729 lines
28 KiB
Python
729 lines
28 KiB
Python
# -*- coding: utf-8 -*-
|
||
"""
|
||
内容分析压缩引擎 v4.0
|
||
从大量爬取内容/文件中提取关键信息，压缩为 2-3 页结构化摘要。
|
||
|
||
核心策略：
|
||
1. 财务指标优先提取（复用 document_analyzer）
|
||
2. 关键词密度排序（TF-IDF 启发式）
|
||
3. 首尾段落加权（倒金字塔结构）
|
||
4. 去重合并（语义相似度检测）
|
||
5. 结构化输出（Markdown 格式）
|
||
|
||
支持用户指定关注维度：财务 / 风险 / 行业 / 政策 / 全面
|
||
"""
|
||
|
||
import re
|
||
import json
|
||
import math
|
||
from pathlib import Path
|
||
from typing import Dict, Any, List, Optional, Tuple, Set
|
||
from datetime import datetime
|
||
from collections import Counter
|
||
from dataclasses import dataclass, field
|
||
|
||
# 复用现有模块
|
||
try:
|
||
from document_analyzer import (
|
||
classify_document, extract_metadata, extract_financial_indicators,
|
||
extract_risk_factors, extract_glossary, extract_sections,
|
||
FINANCIAL_INDICATOR_PATTERNS, DOC_CATEGORIES
|
||
)
|
||
HAS_ANALYZER = True
|
||
except ImportError:
|
||
HAS_ANALYZER = False
|
||
FINANCIAL_INDICATOR_PATTERNS = {}
|
||
DOC_CATEGORIES = []
|
||
|
||
try:
|
||
from enhanced_parser import MultiFormatParser
|
||
HAS_ENHANCED_PARSER = True
|
||
except ImportError:
|
||
HAS_ENHANCED_PARSER = False
|
||
|
||
SKILL_DATA_DIR = Path(__file__).parent.parent / "data"
|
||
SKILL_DATA_DIR.mkdir(parents=True, exist_ok=True)
|
||
|
||
|
||
# ==================== 中文停用词 ====================
|
||
|
||
STOP_WORDS = set([
|
||
'的', '了', '在', '是', '我', '有', '和', '就', '不', '人', '都', '一',
|
||
'一个', '上', '也', '很', '到', '说', '要', '去', '你', '会', '着',
|
||
'没有', '看', '好', '自己', '这', '他', '她', '它', '们', '那', '些',
|
||
'什么', '怎么', '如何', '为什么', '因为', '所以', '但是', '然而',
|
||
'可以', '这个', '那个', '这些', '那些', '已经', '还是', '只是',
|
||
'比较', '非常', '更加', '尤其', '其中', '其他', '以及', '或者',
|
||
'根据', '按照', '通过', '经过', '为了', '对于', '关于', '由于',
|
||
'目前', '现在', '正在', '将', '将把', '被', '让', '把', '向',
|
||
'与', '从', '以', '对', '等', '及', '或', '但', '而', '且',
|
||
'为', '其', '之', '所', '者', '于', '则', '也', '因',
|
||
'年', '月', '日', '时', '分', '万', '亿', '元', '只', '个',
|
||
'该', '此', '本', '各', '每', '另', '另', '第', '前', '后',
|
||
'中', '内', '外', '间', '里', '后', '前', '左右', '上下',
|
||
])
|
||
|
||
|
||
# ==================== 金融关键词权重表 ====================
|
||
|
||
FINANCIAL_KEYWORDS = {
|
||
# 财务指标类 (权重 10)
|
||
"营业收入": 10, "净利润": 10, "总资产": 10, "净资产": 10, "每股收益": 10,
|
||
"市盈率": 10, "市净率": 10, "净资产收益率": 10, "毛利率": 10, "净利率": 10,
|
||
"资产负债率": 10, "现金流": 10, "同比增长": 10, "环比增长": 10,
|
||
"归母净利润": 10, "扣非净利润": 10, "经营性现金流": 10,
|
||
# 市场类 (权重 8)
|
||
"股价": 8, "涨跌幅": 8, "成交量": 8, "成交额": 8, "换手率": 8,
|
||
"市值": 8, "估值": 8, "分红": 8, "股息率": 8,
|
||
# 评级推荐类 (权重 9)
|
||
"买入": 9, "增持": 9, "目标价": 9, "评级": 9, "推荐": 9,
|
||
"上调": 9, "下调": 9,
|
||
# 风险类 (权重 7)
|
||
"风险": 7, "诉讼": 7, "违规": 7, "处罚": 7, "退市": 7,
|
||
"减值": 7, "亏损": 7, "债务违约": 7,
|
||
# 行业政策类 (权重 6)
|
||
"政策": 6, "监管": 6, "改革": 6, "利好": 6, "利空": 6,
