Merge remote-tracking branch 'origin/master' into ji

# Conflicts:
#	frontend-web/src/pages/AgentChatPage.tsx
This commit is contained in:
吉鑫升 committed 2026-05-21 17:44:09 +08:00
commit 643f9b2c76
51 files changed
+2654 -2164

No files matched your search

@@ -8,4 +8,39 @@ def make_ai_writing_session_store(session_factory) -> AIWritingSessionRepository
return AIWritingSessionRepository(session_factory)
__all__ = ["AIWritingSessionRow", "AIWritingSessionRepository", "make_ai_writing_session_store"]
# ── 进程级默认 repo 单例 ──────────────────────────────────────────────────────
#
# AI 写作 graph 注册成 LangGraph Server assistant 后,节点跑在 harness 进程里,
# 拿不到 FastAPI 的 ``request.app.state``;harness → app 又是禁止的 import 方向。
# 这里提供一个 app 启动时注册、harness 节点运行时读取的「跨边界单例」通道。
#
# 写入:app/gateway/deps.py 在创建完 session_factory 之后调
# ``register_default_store(app.state.ai_writing_session_store)``
# 读取:节点里 ``get_default_store()``,None 表示未配置 DB(跳过写入即可)。
_default_store: AIWritingSessionRepository | None = None
def register_default_store(store: AIWritingSessionRepository | None) -> None:
"""登记进程级默认 AI 写作 session store。
幂等:可多次调用,覆盖前值。传 ``None`` 用于显式清理(主要给测试)。
"""
global _default_store
_default_store = store
def get_default_store() -> AIWritingSessionRepository | None:
"""返回进程级默认 AI 写作 session store;未配置时为 ``None``。
节点调用方应在拿到 ``None`` 时跳过 DB 写入(写入是「锦上添花」非致命)。
"""
return _default_store
__all__ = [
"AIWritingSessionRow",
"AIWritingSessionRepository",
"make_ai_writing_session_store",
"register_default_store",
"get_default_store",
]
@@ -4,11 +4,11 @@ from __future__ import annotations
from datetime import UTC, datetime
from sqlalchemy import DateTime, String, Text
from sqlalchemy import String, Text
from sqlalchemy.orm import Mapped, mapped_column
from deerflow.persistence.base import Base
from deerflow.persistence.types import PortableLongText
from deerflow.persistence.types import BeijingDateTime, PortableLongText
class AIWritingSessionRow(Base):
@@ -26,5 +26,13 @@ class AIWritingSessionRow(Base):
# 完整对话时间线(前端组装的 progressEvents + completedInterventions),用于历史回显。
# 用 PortableLongText:MySQL 上映射为 LONGTEXT,避免 TEXT 64KB 上限导致保存 500。
transcript: Mapped[str | None] = mapped_column(PortableLongText, nullable=True) # JSON object
created_at: Mapped[datetime] = mapped_column(DateTime(timezone=True), default=lambda: datetime.now(UTC))
updated_at: Mapped[datetime] = mapped_column(DateTime(timezone=True), default=lambda: datetime.now(UTC))
# 时间列必须用 BeijingDateTime(项目规范,与 threads_meta 等其它表对齐)—— MySQL
# DATETIME 不存时区,raw ``DateTime(timezone=True)`` 会把 UTC wall clock 写入
# 列,读出来比真实北京时间晚 8 小时。``BeijingDateTime`` 在 bind 时 UTC→Beijing、
# 在 result 时 attach +08:00,前后端拿到的就是北京时间且 tz-aware。
created_at: Mapped[datetime] = mapped_column(BeijingDateTime(), default=lambda: datetime.now(UTC))
updated_at: Mapped[datetime] = mapped_column(
BeijingDateTime(),
default=lambda: datetime.now(UTC),
onupdate=lambda: datetime.now(UTC),
)
@@ -63,6 +63,52 @@ class AIWritingSessionRepository:
await session.refresh(row)
return self._row_to_dict(row)
async def upsert_session(
self,
*,
id: str,
title: str,
user_intent: str | None = None,
user_id: str | None | _AutoSentinel = AUTO,
status: str = "in_progress",
) -> dict:
"""幂等:行不存在则创建,行存在则保留原数据(不覆盖 draft / review 等已有写入)。
AI 写作 graph 的 ``intent_parser_node`` 在每次进入时调一次(首启 + resume 时
都会跑),所以必须幂等。resume 时若直接 ``create`` 会因主键冲突抛错;用
``update`` 又找不到行而 no-op。这个方法把两条路合一:行不存在新建,行存在
只刷新 ``updated_at`` 以便列表页排序前置。
"""
resolved_user_id = resolve_user_id(user_id, method_name="AIWritingSessionRepository.upsert_session")
now = datetime.now(UTC)
async with self._sf() as session:
row = await session.get(AIWritingSessionRow, id)
if row is None:
row = AIWritingSessionRow(
id=id,
user_id=resolved_user_id,
title=title,
user_intent=user_intent,
status=status,
draft_title=None,
draft_markdown=None,
completed_interventions=None,
review_result=None,
transcript=None,
created_at=now,
updated_at=now,
)
session.add(row)
else:
# 行已存在(resume 路径):仅刷新 updated_at,保留 user_id / title / draft / review 等。
# 若 user_id 之前为空(migration 兼容场景),用解析出来的填上。
if row.user_id is None and resolved_user_id is not None:
row.user_id = resolved_user_id
row.updated_at = now
await session.commit()
await session.refresh(row)
return self._row_to_dict(row)
async def update(
self,
session_id: str,
@@ -0,0 +1,23 @@
"""Per-user memory configuration override persistence."""
