修改简历提取方式

app/ai/resume_extractor/extractor.py

@@ -1,7 +1,7 @@
 """简历两阶段并行提取
 
-第一阶段：5路并行提取主表信息 + 各子表标识名（极快，输出极短）。
-第二阶段：N路并行提取每条子表记录的详情，description 用字母编号引用原文。
+第一阶段：5路并行提取主表短字段 + 各子表标识名（极快，输出极短）。
+第二阶段：N+1路并行提取每条子表记录的详情（含description原文）+ 个人信息补充（skills/certificates/summary）。
 最终组装为与原方案完全一致的 dict 结构，上下游无感知。
 """
 
@@ -24,31 +24,6 @@ from app.tool.json_helper import parse_llm_json
 _LLM_MODEL = LLM.DOUBAO_LITE_32K
 
 
-# ==================== 文本编号 ====================
-
-def _gen_alpha(n: int):
-    """生成 n 个字母编号：a,b,...,z,aa,ab,...,az,ba,..."""
-    for i in range(n):
-        yield chr(ord('a') + i) if i < 26 else chr(ord('a') + (i // 26 - 1)) + chr(ord('a') + i % 26)
-
-
-def _number_lines(text: str) -> tuple[dict[str, str], str]:
-    """按换行分割、过滤空行、字母编号，返回 (字母→原文dict, 带编号文本)"""
-    raw_lines = [line for line in text.split("\n") if line.strip()]
-    alphas = list(_gen_alpha(len(raw_lines)))
-    line_map = dict(zip(alphas, raw_lines))
-    numbered = "\n".join(f"[{a}] {line}" for a, line in zip(alphas, raw_lines))
-    return line_map, numbered
-
-
-def _resolve_desc(line_map: dict[str, str], desc_str: str | None) -> list[str]:
-    """将逗号分隔的字母编号字符串解析为原文列表"""
-    if not desc_str or not isinstance(desc_str, str):
-        return []
-    keys = [k.strip() for k in desc_str.split(",") if k.strip()]
-    return [line_map[k] for k in keys if k in line_map]
-
-
 # ==================== LLM 调用工具 ====================
 
 def _build_chain(prompt: str):
@@ -77,9 +52,9 @@ _overview_project_chain = _build_chain(OVERVIEW_PROJECT_PROMPT)
 _overview_competition_chain = _build_chain(OVERVIEW_COMPETITION_PROMPT)
 
 
-async def _extract_overview(numbered_text: str) -> dict:
+async def _extract_overview(text: str) -> dict:
     """第一阶段：5路并行提取概览信息"""
-    inp = {"text": numbered_text}
+    inp = {"text": text}
     profile, edu_names, work_names, proj_names, comp_names = await asyncio.gather(
         _safe_invoke(_overview_profile_chain, inp, "概览-个人信息"),
         _safe_invoke(_overview_education_chain, inp, "概览-教育"),
@@ -99,36 +74,36 @@ async def _extract_overview(numbered_text: str) -> dict:
 
 # ==================== 第二阶段：详情 ====================
 
-async def _extract_detail(prompt_tpl: str, name: str, numbered_text: str, label: str) -> dict | None:
-    """单条子表记录详情提取：用 name 替换 prompt 中的 {name}，发送带编号全文"""
+async def _extract_detail(prompt_tpl: str, name: str, text: str, label: str) -> dict | None:
+    """单条子表记录详情提取：用 name 替换 prompt 中的 {name}"""
     prompt = prompt_tpl.replace("{name}", name)
     chain = _build_chain(prompt)
-    return await _safe_invoke(chain, {"text": numbered_text}, label)
+    return await _safe_invoke(chain, {"text": text}, label)
 
 
-async def _extract_all_details(overview: dict, numbered_text: str) -> dict:
-    """第二阶段：根据概览结果，N路并行提取所有子表记录详情 + 个人信息的skills/certificates/summary"""
+async def _extract_all_details(overview: dict, text: str) -> dict:
+    """第二阶段：根据概览结果，N+1路并行提取所有子表记录详情 + 个人信息补充"""
     tasks: list = []
-    task_meta: list[tuple[str, int]] = []  # (模块名, 索引) 用于结果归位
+    task_meta: list[tuple[str, int]] = []
 
