Spaces:
Running
Running
File size: 15,250 Bytes
0a4529c |
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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 |
# DEPENDENCIES
import re
from typing import List
from typing import Optional
from config.models import DocumentChunk
from config.settings import get_settings
from config.models import DocumentMetadata
from config.models import ChunkingStrategy
from config.logging_config import get_logger
from chunking.base_chunker import BaseChunker
from chunking.base_chunker import ChunkerConfig
from chunking.token_counter import TokenCounter
from chunking.overlap_manager import OverlapManager
# Setup Settings and Logging
logger = get_logger(__name__)
settings = get_settings()
class FixedChunker(BaseChunker):
"""
Fixed-size chunking strategy : Splits text into chunks of approximately equal token count with overlap
Best for:
- Small to medium documents (<50K tokens)
- Homogeneous content
- When simplicity is preferred
"""
def __init__(self, chunk_size: int = None, overlap: int = None, respect_sentence_boundaries: bool = True, min_chunk_size: int = 100):
"""
Initialize fixed chunker
Arguments:
----------
chunk_size { int } : Target tokens per chunk (default from settings)
overlap { int } : Overlap tokens between chunks (default from settings)
respect_sentence_boundaries { bool } : Try to break at sentence boundaries
min_chunk_size { int } : Minimum chunk size in tokens
"""
super().__init__(ChunkingStrategy.FIXED)
self.chunk_size = chunk_size or settings.FIXED_CHUNK_SIZE
self.overlap = overlap or settings.FIXED_CHUNK_OVERLAP
self.respect_sentence_boundaries = respect_sentence_boundaries
self.min_chunk_size = min_chunk_size
# Initialize token counter and overlap manager
self.token_counter = TokenCounter()
self.overlap_manager = OverlapManager(overlap_tokens = self.overlap)
# Validate parameters
if (self.overlap >= self.chunk_size):
raise ValueError(f"Overlap ({self.overlap}) must be less than chunk_size ({self.chunk_size})")
self.logger.info(f"Initialized FixedChunker: chunk_size={self.chunk_size}, overlap={self.overlap}, respect_boundaries={self.respect_sentence_boundaries}")
def chunk_text(self, text: str, metadata: Optional[DocumentMetadata] = None) -> List[DocumentChunk]:
"""
Chunk text into fixed-size pieces
Arguments:
----------
text { str } : Input text
metadata { DocumentMetaData } : Document metadata
Returns:
--------
{ list } : List of DocumentChunk objects
"""
if not text or not text.strip():
return []
document_id = metadata.document_id if metadata else "unknown"
# Split into sentences if respecting boundaries
if self.respect_sentence_boundaries:
chunks = self._chunk_with_sentence_boundaries(text = text,
document_id = document_id,
)
else:
chunks = self._chunk_without_boundaries(text = text,
document_id = document_id,
)
# Clean and validate
chunks = [c for c in chunks if (c.token_count >= self.min_chunk_size)]
# Use OverlapManager to add proper overlap
if ((len(chunks) > 1) and (self.overlap > 0)):
chunks = self.overlap_manager.add_overlap(chunks = chunks,
overlap_tokens = self.overlap,
)
self.logger.debug(f"Created {len(chunks)} fixed-size chunks")
return chunks
def _chunk_with_sentence_boundaries(self, text: str, document_id: str) -> List[DocumentChunk]:
"""
Chunk text respecting sentence boundaries
Arguments:
----------
text { str } : Input text
document_id { str } : Document ID
Returns:
--------
{ list } : List of chunks without overlap (overlap added later)
"""
# Split into sentences
sentences = self._split_sentences(text = text)
chunks = list()
current_sentences = list()
current_tokens = 0
start_char = 0
for sentence in sentences:
sentence_tokens = self.token_counter.count_tokens(text = sentence)
# If single sentence exceeds chunk_size, split it
if (sentence_tokens > self.chunk_size):
# Save current chunk if any
if current_sentences:
chunk_text = " ".join(current_sentences)
chunk = self._create_chunk(text = self._clean_chunk_text(chunk_text),
chunk_index = len(chunks),
document_id = document_id,
start_char = start_char,
end_char = start_char + len(chunk_text),
)
chunks.append(chunk)
current_sentences = list()
current_tokens = 0
start_char += len(chunk_text)
# Split long sentence and add as separate chunks
long_sentence_chunks = self._split_long_sentence(sentence = sentence,
document_id = document_id,
start_index = len(chunks),
start_char = start_char,
)
chunks.extend(long_sentence_chunks)
start_char += len(sentence)
continue
# Check if adding this sentence exceeds chunk_size
if (((current_tokens + sentence_tokens) > self.chunk_size) and current_sentences):
# Save current chunk WITHOUT overlap (overlap added later)
chunk_text = " ".join(current_sentences)
chunk = self._create_chunk(text = self._clean_chunk_text(chunk_text),
chunk_index = len(chunks),