from __future__ import annotations
from deerflow.persistence.memory_config.base import UserMemoryConfigStore
from deerflow.persistence.memory_config.memory import MemoryUserMemoryConfigStore
from deerflow.persistence.memory_config.model import UserMemoryConfigRow
from deerflow.persistence.memory_config.sql import SqlUserMemoryConfigRepository
def make_user_memory_config_store(session_factory) -> UserMemoryConfigStore:
if session_factory is None:
return MemoryUserMemoryConfigStore()
return SqlUserMemoryConfigRepository(session_factory)
__all__ = [
"UserMemoryConfigStore",
"UserMemoryConfigRow",
"MemoryUserMemoryConfigStore",
"SqlUserMemoryConfigRepository",
"make_user_memory_config_store",
]
@@ -0,0 +1,32 @@
"""Abstract persistence API for per-user memory config overrides."""
from __future__ import annotations
from abc import ABC, abstractmethod
class UserMemoryConfigStore(ABC):
@abstractmethod
async def get_recall_budget(self, user_id: str) -> str | None:
"""Return the user's recall_budget override, or None if not set."""
raise NotImplementedError
@abstractmethod
async def set_recall_budget(self, user_id: str, value: str) -> None:
"""Upsert the user's recall_budget override."""
raise NotImplementedError
@abstractmethod
async def delete(self, user_id: str) -> None:
"""Remove all per-user overrides for the given user."""
raise NotImplementedError
@abstractmethod
async def load_all(self) -> dict[str, str]:
"""Bulk-load all user recall_budget overrides.
Returns {user_id: recall_budget} for all rows where recall_budget IS NOT NULL.
Called once at startup to warm the in-memory cache.
"""
raise NotImplementedError
@@ -0,0 +1,23 @@
"""In-memory user memory config store (used when database.backend=memory)."""
from __future__ import annotations
from deerflow.persistence.memory_config.base import UserMemoryConfigStore
class MemoryUserMemoryConfigStore(UserMemoryConfigStore):
def __init__(self) -> None:
self._data: dict[str, str] = {}
async def get_recall_budget(self, user_id: str) -> str | None:
return self._data.get(user_id)
async def set_recall_budget(self, user_id: str, value: str) -> None:
self._data[user_id] = value
async def delete(self, user_id: str) -> None:
self._data.pop(user_id, None)
async def load_all(self) -> dict[str, str]:
return dict(self._data)
@@ -0,0 +1,29 @@
"""ORM model for per-user memory configuration overrides."""
from __future__ import annotations
from datetime import UTC, datetime
from sqlalchemy import String
from sqlalchemy.orm import Mapped, mapped_column
from deerflow.persistence.base import Base
from deerflow.persistence.types import BeijingDateTime
class UserMemoryConfigRow(Base):
__tablename__ = "user_memory_configs"
# One row per user; upserted on write.
user_id: Mapped[str] = mapped_column(String(64), primary_key=True)
# The only user-writable field. NULL means "use global default from config.yaml".
# Allowed values: "low" | "mid" | "high".
recall_budget: Mapped[str | None] = mapped_column(String(16), nullable=True)
updated_at: Mapped[datetime] = mapped_column(
BeijingDateTime(),
nullable=False,
default=lambda: datetime.now(UTC),
onupdate=lambda: datetime.now(UTC),
)
@@ -0,0 +1,50 @@
"""SQLAlchemy-backed user memory config repository."""
from __future__ import annotations
from datetime import UTC, datetime
from sqlalchemy import select
from sqlalchemy.ext.asyncio import AsyncSession, async_sessionmaker
from deerflow.persistence.memory_config.base import UserMemoryConfigStore
from deerflow.persistence.memory_config.model import UserMemoryConfigRow
class SqlUserMemoryConfigRepository(UserMemoryConfigStore):
def __init__(self, session_factory: async_sessionmaker[AsyncSession]) -> None:
self._sf = session_factory
async def get_recall_budget(self, user_id: str) -> str | None:
async with self._sf() as session:
row = await session.get(UserMemoryConfigRow, user_id)
return row.recall_budget if row is not None else None
async def set_recall_budget(self, user_id: str, value: str) -> None:
async with self._sf() as session:
row = await session.get(UserMemoryConfigRow, user_id)
if row is None:
row = UserMemoryConfigRow(
user_id=user_id,
recall_budget=value,
updated_at=datetime.now(UTC),
)
session.add(row)
else:
row.recall_budget = value
row.updated_at = datetime.now(UTC)
await session.commit()
async def delete(self, user_id: str) -> None:
async with self._sf() as session:
row = await session.get(UserMemoryConfigRow, user_id)
if row is not None:
await session.delete(row)
await session.commit()
async def load_all(self) -> dict[str, str]:
stmt = select(UserMemoryConfigRow).where(UserMemoryConfigRow.recall_budget.is_not(None))
async with self._sf() as session:
result = await session.execute(stmt)
return {row.user_id: row.recall_budget for row in result.scalars() if row.recall_budget is not None}
@@ -0,0 +1,44 @@
"""Create user_memory_configs table.
