From 3047c479165ce7ffb75458604204cca8e854bef1 Mon Sep 17 00:00:00 2001
From: zk <sim18502043706@163.com>
Date: Fri, 7 Aug 2026 16:12:46 +0800
Subject: [PATCH] =?UTF-8?q?pdf=E6=8F=90=E5=8F=96=E6=8E=92=E5=BA=8F?=
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

---
 app/tool/resume_text_extractor.py | 9 +++++++--
 1 file changed, 7 insertions(+), 2 deletions(-)

diff --git a/app/tool/resume_text_extractor.py b/app/tool/resume_text_extractor.py
index 1a69b72..7ae49a7 100644
--- a/app/tool/resume_text_extractor.py
+++ b/app/tool/resume_text_extractor.py
@@ -37,13 +37,18 @@ async def _parse_txt(content: bytes) -> str:
 
 
 async def _parse_pdf(content: bytes) -> str:
-    """解析 PDF：使用 PyMuPDF 逐页 get_text() 拼接为全文文本"""
+    """解析 PDF：使用 PyMuPDF 逐页 get_text() 拼接为全文文本
+
+    sort=True 按文本块坐标(y,x)排序输出，即视觉阅读顺序。
+    默认的内容流顺序对模板生成的简历完全不可用：这类简历的「公司/职位/时间」表头是绝对定位文本框，
+    会被整批甩到全文最前面，与各自的职责描述相隔几十行，AI 无法还原归属关系。
+    """
 
     def _extract() -> str:
         parts: list[str] = []
         with fitz.open(stream=content, filetype="pdf") as doc:
             for page in doc:
-                parts.append(page.get_text())
+                parts.append(page.get_text("text", sort=True))
         return "\n".join(parts)
 
     return await asyncio.to_thread(_extract)