Files
2026-05-20 12:51:04 +08:00

539 lines
23 KiB
Python
Raw Permalink Blame History

This file contains ambiguous Unicode characters
This file contains Unicode characters that might be confused with other characters. If you think that this is intentional, you can safely ignore this warning. Use the Escape button to reveal them.
"""AI 写作 LangGraph Graph,含 4 个用户干预点。"""
from __future__ import annotations
import logging
from langgraph.graph import END, StateGraph
from langgraph.types import interrupt
from .state import AIWritingState
from .nodes.intent_parser import intent_parser_node
from .nodes.researcher import researcher_node
from .nodes.writer_outline import writer_outline_node
from .nodes.writer_draft import writer_draft_node
from .nodes.editor import editor_node
logger = logging.getLogger(__name__)
# ── 干预节点 ──────────────────────────────────────────────────────────────────
async def pause_material_confirm(state: AIWritingState) -> dict:
"""PAUSE-1:素材确认"""
pkg = state["material_package"]
user_input: dict = interrupt({
"pause_point": "material_confirm",
"message": "素材收集专家已完成素材搜集,请确认素材后继续。",
"materials": pkg["materials"],
"keywords": pkg["keywords"],
"summary": pkg["summary"],
})
action = user_input.get("action", "confirm")
if action == "confirm":
approved_ids = set(user_input.get("approvedMaterialIds") or [m["id"] for m in pkg["materials"]])
filtered = [m for m in pkg["materials"] if m["id"] in approved_ids]
updated_pkg = {**pkg, "materials": filtered}
return {
"material_package": updated_pkg,
"last_user_intervention": user_input,
"status": "writing_outline",
"progress_events": [{"type": "resumed", "agent_name": "系统", "message": "用户已确认素材,开始规划大纲"}],
}
# re_search:重新搜索
extra_kws = user_input.get("extraKeywords") or []
return {
"last_user_intervention": {**user_input, "_extra_keywords": extra_kws},
"status": "researching",
"progress_events": [{"type": "resumed", "agent_name": "系统", "message": "重新搜索中..."}],
}
def _normalize_outline(outline: dict) -> dict:
"""将前端回传的 camelCase 大纲统一为后端 snake_case 格式。"""
if not outline:
return outline
sections = []
for s in outline.get("sections", []):
sections.append({
"section_title": s.get("section_title") or s.get("sectionTitle", ""),
"key_points": s.get("key_points") or s.get("keyPoints", []),
"material_ids": s.get("material_ids") or s.get("materialIds", []),
})
return {"title": outline.get("title", ""), "sections": sections}
def _check_outline_material_support(outline: dict) -> str:
"""轻量启发式:统计未关联任何素材的章节,返回预警文案(空串=无需预警)。
仅做提示、不拦截流程;真正的素材不足会在写作节点被检出并触发求助暂停。
"""
sections = (outline or {}).get("sections", [])
weak = [
(s.get("section_title") or s.get("sectionTitle") or "")
for s in sections
if not (s.get("material_ids") or s.get("materialIds"))
]
weak = [t for t in weak if t]
if not weak:
return ""
return (
f"有 {len(weak)} 个章节未关联检索素材({('、'.join(weak))})。"
"严格依据素材模式下,这些章节在写作时可能因素材不足而暂停向你求助。"
)
async def pause_outline_confirm(state: AIWritingState) -> dict:
"""PAUSE-2:大纲确认"""
user_input: dict = interrupt({
"pause_point": "outline_confirm",
"message": "作家已完成大纲规划,请确认或编辑大纲后继续。",
"outline": state["current_outline"],
})
action = user_input.get("action", "confirm")
if action == "confirm":
raw_outline = user_input.get("editedOutline") or state["current_outline"]
edited = _normalize_outline(raw_outline) if raw_outline else state["current_outline"]
events: list[dict] = [{"type": "resumed", "agent_name": "系统", "message": "用户已确认大纲,开始写作"}]
