File size: 19,277 Bytes
d82600f
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
from langchain_community.document_loaders import PyPDFLoader, PDFPlumberLoader, UnstructuredPDFLoader, PyMuPDFLoader
from langchain_community.document_loaders import UnstructuredPowerPointLoader
from langchain_community.document_loaders import UnstructuredWordDocumentLoader
from langchain_community.document_loaders import TextLoader
from langchain.schema import Document
import os
import fitz  # PyMuPDF
import pytesseract
from PIL import Image
import io
import numpy as np
import cv2
from pdf2image import convert_from_path
import tempfile
import shutil

class DocumentLoader:
    def __init__(self, file_path: str):
        self.file_path = file_path
        self.extension = os.path.splitext(file_path)[1].lower()

    def load(self):
        """Load documents with enhanced PDF processing for scanned documents"""
        try:
            if self.extension == ".pdf":
                return self._load_pdf_with_ocr()
            elif self.extension == ".pptx":
                return UnstructuredPowerPointLoader(self.file_path).load()
            elif self.extension == ".docx":
                return UnstructuredWordDocumentLoader(self.file_path).load()
            elif self.extension == ".txt":
                return TextLoader(self.file_path).load()
            else:
                raise ValueError(f"Unsupported file type: {self.extension}")
        except Exception as e:
            print(f"[ERROR] Document loading failed for {self.file_path}: {e}")
            # Return a basic error document
            return [Document(
                page_content=f"Error loading document: {str(e)}. Please ensure the file is not corrupted and is in a supported format.",
                metadata={"page": 1, "source": self.file_path, "error": str(e)}
            )]

    def _load_pdf_with_ocr(self):
        """Enhanced PDF loading with OCR support for scanned documents"""
        try:
            # First, try to extract text using PyMuPDF (most reliable for text-based PDFs)
            print(f"[INFO] Attempting to extract text using PyMuPDF...")
            documents = self._extract_text_with_pymupdf()
            
            # Check if we got meaningful text content
            total_text = " ".join([doc.page_content for doc in documents])
            if len(total_text.strip()) > 50:  # If we have substantial text, use it
                print(f"[INFO] Successfully extracted {len(total_text)} characters using PyMuPDF")
                return documents
            
            # If text extraction failed or returned minimal content, try OCR
            print(f"[INFO] Text extraction returned minimal content ({len(total_text)} chars). Attempting OCR...")
            documents = self._extract_text_with_ocr()
            
            if documents:
                total_text = " ".join([doc.page_content for doc in documents])
                print(f"[INFO] Successfully extracted {len(total_text)} characters using OCR")
                return documents
            
            # If OCR also fails, try other PDF loaders as fallback
            print(f"[INFO] OCR failed. Trying alternative PDF loaders...")
            documents = self._try_alternative_pdf_loaders()
            
            if documents:
                total_text = " ".join([doc.page_content for doc in documents])
                print(f"[INFO] Successfully extracted {len(total_text)} characters using alternative loaders")
                return documents
            
            # If all methods fail, create a placeholder document with instructions
            print(f"[WARNING] All text extraction methods failed. Creating placeholder document.")
            return [Document(
                page_content="This appears to be a scanned document or image-based PDF. To enable full text extraction, please install Tesseract OCR. For now, you can still use the document for basic operations.",
                metadata={"page": 1, "source": self.file_path, "method": "placeholder"}
            )]
            
        except Exception as e:
            print(f"[ERROR] PDF processing failed: {e}")
            # Final fallback to basic PDF loader
            return PyPDFLoader(self.file_path).load()

    def _extract_text_with_pymupdf(self):
        """Extract text using PyMuPDF (handles most PDF types well)"""
        try:
            doc = fitz.open(self.file_path)
            documents = []
            
            for page_num in range(len(doc)):
                page = doc.load_page(page_num)
                
                # Try to extract text
                text = page.get_text()
                
                # If text is empty or very short, try to get text with more options
                if not text or len(text.strip()) < 10:
                    text = page.get_text("text")
                
                # If still no text, try to get text with layout preservation
                if not text or len(text.strip()) < 10:
                    text = page.get_text("dict")
                    # Extract text from the dict structure
                    if "blocks" in text:
                        text_content = []
                        for block in text["blocks"]:
                            if "lines" in block:
                                for line in block["lines"]:
                                    for span in line["spans"]:
                                        text_content.append(span["text"])
                        text = " ".join(text_content)
                
                if text and len(text.strip()) > 0:
                    documents.append(Document(
                        page_content=text.strip(),
                        metadata={"page": page_num + 1, "source": self.file_path}
                    ))
            
            doc.close()
            return documents
            
        except Exception as e:
            print(f"[WARNING] PyMuPDF extraction failed: {e}")
            return []

    def _extract_text_with_ocr(self):
        """Extract text from scanned PDFs using OCR"""
        try:
            # Check if Tesseract is available and configure it
            try:
                import pytesseract
                
