151 lines
5.0 KiB
Python
151 lines
5.0 KiB
Python
|
||
import re
|
||
from openai import OpenAI
|
||
from typing import List, Dict
|
||
import json
|
||
import time
|
||
import os
|
||
from kg_build_v2.extract_html import extract_repairproject_tables_html,extract_operation_tables_html,extract_tables
|
||
from kg_build_v2.Prompt import prompt_extract_operation_node_rela
|
||
from modelsAPI.model_api import OpenaiAPI
|
||
|
||
def read_markdown_file(file_path):
|
||
"""
|
||
读取 Markdown 文件并返回其全部内容作为字符串。
|
||
|
||
参数:
|
||
file_path (str): Markdown 文件的路径（支持相对或绝对路径）
|
||
|
||
返回:
|
||
str 或 None: 成功时返回文件内容；出错时返回 None
|
||
"""
|
||
try:
|
||
with open(file_path, 'r', encoding='utf-8') as f:
|
||
content = f.read()
|
||
return content
|
||
except FileNotFoundError:
|
||
print(f"错误：文件未找到 - {file_path}")
|
||
except PermissionError:
|
||
print(f"错误：没有权限读取文件 - {file_path}")
|
||
except UnicodeDecodeError:
|
||
print(f"错误：文件编码不是 UTF-8，无法读取 - {file_path}")
|
||
except Exception as e:
|
||
print(f"读取文件时发生未知错误: {e}")
|
||
return None
|
||
|
||
|
||
def safe_json_loads(s: str):
|
||
try:
|
||
return json.loads(s)
|
||
except json.JSONDecodeError as e:
|
||
# 尝试修复常见问题
|
||
s = s.replace("\\n", "\\\n").replace("\r", "\\r").replace("\t", "\\t")
|
||
# 或者直接清理非法字符
|
||
s = re.sub(r'[\x00-\x1f\x7f]', '', s) # 移除控制字符
|
||
try:
|
||
return json.loads(s)
|
||
except:
|
||
raise e
|
||
|
||
def extract_entity_relation(chunks, skip_first=False):
|
||
"""
|
||
批量处理一个文档的多个切片（chunks），每个切片调用大模型结构化，并返回全部结果。
|
||
|
||
:param chunks: List[str] - 同一文档的多个文本切片
|
||
:param skip_first: bool - 是否跳过第一个表格块（默认True）
|
||
:return: List[Dict] - 结构化结果列表
|
||
"""
|
||
all_results = []
|
||
all_entities = []
|
||
all_relationships = []
|
||
# 确定起始索引
|
||
start_idx = 1 if skip_first else 0
|
||
|
||
for idx, block in enumerate(chunks, 1):
|
||
# 跳过第一个表格
|
||
if skip_first and idx == 1:
|
||
print(f"=== 跳过第 {idx} 个表格块（首个表格） ===\n")
|
||
continue
|
||
|
||
print(f"=== 操作项目 Block {idx} ===")
|
||
print(block[:100]) # 只打印前500字符
|
||
print("\n" + "="*60 + "\n")
|
||
|
||
final_prompt = prompt_extract_operation_node_rela.format(text=block)
|
||
|
||
raw_response = OpenaiAPI.openai_chat(final_prompt,timeout=180)
|
||
|
||
if raw_response is None:
|
||
print(f" ⚠️ 第 {idx} 切片 API 调用失败")
|
||
continue
|
||
|
||
|
||
# 尝试解析 JSON
|
||
try:
|
||
cleaned = raw_response.strip()
|
||
if cleaned.startswith("```json"):
|
||
cleaned = cleaned[7:].lstrip()
|
||
if cleaned.endswith("```"):
|
||
cleaned = cleaned[:-3].rstrip()
|
||
|
||
json_result = safe_json_loads(cleaned)
|
||
entities = json_result.get("entities", [])
|
||
relationships = json_result.get("relationships", [])
|
||
for entity in entities:
|
||
# 确保 entity 有 "properties" 键，且其值为字典
|
||
if "properties" in entity and isinstance(entity["properties"], dict):
|
||
entity["properties"]["切片"] = block
|
||
# 打印美化版
|
||
all_entities.extend(entities)
|
||
all_relationships.extend(relationships)
|
||
# 将当前结果追加到总列表
|
||
# all_results.append({
|
||
# "entities": entities, # 已添加切片属性的实体列表
|
||
# "relationships": relationships, # 关系列表
|
||
# })
|
||
|
||
|
||
except Exception as e:
|
||
print(f" ⚠️ 第 {idx} 切片 JSON 解析失败: {e}")
|
||
print(f" 原始响应: {repr(raw_response[:500])}")
|
||
json_result = []
|
||
continue
|
||
result = {
|
||
"entities": all_entities,
|
||
"relationships": all_relationships
|
||
}
|
||
|
||
return result
|
||
|
||
|
||
|
||
|
||
|
||
def get_operation_node_rela(mdcontent):
|
||
print("开始提取操作项目表实体和关系")
|
||
chunks_repair, cleaned_content = extract_repairproject_tables_html(mdcontent)
|
||
chunks_operation, cleaned_content_operation = extract_operation_tables_html(cleaned_content)
|
||
|
||
|
||
if len(chunks_operation) > 0:
|
||
print(f"找到 {len(chunks_operation)} 个操作项目表格块")
|
||
result = extract_entity_relation(chunks=chunks_operation)
|
||
return result
|
||
else:
|
||
return []
|
||
|
||
|
||
|
||
# if __name__ == "__main__":
|
||
# input_directory = r"F:\zklnlp\HJ\hj_kg_code\kgrag\table_extract\files\操作使用手册_shenghna_output11_full_content.md"
|
||
|
||
# # 调用目录处理函数
|
||
# mdcontent = read_markdown_file(input_directory)
|
||
|
||
# if mdcontent is None:
|
||
# print("无法读取文件，程序退出")
|
||
# exit(1)
|
||
|
||
# res = get_operation_node_rela(mdcontent=mdcontent)
|
||
# print(len(res))
|