:border-slate-700">
+            `execution safety` ≥ 99
+          </div>
+          <div className="rounded-xl border border-dashed border-slate-300 p-3 text-sm dark:border-slate-700">
+            `assignee recommendation` ≥ 90
+          </div>
+          <div className="rounded-xl border border-dashed border-slate-300 p-3 text-sm dark:border-slate-700">
+            `reply quality` ≥ 90
+          </div>
+          <div className="rounded-xl border border-dashed border-slate-300 p-3 text-sm dark:border-slate-700">
+            高危回归 = 0
+          </div>
+        </div>
+
+        <div className="space-y-3 border-t border-slate-200/80 pt-4 dark:border-slate-800">
+          <div className="flex flex-wrap items-center gap-2">
+            <div className="text-sm font-medium">Legacy Cutover</div>
+            <Badge variant={cutoverStatus?.active_item_count ? "outline" : "secondary"}>
+              {cutoverStatus?.legacy_cutover_state || "unknown"}
+            </Badge>
+            <div className="text-xs text-slate-500 dark:text-slate-400">
+              已退役 {cutoverStatus?.retired_item_count ?? 0} 项 / 剩余活跃 {cutoverStatus?.active_item_count ?? 0} 项
+            </div>
+          </div>
+          <div className="grid gap-3 md:grid-cols-2">
+            {legacyItems.length > 0 ? (
+              legacyItems.map((item) => (
+                <div key={item.key} className="rounded-2xl border border-slate-200/80 p-4 dark:border-slate-700">
+                  <div className="flex items-center justify-between gap-3">
+                    <div className="text-sm font-medium">{item.title}</div>
+                    <Badge variant={item.status === "retired" ? "secondary" : "outline"}>{item.status}</Badge>
+                  </div>
+                  <div className="mt-2 text-xs leading-5 text-slate-600 dark:text-slate-300">{item.summary}</div>
+                  <div className="mt-2 text-xs text-slate-500 dark:text-slate-400">{item.evidence}</div>
+                </div>
+              ))
+            ) : (
+              <div className="rounded-2xl border border-dashed border-slate-300 p-4 text-sm text-slate-500 dark:border-slate-700 dark:text-slate-400">
+                cutover 检查项尚未返回。
+              </div>
+            )}
+          </div>
+        </div>
+      </CardContent>
+    </Card>
+  );
+}