                # Set Tesseract executable path explicitly
                tesseract_path = r"C:\Program Files\Tesseract-OCR\tesseract.exe"
                if os.path.exists(tesseract_path):
                    pytesseract.pytesseract.tesseract_cmd = tesseract_path
                    print(f"[INFO] Tesseract found at: {tesseract_path}")
                else:
                    # Try to find tesseract in PATH
                    import subprocess
                    try:
                        result = subprocess.run(['tesseract', '--version'], capture_output=True, text=True)
                        if result.returncode == 0:
                            print("[INFO] Tesseract found in PATH")
                        else:
                            raise Exception("Tesseract not found in PATH")
                    except:
                        raise Exception("Tesseract executable not found")
                
                # Test if tesseract is working
                version = pytesseract.get_tesseract_version()
                print(f"[INFO] Tesseract version: {version}")
                
            except Exception as e:
                print(f"[WARNING] Tesseract not available: {e}")
                print("[INFO] Skipping OCR - Tesseract needs to be installed for OCR functionality")
                return []
            
            # Convert PDF to images
            print(f"[INFO] Converting PDF to images for OCR...")
            # Specify the Poppler path explicitly
            poppler_path = r"C:\poppler\poppler-23.11.0\Library\bin"
            images = convert_from_path(self.file_path, dpi=300, poppler_path=poppler_path)
            
            documents = []
            
            for page_num, image in enumerate(images):
                print(f"[INFO] Processing page {page_num + 1} with OCR...")
                
                # Convert PIL image to OpenCV format for preprocessing
                img_cv = cv2.cvtColor(np.array(image), cv2.COLOR_RGB2BGR)
                
                # Preprocess image for better OCR (returns multiple versions)
                processed_images = self._preprocess_image_for_ocr(img_cv)
                
                # Convert all processed images to PIL format
                pil_images = []
                for processed_img in processed_images:
                    try:
                        pil_img = Image.fromarray(processed_img)
                        pil_images.append(pil_img)
                    except:
                        # If conversion fails, use original image
                        pil_images.append(image)
                
                # Perform OCR with multiple attempts and configurations
                best_text = ""
                best_length = 0
                
                # OCR configurations to try (in order of preference)
                ocr_configs = [
                    # Default configuration
                    {"config": "--oem 3 --psm 6", "name": "default"},
                    # Single uniform block of text
                    {"config": "--oem 3 --psm 6 -c tessedit_char_whitelist=ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz0123456789.,!?;:()[]{}'\"- ", "name": "alphanumeric"},
                    # Sparse text with OSD
                    {"config": "--oem 3 --psm 3", "name": "sparse_text"},
                    # Single text line
                    {"config": "--oem 3 --psm 7", "name": "single_line"},
                    # Single word
                    {"config": "--oem 3 --psm 8", "name": "single_word"},
                    # Single word in a circle
                    {"config": "--oem 3 --psm 9", "name": "circular_text"},
                    # Single character
                    {"config": "--oem 3 --psm 10", "name": "single_char"},
                    # Sparse text
                    {"config": "--oem 3 --psm 11", "name": "sparse_text_alt"},
                    # Raw line
                    {"config": "--oem 3 --psm 12", "name": "raw_line"},
                    # Uniform block of text
                    {"config": "--oem 3 --psm 13", "name": "uniform_block"}
                ]
                
                try:
                    # Try OCR on all preprocessed images with all configurations
                    for img_idx, pil_image in enumerate(pil_images):
                        for config in ocr_configs:
                            try:
                                text = pytesseract.image_to_string(
                                    pil_image, 
                                    config=config["config"],
                                    lang='eng'  # Specify English language
                                )
                                
                                # Clean the text
                                cleaned_text = self._clean_ocr_text(text)
                                
                                # Check if this configuration produced better results
                                if len(cleaned_text.strip()) > best_length:
                                    best_text = cleaned_text
                                    best_length = len(cleaned_text.strip())
                                    print(f"[INFO] Better OCR result with image {img_idx+1}, config {config['name']}: {best_length} characters")
                                
                            except Exception as config_error:
                                print(f"[DEBUG] OCR config {config['name']} failed for image {img_idx+1}: {config_error}")
                                continue
                    
                    # Use the best result
                    if best_text and len(best_text.strip()) > 10:
                        documents.append(Document(
                            page_content=best_text.strip(),
                            metadata={"page": page_num + 1, "source": self.file_path, "method": "OCR"}
                        ))
                        print(f"[INFO] OCR extracted {len(best_text)} characters from page {page_num + 1}")
                    else:
                        print(f"[WARNING] OCR returned minimal text for page {page_num + 1} (best: {best_length} chars)")
                        
                except Exception as e:
                    print(f"[WARNING] OCR failed for page {page_num + 1}: {e}")
                    continue
            
            return documents
            
        except Exception as e:
            print(f"[ERROR] OCR processing failed: {e}")
            return []

    def _preprocess_image_for_ocr(self, image):
        """Preprocess image for better OCR results"""
        try:
            # Convert to grayscale
            gray = cv2.cvtColor(image, cv2.COLOR_BGR2GRAY)
            