document_id = document_id,
start_char = start_char,
end_char = start_char + len(chunk_text),
)
chunks.append(chunk)
# OverlapManager will handle the overlap here
current_sentences = [sentence]
current_tokens = sentence_tokens
start_char += len(chunk_text)
else:
# Add sentence to current chunk
current_sentences.append(sentence)
current_tokens += sentence_tokens
# Add final chunk
if current_sentences:
chunk_text = " ".join(current_sentences)
chunk = self._create_chunk(text = self._clean_chunk_text(chunk_text),
chunk_index = len(chunks),
document_id = document_id,
start_char = start_char,
end_char = start_char + len(chunk_text),
)
chunks.append(chunk)
return chunks
def _chunk_without_boundaries(self, text: str, document_id: str) -> List[DocumentChunk]:
"""
Chunk text without respecting boundaries (pure token-based)
Arguments:
----------
text { str } : Input text
document_id { str } : Document ID
Returns:
--------
{ list } : List of chunks WITHOUT overlap
"""
# Use token counter's split method
chunk_texts = self.token_counter.split_into_token_chunks(text,
chunk_size = self.chunk_size,
overlap = 0,
)
chunks = list()
current_pos = 0
for i, chunk_text in enumerate(chunk_texts):
chunk = self._create_chunk(text = self._clean_chunk_text(chunk_text),
chunk_index = i,
document_id = document_id,
start_char = current_pos,
end_char = current_pos + len(chunk_text),
)
chunks.append(chunk)
current_pos += len(chunk_text)
return chunks
def _split_sentences(self, text: str) -> List[str]:
"""
Split text into sentences
Arguments:
----------
text { str } : Input text
Returns:
--------
{ list } : List of sentences
"""
# Handle common abbreviations: Protect them temporarily
protected = text
abbreviations = ['Dr.', 'Mr.', 'Mrs.', 'Ms.', 'Jr.', 'Sr.', 'Prof.', 'Inc.', 'Ltd.', 'Corp.', 'Co.', 'vs.', 'etc.', 'e.g.', 'i.e.', 'Ph.D.', 'M.D.', 'B.A.', 'M.A.', 'U.S.', 'U.K.']
for abbr in abbreviations:
protected = protected.replace(abbr, abbr.replace('.', '<DOT>'))
# Split on sentence boundaries
# - Pattern: period/question/exclamation followed by space and capital letter
sentence_pattern = r'(?<=[.!?])\s+(?=[A-Z])'
sentences = re.split(sentence_pattern, protected)
# Restore abbreviations
sentences = [s.replace('<DOT>', '.').strip() for s in sentences]
# Filter empty
sentences = [s for s in sentences if s]
return sentences
def _split_long_sentence(self, sentence: str, document_id: str, start_index: int, start_char: int) -> List[DocumentChunk]:
"""
Split a sentence that's longer than chunk_size
Arguments:
----------
sentence { str } : Long sentence
document_id { str } : Document ID
start_index { str } : Starting chunk index
start_char { int } : Starting character position
Returns:
--------
{ list } : List of chunks
"""
# Split by commas, semicolons, or just by tokens
parts = re.split(r'[,;]', sentence)
chunks = list()
current_text = list()
current_tokens = 0
for part in parts:
part = part.strip()
if not part:
continue
part_tokens = self.token_counter.count_tokens(part)
if (((current_tokens + part_tokens) > self.chunk_size) and current_text):
# Save current chunk
chunk_text = " ".join(current_text)
chunk = self._create_chunk(text = self._clean_chunk_text(chunk_text),
chunk_index = start_index + len(chunks),
document_id = document_id,
start_char = start_char,
end_char = start_char + len(chunk_text),
)
chunks.append(chunk)
start_char += len(chunk_text)
current_text = []
current_tokens = 0
current_text.append(part)
current_tokens += part_tokens
# Add final part
if current_text:
chunk_text = " ".join(current_text)
chunk = self._create_chunk(text = self._clean_chunk_text(chunk_text),
chunk_index = start_index + len(chunks),
document_id = document_id,
start_char = start_char,
end_char = start_char + len(chunk_text),
)
chunks.append(chunk)
return chunks
def _get_overlap_sentences(self, sentences: List[str], overlap_tokens: int) -> List[str]:
"""
Get last few sentences that fit in overlap window
Arguments:
----------
sentences { list } : List of sentences
overlap_tokens { int } : Target overlap tokens
Returns:
--------
{ list } : List of overlap sentences
"""
overlap = list()
tokens = 0
# Add sentences from the end until we reach overlap size
for sentence in reversed(sentences):
sentence_tokens = self.token_counter.count_tokens(sentence)
if ((tokens + sentence_tokens) <= overlap_tokens):
overlap.insert(0, sentence)
tokens += sentence_tokens
else:
break
return overlap
@classmethod
def from_config(cls, config: ChunkerConfig) -> 'FixedChunker':
"""
Create FixedChunker from configuration
Arguments:
----------
config { ChunkerConfig } : ChunkerConfig object
Returns:
--------
FixedChunker instance
"""
return cls(chunk_size = config.chunk_size,
overlap = config.overlap,
respect_sentence_boundaries = config.respect_boundaries,
min_chunk_size = config.min_chunk_size,
) |