Spaces:
Running
Running
File size: 10,252 Bytes
500cf95 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 |
"""
PDF Parser Service for RAG Chatbot
Extracts text from PDF and splits into chunks for indexing
"""
import pypdfium2 as pdfium
from typing import List, Dict, Optional
import re
from dataclasses import dataclass
@dataclass
class PDFChunk:
"""Represents a chunk of text from PDF"""
text: str
page_number: int
chunk_index: int
metadata: Dict
class PDFParser:
"""Parse PDF files and prepare for RAG indexing"""
def __init__(
self,
chunk_size: int = 500, # words per chunk
chunk_overlap: int = 50, # words overlap between chunks
min_chunk_size: int = 50 # minimum words in a chunk
):
self.chunk_size = chunk_size
self.chunk_overlap = chunk_overlap
self.min_chunk_size = min_chunk_size
def extract_text_from_pdf(self, pdf_path: str) -> Dict[int, str]:
"""
Extract text from PDF file
Args:
pdf_path: Path to PDF file
Returns:
Dictionary mapping page number to text content
"""
pdf_text = {}
try:
pdf = pdfium.PdfDocument(pdf_path)
for page_num in range(len(pdf)):
page = pdf[page_num]
textpage = page.get_textpage()
text = textpage.get_text_range()
# Clean text
text = self._clean_text(text)
pdf_text[page_num + 1] = text # 1-indexed pages
return pdf_text
except Exception as e:
raise Exception(f"Error reading PDF: {str(e)}")
def _clean_text(self, text: str) -> str:
"""Clean extracted text"""
# Remove excessive whitespace
text = re.sub(r'\s+', ' ', text)
# Remove special characters that might cause issues
text = text.replace('\x00', '')
return text.strip()
def chunk_text(self, text: str, page_number: int) -> List[PDFChunk]:
"""
Split text into overlapping chunks
Args:
text: Text to chunk
page_number: Page number this text came from
Returns:
List of PDFChunk objects
"""
# Split into words
words = text.split()
if len(words) < self.min_chunk_size:
# Text too short, return as single chunk
if len(words) > 0:
return [PDFChunk(
text=text,
page_number=page_number,
chunk_index=0,
metadata={'page': page_number, 'chunk': 0}
)]
return []
chunks = []
chunk_index = 0
start = 0
while start < len(words):
# Get chunk
end = min(start + self.chunk_size, len(words))
chunk_words = words[start:end]
chunk_text = ' '.join(chunk_words)
chunks.append(PDFChunk(
text=chunk_text,
page_number=page_number,
chunk_index=chunk_index,
metadata={
'page': page_number,
'chunk': chunk_index,
'start_word': start,
'end_word': end
}
))
chunk_index += 1
# Move start position with overlap
start = end - self.chunk_overlap
# Avoid infinite loop
if start >= len(words) - self.min_chunk_size:
break
return chunks
def parse_pdf(
self,
pdf_path: str,
document_metadata: Optional[Dict] = None
) -> List[PDFChunk]:
"""
Parse entire PDF into chunks
Args:
pdf_path: Path to PDF file
document_metadata: Additional metadata for the document
Returns:
List of all chunks from the PDF
"""
# Extract text from all pages
pages_text = self.extract_text_from_pdf(pdf_path)
# Chunk each page
all_chunks = []
for page_num, text in pages_text.items():
chunks = self.chunk_text(text, page_num)
# Add document metadata
if document_metadata:
for chunk in chunks:
chunk.metadata.update(document_metadata)
all_chunks.extend(chunks)
return all_chunks
def parse_pdf_bytes(
self,
pdf_bytes: bytes,
document_metadata: Optional[Dict] = None
) -> List[PDFChunk]:
"""
Parse PDF from bytes (for uploaded files)
Args:
pdf_bytes: PDF file as bytes
document_metadata: Additional metadata
Returns:
List of chunks
"""
import tempfile
import os
# Save to temp file
with tempfile.NamedTemporaryFile(delete=False, suffix='.pdf') as tmp:
tmp.write(pdf_bytes)
tmp_path = tmp.name
try:
chunks = self.parse_pdf(tmp_path, document_metadata)
return chunks
finally:
# Clean up temp file
if os.path.exists(tmp_path):
os.unlink(tmp_path)
def get_pdf_info(self, pdf_path: str) -> Dict:
"""
Get basic info about PDF
Args:
pdf_path: Path to PDF file
Returns:
Dictionary with PDF information
"""
try:
pdf = pdfium.PdfDocument(pdf_path)
info = {
'num_pages': len(pdf),
'file_path': pdf_path,
}
return info
except Exception as e:
raise Exception(f"Error reading PDF info: {str(e)}")
class PDFIndexer:
"""Index PDF chunks into RAG system"""
def __init__(self, embedding_service, qdrant_service, documents_collection):
self.embedding_service = embedding_service
self.qdrant_service = qdrant_service
self.documents_collection = documents_collection
self.parser = PDFParser()
def index_pdf(
self,
pdf_path: str,
document_id: str,
document_metadata: Optional[Dict] = None
) -> Dict:
"""
Index entire PDF into RAG system
Args:
pdf_path: Path to PDF file
document_id: Unique ID for this document
document_metadata: Additional metadata (title, author, etc.)