            # Apply noise reduction
            denoised = cv2.fastNlMeansDenoising(gray)
            
            # Try multiple preprocessing approaches
            processed_images = []
            
            # Approach 1: Adaptive thresholding
            try:
                thresh1 = cv2.adaptiveThreshold(
                    denoised, 255, cv2.ADAPTIVE_THRESH_GAUSSIAN_C, cv2.THRESH_BINARY, 11, 2
                )
                processed_images.append(thresh1)
            except:
                pass
            
            # Approach 2: Otsu thresholding
            try:
                _, thresh2 = cv2.threshold(denoised, 0, 255, cv2.THRESH_BINARY + cv2.THRESH_OTSU)
                processed_images.append(thresh2)
            except:
                pass
            
            # Approach 3: Simple thresholding
            try:
                _, thresh3 = cv2.threshold(denoised, 127, 255, cv2.THRESH_BINARY)
                processed_images.append(thresh3)
            except:
                pass
            
            # Approach 4: Original grayscale (sometimes works better)
            processed_images.append(denoised)
            
            # Apply morphological operations to clean up
            cleaned_images = []
            for img in processed_images:
                try:
                    # Small kernel for fine details
                    kernel_small = cv2.getStructuringElement(cv2.MORPH_RECT, (1, 1))
                    cleaned_small = cv2.morphologyEx(img, cv2.MORPH_CLOSE, kernel_small)
                    cleaned_images.append(cleaned_small)
                    
                    # Medium kernel for general cleaning
                    kernel_medium = cv2.getStructuringElement(cv2.MORPH_RECT, (2, 2))
                    cleaned_medium = cv2.morphologyEx(img, cv2.MORPH_CLOSE, kernel_medium)
                    cleaned_images.append(cleaned_medium)
                    
                except:
                    cleaned_images.append(img)
            
            # Return all processed images for testing
            return cleaned_images
            
        except Exception as e:
            print(f"[WARNING] Image preprocessing failed: {e}")
            return [image]

    def _clean_ocr_text(self, text):
        """Clean and improve OCR text"""
        if not text:
            return text
        
        # Remove excessive whitespace
        text = ' '.join(text.split())
        
        # Fix common OCR errors
        text = text.replace('|', 'I')  # Common OCR error
        text = text.replace('0', 'O')  # Sometimes numbers are confused with letters
        text = text.replace('1', 'l')  # Sometimes 1 is confused with l
        text = text.replace('l', 'I')  # Sometimes l is confused with I
        text = text.replace('rn', 'm')  # Common OCR error
        text = text.replace('cl', 'd')  # Common OCR error
        text = text.replace('vv', 'w')  # Common OCR error
        
        # Remove lines that are likely noise (very short lines)
        lines = text.split('\n')
        cleaned_lines = []
        for line in lines:
            line = line.strip()
            # Keep lines with more than 2 characters and not just punctuation
            if len(line) > 2 and not all(c in '.,!?;:()[]{}' for c in line):
                cleaned_lines.append(line)
        
        # Join lines and clean up
        result = '\n'.join(cleaned_lines)
        
        # Remove excessive newlines
        result = '\n'.join(line for line in result.split('\n') if line.strip())
        
        return result

    def _try_alternative_pdf_loaders(self):
        """Try alternative PDF loaders if primary methods fail"""
        loaders = [
            ("PDFPlumberLoader", lambda: PDFPlumberLoader(self.file_path).load()),
            ("UnstructuredPDFLoader", lambda: UnstructuredPDFLoader(self.file_path).load()),
            ("PyPDFLoader", lambda: PyPDFLoader(self.file_path).load())
        ]
        
        for loader_name, loader_func in loaders:
            try:
                print(f"[INFO] Trying {loader_name}...")
                documents = loader_func()
                total_text = " ".join([doc.page_content for doc in documents])
                if len(total_text.strip()) > 10:
                    print(f"[INFO] {loader_name} successfully extracted {len(total_text)} characters")
                    return documents
            except Exception as e:
                print(f"[WARNING] {loader_name} failed: {e}")
                continue
        
        return []

    def get_page_count(self):
        """Get page count for different document types"""
        if self.extension == ".pdf":
            try:
                # Try PyMuPDF first (most reliable)
                doc = fitz.open(self.file_path)
                page_count = len(doc)
                doc.close()
                return page_count
            except Exception:
                try:
                    # Fallback to PyPDF2
                    import PyPDF2
                    with open(self.file_path, "rb") as f:
                        reader = PyPDF2.PdfReader(f)
                        return len(reader.pages)
                except Exception:
                    return None
        elif self.extension == ".pptx":
            try:
                from pptx import Presentation
                prs = Presentation(self.file_path)
                return len(prs.slides)
            except Exception:
                return None
        elif self.extension == ".docx":
            try:
                from docx import Document as DocxDocument
                doc = DocxDocument(self.file_path)
                # DOCX doesn't have strict pages, but we can estimate by section breaks or paragraphs
                return max(1, len(doc.paragraphs) // 30)  # Rough estimate: 30 paragraphs per page
            except Exception:
                return None
        elif self.extension == ".txt":
            try:
                with open(self.file_path, "r", encoding="utf-8") as f:
                    words = f.read().split()
                    return max(1, len(words) // 500)
            except Exception:
                return None
        else:
            return None