admin/MuMuAINovel
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

update: 优化剧情分析与章节规划算法，集成伏笔上下文追踪；完善章节删除时的级联清理逻辑

Browse Source
This commit is contained in:
xiamuceer-j
2026-01-19 17:23:50 +08:00
parent dc3dbaaf2c
commit 927072d16f
3 changed files with 308 additions and 31 deletions
Download Patch File Download Diff File Expand all files Collapse all files
backend/app/services/chapter_context_service.py
+175 -10
View File
@@ -12,6 +12,7 @@ from app.models.outline import Outline
from app.models.character import Character
from app.models.career import Career, CharacterCareer
from app.models.memory import StoryMemory
from app.models.foreshadow import Foreshadow
from app.logger import get_logger
logger = get_logger(__name__)
@@ -25,12 +26,14 @@ class ChapterContext:
采用RTCO框架的分层设计：
- P0-核心（必须）：大纲、衔接点、字数要求
- P1-重要（按需）：角色、情感基调、风格
- P2-参考（条件触发）：记忆、故事骨架、MCP资料
- P2-参考（条件触发）：记忆、故事骨架、MCP资料、伏笔提醒
"""
# === P0-核心信息（必须包含）===
chapter_outline: str = "" # 本章大纲
continuation_point: Optional[str] = None # 衔接锚点（上一章结尾）
continuation_point: Optional[str] = None # 衔接锚点（增强版：含上一章摘要和结尾）
previous_chapter_summary: Optional[str] = None # 🔧 新增：上一章剧情摘要
previous_chapter_events: Optional[List[str]] = None # 🔧 新增：上一章关键事件
target_word_count: int = 3000 # 目标字数
min_word_count: int = 2500 # 最小字数
max_word_count: int = 4000 # 最大字数
@@ -54,6 +57,7 @@ class ChapterContext:
relevant_memories: Optional[str] = None # 相关记忆（精简版）
story_skeleton: Optional[str] = None # 故事骨架（50章+启用）
mcp_references: Optional[str] = None # MCP参考资料
foreshadow_reminders: Optional[str] = None # 伏笔提醒（新增）
# === 元信息 ===
context_stats: Dict[str, Any] = field(default_factory=dict) # 统计信息
@@ -62,7 +66,8 @@ class ChapterContext:
"""计算总上下文长度"""
total = 0
for field_name in ['chapter_outline', 'continuation_point', 'chapter_characters',
'relevant_memories', 'story_skeleton', 'style_instruction']:
'relevant_memories', 'story_skeleton', 'style_instruction',
'foreshadow_reminders', 'previous_chapter_summary']:
value = getattr(self, field_name, None)
if value:
total += len(value)
@@ -91,14 +96,16 @@ class ChapterContextBuilder:
STYLE_MAX_LENGTH = 200 # 风格描述最大长度
MAX_CONTEXT_LENGTH = 3000 # 总上下文最大字符数
def __init__(self, memory_service=None):
def __init__(self, memory_service=None, foreshadow_service=None):
"""
初始化构建器
Args:
memory_service: 记忆服务实例（可选，用于检索相关记忆）
foreshadow_service: 伏笔服务实例（可选，用于获取伏笔提醒）
"""
self.memory_service = memory_service
self.foreshadow_service = foreshadow_service
async def build(
self,
@@ -155,20 +162,28 @@ class ChapterContextBuilder:
chapter, outline, project.outline_mode
)
# === 衔接锚点（根据章节调整长度）===
# === 衔接锚点（根据章节调整长度，增强版含摘要和事件）===
if chapter_number == 1:
context.continuation_point = None
context.previous_chapter_summary = None
context.previous_chapter_events = None
logger.info(" ✅ 第1章无需衔接锚点")
elif chapter_number <= 10:
context.continuation_point = await self._get_last_ending(
ending_info = await self._get_last_ending_enhanced(
chapter, db, self.ENDING_LENGTH_SHORT
)
logger.info(f" ✅ 衔接锚点（短）: {len(context.continuation_point or '')}字符")
context.continuation_point = ending_info.get('ending_text')
context.previous_chapter_summary = ending_info.get('summary')
context.previous_chapter_events = ending_info.get('key_events')
logger.info(f" ✅ 衔接锚点（短）: {len(context.continuation_point or '')}字符, 摘要: {len(context.previous_chapter_summary or '')}字符")
else:
context.continuation_point = await self._get_last_ending(
ending_info = await self._get_last_ending_enhanced(
chapter, db, self.ENDING_LENGTH_NORMAL
)
logger.info(f" ✅ 衔接锚点（标准）: {len(context.continuation_point or '')}字符")
context.continuation_point = ending_info.get('ending_text')
context.previous_chapter_summary = ending_info.get('summary')
context.previous_chapter_events = ending_info.get('key_events')
logger.info(f" ✅ 衔接锚点（标准）: {len(context.continuation_point or '')}字符, 摘要: {len(context.previous_chapter_summary or '')}字符")
# === P1-重要信息 ===
context.chapter_characters = await self._build_chapter_characters(
@@ -200,6 +215,14 @@ class ChapterContextBuilder:
)
logger.info(f" ✅ 故事骨架: {len(context.story_skeleton or '')}字符")
