perf(storage): optimize database hot paths

DESIGN.md

@@ -217,6 +217,7 @@ tasks(task_id pk, user_id fk, name NOT NULL, auto_title_pending default false,
       channel default 'web',    -- 0013 渠道来源,仅 INSERT 写定
       run_status default 'idle', run_error,   -- 0004 合并 runs 表
       run_owner,                -- 0020 蓝绿实例归属,reaper 只收自己色;单实例 NULL
+      next_message_idx,         -- 0029 task 行锁下原子分配 messages.idx
       scheduled_job_id,         -- 0017 定时任务归属,普通列表过滤
       context_base_idx,         -- 0019 §8.8 软重置窗口起点
       deleted_at,               -- 0010 软删
@@ -232,6 +233,7 @@ scheduled_jobs(§8.5)   channel_bindings(§8.7,判别列+JSONB)
 - working_dir 存相对 ROOT posix 串,读写统一过 `core/paths.py`;入口 `validate_task_name` 拒空/`/\NUL`/`.` 起头。
 - `auto_title_pending`(0023)只是一轮 UI 命名闸，不是 task 状态机；旧创建入口/存量行恒 false，快速入口首发后消费，人工改名优先清闸。
 - **0004 简化**:runs 表只写不读、run_id 单活 run 下全冗余 → 合并 `run_status/run_error` 入 tasks。**0006**:`tasks.model_profile` 为 source-of-truth(PATCH 切、下条 send 生效);usage_events 重建 v2 多态形态,统计 source-of-truth;tasks 三列保留作粗概览。run_status 终态:ok 收回 idle,error(出错)与 cancelled(用户停止)是持久终态 —— 前端 `renderPersistedRunTerminal` 据此在每次重渲后补持久卡(扛过收尾 loadMessages 整屏重建),刷新/切任务仍在;下次起新 run(post_message 写 running)覆盖清掉。
+- **0029 消息序号**:`tasks.next_message_idx` 在 task 行锁下统一分配 `messages.idx`，Web、agent 与渠道追加不再各自维护序号或依赖冲突重试；分配时仍与 `max(idx)` 校准，允许蓝绿发布窗口内旧实例继续写入。清空消息与计数器在同一事务归零。
 - **No-subtask**:同 user 下前缀互含即拒(归一 posix 后 Python 端比对);同 working_dir 允许。
 - **文件面板先备料**:user_root 与普通目录都可显式新建直接子目录；创建成功后前端进入该目录，用户可先上传/选入资料，再把顶层空目录选作新对话 working_dir。目录 leaf 复用 `validate_task_name`，不允许借 UI 创建点目录或路径式名称。
 - **DB-aware service 是顶层 working_dir mutation 的唯一原语,DB-FS 一致性服务端内化**:文件面板 rename 与对话 `rename_working_dir` 共用同一服务(事务锁关联 task、running→409、DB UPDATE 先于 FS)；对话工具只登记本轮内存动作，等 agent 正常回复完、当前 task 退出 running 后执行，避免 executor/system prompt 仍握旧 cwd。服务在收尾前退出时动作丢失但目录不变，不引持久队列或 migration。delete 仍仅走 files API，被 task 引用时 409。

PROGRESS.md

@@ -2,7 +2,7 @@
 
 > 配合 `DESIGN.md`。本文件只记 phase 状态、决策偏差、文件量、下一步。每条 1-2 句:做了啥 + 关键判断;细节查 `git log` / `git diff` / `DESIGN §7.9`。
 
-最后更新:2026-08-12(PDF/PPT 连续预览滚轮与触摸修复,bump 0.64.1)
+最后更新:2026-08-12(数据库热路径与消息序号优化,bump 0.64.2)
 