|
||
"行业前景": 6, "市场竞争": 6, "技术创新": 6,
|
||
# 事件类 (权重 5)
|
||
"收购": 5, "并购": 5, "重组": 5, "上市": 5, "融资": 5,
|
||
"增发": 5, "回购": 5, "重大合同": 5, "战略合作": 5,
|
||
}
|
||
|
||
# 按维度分类的关键词
|
||
DIMENSION_KEYWORDS = {
|
||
"财务": ["营业收入", "净利润", "总资产", "净资产", "每股收益", "市盈率", "市净率",
|
||
"净资产收益率", "毛利率", "净利率", "资产负债率", "现金流", "同比增长",
|
||
"归母净利润", "扣非净利润", "经营性现金流"],
|
||
"风险": ["风险", "诉讼", "违规", "处罚", "退市", "减值", "亏损", "债务违约",
|
||
"担保", "质押", "冻结", "警示", "问询", "调查"],
|
||
"行业": ["行业", "市场", "竞争", "格局", "份额", "趋势", "前景", "产业链",
|
||
"上下游", "供给", "需求", "产能", "景气"],
|
||
"政策": ["政策", "监管", "法规", "改革", "合规", "牌照", "审批", "指导意见",
|
||
"通知", "规定", "办法"],
|
||
"事件": ["收购", "并购", "重组", "上市", "融资", "增发", "回购", "分红",
|
||
"重大合同", "战略合作", "投资", "减持", "增持", "解禁"],
|
||
}
|
||
|
||
|
||
# ==================== 数据结构 ====================
|
||
|
||
@dataclass
|
||
class CompressConfig:
|
||
"""压缩配置"""
|
||
focus: str = "全面" # 财务 / 风险 / 行业 / 政策 / 事件 / 全面
|
||
max_pages: int = 3 # 目标页数 (约 1000 字/页)
|
||
max_chars: int = 3000 # 最大输出字符数
|
||
include_tables: bool = True # 是否包含表格
|
||
include_charts_hint: bool = True # 是否包含图表建议
|
||
language: str = "zh" # 输出语言
|
||
|
||
|
||
@dataclass
|
||
class CompressResult:
|
||
"""压缩结果"""
|
||
title: str
|
||
summary: str # 一句话摘要
|
||
key_points: List[str] # 关键要点 (3-5 条)
|
||
financial_highlights: Dict[str, Any] # 财务亮点
|
||
risk_summary: List[str] # 风险摘要
|
||
structured_report: str # 结构化 Markdown 报告
|
||
focus_dimensions: List[str] # 实际覆盖的维度
|
||
stats: Dict[str, Any] # 统计信息
|
||
source_info: Dict[str, Any] # 来源信息
|
||
|
||
|
||
# ==================== 内容压缩引擎 ====================
|
||
|
||
class ContentCompressor:
|
||
"""内容分析压缩引擎。"""
|
||
|
||
def __init__(self):
|
||
self._parser = None
|
||
if HAS_ENHANCED_PARSER:
|
||
self._parser = MultiFormatParser()
|
||
|
||
def compress(self, source: Any, config: Optional[CompressConfig] = None) -> CompressResult:
|
||
"""
|
||
压缩内容为 2-3 页摘要。
|
||
|
||
Args:
|
||
source: 文本内容(str)、文件路径(str)、或解析结果(dict)
|
||
config: 压缩配置
|
||
|
||
返回: CompressResult
|
||
"""
|
||
if config is None:
|
||
config = CompressConfig()
|
||
|
||
# 1. 标准化输入
|
||
text_content, source_info = self._normalize_source(source)
|
||
if not text_content:
|
||
return CompressResult(
|
||
title="内容为空", summary="", key_points=[],
|
||
financial_highlights={}, risk_summary=[],
|
||
structured_report="# 内容压缩失败\n\n原始内容为空，无法提取有效信息。",
|
||
focus_dimensions=[], stats={}, source_info=source_info
|
||
)
|
||
|
||
# 2. 提取结构化信息
|
||
metadata = self._extract_metadata(text_content, source_info)
|
||
financial = self._extract_financial(text_content)
|
||
risks = self._extract_risks(text_content)
|
||
sections = self._extract_sections(text_content)
|
||
keywords = self._extract_keywords(text_content, config.focus)
|
||
|
||
# 3. 生成关键要点（按维度筛选）
|
||
key_points = self._generate_key_points(
