fix: prevent timeout in file encoding detection for large files

- Add sample_size parameter to detect_file_encodings function
- Read only a sample of the file instead of the entire file for encoding detection
- This prevents timeout errors on large files while maintaining accurate encoding detection
- Default sample size is 1MB which is sufficient for most encoding detection scenarios

Fixes #21327

api/core/rag/extractor/helpers.py

@@ -1,7 +1,7 @@
 """Document loader helpers."""
 
 import concurrent.futures
-from pathlib import Path
+import os
 from typing import NamedTuple, Optional, cast
 
 
@@ -16,7 +16,7 @@ class FileEncoding(NamedTuple):
     """The language of the file."""
 
 
-def detect_file_encodings(file_path: str, timeout: int = 5) -> list[FileEncoding]:
+def detect_file_encodings(file_path: str, timeout: int = 5, sample_size: int = 1024 * 1024) -> list[FileEncoding]:
     """Try to detect the file encoding.
 
     Returns a list of `FileEncoding` tuples with the detected encodings ordered
@@ -25,11 +25,16 @@ def detect_file_encodings(file_path: str, timeout: int = 5) -> list[FileEncoding
     Args:
         file_path: The path to the file to detect the encoding for.
         timeout: The timeout in seconds for the encoding detection.
+        sample_size: The number of bytes to read for encoding detection. Default is 1MB.
+                    For large files, reading only a sample is sufficient and prevents timeout.
     """
     import chardet
 
     def read_and_detect(file_path: str) -> list[dict]:
-        rawdata = Path(file_path).read_bytes()
+        with open(file_path, "rb") as f:
+            # Read only a sample of the file for encoding detection
+            # This prevents timeout on large files while still providing accurate encoding detection
+            rawdata = f.read(sample_size)
         return cast(list[dict], chardet.detect_all(rawdata))
 
     with concurrent.futures.ThreadPoolExecutor() as executor:

api/core/rag/extractor/text_extractor.py

@@ -36,8 +36,16 @@ class TextExtractor(BaseExtractor):
                         break
                     except UnicodeDecodeError:
                         continue
+                else:
+                    raise RuntimeError(
+                        f"Decode failed: {self._file_path}, all detected encodings failed. "
+                        f"Original error: {e}"
+                    )
             else:
-                raise RuntimeError(f"Error loading {self._file_path}") from e
+                raise RuntimeError(
+                    f"Decode failed: {self._file_path}, specified encoding failed. "
+                    f"Original error: {e}"
+                )
         except Exception as e:
             raise RuntimeError(f"Error loading {self._file_path}") from e