# 阶段四:严格模式下对大纲做轻量素材充分性预警(仅提示,不拦截流程)
if state.get("writing_mode") == "strict":
warning = _check_outline_material_support(edited)
if warning:
events.append({
"type": "notice", "agent_name": "系统",
"title": "素材充分性提示", "detail": warning,
})
return {
"current_outline": edited,
"last_user_intervention": user_input,
"status": "writing_draft",
"progress_events": events,
}
# re_outline
feedback = user_input.get("outlineFeedback", "")
updates: dict = {
"last_user_intervention": {**user_input, "_feedback": feedback},
"status": "writing_outline",
"progress_events": [{"type": "resumed", "agent_name": "系统", "message": f"重新规划大纲({feedback[:30]})"}],
}
# 用户在卡片里手动改过大纲 → 以手动版作为本次重规划的编辑基准
raw_edited = user_input.get("editedOutline")
if raw_edited:
updates["current_outline"] = _normalize_outline(raw_edited)
return updates
async def pause_draft_confirm(state: AIWritingState) -> dict:
"""PAUSE-3:草稿确认"""
draft = state["current_draft"]
user_input: dict = interrupt({
"pause_point": "draft_confirm",
"message": "草稿已完成,请确认或给出修改意见。",
"draft_title": draft["title"],
"revision_count": draft["revision_count"],
"review_result": state.get("review_result"),
})
action = user_input.get("action", "to_editor")
if action == "finalize":
return {
"last_user_intervention": user_input,
"status": "done",
"progress_events": [{"type": "resumed", "agent_name": "系统", "message": "用户确认定稿"}],
}
if action == "user_revise":
notes = user_input.get("userRevisionNotes", "")
return {
"last_user_intervention": {**user_input, "_user_notes": notes},
"revision_count": state.get("revision_count", 0) + 1,
"status": "revising",
"progress_events": [{"type": "resumed", "agent_name": "系统", "message": f"按用户意见重写({notes[:30]})"}],
}
# to_editor
return {
"last_user_intervention": user_input,
"status": "reviewing",
"progress_events": [{"type": "resumed", "agent_name": "系统", "message": "提交编辑审核"}],
}
async def pause_section_help(state: AIWritingState) -> dict:
"""素材不足求助点:严格模式下作家发现章节缺素材,让用户逐章决定如何处理。
用户对每个被阻塞章节选择:
- supplement:补充检索素材(整体回到检索流程重走)
- delete:删除该章节
- loose:按通用知识续写该章节
"""
blocked = state.get("blocked_sections") or []
user_input: dict = interrupt({
"pause_point": "section_help",
"message": "部分章节缺少足够素材支撑写作,请决定如何处理。",
"blocked_sections": blocked,
})
# sectionDecisions: [{"sectionTitle": "...", "decision": "supplement|delete|loose"}]
by_title: dict = {}
for d in user_input.get("sectionDecisions") or []:
title = d.get("sectionTitle") or d.get("section_title")
if title:
by_title[title] = d.get("decision", "loose")
# 任一章节选「补充检索」→ 整体回到检索,重走「检索 → 素材确认 → 大纲 → 写作」
supplement_reasons = [
b["reason"] for b in blocked if by_title.get(b["section_title"]) == "supplement"
]
if supplement_reasons:
return {
# supplement_search:补充检索(保留已有素材、追加新素材),完成后直接回写作、不重排大纲
"last_user_intervention": {**user_input, "action": "supplement_search", "_extra_keywords": supplement_reasons},
"status": "researching",
"blocked_sections": None,
"progress_events": [{"type": "resumed", "agent_name": "系统", "message": "补充检索素材中…"}],
}
# 否则:删除标记章节、记录「通用知识续写」章节,直接回写作
outline = state.get("current_outline") or {"title": "", "sections": []}
kept, loose_titles = [], []
for s in outline.get("sections", []):
title = s.get("section_title") or s.get("sectionTitle", "")
decision = by_title.get(title)
if decision == "delete":
continue
if decision == "loose":
loose_titles.append(title)
kept.append(s)
# 防御:用户把所有章节都删了 → 保留原大纲并全部按通用知识写,避免空文章
if not kept:
kept = outline.get("sections", [])
