RAG_ChatBot / src /document_processor.py
Jialun He
1st version
11d9dfb
"""
Document processing module for parsing and chunking various document formats.
"""
import re
from pathlib import Path
from typing import Dict, List, Optional, Tuple, Any
import hashlib
import mimetypes
# Document parsing imports
import PyPDF2
from docx import Document as DocxDocument
from io import BytesIO
from .error_handler import DocumentProcessingError, validate_file_upload
class DocumentChunk:
"""Represents a chunk of processed document content."""
def __init__(
self,
content: str,
metadata: Dict[str, Any],
chunk_id: str = None
):
self.content = content.strip()
self.metadata = metadata
self.chunk_id = chunk_id or self._generate_chunk_id()
def _generate_chunk_id(self) -> str:
"""Generate unique chunk ID based on content hash."""
content_hash = hashlib.md5(self.content.encode()).hexdigest()[:8]
source = self.metadata.get("source", "unknown")
page = self.metadata.get("page", 0)
return f"{Path(source).stem}_{page}_{content_hash}"
def to_dict(self) -> Dict[str, Any]:
"""Convert chunk to dictionary representation."""
return {
"chunk_id": self.chunk_id,
"content": self.content,
"metadata": self.metadata
}
class DocumentProcessor:
"""Main document processing class supporting multiple formats."""
def __init__(self, config: Dict[str, Any]):
self.config = config
self.processing_config = config.get("processing", {})
self.chunk_size = self.processing_config.get("chunk_size", 512)
self.chunk_overlap = self.processing_config.get("chunk_overlap", 50)
self.min_chunk_size = self.processing_config.get("min_chunk_size", 100)
self.max_chunks_per_doc = self.processing_config.get("max_chunks_per_doc", 1000)
self.supported_formats = self.processing_config.get("supported_formats", ["pdf", "docx", "txt"])
def process_document(
self,
file_path: str,
filename: Optional[str] = None
) -> List[DocumentChunk]:
"""
Process a document and return list of chunks.
Args:
file_path: Path to the document file
filename: Optional original filename
Returns:
List of DocumentChunk objects
"""
# Validate file
max_size = self.config.get("app", {}).get("max_upload_size", 50) * 1024 * 1024
allowed_extensions = [f".{fmt}" for fmt in self.supported_formats]
validate_file_upload(file_path, max_size, allowed_extensions)
file_path = Path(file_path)
filename = filename or file_path.name
# Detect file type and extract text
try:
text_content, metadata = self._extract_text(file_path, filename)
if not text_content.strip():
raise DocumentProcessingError("Document appears to be empty or contains no extractable text")
# Create chunks
chunks = self._create_chunks(text_content, metadata)
if not chunks:
raise DocumentProcessingError("Failed to create any valid chunks from document")
if len(chunks) > self.max_chunks_per_doc:
raise DocumentProcessingError(
f"Document too large. Generated {len(chunks)} chunks, "
f"maximum allowed is {self.max_chunks_per_doc}"
)
return chunks
except Exception as e:
if isinstance(e, DocumentProcessingError):
raise
else:
raise DocumentProcessingError(f"Failed to process document: {str(e)}") from e
def _extract_text(self, file_path: Path, filename: str) -> Tuple[str, Dict[str, Any]]:
"""Extract text from document based on file type."""
file_extension = file_path.suffix.lower()
# Base metadata
metadata = {
"source": str(file_path),
"filename": filename,
"file_type": file_extension,
"file_size": file_path.stat().st_size
}
if file_extension == ".pdf":
text, pdf_metadata = self._extract_pdf_text(file_path)
metadata.update(pdf_metadata)
elif file_extension == ".docx":
text, docx_metadata = self._extract_docx_text(file_path)
metadata.update(docx_metadata)
elif file_extension == ".txt":
text, txt_metadata = self._extract_txt_text(file_path)
metadata.update(txt_metadata)
else:
raise DocumentProcessingError(f"Unsupported file format: {file_extension}")
return text, metadata
def _extract_pdf_text(self, file_path: Path) -> Tuple[str, Dict[str, Any]]:
"""Extract text from PDF file."""
try:
with open(file_path, "rb") as file:
pdf_reader = PyPDF2.PdfReader(file)
if len(pdf_reader.pages) == 0:
raise DocumentProcessingError("PDF file contains no pages")
text_parts = []
for page_num, page in enumerate(pdf_reader.pages):
try:
page_text = page.extract_text()
if page_text.strip():
text_parts.append(f"\n\n--- Page {page_num + 1} ---\n\n{page_text}")
except Exception as e:
# Log warning but continue with other pages
print(f"Warning: Could not extract text from page {page_num + 1}: {e}")
if not text_parts:
raise DocumentProcessingError("Could not extract any text from PDF")
# Extract metadata
pdf_metadata = {
"page_count": len(pdf_reader.pages),
"pdf_metadata": {}
}
if pdf_reader.metadata:
pdf_metadata["pdf_metadata"] = {
"title": pdf_reader.metadata.get("/Title", ""),
"author": pdf_reader.metadata.get("/Author", ""),
"subject": pdf_reader.metadata.get("/Subject", ""),
"creator": pdf_reader.metadata.get("/Creator", "")
}
return "\n".join(text_parts), pdf_metadata
except Exception as e:
if isinstance(e, DocumentProcessingError):
raise
else:
raise DocumentProcessingError(f"Failed to read PDF file: {str(e)}") from e
def _extract_docx_text(self, file_path: Path) -> Tuple[str, Dict[str, Any]]:
"""Extract text from DOCX file."""