|
||
text_content, metadata, financial, risks, keywords, config
|
||
)
|
||
|
||
# 4. 构建结构化报告
|
||
structured_report = self._build_report(
|
||
metadata, financial, risks, key_points, sections, config
|
||
)
|
||
|
||
# 5. 组装结果
|
||
title = metadata.get("title") or source_info.get("name", "未命名内容")
|
||
summary = self._generate_one_line_summary(text_content, metadata, financial)
|
||
|
||
return CompressResult(
|
||
title=title,
|
||
summary=summary,
|
||
key_points=key_points,
|
||
financial_highlights=financial,
|
||
risk_summary=risks[:5],
|
||
structured_report=structured_report,
|
||
focus_dimensions=self._get_active_dimensions(config.focus),
|
||
stats={
|
||
"source_chars": len(text_content),
|
||
"output_chars": len(structured_report),
|
||
"compression_ratio": f"{len(structured_report)/max(len(text_content),1)*100:.1f}%",
|
||
"key_points_count": len(key_points),
|
||
"financial_indicators_count": len(financial),
|
||
"risk_factors_count": len(risks),
|
||
"compressed_at": datetime.now().strftime("%Y-%m-%d %H:%M:%S"),
|
||
},
|
||
source_info=source_info,
|
||
)
|
||
|
||
def compress_multiple(self, sources: List[Any],
|
||
config: Optional[CompressConfig] = None) -> CompressResult:
|
||
"""
|
||
批量压缩多个来源的内容，合并为一个统一摘要。
|
||
|
||
Args:
|
||
sources: 多个来源（文本、文件路径、解析结果）
|
||
config: 压缩配置
|
||
|
||
返回: 合并后的 CompressResult
|
||
"""
|
||
if config is None:
|
||
config = CompressConfig()
|
||
|
||
results = []
|
||
for src in sources:
|
||
try:
|
||
r = self.compress(src, config)
|
||
results.append(r)
|
||
except Exception:
|
||
continue
|
||
|
||
if not results:
|
||
return CompressResult(
|
||
title="合并压缩失败", summary="", key_points=[],
|
||
financial_highlights={}, risk_summary=[],
|
||
structured_report="# 合并压缩失败\n\n所有来源均无法提取有效信息。",
|
||
focus_dimensions=[], stats={}, source_info={}
|
||
)
|
||
|
||
# 合并
|
||
all_points = []
|
||
all_fin = {}
|
||
all_risks = []
|
||
for r in results:
|
||
all_points.extend(r.key_points)
|
||
all_fin.update(r.financial_highlights)
|
||
all_risks.extend(r.risk_summary)
|
||
|
||
# 去重关键要点
|
||
unique_points = []
|
||
seen = set()
|
||
for p in all_points:
|
||
norm = re.sub(r'\s+', '', p)[:30]
|
||
if norm not in seen:
|
||
seen.add(norm)
|
||
unique_points.append(p)
|
||
|
||
# 构建合并报告
|
||
merged_meta = {"title": f"多源内容合并摘要 ({len(results)} 个来源)"}
|
||
return CompressResult(
|
||
title=merged_meta["title"],
|
||
summary=f"合并了 {len(results)} 个来源的内容，提取 {len(unique_points)} 条关键要点。",
|
||
key_points=unique_points[:8],
|
||
financial_highlights=dict(list(all_fin.items())[:15]),
|
||
risk_summary=list(dict.fromkeys(all_risks))[:8],
|
||
structured_report=self._build_merged_report(results, config),
|
||
focus_dimensions=self._get_active_dimensions(config.focus),
|
||
stats={
|
||
"source_count": len(results),
|
||
"total_source_chars": sum(r.stats.get("source_chars", 0) for r in results),
