fix(knowledge): read PDF bytes via get_buffer for zip members

apps/common/handle/impl/pdf_split_handle.py

@@ -44,10 +44,11 @@ def check_links_in_pdf(doc):
 
 class PdfSplitHandle(BaseSplitHandle):
     def handle(self, file, pattern_list: List, with_filter: bool, limit: int, get_buffer, save_image):
+        # Prefer get_buffer so ZipExtFile members (no .chunks()) work the same as UploadedFile.
+        buffer = get_buffer(file)
         with tempfile.NamedTemporaryFile(delete=False) as temp_file:
             # 将上传的文件保存到临时文件中
-            for chunk in file.chunks():
-                temp_file.write(chunk)
+            temp_file.write(buffer)
             # 获取临时文件的路径
             temp_file_path = temp_file.name
 

apps/common/handle/impl/tests/test_pdf_zip_buffer.py

@@ -0,0 +1,55 @@
+"""Regression for MaxKB #7436: PDF members inside zip have no .chunks()."""
+
+from io import BytesIO
+
+
+class _ZipLikePdf:
+    """Minimal stand-in for zipfile.ZipExtFile (read-only, no chunks)."""
+
+    def __init__(self, name: str, data: bytes):
+        self.name = name
+        self._buf = BytesIO(data)
+
+    def read(self, *args, **kwargs):
+        return self._buf.read(*args, **kwargs)
+
+
+def test_pdf_handle_reads_via_get_buffer_without_chunks(monkeypatch, tmp_path):
+    import types
+    import sys
+
+    # Avoid importing the real fitz-heavy module graph; exercise the buffer write path.
+    from common.handle.impl import pdf_split_handle as mod
+
+    calls = {}
+
+    class FakeDoc:
+        def close(self):
+            pass
+
+    def fake_open(path):
+        calls["path"] = path
+        with open(path, "rb") as f:
+            calls["bytes"] = f.read()
+        return FakeDoc()
+
+    monkeypatch.setattr(mod.fitz, "open", fake_open)
+    monkeypatch.setattr(
+        mod.PdfSplitHandle,
+        "handle_toc",
+        staticmethod(lambda *a, **k: [{"title": "t", "content": "c"}]),
+    )
+
+    payload = b"%PDF-1.4 fake"
+    file_obj = _ZipLikePdf("doc.pdf", payload)
+
+    def get_buffer(f):
+        return f.read()
+
+    result = mod.PdfSplitHandle().handle(
+        file_obj, None, False, 500, get_buffer, lambda images: None
+    )
+
+    assert result["name"] == "doc.pdf"
+    assert calls["bytes"] == payload
+    assert not hasattr(file_obj, "chunks")

apps/common/handle/impl/zip_split_handle.py

@@ -7,6 +7,7 @@
     @desc:
 """
 import io
+import logging
 import os
 import re
 import uuid
@@ -29,6 +30,8 @@
 from dataset.models import Image
 from django.utils.translation import gettext_lazy as _
 
+max_kb = logging.getLogger("max_kb")
+
 
 class FileBufferHandle:
     buffer = None
@@ -138,8 +141,8 @@ def handle(self, file, pattern_list: List, with_filter: bool, limit: int, get_bu
                             result = [*result, *value]
                         else:
                             result.append(value)
-                    except Exception:
-                        pass
+                    except Exception as e:
+                        max_kb.warning(f"Zip member: {file}, split error: {e}")
             image_list = get_image_list(result, files)
             result = filter_image_file(result, image_list)
             image_mode_list = []