loose_titles = [s.get("section_title") or s.get("sectionTitle", "") for s in kept]
return {
"current_outline": {**outline, "sections": kept},
"last_user_intervention": {**user_input, "_loose_sections": loose_titles},
"status": "writing_draft",
"blocked_sections": None,
"progress_events": [{"type": "resumed", "agent_name": "系统", "message": "已按你的选择继续写作…"}],
}
async def pause_review_confirm(state: AIWritingState) -> dict:
"""PAUSE-4:编辑打回确认"""
user_input: dict = interrupt({
"pause_point": "review_confirm",
"message": "编辑认为文章需要修改,请决定下一步操作。",
"review_result": state["review_result"],
"revision_count": state.get("revision_count", 0),
"max_revisions": state.get("max_revisions", 3),
})
if user_input.get("overrideVerdict") or user_input.get("action") == "force_finalize":
return {
"last_user_intervention": user_input,
"status": "done",
"progress_events": [{"type": "resumed", "agent_name": "系统", "message": "用户强制定稿"}],
}
action = user_input.get("action", "accept_review")
extra_notes = user_input.get("userRevisionNotes", "")
return {
"last_user_intervention": {**user_input, "_extra_notes": extra_notes, "action": action},
"revision_count": state.get("revision_count", 0) + 1,
"status": "revising",
"progress_events": [{"type": "resumed", "agent_name": "系统", "message": "按编辑意见重写"}],
}
# ── 条件路由 ──────────────────────────────────────────────────────────────────
def after_researcher(state: AIWritingState) -> str:
"""检索完成后:补充检索(素材不足求助)直接回写作,否则去素材确认。"""
return "writer_draft" if state.get("status") == "writing_draft" else "pause_material"
def after_material_pause(state: AIWritingState) -> str:
return "researcher" if state["status"] == "researching" else "writer_outline"
def after_outline_pause(state: AIWritingState) -> str:
return "writer_outline" if state["status"] == "writing_outline" else "writer_draft"
def after_writer_draft(state: AIWritingState) -> str:
"""写作完成后:素材不足则去求助暂停点,否则去草稿确认。"""
return "pause_section_help" if state["status"] == "awaiting_section_help" else "pause_draft"
def after_section_help_pause(state: AIWritingState) -> str:
"""求助暂停点之后:补充检索回 researcher,否则回写作。"""
return "researcher" if state["status"] == "researching" else "writer_draft"
def after_draft_pause(state: AIWritingState) -> str:
status = state["status"]
if status == "done":
return "end"
if status == "revising":
return "writer_draft"
return "editor"
def after_review(state: AIWritingState) -> str:
review = state.get("review_result") or {}
if review.get("verdict") == "pass":
return "pause_draft"
rev_count = state.get("revision_count", 0)
max_rev = state.get("max_revisions", 3)
if rev_count >= max_rev:
return "pause_draft" # 超出上限,强制到草稿确认
return "pause_review"
def after_review_pause(state: AIWritingState) -> str:
return "end" if state["status"] == "done" else "writer_draft"
# ── 组装 Graph ────────────────────────────────────────────────────────────────
def build_ai_writing_graph(checkpointer=None):
"""构建并 compile AI 写作 graph。
Args:
checkpointer: 可选的 LangGraph checkpointer。
- ``None`` → 退回 ``MemorySaver`` —— 仅供测试 / 开发使用。
进程或容器重启后会话状态全部丢失,前端拿着旧 session_id 调
``/resume`` 会得到 404。
- 其他 → 传入的 (Async)BaseCheckpointSaver 实例,用于持久化。
"""
g = StateGraph(AIWritingState)
g.add_node("intent_parser", intent_parser_node)
g.add_node("researcher", researcher_node)
g.add_node("pause_material", pause_material_confirm)
g.add_node("writer_outline", writer_outline_node)
g.add_node("pause_outline", pause_outline_confirm)
g.add_node("writer_draft", writer_draft_node)