try:
doc = DocxDocument(file_path)
# Extract paragraphs
paragraphs = []
for paragraph in doc.paragraphs:
text = paragraph.text.strip()
if text:
paragraphs.append(text)
# Extract tables
table_texts = []
for table in doc.tables:
table_data = []
for row in table.rows:
row_data = [cell.text.strip() for cell in row.cells if cell.text.strip()]
if row_data:
table_data.append(" | ".join(row_data))
if table_data:
table_texts.append("Table:\n" + "\n".join(table_data))
all_text = "\n\n".join(paragraphs + table_texts)
if not all_text.strip():
raise DocumentProcessingError("DOCX file contains no extractable text")
# Metadata
docx_metadata = {
"paragraph_count": len(paragraphs),
"table_count": len(table_texts)
}
# Core properties
if hasattr(doc, "core_properties"):
props = doc.core_properties
docx_metadata["docx_metadata"] = {
"title": props.title or "",
"author": props.author or "",
"subject": props.subject or "",
"created": str(props.created) if props.created else ""
}
return all_text, docx_metadata
except Exception as e:
if isinstance(e, DocumentProcessingError):
raise
else:
raise DocumentProcessingError(f"Failed to read DOCX file: {str(e)}") from e
def _extract_txt_text(self, file_path: Path) -> Tuple[str, Dict[str, Any]]:
"""Extract text from TXT file."""
try:
# Try different encodings
encodings = ["utf-8", "utf-8-sig", "latin1", "cp1252"]
text = None
encoding_used = None
for encoding in encodings:
try:
with open(file_path, "r", encoding=encoding) as file:
text = file.read()
encoding_used = encoding
break
except UnicodeDecodeError:
continue
if text is None:
raise DocumentProcessingError("Could not decode text file with any supported encoding")
if not text.strip():
raise DocumentProcessingError("Text file is empty")
# Basic text statistics
lines = text.split("\n")
txt_metadata = {
"encoding": encoding_used,
"line_count": len(lines),
"char_count": len(text)
}
return text, txt_metadata
except Exception as e:
if isinstance(e, DocumentProcessingError):
raise
else:
raise DocumentProcessingError(f"Failed to read text file: {str(e)}") from e
def _create_chunks(self, text: str, base_metadata: Dict[str, Any]) -> List[DocumentChunk]:
"""Create overlapping chunks from text."""
# Clean and normalize text
text = self._clean_text(text)
# Split into sentences for better chunk boundaries
sentences = self._split_into_sentences(text)
if not sentences:
return []
chunks = []
current_chunk = []
current_length = 0
for sentence in sentences:
sentence_length = len(sentence)
# If adding this sentence would exceed chunk size
if current_length + sentence_length > self.chunk_size and current_chunk:
# Create chunk from current sentences
chunk_text = " ".join(current_chunk)
if len(chunk_text) >= self.min_chunk_size:
chunk_metadata = {
**base_metadata,
"chunk_index": len(chunks),
"char_count": len(chunk_text),
"sentence_count": len(current_chunk)
}
chunks.append(DocumentChunk(chunk_text, chunk_metadata))
# Start new chunk with overlap
if self.chunk_overlap > 0:
overlap_sentences = self._get_overlap_sentences(current_chunk)
current_chunk = overlap_sentences
current_length = sum(len(s) for s in overlap_sentences)
else:
current_chunk = []
current_length = 0
# Add current sentence
current_chunk.append(sentence)
current_length += sentence_length
# Create final chunk
if current_chunk:
chunk_text = " ".join(current_chunk)
if len(chunk_text) >= self.min_chunk_size:
chunk_metadata = {
**base_metadata,
"chunk_index": len(chunks),
"char_count": len(chunk_text),
"sentence_count": len(current_chunk)
}
chunks.append(DocumentChunk(chunk_text, chunk_metadata))
return chunks
def _clean_text(self, text: str) -> str:
"""Clean and normalize text."""
# Remove excessive whitespace
text = re.sub(r'\s+', ' ', text)
# Remove page markers (from PDF extraction)
text = re.sub(r'\n--- Page \d+ ---\n', '\n', text)
# Fix common OCR errors and formatting issues
text = re.sub(r'([a-z])([A-Z])', r'\1 \2', text) # Add space between camelCase
text = re.sub(r'([.!?])([A-Z])', r'\1 \2', text) # Add space after punctuation
return text.strip()
def _split_into_sentences(self, text: str) -> List[str]:
"""Split text into sentences using simple heuristics."""
# Simple sentence splitting - can be enhanced with NLTK if needed
sentences = re.split(r'[.!?]+', text)
# Clean up sentences
cleaned_sentences = []
for sentence in sentences:
sentence = sentence.strip()
if len(sentence) >= 10: # Minimum sentence length
cleaned_sentences.append(sentence)
return cleaned_sentences
def _get_overlap_sentences(self, sentences: List[str]) -> List[str]:
"""Get sentences for overlap based on character count."""
overlap_sentences = []
overlap_length = 0
# Take sentences from the end up to the overlap size
for sentence in reversed(sentences):
if overlap_length + len(sentence) <= self.chunk_overlap:
overlap_sentences.insert(0, sentence)
overlap_length += len(sentence)
else:
break
return overlap_sentences
def get_document_stats(self, chunks: List[DocumentChunk]) -> Dict[str, Any]:
"""Get statistics about processed document."""
if not chunks:
return {"chunk_count": 0, "total_chars": 0, "avg_chunk_size": 0}
total_chars = sum(len(chunk.content) for chunk in chunks)
return {
"chunk_count": len(chunks),
"total_chars": total_chars,
"avg_chunk_size": total_chars / len(chunks),
"min_chunk_size": min(len(chunk.content) for chunk in chunks),
"max_chunk_size": max(len(chunk.content) for chunk in chunks),
"source_file": chunks[0].metadata.get("filename", "unknown"),
"file_type": chunks[0].metadata.get("file_type", "unknown")
}