-    # profile 的 skills/certificates/summaryLines
-    tasks.append(_extract_detail(DETAIL_PROFILE_PROMPT, "", numbered_text, "详情-个人信息补充"))
+    # profile 补充：skills/certificates/summary
+    tasks.append(_extract_detail(DETAIL_PROFILE_PROMPT, "", text, "详情-个人信息补充"))
     task_meta.append(("profile_extra", 0))
 
     for i, name in enumerate(overview["education"]):
-        tasks.append(_extract_detail(DETAIL_EDUCATION_PROMPT, name, numbered_text, f"详情-教育-{name}"))
+        tasks.append(_extract_detail(DETAIL_EDUCATION_PROMPT, name, text, f"详情-教育-{name}"))
         task_meta.append(("education", i))
     for i, name in enumerate(overview["work"]):
-        tasks.append(_extract_detail(DETAIL_WORK_PROMPT, name, numbered_text, f"详情-工作-{name}"))
+        tasks.append(_extract_detail(DETAIL_WORK_PROMPT, name, text, f"详情-工作-{name}"))
         task_meta.append(("work", i))
     for i, name in enumerate(overview["internship"]):
-        tasks.append(_extract_detail(DETAIL_INTERNSHIP_PROMPT, name, numbered_text, f"详情-实习-{name}"))
+        tasks.append(_extract_detail(DETAIL_INTERNSHIP_PROMPT, name, text, f"详情-实习-{name}"))
         task_meta.append(("internship", i))
     for i, name in enumerate(overview["project"]):
-        tasks.append(_extract_detail(DETAIL_PROJECT_PROMPT, name, numbered_text, f"详情-项目-{name}"))
+        tasks.append(_extract_detail(DETAIL_PROJECT_PROMPT, name, text, f"详情-项目-{name}"))
         task_meta.append(("project", i))
     for i, name in enumerate(overview["competition"]):
-        tasks.append(_extract_detail(DETAIL_COMPETITION_PROMPT, name, numbered_text, f"详情-竞赛-{name}"))
+        tasks.append(_extract_detail(DETAIL_COMPETITION_PROMPT, name, text, f"详情-竞赛-{name}"))
         task_meta.append(("competition", i))
 
     results = await asyncio.gather(*tasks)
@@ -140,22 +115,19 @@ async def _extract_all_details(overview: dict, numbered_text: str) -> dict:
 
 # ==================== 组装 ====================
 
-def _assemble(overview: dict, details: dict, line_map: dict[str, str]) -> dict:
+def _assemble(overview: dict, details: dict) -> dict:
     """将两阶段结果组装为与原方案一致的 dict 结构"""
     profile = overview["profile"]
-    # 合并第二阶段提取的 skills/certificates/summaryLines
     profile_extra = details.get("profile_extra", [{}])[0] if details.get("profile_extra") else {}
     profile["skills"] = profile_extra.get("skills") or []
     profile["certificates"] = profile_extra.get("certificates") or []
-    summary_str = profile_extra.get("summaryLines")
-    summary_texts = _resolve_desc(line_map, summary_str)
-    profile["summary"] = "\n".join(summary_texts) if summary_texts else None
+    profile["summary"] = profile_extra.get("summary")
     result = dict(profile)
     for module in ("education", "work", "internship", "project", "competition"):
         items = []
         for item in details.get(module, []):
-            desc_str = item.pop("descLines", None)
-            item["description"] = _resolve_desc(line_map, desc_str)
+            if not item.get("description"):
+                item["description"] = []
             items.append(item)
         result[module] = items
     return result
@@ -165,17 +137,14 @@ def _assemble(overview: dict, details: dict, line_map: dict[str, str]) -> dict:
 
 async def extract_all(text: str) -> dict:
     """两阶段并行提取简历，返回与原方案完全一致的结构化数据"""
-    line_map, numbered_text = _number_lines(text)
-    log.info(f"文本编号完成，共 {len(line_map)} 行")
-
     log.info("第一阶段：5路并行概览提取")
-    overview = await _extract_overview(numbered_text)
+    overview = await _extract_overview(text)
     log.info(f"概览完成 - 教育:{len(overview['education'])} 工作:{len(overview['work'])} 实习:{len(overview['internship'])} 项目:{len(overview['project'])} 竞赛:{len(overview['competition'])}")
 
     total = sum(len(overview[m]) for m in ("education", "work", "internship", "project", "competition"))
-    log.info(f"第二阶段：{total}路并行详情提取")
-    details = await _extract_all_details(overview, numbered_text)
+    log.info(f"第二阶段：{total + 1}路并行详情提取")
+    details = await _extract_all_details(overview, text)
 
-    result = _assemble(overview, details, line_map)
+    result = _assemble(overview, details)
     log.info("两阶段提取完成，数据组装完毕")
     return result