g.add_node("pause_section_help", pause_section_help)
g.add_node("pause_draft", pause_draft_confirm)
g.add_node("editor", editor_node)
g.add_node("pause_review", pause_review_confirm)
g.set_entry_point("intent_parser")
g.add_edge("intent_parser", "researcher")
g.add_conditional_edges("researcher", after_researcher,
{"pause_material": "pause_material", "writer_draft": "writer_draft"})
g.add_conditional_edges("pause_material", after_material_pause,
{"researcher": "researcher", "writer_outline": "writer_outline"})
g.add_edge("writer_outline", "pause_outline")
g.add_conditional_edges("pause_outline", after_outline_pause,
{"writer_outline": "writer_outline", "writer_draft": "writer_draft"})
g.add_conditional_edges("writer_draft", after_writer_draft,
{"pause_section_help": "pause_section_help", "pause_draft": "pause_draft"})
g.add_conditional_edges("pause_section_help", after_section_help_pause,
{"researcher": "researcher", "writer_draft": "writer_draft"})
g.add_conditional_edges("pause_draft", after_draft_pause,
{"end": END, "writer_draft": "writer_draft", "editor": "editor"})
g.add_conditional_edges("editor", after_review,
{"pause_draft": "pause_draft", "pause_review": "pause_review"})
g.add_conditional_edges("pause_review", after_review_pause,
{"end": END, "writer_draft": "writer_draft"})
if checkpointer is None:
# MemorySaver 仅作为 fallback —— 提示一下,正式部署应通过
# get_ai_writing_graph() 拿到带持久化 checkpointer 的 graph。
from langgraph.checkpoint.memory import MemorySaver
checkpointer = MemorySaver()
return g.compile(checkpointer=checkpointer)
# ── 持久化 checkpointer 单例(懒加载)─────────────────────────────────────────
# 模块级 graph 槽位。import 时为 None;首次 ``get_ai_writing_graph()`` 调用时
# 按 config.yaml 的 checkpointer 段懒构建出真正的 graph。
# 测试可通过 ``monkeypatch.setattr(graph_mod, "AI_WRITING_GRAPH", stub)``
# 直接注入桩;本模块所有访问点都走 ``get_ai_writing_graph()`` 这一入口,会优先
# 返回 module attribute 的值,保留测试桩注入路径。
AI_WRITING_GRAPH = None
# 持有 async checkpointer(底层 aiosqlite Connection 或 psycopg pool 的所有者)。
# 进程生命周期内常驻;shutdown 时由 ``shutdown_ai_writing_graph`` 显式关闭。
_async_checkpointer_singleton = None
def _fallback_to_memory_saver(reason: str):
"""退回 InMemorySaver 的统一兜底入口。
把 reason 打到 warning 日志里方便运维定位。InMemorySaver 进程重启会丢
状态,但**当次写作能跑完** —— 我们的首要目标是「写作流程能继续」,
持久化是锦上添花。
"""
from langgraph.checkpoint.memory import InMemorySaver
logger.warning(
"ai_writing: 持久化 checkpointer 不可用,退回 InMemorySaver。原因: %s",
reason,
)
return InMemorySaver()
async def _get_async_checkpointer():
"""按 config.yaml 的 checkpointer 段返回一个 async checkpointer 实例。
**绝不抛异常** —— 任何失败一律退回 InMemorySaver。这是为了保证「写作
流程能继续」这个首要目标:持久化坏了用户至少还能写完当次稿,重启会丢
会话但不影响当下创作。
优先级:
1. ``checkpointer:`` 段(``type=sqlite|postgres|memory``)
2. fallback 到 ``InMemorySaver`` 并打 warning
返回的 saver 与主 agent 公用同一 sqlite 文件 —— LangGraph 的 schema
用 ``(thread_id, checkpoint_ns)`` 做主键,AI 写作的 thread_id 是
会话 uuid,跟主 agent 的 LangGraph thread_id 不会撞。
"""
global _async_checkpointer_singleton
if _async_checkpointer_singleton is not None:
return _async_checkpointer_singleton
try:
from deerflow.config.checkpointer_config import get_checkpointer_config
cp_cfg = get_checkpointer_config()
except Exception:
logger.exception("ai_writing: 加载 checkpointer 配置失败")
_async_checkpointer_singleton = _fallback_to_memory_saver("加载 config 抛异常")
return _async_checkpointer_singleton
if cp_cfg is None:
_async_checkpointer_singleton = _fallback_to_memory_saver(
"config.yaml 未配置 checkpointer 段"
)