|
||
"key_points_count": len(unique_points),
|
||
"compressed_at": datetime.now().strftime("%Y-%m-%d %H:%M:%S"),
|
||
},
|
||
source_info={"sources": [r.source_info for r in results]},
|
||
)
|
||
|
||
def _normalize_source(self, source: Any) -> Tuple[str, Dict[str, Any]]:
|
||
"""标准化输入为纯文本。"""
|
||
source_info = {}
|
||
|
||
if isinstance(source, str) and len(source) < 500 and os.path.isfile(source):
|
||
# 文件路径
|
||
source_info = {"type": "file", "path": source, "name": Path(source).name}
|
||
result = self._parse_file(source)
|
||
text = result.get("text_content", "") or ""
|
||
source_info.update({k: v for k, v in result.items()
|
||
if k not in ("text_content", "all_text", "slides")})
|
||
return text, source_info
|
||
|
||
elif isinstance(source, dict):
|
||
# 解析结果
|
||
source_info = {
|
||
"type": source.get("file_type", "dict"),
|
||
"name": source.get("file_path", source.get("title", "未命名")),
|
||
}
|
||
text = source.get("text_content", source.get("all_text", "")) or ""
|
||
if not text:
|
||
text = json.dumps(source, ensure_ascii=False)
|
||
return text, source_info
|
||
|
||
elif isinstance(source, str):
|
||
# 纯文本
|
||
source_info = {"type": "text", "name": "文本内容"}
|
||
return source, source_info
|
||
|
||
else:
|
||
# 尝试转字符串
|
||
source_info = {"type": "unknown", "name": str(type(source).__name__)}
|
||
return str(source), source_info
|
||
|
||
def _parse_file(self, file_path: str) -> Dict[str, Any]:
|
||
"""使用增强解析器解析文件。"""
|
||
if self._parser:
|
||
try:
|
||
return self._parser.parse(file_path)
|
||
except Exception:
|
||
pass
|
||
|
||
# fallback: 直接读文本
|
||
result = {"file_type": Path(file_path).suffix, "file_path": file_path}
|
||
try:
|
||
with open(file_path, 'r', encoding='utf-8') as f:
|
||
result["text_content"] = f.read()
|
||
except (UnicodeDecodeError, Exception):
|
||
try:
|
||
with open(file_path, 'rb') as f:
|
||
raw = f.read()
|
||
# 尝试常见编码
|
||
for enc in ['gbk', 'gb2312', 'gb18030']:
|
||
try:
|
||
result["text_content"] = raw.decode(enc)
|
||
break
|
||
except UnicodeDecodeError:
|
||
continue
|
||
if "text_content" not in result:
|
||
result["text_content"] = raw.decode('utf-8', errors='ignore')
|
||
except Exception:
|
||
result["text_content"] = ""
|
||
return result
|
||
|
||
def _extract_metadata(self, text: str, source_info: Dict) -> Dict[str, str]:
|
||
"""提取内容元数据。"""
|
||
if HAS_ANALYZER:
|
||
meta = extract_metadata(text)
|
||
if source_info.get("name"):
|
||
meta.setdefault("title", source_info["name"])
|
||
return meta
|
||
|
||
# 简易提取
|
||
meta = {"title": source_info.get("name", ""), "institution": "", "date": "", "stock_code": ""}
|
||
lines = [l.strip() for l in text.split('\n') if l.strip()]
|
||
if lines:
|
||
meta["title"] = lines[0][:120]
|
||
# 日期
|
||
dm = re.search(r'(\d{4})[-/年](\d{1,2})[-/月](\d{1,2})', text[:2000])