Replaces the per-user file-based memory config override
({base_dir}/users/{user_id}/memory_config.json) with a DB row.
Stores only hindsight.recall_budget — the single user-writable field
after the memory config simplification.
Revision ID: 20260521_01
Revises: 20260519_02
Create Date: 2026-05-21
"""
from __future__ import annotations
import sqlalchemy as sa
from alembic import op
from sqlalchemy import inspect
revision = "20260521_01"
down_revision = "20260519_02"
branch_labels = None
depends_on = None
def _has_table(table_name: str) -> bool:
inspector = inspect(op.get_bind())
return table_name in inspector.get_table_names()
def upgrade() -> None:
if _has_table("user_memory_configs"):
return
op.create_table(
"user_memory_configs",
sa.Column("user_id", sa.String(64), primary_key=True),
sa.Column("recall_budget", sa.String(16), nullable=True),
sa.Column("updated_at", sa.DateTime(timezone=True), nullable=False),
)
def downgrade() -> None:
if _has_table("user_memory_configs"):
op.drop_table("user_memory_configs")
@@ -16,6 +16,7 @@ there is no matching entity directory.
"""
from deerflow.persistence.agents.model import AgentRow
from deerflow.persistence.memory_config.model import UserMemoryConfigRow
from deerflow.persistence.feedback.model import FeedbackRow
from deerflow.persistence.llm_metrics.model import LlmCallMetricRow
from deerflow.persistence.models.run_event import RunEventRow
@@ -55,5 +56,6 @@ __all__ = [
"ThreadMetaRow",
"ThreadShareRow",
"ToolCallMetricRow",
"UserMemoryConfigRow",
"UserRow",
]
@@ -56,6 +56,17 @@ class ThreadMetaStore(abc.ABC):
async def update_status(self, thread_id: str, status: str, *, user_id: str | None | _AutoSentinel = AUTO) -> None:
pass
@abc.abstractmethod
async def update_assistant_id(self, thread_id: str, assistant_id: str, *, user_id: str | None | _AutoSentinel = AUTO) -> None:
"""Backfill ``assistant_id`` for a thread row.
Used by ``start_run`` when the row was pre-created via ``POST /threads``
without an assistant_id (the SDK's ``client.threads.create()`` doesn't
pass it), and the first run on the thread reveals the assistant. No-op
if the thread does not exist or the owner check fails.
"""
pass
@abc.abstractmethod
async def update_metadata(self, thread_id: str, metadata: dict, *, user_id: str | None | _AutoSentinel = AUTO) -> None:
"""Merge ``metadata`` into the thread's metadata field.
@@ -117,6 +117,14 @@ class MemoryThreadMetaStore(ThreadMetaStore):
record["updated_at"] = now_iso()
await self._store.aput(THREADS_NS, thread_id, record)
async def update_assistant_id(self, thread_id: str, assistant_id: str, *, user_id: str | None | _AutoSentinel = AUTO) -> None:
record = await self._get_owned_record(thread_id, user_id, "MemoryThreadMetaStore.update_assistant_id")
if record is None:
return
record["assistant_id"] = assistant_id
record["updated_at"] = now_iso()
await self._store.aput(THREADS_NS, thread_id, record)
async def update_metadata(self, thread_id: str, metadata: dict, *, user_id: str | None | _AutoSentinel = AUTO) -> None:
record = await self._get_owned_record(thread_id, user_id, "MemoryThreadMetaStore.update_metadata")
if record is None:
@@ -174,6 +174,20 @@ class ThreadMetaRepository(ThreadMetaStore):
await session.execute(update(ThreadMetaRow).where(ThreadMetaRow.thread_id == thread_id).values(status=status, updated_at=datetime.now(UTC)))
await session.commit()
async def update_assistant_id(
self,
thread_id: str,
assistant_id: str,
*,
user_id: str | None | _AutoSentinel = AUTO,
) -> None:
resolved_user_id = resolve_user_id(user_id, method_name="ThreadMetaRepository.update_assistant_id")
async with self._sf() as session:
if not await self._check_ownership(session, thread_id, resolved_user_id):
return
await session.execute(update(ThreadMetaRow).where(ThreadMetaRow.thread_id == thread_id).values(assistant_id=assistant_id, updated_at=datetime.now(UTC)))
await session.commit()
async def update_metadata(
self,
thread_id: str,