 ---
 
@@ -23,6 +23,8 @@
 
 ### 2026-08-12
 
+- **08-12 / 0.64.2 / 数据库热路径 + 消息序号原子化**:新增 0029 migration 平滑回填 `tasks.next_message_idx`，Web、agent 与渠道消息统一在 task 行锁下分配序号，并以实际 `max(idx)` 兼容蓝绿旧实例并行；任务列表、调度扫描/历史和 usage 时间序列补部分/BRIN 索引，列表追加 UUID 稳定排序，artifact 目录操作只查询并锁定目标子树。完整 532 项 unittest 全绿（17 skip），0029 PostgreSQL DDL 编译、Alembic 单 head、Ruff 致命规则及 diff 检查通过；未配置显式测试库，DB 集成测试安全跳过，未连接或写入生产 DB。
+
 - **08-12 / 0.64.1 / PDF/PPT 连续预览滚轮与触摸修复**:外层预览的防滚动穿透逻辑不再拦截来自 PDF 内层 viewport 的普通 wheel 事件，保留图片 Ctrl+滚轮缩放和其他预览的边界隔离；PDF viewport 同步声明触摸平移与 WebView/iOS 惯性滚动，使鼠标滚轮、触控板和手机上下滑动都直接驱动连续页列表。Node 前端预览 11 项、JavaScript 语法及 diff 检查通过；无 schema、migration、HTTP API、依赖或运行方式变化，未连接生产 DB。
 
 - **08-12 / 0.64.0 / artifact 稳定身份 + 隐藏回收**:新增 0028 `artifacts` 生命周期表并回填存量结构化引用；新发布消息写带 `artifact_id` 的 v2 引用，移动/重命名保持身份，复制创建独立身份并记录来源，历史卡片按身份解析最新路径。删除已发布产物时移动到用户隐藏目录 `.zcbot_artifact_trash/` 并软删记录，普通文件仍物理删除；Python 35 项（测试库门控 1 skip）、Node 前端 11 项、mypy、Alembic 单 head、编译、Ruff 致命规则及 diff 检查通过，未连接或写入生产 DB。