|
||
if dm:
|
||
meta["date"] = f"{dm.group(1)}-{int(dm.group(2)):02d}-{int(dm.group(3)):02d}"
|
||
# 股票代码
|
||
cm = re.search(r'[\(（](\d{6})[\)）]', text[:500])
|
||
if cm:
|
||
meta["stock_code"] = cm.group(1)
|
||
return meta
|
||
|
||
def _extract_financial(self, text: str) -> Dict[str, Any]:
|
||
"""提取财务指标。"""
|
||
if HAS_ANALYZER:
|
||
return extract_financial_indicators(text)
|
||
|
||
# 简易财务提取
|
||
fin: Dict[str, Any] = {}
|
||
for key, pat in FINANCIAL_INDICATOR_PATTERNS.items():
|
||
m = re.search(pat, text)
|
||
if m:
|
||
try:
|
||
val = m.group(1).replace(",", "")
|
||
fin[key] = float(val) if '.' in val else int(val)
|
||
except (ValueError, AttributeError):
|
||
fin[key] = m.group(1)
|
||
return fin
|
||
|
||
def _extract_risks(self, text: str) -> List[str]:
|
||
"""提取风险因素。"""
|
||
if HAS_ANALYZER:
|
||
return extract_risk_factors(text)
|
||
|
||
risks = []
|
||
risk_section = re.search(
|
||
r'(?:风险因素|风险提示|风险分析|风险警示)(.{0,3000}?)(?=第[一二三四五六七八九十\d]+[章节]|\Z)',
|
||
text, re.DOTALL
|
||
)
|
||
if risk_section:
|
||
chunk = risk_section.group(0)
|
||
for m in re.finditer(r'[•·●◆■\d+\.、．)]\s*([^\n•·●◆■]{10,200})', chunk):
|
||
item = m.group(1).strip()
|
||
if len(item) > 10 and item not in risks:
|
||
risks.append(item[:150])
|
||
return risks[:15]
|
||
|
||
def _extract_sections(self, text: str) -> List[Dict[str, Any]]:
|
||
"""提取章节结构。"""
|
||
if HAS_ANALYZER:
|
||
return extract_sections(text)
|
||
|
||
sections = []
|
||
for m in re.finditer(
|
||
r'^[第]*[一二三四五六七八九十\d]+[章节篇、.]\s*([^\n]{3,80})$',
|
||
text, re.MULTILINE
|
||
):
|
||
sections.append({"title": m.group(0).strip(), "level": 1})
|
||
return sections[:20]
|
||
|
||
def _extract_keywords(self, text: str, focus: str) -> Dict[str, float]:
|
||
"""
|
||
提取关键词及其重要性得分（TF-IDF 启发式）。
|
||
"""
|
||
# 分词（简单字符级 n-gram）
|
||
words = []
|
||
for m in re.finditer(r'[\u4e00-\u9fff]{2,8}|[A-Za-z]{2,20}|[+-]?\d+[\.]?\d*%?', text):
|
||
words.append(m.group())
|
||
|
||
total = len(words) if words else 1
|
||
|
||
# TF 计算
|
||
word_counts = Counter(words)
|
||
|
||
# 结合预定义金融关键词权重
|
||
keywords: Dict[str, float] = {}
|
||
active_dims = self._get_active_dimensions(focus)
|
||
|
||
for word, count in word_counts.items():
|
||
if word in STOP_WORDS or len(word) < 2:
|
||
continue
|
||
|
||
tf = count / total
|
||
bonus = 0
|
||
|
||
# 金融关键词加权
|
||
if word in FINANCIAL_KEYWORDS:
|
||
bonus = FINANCIAL_KEYWORDS[word] * 0.1
|
||
|
||
# 维度匹配加权
|
||
for dim in active_dims:
|
||
if dim in DIMENSION_KEYWORDS and word in DIMENSION_KEYWORDS[dim]:
|
||
bonus += 0.5
|
||
|
||
score = tf * (1 + bonus)
|
||
if score > 0.001:
|
||
keywords[word] = score
|
||
|
||
# 排序取 Top 20
|
||
return dict(sorted(keywords.items(), key=lambda x: -x[1])[:20])
|
||
|
||
def _generate_key_points(self, text: str, metadata: Dict, financial: Dict,
