抽象json提取风格

app/ai/resume_diagnoser/diagnoser.py

@@ -1,9 +1,6 @@
 """简历诊断 AI 引擎：并行诊断 + 汇总评价"""
 
 import asyncio
-import re
-
-from json_repair import repair_json
 
 from langchain_core.output_parsers import StrOutputParser
 from langchain_core.prompts import ChatPromptTemplate
@@ -11,13 +8,7 @@ from langchain_core.prompts import ChatPromptTemplate
 from app.ai.models import LLM
 from app.ai.resume_diagnoser.prompts import DIAGNOSE_MODULE_PROMPT, SUMMARY_PROMPT, POLISH_PROMPT
 from app.core.logger import log
-
-
-def _parse_json(text: str) -> dict:
-    """解析 AI 输出的 JSON，自动去除 markdown 代码块包裹，容错处理"""
-    cleaned = re.sub(r"^```(?:json)?\s*\n?", "", text.strip())
-    cleaned = re.sub(r"\n?```\s*$", "", cleaned)
-    return repair_json(cleaned, return_objects=True)
+from app.tool.json_helper import parse_llm_json
 
 
 # 诊断链（StrOutputParser 拿原始文本，再手动解析 JSON，避免 markdown 代码块导致解析失败）
@@ -92,7 +83,7 @@ async def polish_content(module_type: str, reference_content: list[dict] | str |
     }
     try:
         raw = await _polish_chain.ainvoke(inp)
-        result = _parse_json(raw)
+        result = parse_llm_json(raw)
         if isinstance(result, list):
             return [str(item) for item in result]
         return [str(result)]
@@ -106,7 +97,7 @@ async def _safe_invoke(task: dict) -> dict:
     raw = ""
     try:
         raw = await _diagnose_chain.ainvoke(task)
-        return _parse_json(raw)
+        return parse_llm_json(raw)
     except Exception as e:
         log.warning(f"AI诊断[{task.get('module_type', '')}]失败: {e}\n原始输出: {raw[:500]}")
         return _empty_result()

app/ai/resume_extractor/extractor.py

@@ -1,9 +1,7 @@
 """简历并行提取：将完整简历文本拆分为5个AI任务并行提取"""
 
 import asyncio
-import re
 
-from json_repair import repair_json
 from langchain_core.output_parsers import StrOutputParser
 from langchain_core.prompts import ChatPromptTemplate
 
@@ -13,13 +11,7 @@ from app.ai.resume_extractor.prompts import (
     PROJECT_PROMPT, COMPETITION_PROMPT,
 )
 from app.core.logger import log
-
-
-def _parse_json(text: str) -> dict:
-    """解析 AI 输出的 JSON，自动去除 markdown 代码块包裹，容错处理"""
-    cleaned = re.sub(r"^```(?:json)?\s*\n?", "", text.strip())
-    cleaned = re.sub(r"\n?```\s*$", "", cleaned)
-    return repair_json(cleaned, return_objects=True)
+from app.tool.json_helper import parse_llm_json
 
 
 def _build_chain(prompt: str):
@@ -65,7 +57,7 @@ async def _safe_invoke(chain, inp: dict, label: str):
     """单个链调用，失败返回空"""
     try:
         raw = await chain.ainvoke(inp)
-        return _parse_json(raw)
+        return parse_llm_json(raw)
     except Exception as e:
         log.warning(f"AI提取[{label}]失败: {e}")
         return {} if "个人信息" in label else []

app/ai/skill_gap_analyzer/analyzer.py

@@ -5,9 +5,7 @@
 """
 
 import asyncio
-import re
 
-from json_repair import repair_json
 from langchain_core.output_parsers import StrOutputParser
 from langchain_core.prompts import ChatPromptTemplate
 
@@ -17,13 +15,7 @@ from app.ai.skill_gap_analyzer.prompts import (
     AGENT_PLAN_PROMPT, AGENT_MODULE_EDIT_PROMPT, AGENT_MODULE_ADD_PROMPT, MODULE_SCHEMAS,
 )
 from app.core.logger import log
-
-
-def _parse_json(text: str):
-    """解析 AI 输出的 JSON，自动去除 markdown 代码块包裹，容错处理"""
-    cleaned = re.sub(r"^```(?:json)?\s*\n?", "", text.strip())
-    cleaned = re.sub(r"\n?```\s*$", "", cleaned)
-    return repair_json(cleaned, return_objects=True)
+from app.tool.json_helper import parse_llm_json
 
 
 # ===== 差距分析 =====
@@ -39,7 +31,7 @@ async def analyze_skill_gap(skill_tags: list[str], resume_json: str) -> list[str
     """分析技能差距，返回缺失技能列表"""
     try:
         raw = await _skill_gap_chain.ainvoke({"skill_tags": str(skill_tags), "resume_json": resume_json})
-        result = _parse_json(raw)
+        result = parse_llm_json(raw)
         if isinstance(result, list):
             return [s for s in result if isinstance(s, str) and s in skill_tags]
         return skill_tags  # 解析异常降级：全部标记缺失
@@ -85,7 +77,7 @@ async def optimize_module(job_title: str, job_description: str, module_data: str
             "job_title": job_title, "job_description": job_description or "",
             "original_module_data": module_data,
         })
-        return _parse_json(raw)
+        return parse_llm_json(raw)
     except Exception as e:
         log.warning(f"AI优化经历模块失败: {e}")
         return None
@@ -109,7 +101,7 @@ async def plan_edit(job_title: str, job_description: str, resume_json: str,
             "resume_json": resume_json,
             "chat_history": chat_history, "instruction": instruction,
         })
-        result = _parse_json(raw)
+        result = parse_llm_json(raw)
         return result if isinstance(result, dict) else None
     except Exception as e:
         log.warning(f"AI规划失败: {e}")
@@ -135,7 +127,7 @@ async def execute_record_edit(job_title: str, job_description: str, instruction:
             "instruction": instruction, "chat_history": chat_history,
             "module_schema": module_schema, "record_data": record_data,
         })
-        return _parse_json(raw)
+        return parse_llm_json(raw)
     except Exception as e:
         log.warning(f"AI单条记录修改失败: {e}")
         return None
@@ -159,7 +151,7 @@ async def execute_record_add(job_title: str, job_description: str, instruction:
             "instruction": instruction, "chat_history": chat_history,
             "module_schema": module_schema,
         })
-        result = _parse_json(raw)
+        result = parse_llm_json(raw)
         return result if isinstance(result, dict) else None
     except Exception as e:
         log.warning(f"AI新增记录失败: {e}")

app/tool/json_helper.py

@@ -0,0 +1,15 @@
+"""AI 输出 JSON 解析工具
+
+将 LLM 返回的可能带 markdown 代码块包裹的文本解析为 Python 对象。
+"""
+
+import re
+
+from json_repair import repair_json
+
+
+def parse_llm_json(text: str):
+    """解析 AI 输出的 JSON，自动去除 markdown 代码块包裹，容错处理"""
+    cleaned = re.sub(r"^```(?:json)?\s*\n?", "", text.strip())
+    cleaned = re.sub(r"\n?```\s*$", "", cleaned)
+    return repair_json(cleaned, return_objects=True)