Returns:
Indexing results
"""
# Parse PDF
chunks = self.parser.parse_pdf(pdf_path, document_metadata)
# Index each chunk
indexed_count = 0
chunk_ids = []
for chunk in chunks:
# Generate unique ID for chunk
chunk_id = f"{document_id}_p{chunk.page_number}_c{chunk.chunk_index}"
# Generate embedding
embedding = self.embedding_service.encode_text(chunk.text)
# Prepare metadata
metadata = {
'text': chunk.text,
'document_id': document_id,
'page': chunk.page_number,
'chunk_index': chunk.chunk_index,
'source': 'pdf',
**chunk.metadata
}
# Index to Qdrant
self.qdrant_service.index_data(
doc_id=chunk_id,
embedding=embedding,
metadata=metadata
)
chunk_ids.append(chunk_id)
indexed_count += 1
# Save document info to MongoDB
doc_info = {
'document_id': document_id,
'type': 'pdf',
'file_path': pdf_path,
'num_chunks': indexed_count,
'chunk_ids': chunk_ids,
'metadata': document_metadata or {},
'pdf_info': self.parser.get_pdf_info(pdf_path)
}
self.documents_collection.insert_one(doc_info)
return {
'success': True,
'document_id': document_id,
'chunks_indexed': indexed_count,
'chunk_ids': chunk_ids[:5] # Return first 5 as sample
}
def index_pdf_bytes(
self,
pdf_bytes: bytes,
document_id: str,
filename: str,
document_metadata: Optional[Dict] = None
) -> Dict:
"""
Index PDF from bytes (for uploaded files)
Args:
pdf_bytes: PDF file as bytes
document_id: Unique ID for this document
filename: Original filename
document_metadata: Additional metadata
Returns:
Indexing results
"""
# Parse PDF
metadata = document_metadata or {}
metadata['filename'] = filename
chunks = self.parser.parse_pdf_bytes(pdf_bytes, metadata)
# Index each chunk
indexed_count = 0
chunk_ids = []
for chunk in chunks:
# Generate unique ID for chunk
chunk_id = f"{document_id}_p{chunk.page_number}_c{chunk.chunk_index}"
# Generate embedding
embedding = self.embedding_service.encode_text(chunk.text)
# Prepare metadata
metadata = {
'text': chunk.text,
'document_id': document_id,
'page': chunk.page_number,
'chunk_index': chunk.chunk_index,
'source': 'pdf',
'filename': filename,
**chunk.metadata
}
# Index to Qdrant
self.qdrant_service.index_data(
doc_id=chunk_id,
embedding=embedding,
metadata=metadata
)
chunk_ids.append(chunk_id)
indexed_count += 1
# Save document info to MongoDB
doc_info = {
'document_id': document_id,
'type': 'pdf',
'filename': filename,
'num_chunks': indexed_count,
'chunk_ids': chunk_ids,
'metadata': metadata
}
self.documents_collection.insert_one(doc_info)
return {
'success': True,
'document_id': document_id,
'filename': filename,
'chunks_indexed': indexed_count,
'chunk_ids': chunk_ids[:5]
}
|