|
||
risks: List, keywords: Dict, config: CompressConfig) -> List[str]:
|
||
"""生成关键要点列表。"""
|
||
points = []
|
||
|
||
# 1. 财务要点
|
||
if financial:
|
||
fin_items = list(financial.items())
|
||
for key, val in fin_items[:5]:
|
||
label = key.replace('_', ' ')
|
||
points.append(f"📊 {label}: {val}")
|
||
|
||
# 2. 关键词要点
|
||
top_kw = list(keywords.keys())[:5]
|
||
if top_kw:
|
||
# 查找这些关键词所在的句子
|
||
sentences = re.split(r'[。！；\n]', text)
|
||
for kw in top_kw[:3]:
|
||
for sent in sentences:
|
||
if kw in sent and len(sent) > 15:
|
||
sent_clean = sent.strip()[:120]
|
||
if sent_clean not in points and sent_clean not in ' '.join(points):
|
||
points.append(f"🔑 [{kw}] {sent_clean}")
|
||
break
|
||
|
||
# 3. 风险要点
|
||
if risks and config.focus in ("风险", "全面"):
|
||
for risk in risks[:3]:
|
||
if risk not in ' '.join(points):
|
||
points.append(f"⚠️ 风险提示: {risk[:100]}")
|
||
|
||
# 4. 标题/元数据要点
|
||
if metadata.get("title"):
|
||
points.insert(0, f"📄 标题: {metadata['title']}")
|
||
if metadata.get("date"):
|
||
points.insert(1, f"📅 日期: {metadata['date']}")
|
||
if metadata.get("institution"):
|
||
points.insert(2, f"🏢 机构: {metadata['institution']}")
|
||
|
||
# 确保至少 3 条
|
||
if len(points) < 3:
|
||
# 取文本首尾关键句
|
||
sentences = [s.strip() for s in re.split(r'[。！；\n]', text) if len(s.strip()) > 20]
|
||
for sent in sentences[:5]:
|
||
if sent not in ' '.join(points):
|
||
points.append(f"💡 {sent[:120]}")
|
||
if len(points) >= 5:
|
||
break
|
||
|
||
return points[:8]
|
||
|
||
def _generate_one_line_summary(self, text: str, metadata: Dict,
|
||
financial: Dict) -> str:
|
||
"""生成一句话摘要。"""
|
||
parts = []
|
||
if metadata.get("title"):
|
||
parts.append(metadata["title"])
|
||
if metadata.get("institution"):
|
||
parts.append(f"({metadata['institution']})")
|
||
if metadata.get("date"):
|
||
parts.append(f"[{metadata['date']}]")
|
||
|
||
# 取第一段关键词句
|
||
lines = [l.strip() for l in text.split('\n') if len(l.strip()) > 30]
|
||
if lines:
|
||
first_line = lines[0][:200]
|
||
parts.append(f"— {first_line}")
|
||
|
||
return ' '.join(parts)[:300]
|
||
|
||
def _build_report(self, metadata: Dict, financial: Dict, risks: List,
|
||
key_points: List, sections: List,
|
||
config: CompressConfig) -> str:
|
||
"""构建结构化 Markdown 报告。"""
|
||
lines = []
|
||
|
||
# 标题
|
||
title = metadata.get("title") or "内容压缩报告"
|
||
lines.append(f"# {title}\n")
|
||
|
||
# 元数据栏
|
||
meta_parts = []
|
||
if metadata.get("date"):
|
||
meta_parts.append(f"📅 {metadata['date']}")
|
||
if metadata.get("institution"):
|
||
meta_parts.append(f"🏢 {metadata['institution']}")
|
||
if metadata.get("stock_code"):
|
||
meta_parts.append(f"📈 {metadata['stock_code']}")
|
||
if meta_parts:
|
||
lines.append(" | ".join(meta_parts))
|
||
lines.append("")
|
||
|
||
# 一句话摘要
|
||
lines.append("## 📋 摘要\n")
|
||