RUN.md

@@ -750,6 +750,8 @@ systemctl restart zcbot
 .venv/bin/python main.py db upgrade head
 ```
 
+0.64.2 部署必须先升级到 migration `0029`；它会回填消息序号计数器并创建任务、调度和用量热路径索引，不删除或重置存量数据。
+
 启用 docker backend 重启 web(确保 `.env` 有 `ZCBOT_SANDBOX_BACKEND=docker` 或 systemd
 unit 已设):
 

core/__init__.py

@@ -1,3 +1,3 @@
 # zcbot 版本号单一事实源:web/app.py 的 FastAPI version、/healthz 返回、前端展示都引这里。
 # 改版本只动这一行。
-__version__ = "0.64.1"
+__version__ = "0.64.2"

core/artifact_lifecycle.py

@@ -9,7 +9,7 @@ from datetime import datetime, timezone
 from pathlib import Path
 from uuid import UUID, uuid4
 
-from sqlalchemy import select
+from sqlalchemy import or_, select
 from sqlalchemy.dialects.postgresql import insert as pg_insert
 
 from .artifacts import ARTIFACT_TRASH_DIR, ArtifactRef
@@ -29,6 +29,15 @@ def _hash_file(path: Path) -> str:
     return digest.hexdigest()
 
 
+def artifact_subtree_clause(path: str):
+    """精确匹配文件/目录自身及其子树，并转义 LIKE 元字符。"""
+    escaped = path.replace("\\", "\\\\").replace("%", "\\%").replace("_", "\\_")
+    return or_(
+        Artifact.current_path == path,
+        Artifact.current_path.like(escaped + "/%", escape="\\"),
+    )
+
+
 def register_published_artifacts(
     *,
     user_id: UUID,
@@ -98,6 +107,7 @@ def rename_active_artifacts(
             select(Artifact).where(
                 Artifact.user_id == user_id,
                 Artifact.status == "active",
+                artifact_subtree_clause(old_rel),
             )
         ).scalars().all()
         changed = 0
@@ -129,6 +139,7 @@ def copy_active_artifacts(
             select(Artifact).where(
                 Artifact.user_id == user_id,
                 Artifact.status == "active",
+                artifact_subtree_clause(source_rel),
             )
         ).scalars().all()
         created = 0
@@ -176,6 +187,7 @@ def trash_active_artifacts(
                 select(Artifact).where(
                     Artifact.user_id == user_id,
                     Artifact.status == "active",
+                    artifact_subtree_clause(source_rel),
                 ).with_for_update()
             ).scalars().all()
             matches = [

core/session.py

@@ -10,16 +10,15 @@ system prompt **不入库** —— 每次 build_agent 重建拼到 messages[0](
 """
 from __future__ import annotations
 
-from pathlib import Path
 from typing import Any, Dict, List, Optional
 from uuid import UUID
 
-from sqlalchemy import delete, func, select
+from sqlalchemy import delete, func, select, update
 
-from .storage import session_scope
-from .storage.models import Message, Task, sanitize_jsonb_nul
-from .file_store import atomic_write_text
 from .markdown_guard import normalize_markdown_fences
+from .storage import session_scope
+from .storage.message_index import allocate_message_idx
+from .storage.models import Message, Task, sanitize_jsonb_nul
 
 
 def _to_dict(msg: Any) -> Any:
@@ -97,16 +96,17 @@ class Session:
             return None
 
         with session_scope() as s:
+            idx = allocate_message_idx(s, self.task_id)
             row = Message(
                 task_id=self.task_id,
-                idx=self._db_idx,
+                idx=idx,
                 payload=msg_dict,
                 artifact_refs=artifact_refs,
             )
             s.add(row)
             s.flush()  # 触发 INSERT 拿到 server-default 生成的 message_id
             msg_id = row.message_id
-        self._db_idx += 1
+        self._db_idx = idx + 1
         return msg_id
 
     def last_measured_usage(self) -> Optional[tuple]:
@@ -182,6 +182,9 @@ class Session:
             self._n_head = 0
         with session_scope() as s:
             s.execute(delete(Message).where(Message.task_id == self.task_id))
+            s.execute(
+                update(Task).where(Task.task_id == self.task_id).values(next_message_idx=0)
+            )
         self._db_idx = 0
         self._base_idx = 0
 

core/storage/message_index.py

@@ -0,0 +1,35 @@
+"""同一 task 内消息序号的统一事务分配入口。"""
+from __future__ import annotations
+
+from typing import Optional
+from uuid import UUID
+
+from sqlalchemy import func, select
+from sqlalchemy.orm import Session as OrmSession
+
+from .models import Message, Task
+
+
+def allocate_message_idx(
+    session: OrmSession,
+    task_id: UUID,
+    *,
+    locked_task: Optional[Task] = None,
+) -> int:
+    """锁定 task 行并返回唯一递增 idx；max 校准兼容蓝绿旧实例。"""