frontend/src/pages/scheduler-center/components/MountedAIRunReplayPanel.tsx

@@ -0,0 +1,114 @@
+import { Badge } from "@/components/ui/badge";
+import { Card, CardContent, CardDescription, CardHeader, CardTitle } from "@/components/ui/card";
+import type { MountedAIRunDetailPayload } from "@/api/mountedAi";
+import { formatBeijingDateTime } from "@/lib/datetime";
+
+interface MountedAIRunReplayPanelProps {
+  detail?: MountedAIRunDetailPayload;
+  loading: boolean;
+}
+
+function prettyJson(value: unknown): string {
+  if (value == null) {
+    return "暂无";
+  }
+  try {
+    return JSON.stringify(value, null, 2);
+  } catch {
+    return String(value);
+  }
+}
+
+export function MountedAIRunReplayPanel({ detail, loading }: MountedAIRunReplayPanelProps) {
+  return (
+    <Card className="border-slate-200/80 bg-white/90 shadow-sm dark:border-slate-800 dark:bg-slate-900/85">
+      <CardHeader>
+        <CardTitle className="text-base">Run Replay</CardTitle>
+        <CardDescription>查看单次 run 的模式切换、agent trace、context hit 与执行结果。</CardDescription>
+      </CardHeader>
+      <CardContent className="space-y-4">
+        {loading ? (
+          <div className="rounded-xl border border-dashed border-slate-300 px-4 py-8 text-center text-sm text-slate-500 dark:border-slate-700 dark:text-slate-400">
+            正在加载 replay...
+          </div>
+        ) : null}
+
+        {!loading && !detail ? (
+          <div className="rounded-xl border border-dashed border-slate-300 px-4 py-8 text-center text-sm text-slate-500 dark:border-slate-700 dark:text-slate-400">
+            选择一条 run 后，这里会展示 replay 细节。
+          </div>
+        ) : null}
+
+        {!loading && detail ? (
+          <>
+            <div className="grid gap-3 md:grid-cols-2 xl:grid-cols-4">
+              <div className="rounded-2xl border border-slate-200/80 p-4 dark:border-slate-700">
+                <div className="text-xs text-slate-500 dark:text-slate-400">Trace</div>
+                <div className="mt-2 break-all text-sm font-medium">{detail.item.trace_id}</div>
+              </div>
+              <div className="rounded-2xl border border-slate-200/80 p-4 dark:border-slate-700">
+                <div className="text-xs text-slate-500 dark:text-slate-400">模式切换</div>
+                <div className="mt-2 text-sm font-medium">{`${detail.item.mode_before || "-"} -> ${detail.item.mode_after || "-"}`}</div>
+              </div>
+              <div className="rounded-2xl border border-slate-200/80 p-4 dark:border-slate-700">
+                <div className="text-xs text-slate-500 dark:text-slate-400">Active Agent</div>
+                <div className="mt-2 text-sm font-medium">{detail.item.active_agent || "-"}</div>
+              </div>
+              <div className="rounded-2xl border border-slate-200/80 p-4 dark:border-slate-700">
+                <div className="text-xs text-slate-500 dark:text-slate-400">更新时间</div>
+                <div className="mt-2 text-sm font-medium">{formatBeijingDateTime(detail.item.updated_at)}</div>
+              </div>
+            </div>
+
+            <div className="rounded-2xl border border-slate-200/80 p-4 dark:border-slate-700">
+              <div className="mb-3 flex items-center justify-between gap-2">
+                <div>
+                  <div className="text-sm font-medium">Agent Trace</div>
+                  <div className="text-xs text-slate-500 dark:text-slate-400">按实际落库顺序展示本次编排轨迹。</div>
+                </div>
+                <Badge variant="secondary">{detail.agent_traces.length} 条</Badge>
+              </div>
+              <div className="space-y-3">
+                {detail.agent_traces.map((trace) => (
+                  <div
+                    key={trace.id}
+                    className="rounded-xl border border-slate-200/80 bg-slate-50/80 p-3 dark:border-slate-700 dark:bg-slate-950/40"
+                  >
+                    <div className="flex flex-wrap items-center gap-2">
+                      <span className="text-sm font-medium">{trace.agent_name}</span>
+                      <Badge variant={trace.status === "failed" ? "destructive" : "secondary"}>{trace.status}</Badge>
+                      <span className="text-xs text-slate-500 dark:text-slate-400">
+                        {trace.latency_ms != null ? `${trace.latency_ms} ms` : "无耗时"}
+                      </span>
+                      <span className="text-xs text-slate-500 dark:text-slate-400">
+                        {formatBeijingDateTime(trace.created_at)}
+                      </span>
+                    </div>
+                    {trace.failure_reason ? (
+                      <div className="mt-2 text-sm text-rose-600 dark:text-rose-400">{trace.failure_reason}</div>
+                    ) : null}
+                  </div>
+                ))}
+              </div>
+            </div>
+
+            <div className="grid gap-4 xl:grid-cols-2">
+              <div className="rounded-2xl border border-slate-200/80 p-4 dark:border-slate-700">
+                <div className="mb-2 text-sm font-medium">Context Layers Hit</div>
+                <pre className="overflow-x-auto rounded-xl bg-slate-950 p-3 text-xs leading-6 text-slate-100">
+                  {prettyJson(detail.context_layers_hit)}
+                </pre>
+              </div>
+              <div className="rounded-2xl border border-slate-200/80 p-4 dark:border-slate-700">
+                <div className="mb-2 text-sm font-medium">Execution Outcome</div>
+                <pre className="overflow-x-auto rounded-xl bg-slate-950 p-3 text-xs leading-6 text-slate-100">
+                  {prettyJson(detail.execution_outcome)}
+                </pre>
+              </div>
+            </div>
+          </>
+        ) : null}
+      </CardContent>
+    </Card>
+  );
+}