lines.append(f"> {self._generate_one_line_summary('', metadata, financial)}\n")
|
||
|
||
# 关键要点
|
||
if key_points:
|
||
lines.append("## 🔑 关键要点\n")
|
||
for i, point in enumerate(key_points, 1):
|
||
lines.append(f"{i}. {point}")
|
||
lines.append("")
|
||
|
||
# 财务数据
|
||
if financial and config.include_tables:
|
||
lines.append("## 💰 核心财务数据\n")
|
||
lines.append("| 指标 | 数值 |")
|
||
lines.append("|------|------|")
|
||
for key, val in list(financial.items())[:10]:
|
||
label = key.replace('_', ' ').replace(' ', ' ')
|
||
lines.append(f"| {label} | {val} |")
|
||
lines.append("")
|
||
|
||
# 风险提示
|
||
if risks and config.focus in ("风险", "全面"):
|
||
lines.append("## ⚠️ 风险因素\n")
|
||
for i, risk in enumerate(risks[:5], 1):
|
||
lines.append(f"{i}. {risk}")
|
||
lines.append("")
|
||
|
||
# 章节结构（如果有）
|
||
if sections and len(sections) >= 3:
|
||
lines.append("## 📑 内容结构\n")
|
||
for sec in sections[:10]:
|
||
indent = " " * (sec.get("level", 1) - 1)
|
||
lines.append(f"{indent}- {sec['title']}")
|
||
lines.append("")
|
||
|
||
# 页脚
|
||
lines.append("---")
|
||
lines.append(f"*压缩时间: {datetime.now().strftime('%Y-%m-%d %H:%M:%S')} | "
|
||
f"关注维度: {config.focus}*")
|
||
|
||
return '\n'.join(lines)
|
||
|
||
def _build_merged_report(self, results: List[CompressResult],
|
||
config: CompressConfig) -> str:
|
||
"""构建多源合并报告。"""
|
||
lines = [f"# 多源内容合并摘要 ({len(results)} 个来源)\n"]
|
||
|
||
lines.append("## 📋 来源列表\n")
|
||
for i, r in enumerate(results, 1):
|
||
lines.append(f"{i}. [{r.source_info.get('type', '?')}] {r.title}")
|
||
lines.append("")
|
||
|
||
# 合并关键要点
|
||
all_points = []
|
||
seen = set()
|
||
for r in results:
|
||
for p in r.key_points:
|
||
norm = re.sub(r'\s+', '', p)[:30]
|
||
if norm not in seen:
|
||
seen.add(norm)
|
||
all_points.append(p)
|
||
|
||
if all_points:
|
||
lines.append("## 🔑 综合关键要点\n")
|
||
for i, point in enumerate(all_points[:10], 1):
|
||
lines.append(f"{i}. {point}")
|
||
lines.append("")
|
||
|
||
# 合并财务数据
|
||
all_fin: Dict[str, Any] = {}
|
||
for r in results:
|
||
all_fin.update(r.financial_highlights)
|
||
|
||
if all_fin:
|
||
lines.append("## 💰 财务数据汇总\n")
|
||
lines.append("| 指标 | 数值 |")
|
||
lines.append("|------|------|")
|
||
for key, val in list(all_fin.items())[:12]:
|
||
label = key.replace('_', ' ').replace(' ', ' ')
|
||
lines.append(f"| {label} | {val} |")
|
||
lines.append("")
|
||
|
||
lines.append("---")
|
||
lines.append(f"*合并压缩时间: {datetime.now().strftime('%Y-%m-%d %H:%M:%S')}*")
|
||
return '\n'.join(lines)
|
||
|
||
def _get_active_dimensions(self, focus: str) -> List[str]:
|
||
"""获取活跃的分析维度。"""
|
||
if focus == "全面":
|
||
return list(DIMENSION_KEYWORDS.keys())
|
||
elif focus in DIMENSION_KEYWORDS:
|
||
return [focus]
|
||
return ["全面"]
|
||
|
||
def to_text(self, result: CompressResult) -> str:
|
||
"""将压缩结果转为纯文本。"""
|
||
return result.structured_report