+    task = locked_task
+    if task is None:
+        task = session.execute(
+            select(Task).where(Task.task_id == task_id).with_for_update()
+        ).scalar_one()
+    elif task.task_id != task_id:
+        raise ValueError("locked_task 与 task_id 不一致")
+
+    max_idx = session.execute(
+        select(func.max(Message.idx)).where(Message.task_id == task_id)
+    ).scalar_one()
+    persisted_next = int(getattr(task, "next_message_idx", 0) or 0)
+    observed_next = -1 if max_idx is None else int(max_idx)
+    next_idx = max(persisted_next, observed_next + 1)
+    task.next_message_idx = next_idx + 1
+    return next_idx

core/storage/models.py

@@ -33,7 +33,8 @@ from sqlalchemy import (
     func,
     text,
 )
-from sqlalchemy.dialects.postgresql import JSONB, UUID as PG_UUID
+from sqlalchemy.dialects.postgresql import JSONB
+from sqlalchemy.dialects.postgresql import UUID as PG_UUID
 from sqlalchemy.orm import DeclarativeBase, Mapped, mapped_column, validates
 
 
@@ -85,6 +86,16 @@ class User(Base):
 
 class Task(Base):
     __tablename__ = "tasks"
+    __table_args__ = (
+        Index(
+            "ix_tasks_user_active_updated", "user_id", "updated_at", "task_id",
+            postgresql_where=text("deleted_at IS NULL AND scheduled_job_id IS NULL"),
+        ),
+        Index(
+            "ix_tasks_schedule_created", "scheduled_job_id", "created_at", "task_id",
+            postgresql_where=text("deleted_at IS NULL AND scheduled_job_id IS NOT NULL"),
+        ),
+    )
 
     task_id: Mapped[UUID] = mapped_column(PG_UUID(as_uuid=True), primary_key=True, default=uuid4)
     user_id: Mapped[UUID] = mapped_column(
@@ -130,6 +141,10 @@ class Task(Base):
     # (blue/green),启动 reaper 只收自己实例的孤儿。单实例部署恒 NULL,行为不变。
     # 只在 running/cancelling 期间有意义,run 收尾不清(reaper 只看活跃行)。
     run_owner: Mapped[Optional[str]] = mapped_column(Text, nullable=True)
+    # 0029:同一 task 的下一条消息序号；所有消息入口以 task 行锁串行分配。
+    next_message_idx: Mapped[int] = mapped_column(
+        Integer, nullable=False, default=0, server_default="0"
+    )
     # 喂给模型的上下文窗口起点(0019,channel 长会话软重置)。Session.load 只把 idx >=
     # context_base_idx 的消息装进 LLM 上下文;之前的历史仍全量留 messages 表(web 翻得到)。
     # web 普通任务恒 0 = 喂全量;channel 入站按 gap / 「新话题」推进。详 DESIGN §8.7。
@@ -255,6 +270,9 @@ class UsageEvent(Base):
     按 user 聚合的统计 query 走 (user_id, created_at) 索引,不 JOIN tasks 表。
     """
     __tablename__ = "usage_events"
+    __table_args__ = (
+        Index("ix_usage_created_brin", "created_at", postgresql_using="brin"),
+    )
 
     event_id: Mapped[UUID] = mapped_column(PG_UUID(as_uuid=True), primary_key=True, default=uuid4)
     user_id: Mapped[UUID] = mapped_column(
@@ -321,6 +339,16 @@ class ScheduledJob(Base):
     """
 
     __tablename__ = "scheduled_jobs"
+    __table_args__ = (
+        Index(
+            "ix_scheduled_jobs_due_active", "next_run_at",
+            postgresql_where=text("enabled IS TRUE AND deleted_at IS NULL"),
+        ),
+        Index(
+            "ix_scheduled_jobs_user_active_created", "user_id", "created_at",
+            postgresql_where=text("deleted_at IS NULL"),
+        ),
+    )
 
     job_id: Mapped[UUID] = mapped_column(PG_UUID(as_uuid=True), primary_key=True, default=uuid4)
     user_id: Mapped[UUID] = mapped_column(

core/storage/utils.py

@@ -6,9 +6,9 @@ from uuid import UUID
 
 from sqlalchemy import func, select, update
 from sqlalchemy.dialects.postgresql import insert
-from sqlalchemy.exc import IntegrityError
 
 from .engine import session_scope
+from .message_index import allocate_message_idx
 from .models import Message, Task
 
 
@@ -64,26 +64,15 @@ def ensure_local_task_row(
 def append_channel_message(
     task_id: UUID, content: str, *, role: str = "assistant", kind: Optional[str] = None
 ) -> None:
-    """往 task 追加一条非 agent-run 产生的消息(push 出站记录等)。原子算 idx
+    """往 task 追加一条非 agent-run 产生的消息(push 出站记录等)。
-    (SELECT max(idx)+1)+INSERT;撞 uq_messages_task_idx(与入站 agent run 并发
+
-    append)→ 重试。payload 形态同 Session.append 的 {role, content};不设
+    与 agent/Web 入口共用 task 行锁下的序号分配。payload 形态同 Session.append；不设
     model_profile / tokens_*(非模型产出,usage 不计)。kind 写 messages.kind 列
     (独立列,不进 payload):"push" 标记 push 记录,extract_last_assistant_text 据此跳过。"""
     payload = {"role": role, "content": content}
-    last_err: Optional[Exception] = None
+    with session_scope() as s:
-    for _ in range(3):
+        next_idx = allocate_message_idx(s, task_id)
-        try:
+        s.add(Message(task_id=task_id, idx=next_idx, payload=payload, kind=kind))
-            with session_scope() as s:
-                max_idx = s.execute(
-                    select(func.max(Message.idx)).where(Message.task_id == task_id)
-                ).scalar()
-                next_idx = (max_idx if max_idx is not None else -1) + 1
-                s.add(Message(task_id=task_id, idx=next_idx, payload=payload, kind=kind))
-            return
-        except IntegrityError as e:
-            last_err = e
-            continue
-    raise RuntimeError(f"append_channel_message: idx 冲突重试耗尽: {last_err}")
 
 
 def upsert_task(

core/working_dirs.py

@@ -11,6 +11,7 @@ from uuid import UUID
 
 from sqlalchemy import select, update
 
+from .artifact_lifecycle import artifact_subtree_clause
 from .paths import to_db_path
 from .storage import NoSubtaskError, check_no_subtask, session_scope
 from .storage.models import Artifact, Task
@@ -100,6 +101,7 @@ def rename_working_dir(
             select(Artifact).where(
                 Artifact.user_id == user_id,
                 Artifact.status == "active",
+                artifact_subtree_clause(old_rel),
             )
         ).scalars().all()
         for artifact in artifacts:

db/migrations/versions/20260812_1800_0029_storage_hot_paths.py

@@ -0,0 +1,79 @@
+"""Add atomic message indexes and storage hot-path indexes.
+
+Revision ID: 0029
+Revises: 0028
+Create Date: 2026-08-12
+"""
+from typing import Sequence, Union
+
+import sqlalchemy as sa
+from alembic import op
+
+
+revision: str = "0029"
+down_revision: Union[str, None] = "0028"
+branch_labels: Union[str, Sequence[str], None] = None
+depends_on: Union[str, Sequence[str], None] = None
+
+
+def upgrade() -> None:
+    op.add_column(
+        "tasks",
+        sa.Column("next_message_idx", sa.Integer(), server_default="0", nullable=False),
+    )
+    op.execute(
+        """
+        UPDATE tasks AS t
+        SET next_message_idx = COALESCE(m.next_idx, 0)
+        FROM (
+            SELECT task_id, MAX(idx) + 1 AS next_idx
+            FROM messages
+            GROUP BY task_id
+        ) AS m
+        WHERE m.task_id = t.task_id
+        """
+    )
+
+    op.create_index(
+        "ix_tasks_user_active_updated",
+        "tasks",
+        ["user_id", "updated_at", "task_id"],
+        postgresql_where=sa.text("deleted_at IS NULL AND scheduled_job_id IS NULL"),
+    )
+    op.create_index(
+        "ix_tasks_schedule_created",
+        "tasks",
+        ["scheduled_job_id", "created_at", "task_id"],
+        postgresql_where=sa.text(
+            "deleted_at IS NULL AND scheduled_job_id IS NOT NULL"
+        ),
+    )
+    op.create_index(
+        "ix_scheduled_jobs_due_active",
+        "scheduled_jobs",
+        ["next_run_at"],
+        postgresql_where=sa.text("enabled IS TRUE AND deleted_at IS NULL"),
+    )
+    op.create_index(
+        "ix_scheduled_jobs_user_active_created",
+        "scheduled_jobs",
+        ["user_id", "created_at"],
+        postgresql_where=sa.text("deleted_at IS NULL"),
+    )
+    op.create_index(
+        "ix_usage_created_brin",
+        "usage_events",
+        ["created_at"],
+        postgresql_using="brin",
+    )
+
+
+def downgrade() -> None:
+    op.drop_index("ix_usage_created_brin", table_name="usage_events")
+    op.drop_index(
+        "ix_scheduled_jobs_user_active_created", table_name="scheduled_jobs"
+    )
+    op.drop_index("ix_scheduled_jobs_due_active", table_name="scheduled_jobs")
+    op.drop_index("ix_tasks_schedule_created", table_name="tasks")
+    op.drop_index("ix_tasks_user_active_updated", table_name="tasks")
+    op.drop_column("tasks", "next_message_idx")

tests/test_artifacts.py

@@ -6,15 +6,15 @@ from types import SimpleNamespace
 from unittest.mock import MagicMock, patch
 from uuid import uuid4
 
+from core.artifact_lifecycle import artifact_subtree_clause, trash_active_artifacts
 from core.artifacts import (
-    ArtifactRef,
     ArtifactPathError,
+    ArtifactRef,
     ToolExecutionResult,
     resolve_artifact_path,
 )
 from core.executor import ExecCtx
 from core.executor_host import HostExecutor
-from core.artifact_lifecycle import trash_active_artifacts
 from tools.publish_artifacts import PublishArtifactsTool