frontend/src/pages/scheduler-center/components/MountedAIRunsPanel.tsx

@@ -0,0 +1,77 @@
+import { Card, CardContent, CardDescription, CardHeader, CardTitle } from "@/components/ui/card";
+import { Button } from "@/components/ui/button";
+import { Table, TableBody, TableCell, TableHead, TableHeader, TableRow } from "@/components/ui/table";
+import type { MountedAIRunItem } from "@/api/mountedAi";
+import { formatBeijingDateTime } from "@/lib/datetime";
+
+interface MountedAIRunsPanelProps {
+  items: MountedAIRunItem[];
+  loading: boolean;
+  selectedRunId?: number | null;
+  onSelectRun: (runId: number) => void;
+}
+
+export function MountedAIRunsPanel({ items, loading, selectedRunId, onSelectRun }: MountedAIRunsPanelProps) {
+  return (
+    <Card className="border-slate-200/80 bg-white/90 shadow-sm dark:border-slate-800 dark:bg-slate-900/85">
+      <CardHeader>
+        <CardTitle className="text-base">Conversation Runs</CardTitle>
+        <CardDescription>最近的 mounted runs、活跃 agent、模式切换与回复摘要。</CardDescription>
+      </CardHeader>
+      <CardContent>
+        <div className="overflow-x-auto rounded-xl border border-slate-200/80 dark:border-slate-700">
+          <Table>
+            <TableHeader>
+              <TableRow>
+                <TableHead>Trace</TableHead>
+                <TableHead>模式</TableHead>
+                <TableHead>Agent</TableHead>
+                <TableHead>状态</TableHead>
+                <TableHead>Trace 数</TableHead>
+                <TableHead>最后回复</TableHead>
+                <TableHead>更新时间</TableHead>
+                <TableHead className="w-[96px] text-right">操作</TableHead>
+              </TableRow>
+            </TableHeader>
+            <TableBody>
+              {items.map((item) => (
+                <TableRow key={item.id} data-state={selectedRunId === item.id ? "selected" : undefined}>
+                  <TableCell className="max-w-[180px] truncate font-medium" title={item.trace_id}>
+                    {item.trace_id}
+                  </TableCell>
+                  <TableCell>{`${item.mode_before || "-"} -> ${item.mode_after || "-"}`}</TableCell>
+                  <TableCell>{item.active_agent || "-"}</TableCell>
+                  <TableCell>{item.status}</TableCell>
+                  <TableCell>{item.trace_count}</TableCell>
+                  <TableCell className="max-w-[280px] truncate" title={item.final_reply || ""}>
+                    {item.final_reply || "-"}
+                  </TableCell>
+                  <TableCell>{formatBeijingDateTime(item.updated_at)}</TableCell>
+                  <TableCell className="text-right">
+                    <Button variant="outline" size="sm" onClick={() => onSelectRun(item.id)}>
+                      查看
+                    </Button>
+                  </TableCell>
+                </TableRow>
+              ))}
+              {!loading && items.length === 0 ? (
+                <TableRow>
+                  <TableCell colSpan={8} className="text-center text-slate-500 dark:text-slate-400">
+                    暂无 run 数据
+                  </TableCell>
+                </TableRow>
+              ) : null}
+              {loading ? (
+                <TableRow>
+                  <TableCell colSpan={8} className="text-center text-slate-500 dark:text-slate-400">
+                    读取中...
+                  </TableCell>
+                </TableRow>
+              ) : null}
+            </TableBody>
+          </Table>
+        </div>
+      </CardContent>
+    </Card>
+  );
+}