return _async_checkpointer_singleton
if cp_cfg.type == "memory":
from langgraph.checkpoint.memory import InMemorySaver
_async_checkpointer_singleton = InMemorySaver()
logger.info("ai_writing: 配置 type=memory,使用 InMemorySaver")
return _async_checkpointer_singleton
if cp_cfg.type == "sqlite":
try:
import aiosqlite
from langgraph.checkpoint.sqlite.aio import AsyncSqliteSaver
from deerflow.runtime.store._sqlite_utils import (
ensure_sqlite_parent_dir,
resolve_sqlite_conn_str,
)
conn_str = resolve_sqlite_conn_str(
cp_cfg.connection_string or "ai_writing_checkpoints.db"
)
ensure_sqlite_parent_dir(conn_str)
# 直接持有 aiosqlite.Connection —— 进程生命周期,shutdown 时显式关。
conn = await aiosqlite.connect(conn_str)
saver = AsyncSqliteSaver(conn)
await saver.setup()
_async_checkpointer_singleton = saver
logger.info("ai_writing: 使用 AsyncSqliteSaver (%s)", conn_str)
return saver
except Exception as exc:
# sqlite 失败原因可能是:依赖未装 / 磁盘满 / 文件权限 / 路径不存在 /
# 文件被外部进程独占锁。任何一个都不能让 AI 写作整个不可用。
logger.exception(
"ai_writing: AsyncSqliteSaver 初始化失败,conn_str=%r",
getattr(cp_cfg, "connection_string", None),
)
_async_checkpointer_singleton = _fallback_to_memory_saver(
f"sqlite 初始化失败: {type(exc).__name__}: {exc}"
)
return _async_checkpointer_singleton
if cp_cfg.type == "postgres":
try:
from langgraph.checkpoint.postgres.aio import AsyncPostgresSaver
from psycopg_pool import AsyncConnectionPool
if not cp_cfg.connection_string:
raise ValueError("checkpointer.connection_string 是 postgres 必填项")
pool = AsyncConnectionPool(cp_cfg.connection_string, open=False)
await pool.open()
saver = AsyncPostgresSaver(pool)
await saver.setup()
_async_checkpointer_singleton = saver
logger.info("ai_writing: 使用 AsyncPostgresSaver")
return saver
except Exception as exc:
logger.exception("ai_writing: AsyncPostgresSaver 初始化失败")
_async_checkpointer_singleton = _fallback_to_memory_saver(
f"postgres 初始化失败: {type(exc).__name__}: {exc}"
)
return _async_checkpointer_singleton
# 未知 type —— config 配错也不能让服务挂
_async_checkpointer_singleton = _fallback_to_memory_saver(
f"未知 checkpointer.type={cp_cfg.type!r}"
)
return _async_checkpointer_singleton
async def get_ai_writing_graph():
"""懒加载并返回 AI 写作 graph 单例。
第一次调用按 config 选定的 checkpointer 构建 graph 并缓存;后续调用直接
返回缓存。测试可在调用前通过 ``monkeypatch.setattr`` 直接替换模块级
``AI_WRITING_GRAPH`` —— 本函数会优先返回已设的值。
"""
global AI_WRITING_GRAPH
if AI_WRITING_GRAPH is not None:
return AI_WRITING_GRAPH
checkpointer = await _get_async_checkpointer()
AI_WRITING_GRAPH = build_ai_writing_graph(checkpointer=checkpointer)
return AI_WRITING_GRAPH
def get_ai_writing_checkpointer():
"""返回当前已实例化的 checkpointer(清理任务用它来删 thread)。
若尚未懒构建过则返回 None。
"""
return _async_checkpointer_singleton
async def shutdown_ai_writing_graph() -> None:
"""关闭 saver 持有的底层连接,并清空 graph 单例。Lifespan shutdown 调用。"""
global AI_WRITING_GRAPH, _async_checkpointer_singleton
saver = _async_checkpointer_singleton
_async_checkpointer_singleton = None
AI_WRITING_GRAPH = None
if saver is None:
return
# aiosqlite saver: saver.conn 是 aiosqlite.Connection
conn = getattr(saver, "conn", None)
if conn is not None and hasattr(conn, "close"):
try:
await conn.close()
except Exception:
logger.warning("ai_writing: 关闭 checkpointer connection 失败", exc_info=True)
# postgres saver: saver.pool 是 AsyncConnectionPool
pool = getattr(saver, "pool", None)
if pool is not None and hasattr(pool, "close"):
try:
await pool.close()
except Exception:
logger.warning("ai_writing: 关闭 checkpointer pool 失败", exc_info=True)