# === P2-伏笔提醒（新增）===
if self.foreshadow_service:
context.foreshadow_reminders = await self._get_foreshadow_reminders(
project.id, chapter_number, db
)
if context.foreshadow_reminders:
logger.info(f" ✅ 伏笔提醒: {len(context.foreshadow_reminders)}字符")
# === 统计信息 ===
context.context_stats = {
"chapter_number": chapter_number,
@@ -208,6 +231,7 @@ class ChapterContextBuilder:
"characters_length": len(context.chapter_characters),
"memories_length": len(context.relevant_memories or ""),
"skeleton_length": len(context.story_skeleton or ""),
"foreshadow_length": len(context.foreshadow_reminders or ""),
"total_length": context.get_total_context_length()
}
@@ -263,7 +287,7 @@ class ChapterContextBuilder:
max_length: int
) -> Optional[str]:
"""
获取上一章结尾内容作为衔接锚点
获取上一章结尾内容作为衔接锚点（旧版本，保留兼容性）
Args:
chapter: 当前章节
@@ -294,6 +318,105 @@ class ChapterContextBuilder:
return content[-max_length:]
async def _get_last_ending_enhanced(
self,
chapter: Chapter,
db: AsyncSession,
max_length: int
) -> Dict[str, Any]:
"""
获取增强版衔接锚点（含上一章摘要和关键事件）
🔧 新增功能：
1. 提取上一章结尾文本
2. 获取上一章剧情摘要（从记忆或expansion_plan）
3. 提取上一章关键事件
Args:
chapter: 当前章节
db: 数据库会话
max_length: 最大长度
Returns:
包含 ending_text, summary, key_events 的字典
"""
result_info = {
'ending_text': None,
'summary': None,
'key_events': []
}
if chapter.chapter_number <= 1:
return result_info
# 查询上一章
result = await db.execute(
select(Chapter)
.where(Chapter.project_id == chapter.project_id)
.where(Chapter.chapter_number == chapter.chapter_number - 1)
)
prev_chapter = result.scalar_one_or_none()
if not prev_chapter:
return result_info
# 1. 提取结尾内容
if prev_chapter.content:
content = prev_chapter.content.strip()
if len(content) <= max_length:
result_info['ending_text'] = content
else:
result_info['ending_text'] = content[-max_length:]
# 2. 获取上一章摘要
# 优先从记忆中获取 chapter_summary
summary_result = await db.execute(
select(StoryMemory.content)
.where(StoryMemory.project_id == chapter.project_id)
.where(StoryMemory.chapter_id == prev_chapter.id)
.where(StoryMemory.memory_type == 'chapter_summary')
.limit(1)
)
summary_mem = summary_result.scalar_one_or_none()
if summary_mem:
result_info['summary'] = summary_mem[:300] # 限制长度
elif prev_chapter.summary:
# 回退到章节的summary字段
result_info['summary'] = prev_chapter.summary[:300]
elif prev_chapter.expansion_plan:
# 再回退到expansion_plan中的plot_summary
try:
plan = json.loads(prev_chapter.expansion_plan)
result_info['summary'] = plan.get('plot_summary', '')[:300]
except json.JSONDecodeError:
pass
# 3. 提取上一章关键事件
if prev_chapter.expansion_plan:
try:
plan = json.loads(prev_chapter.expansion_plan)
key_events = plan.get('key_events', [])
if key_events:
result_info['key_events'] = key_events[:5] # 最多5个事件
except json.JSONDecodeError:
pass
# 如果没有从expansion_plan获取到，尝试从记忆中获取
if not result_info['key_events']:
events_result = await db.execute(
select(StoryMemory.content)
.where(StoryMemory.project_id == chapter.project_id)
.where(StoryMemory.chapter_id == prev_chapter.id)
.where(StoryMemory.memory_type == 'plot_point')
.limit(5)
)
event_mems = events_result.scalars().all()
if event_mems:
result_info['key_events'] = [e[:100] for e in event_mems]
return result_info
async def _build_chapter_characters(
self,
chapter: Chapter,
@@ -564,6 +687,48 @@ class ChapterContextBuilder:
return "\n".join(lines) if lines else None
async def _get_foreshadow_reminders(
self,
project_id: str,
chapter_number: int,
db: AsyncSession
) -> Optional[str]:
"""
获取伏笔提醒信息用于章节生成
策略：
1. 获取计划在本章或之前回收但未回收的伏笔（超期提醒）
2. 获取已埋入且接近需要回收的伏笔（提前提醒）
3. 获取本章计划埋入的伏笔（埋入提醒）
Args:
project_id: 项目ID
chapter_number: 当前章节号
db: 数据库会话
Returns:
格式化的伏笔提醒文本
"""
if not self.foreshadow_service:
return None
try:
context_result = await self.foreshadow_service.build_chapter_context(
db=db,
project_id=project_id,
chapter_number=chapter_number,
include_pending=True,
include_overdue=True,
lookahead=5
)
context_text = context_result.get("context_text", "")
return context_text if context_text else None
except Exception as e:
logger.error(f"❌ 获取伏笔提醒失败: {str(e)}")
return None
async def _build_story_skeleton(
self,
project_id: str,