|
||
|
||
def to_json(self, result: CompressResult) -> str:
|
||
"""将压缩结果序列化为 JSON。"""
|
||
data = {
|
||
"title": result.title,
|
||
"summary": result.summary,
|
||
"key_points": result.key_points,
|
||
"financial_highlights": result.financial_highlights,
|
||
"risk_summary": result.risk_summary,
|
||
"focus_dimensions": result.focus_dimensions,
|
||
"stats": result.stats,
|
||
"source_info": result.source_info,
|
||
"report": result.structured_report,
|
||
}
|
||
return json.dumps(data, ensure_ascii=False, indent=2)
|
||
|
||
def save_report(self, result: CompressResult, output_path: str = "") -> str:
|
||
"""保存压缩报告到文件。"""
|
||
if not output_path:
|
||
timestamp = datetime.now().strftime("%Y%m%d_%H%M%S")
|
||
safe_title = re.sub(r'[^\w\u4e00-\u9fff]', '_', result.title)[:50]
|
||
output_path = str(SKILL_DATA_DIR / "compressed" / f"{safe_title}_{timestamp}.md")
|
||
|
||
Path(output_path).parent.mkdir(parents=True, exist_ok=True)
|
||
with open(output_path, 'w', encoding='utf-8') as f:
|
||
f.write(result.structured_report)
|
||
return output_path
|
||
|
||
|
||
# ==================== 便捷函数 ====================
|
||
|
||
_compressor = ContentCompressor()
|
||
|
||
|
||
def compress_content(source: Any, focus: str = "全面",
|
||
max_pages: int = 3) -> CompressResult:
|
||
"""
|
||
一键压缩内容为 2-3 页摘要。
|
||
|
||
Args:
|
||
source: 文本、文件路径或解析结果
|
||
focus: 关注维度 — 财务 / 风险 / 行业 / 政策 / 事件 / 全面
|
||
max_pages: 目标页数
|
||
|
||
返回: CompressResult
|
||
"""
|
||
config = CompressConfig(focus=focus, max_pages=max_pages,
|
||
max_chars=max_pages * 1000)
|
||
return _compressor.compress(source, config)
|
||
|
||
|
||
def compress_multiple(sources: List[Any], focus: str = "全面",
|
||
max_pages: int = 3) -> CompressResult:
|
||
"""批量压缩多个来源。"""
|
||
config = CompressConfig(focus=focus, max_pages=max_pages,
|
||
max_chars=max_pages * 1000)
|
||
return _compressor.compress_multiple(sources, config)
|
||
|
||
|
||
# ==================== CLI 入口 ====================
|
||
|
||
if __name__ == "__main__":
|
||
import sys
|
||
|
||
if len(sys.argv) < 2:
|
||
print("用法: python content_compressor.py <文件路径> [关注维度] [页数]")
|
||
print(" 关注维度: 财务 / 风险 / 行业 / 政策 / 事件 / 全面(默认)")
|
||
print(" 页数: 1-5 (默认3)")
|
||
print()
|
||
print("示例:")
|
||
print(" python content_compressor.py report.pdf 财务 2")
|
||
print(" python content_compressor.py \"这是一段长文本...\" 全面")
|
||
sys.exit(1)
|
||
|
||
source = sys.argv[1]
|
||
focus = sys.argv[2] if len(sys.argv) > 2 else "全面"
|
||
pages = int(sys.argv[3]) if len(sys.argv) > 3 else 3
|
||
|
||
result = compress_content(source, focus=focus, max_pages=pages)
|
||
|
||
print(result.structured_report)
|
||
print(f"\n{'='*60}")
|
||
print(f"📊 压缩统计: {json.dumps(result.stats, ensure_ascii=False)}")
|
||
|
||
# 自动保存
|
||
saved = _compressor.save_report(result)
|
||
print(f"💾 已保存: {saved}")
|