 from web.routers.files import _task_file_target
 
@@ -86,6 +86,11 @@ class ArtifactPathTests(unittest.TestCase):
         self.assertTrue(trashed.is_file())
         self.assertEqual(trashed.read_bytes(), b"pdf")
 
+    def test_subtree_query_escapes_like_wildcards(self) -> None:
+        clause = artifact_subtree_clause("项目_100%/报告")
+        compiled = clause.compile().params
+        self.assertIn("项目\\_100\\%/报告/%", compiled.values())
+
     def test_explicit_dot_slash_disambiguates_same_named_subdirectory(self) -> None:
         nested = self.wd / "技术讨论" / "nested.html"
         nested.parent.mkdir()

tests/test_message_index.py

@@ -0,0 +1,39 @@
+import unittest
+from types import SimpleNamespace
+from unittest.mock import MagicMock
+from uuid import uuid4
+
+from core.storage.message_index import allocate_message_idx
+
+
+class MessageIndexTests(unittest.TestCase):
+    def test_allocator_uses_persisted_counter_when_ahead(self) -> None:
+        tid = uuid4()
+        task = SimpleNamespace(task_id=tid, next_message_idx=8)
+        session = MagicMock()
+        session.execute.return_value.scalar_one.return_value = 5
+
+        self.assertEqual(allocate_message_idx(session, tid, locked_task=task), 8)
+        self.assertEqual(task.next_message_idx, 9)
+
+    def test_allocator_catches_up_with_legacy_writer(self) -> None:
+        tid = uuid4()
+        task = SimpleNamespace(task_id=tid, next_message_idx=3)
+        session = MagicMock()
+        session.execute.return_value.scalar_one.return_value = 6
+
+        self.assertEqual(allocate_message_idx(session, tid, locked_task=task), 7)
+        self.assertEqual(task.next_message_idx, 8)
+
+    def test_empty_task_starts_at_zero(self) -> None:
+        tid = uuid4()
+        task = SimpleNamespace(task_id=tid, next_message_idx=0)
+        session = MagicMock()
+        session.execute.return_value.scalar_one.return_value = None
+
+        self.assertEqual(allocate_message_idx(session, tid, locked_task=task), 0)
+        self.assertEqual(task.next_message_idx, 1)
+
+
+if __name__ == "__main__":
+    unittest.main()

tests/test_message_payload.py

@@ -56,7 +56,10 @@ class MessagePayloadSanitizationTests(unittest.TestCase):
             yield fake_db
 
         session = Session(task_id=uuid4())
-        with patch("core.session.session_scope", side_effect=fake_session_scope):
+        with (
+            patch("core.session.session_scope", side_effect=fake_session_scope),
+            patch("core.session.allocate_message_idx", return_value=0),
+        ):
             session.append({"role": "tool", "content": "before\x00after"})
 
         expected = {"role": "tool", "content": "beforeafter"}
@@ -81,7 +84,10 @@ class MessagePayloadSanitizationTests(unittest.TestCase):
 
         broken = "```markdown\n```mermaid\nA --> B\n```\n```\n正文\n"
         session = Session(task_id=uuid4())
-        with patch("core.session.session_scope", side_effect=fake_session_scope):
+        with (
+            patch("core.session.session_scope", side_effect=fake_session_scope),
+            patch("core.session.allocate_message_idx", return_value=0),
+        ):
             session.append({"role": "assistant", "content": broken})
 
         stored = session.messages[0]["content"]

tests/test_rename_working_dir.py

@@ -189,6 +189,7 @@ class WorkingDirServiceTests(unittest.TestCase):
         session.execute.side_effect = [
             SimpleNamespace(all=lambda: rows),
             MagicMock(),
+            SimpleNamespace(scalars=lambda: SimpleNamespace(all=lambda: [])),
         ]
 
         @contextmanager
@@ -214,7 +215,7 @@ class WorkingDirServiceTests(unittest.TestCase):
             self.assertEqual(result.tasks_updated, 2)
             self.assertFalse(old.exists())
             self.assertEqual((new / "产物.txt").read_text(encoding="utf-8"), "ok")
-            self.assertEqual(session.execute.call_count, 2)
+            self.assertEqual(session.execute.call_count, 3)
 
     def test_active_associated_task_blocks_before_fs_change(self) -> None:
         uid = uuid4()

tests/test_run_lifecycle.py

@@ -23,7 +23,7 @@ class ClaimRunTests(unittest.TestCase):
         session = MagicMock()
         session.execute.side_effect = [
             SimpleNamespace(scalar_one_or_none=lambda: task),
-            SimpleNamespace(scalar_one=lambda: 7),