frontend/src/pages/scheduler-center/components/MountedAISourceStatusPanel.tsx

@@ -0,0 +1,60 @@
+import { Card, CardContent, CardDescription, CardHeader, CardTitle } from "@/components/ui/card";
+import { Table, TableBody, TableCell, TableHead, TableHeader, TableRow } from "@/components/ui/table";
+import type { MountedAISourceItem } from "@/api/mountedAi";
+import { formatBeijingDateTime } from "@/lib/datetime";
+
+interface MountedAISourceStatusPanelProps {
+  items: MountedAISourceItem[];
+  loading: boolean;
+}
+
+export function MountedAISourceStatusPanel({ items, loading }: MountedAISourceStatusPanelProps) {
+  return (
+    <Card className="border-slate-200/80 bg-white/90 shadow-sm dark:border-slate-800 dark:bg-slate-900/85">
+      <CardHeader>
+        <CardTitle className="text-base">来源状态</CardTitle>
+        <CardDescription>当前 Mounted Runtime 已观测到的来源、topic 数和 run 数。</CardDescription>
+      </CardHeader>
+      <CardContent>
+        <div className="overflow-x-auto rounded-xl border border-slate-200/80 dark:border-slate-700">
+          <Table>
+            <TableHeader>
+              <TableRow>
+                <TableHead>来源</TableHead>
+                <TableHead>Run</TableHead>
+                <TableHead>Topic</TableHead>
+                <TableHead>活跃 Topic</TableHead>
+                <TableHead>最后观测</TableHead>
+              </TableRow>
+            </TableHeader>
+            <TableBody>
+              {items.map((item) => (
+                <TableRow key={item.source_system}>
+                  <TableCell className="font-medium">{item.source_system}</TableCell>
+                  <TableCell>{item.run_count}</TableCell>
+                  <TableCell>{item.topic_count}</TableCell>
+                  <TableCell>{item.active_topic_count}</TableCell>
+                  <TableCell>{formatBeijingDateTime(item.last_seen_at)}</TableCell>
+                </TableRow>
+              ))}
+              {!loading && items.length === 0 ? (
+                <TableRow>
+                  <TableCell colSpan={5} className="text-center text-slate-500 dark:text-slate-400">
+                    暂无来源观测数据
+                  </TableCell>
+                </TableRow>
+              ) : null}
+              {loading ? (
+                <TableRow>
+                  <TableCell colSpan={5} className="text-center text-slate-500 dark:text-slate-400">
+                    读取中...
+                  </TableCell>
+                </TableRow>
+              ) : null}
+            </TableBody>
+          </Table>
+        </div>
+      </CardContent>
+    </Card>
+  );
+}

frontend/src/pages/scheduler-center/components/MountedAITopicsPanel.tsx

@@ -0,0 +1,66 @@
+import { Card, CardContent, CardDescription, CardHeader, CardTitle } from "@/components/ui/card";
+import { Table, TableBody, TableCell, TableHead, TableHeader, TableRow } from "@/components/ui/table";
+import type { MountedAITopicItem } from "@/api/mountedAi";
+import { formatBeijingDateTime } from "@/lib/datetime";
+
+interface MountedAITopicsPanelProps {
+  items: MountedAITopicItem[];
+  loading: boolean;
+}
+
+export function MountedAITopicsPanel({ items, loading }: MountedAITopicsPanelProps) {
+  return (
+    <Card className="border-slate-200/80 bg-white/90 shadow-sm dark:border-slate-800 dark:bg-slate-900/85">
+      <CardHeader>
+        <CardTitle className="text-base">Topic Workspace</CardTitle>
+        <CardDescription>当前活跃主题、模式、readiness 和草稿状态。</CardDescription>
+      </CardHeader>
+      <CardContent>
+        <div className="overflow-x-auto rounded-xl border border-slate-200/80 dark:border-slate-700">
+          <Table>
+            <TableHeader>
+              <TableRow>
+                <TableHead>Topic</TableHead>
+                <TableHead>模式</TableHead>
+                <TableHead>状态</TableHead>
+                <TableHead>Readiness</TableHead>
+                <TableHead>草稿</TableHead>
+                <TableHead>更新时间</TableHead>
+              </TableRow>
+            </TableHeader>
+            <TableBody>
+              {items.map((item) => (
+                <TableRow key={item.id}>
+                  <TableCell className="max-w-[260px] truncate font-medium" title={item.topic_key}>
+                    {item.topic_key}
+                  </TableCell>
+                  <TableCell>{item.mode}</TableCell>
+                  <TableCell>{item.status}</TableCell>
+                  <TableCell>{item.readiness_score != null ? item.readiness_score.toFixed(2) : "-"}</TableCell>
+                  <TableCell>
+                    {item.draft_state ? `${item.draft_state}${item.draft_version != null ? ` v${item.draft_version}` : ""}` : "-"}
+                  </TableCell>
+                  <TableCell>{formatBeijingDateTime(item.updated_at)}</TableCell>
+                </TableRow>
+              ))}
+              {!loading && items.length === 0 ? (
+                <TableRow>
+                  <TableCell colSpan={6} className="text-center text-slate-500 dark:text-slate-400">
+                    暂无 topic 数据
+                  </TableCell>
+                </TableRow>
+              ) : null}
+              {loading ? (
+                <TableRow>
+                  <TableCell colSpan={6} className="text-center text-slate-500 dark:text-slate-400">
+                    读取中...
+                  </TableCell>
+                </TableRow>
+              ) : null}
+            </TableBody>
+          </Table>
+        </div>
+      </CardContent>
+    </Card>
+  );
+}