+            SimpleNamespace(scalar_one=lambda: 6),
             MagicMock(),
         ]
 

tests/test_storage_migration.py

@@ -0,0 +1,37 @@
+from __future__ import annotations
+
+import importlib
+import unittest
+from unittest.mock import patch
+
+from alembic.migration import MigrationContext
+from alembic.operations import Operations
+from sqlalchemy import create_mock_engine
+from sqlalchemy.dialects import postgresql
+
+
+class StorageMigrationTests(unittest.TestCase):
+    def test_0029_upgrade_compiles_as_postgresql_ddl(self) -> None:
+        statements: list[str] = []
+
+        def capture(sql, *multiparams, **params):
+            statements.append(str(sql.compile(dialect=postgresql.dialect())))
+
+        engine = create_mock_engine("postgresql+psycopg://", capture)
+        operations = Operations(MigrationContext.configure(engine.connect()))
+        migration = importlib.import_module(
+            "db.migrations.versions.20260812_1800_0029_storage_hot_paths"
+        )
+        with patch.object(migration, "op", operations):
+            migration.upgrade()
+
+        rendered = "\n".join(statements)
+        self.assertIn("next_message_idx", rendered)
+        self.assertIn("MAX(idx) + 1", rendered)
+        self.assertIn("ix_tasks_user_active_updated", rendered)
+        self.assertIn("ix_scheduled_jobs_due_active", rendered)
+        self.assertIn("USING brin", rendered)
+
+
+if __name__ == "__main__":
+    unittest.main()

tests/test_web_common.py

@@ -7,17 +7,26 @@ class TaskOrderingTests(unittest.TestCase):
     def test_default_ordering_uses_latest_update_first(self):
         ordering = parse_ordering(None)
 
-        self.assertEqual([str(clause) for clause in ordering], ["tasks.updated_at DESC"])
+        self.assertEqual(
+            [str(clause) for clause in ordering],
+            ["tasks.updated_at DESC", "tasks.task_id DESC"],
+        )
 
     def test_invalid_ordering_falls_back_to_latest_update_first(self):
         ordering = parse_ordering("not_a_field")
 
-        self.assertEqual([str(clause) for clause in ordering], ["tasks.updated_at DESC"])
+        self.assertEqual(
+            [str(clause) for clause in ordering],
+            ["tasks.updated_at DESC", "tasks.task_id DESC"],
+        )
 
     def test_explicit_created_at_ordering_remains_supported(self):
         ordering = parse_ordering("-created_at")
 
-        self.assertEqual([str(clause) for clause in ordering], ["tasks.created_at DESC"])
+        self.assertEqual(
+            [str(clause) for clause in ordering],
+            ["tasks.created_at DESC", "tasks.task_id DESC"],
+        )
 
 
 if __name__ == "__main__":

web/common.py

@@ -76,6 +76,9 @@ def parse_ordering(s: Optional[str]) -> list:
     if not cols:
         # 用户传了全无效字段 → fallback 默认
         cols = [Task.updated_at.desc()]
+    # 所有允许字段都可能重复；UUID 兜底保证 OFFSET 分页在同一快照内稳定。
+    last_desc = str(cols[-1]).upper().endswith(" DESC")
+    cols.append(Task.task_id.desc() if last_desc else Task.task_id.asc())
     return cols
 
 

web/routers/messages.py

@@ -408,6 +408,7 @@ def register_message_routes(app, *, require_user) -> None:
                 "cost_cny": 0,
                 "run_status": "idle",
                 "run_error": None,
+                "next_message_idx": 0,
                 # 全删后 idx 从 0 重起,base 必须归零否则 load 窗口起点悬空(0019);
                 # 摘要一并清,否则清空后的对话还会被注入旧前情摘要(0021)
                 "context_base_idx": 0,

web/run_lifecycle.py

@@ -10,9 +10,10 @@ from dataclasses import dataclass, field
 from typing import Any, Callable, Optional
 from uuid import UUID
 
-from sqlalchemy import func, select, update
+from sqlalchemy import select, update
 
 from core.storage import session_scope
+from core.storage.message_index import allocate_message_idx
 from core.storage.models import Message, Task
 
 from .broker import broker
@@ -76,10 +77,7 @@ def claim_run_with_message(
             extra_values, metadata = prepare(s, task)
 
         # task 行锁串行化同一 task 的 idx 分配，无需额外序列表或 advisory lock。
-        next_idx = s.execute(
+        next_idx = allocate_message_idx(s, task_id, locked_task=task)
-            select(func.coalesce(func.max(Message.idx), -1) + 1)
-            .where(Message.task_id == task_id)
-        ).scalar_one()
         s.add(Message(
             task_id=task_id,
             idx=int(next_idx),