chaulmark/DocuLens
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

0.5.2

Browse source 
Bug fixes.
This commit is contained in:
Drakonis96 2025-03-11 14:38:12 +01:00
parent ee77b2989f
commit a3ef26ff29
5 changed files with 132 additions and 117 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

BIN
.DS_Store vendored Normal file
View file

Binary file not shown.

6
Dockerfile
View file

@ -1,5 +1,5 @@
# Stage 1: Build the frontend # Stage 1: Build the frontend
FROM node:16-alpine as frontend-build FROM node:16-alpine AS frontend-build
WORKDIR /app/frontend WORKDIR /app/frontend
COPY frontend/package.json frontend/package-lock.json* ./ COPY frontend/package.json frontend/package-lock.json* ./
RUN npm install RUN npm install
@ -16,6 +16,10 @@ RUN apt-get update && apt-get install -y --no-install-recommends \
poppler-utils \ poppler-utils \
tesseract-ocr \ tesseract-ocr \
ghostscript \ ghostscript \
qpdf \
unpaper \
libxml2 \
libxslt1.1 \
libffi-dev \ libffi-dev \
libssl-dev \ libssl-dev \
&& rm -rf /var/lib/apt/lists/* && rm -rf /var/lib/apt/lists/*

6
README.md
View file

@ -1,6 +1,6 @@
<center> <div align="center">
<img src="logo.png" alt="ocrAI Logo" width="150"> <img src="/backend/static/logo.png" alt="ocrAI Logo" width="250">
</center> </div>
# ocrAI 🤖 # ocrAI 🤖

BIN
backend/.DS_Store vendored
View file

Binary file not shown.

233
backend/utils.py
View file

@ -29,8 +29,9 @@ def encode_image(file_path):
def run_tesseract(file_path): def run_tesseract(file_path):
""" """
If the file is a PDF, perform OCR page by page and add a "Page X:" header; Si el archivo es un PDF, se realiza OCR página a página añadiendo al principio de cada una
otherwise, perform OCR normally. la cabecera con el formato [Page 0001], [Page 0002], etc.
Para otros formatos se realiza OCR normal.
""" """
extracted_text = "" extracted_text = ""
if file_path.lower().endswith(".pdf"): if file_path.lower().endswith(".pdf"):
@ -38,7 +39,7 @@ def run_tesseract(file_path):
pages = convert_from_path(file_path) pages = convert_from_path(file_path)
for i, page in enumerate(pages, start=1): for i, page in enumerate(pages, start=1):
page_text = pytesseract.image_to_string(page, lang='eng') page_text = pytesseract.image_to_string(page, lang='eng')
extracted_text += f"Page {i}:\n{page_text}\n\n" extracted_text += f"[Page {i:04d}]\n{page_text}\n\n"
except Exception as e: except Exception as e:
extracted_text = f"❌ Error processing PDF: {str(e)}" extracted_text = f"❌ Error processing PDF: {str(e)}"
else: else:
@ -115,7 +116,7 @@ def ocr_file_by_pages(file_path, api, model, prompt_key, update_progress, is_can
temp_filename = os.path.join(OUTPUT_FOLDER, f"temp_page_{uuid.uuid4().hex}.png") temp_filename = os.path.join(OUTPUT_FOLDER, f"temp_page_{uuid.uuid4().hex}.png")
page.save(temp_filename, "PNG") page.save(temp_filename, "PNG")
page_text = call_api_ocr(api, model, temp_filename, prompt_key) page_text = call_api_ocr(api, model, temp_filename, prompt_key)
final_text += f"Page {i}:\n{page_text}\n\n" final_text += f"[Page {i:04d}]\n{page_text}\n\n"
os.remove(temp_filename) os.remove(temp_filename)
progress = int((i / total) * 100) progress = int((i / total) * 100)
update_progress(progress, f"📄 Processed page {i} of {total}.") update_progress(progress, f"📄 Processed page {i} of {total}.")
@ -140,7 +141,7 @@ def translate_file_by_pages(file_path, api, model, target_language, prompt_key,
page.save(temp_filename, "PNG") page.save(temp_filename, "PNG")
page_text = pytesseract.image_to_string(page, lang='eng') page_text = pytesseract.image_to_string(page, lang='eng')
translated_page = call_api_translation(api, model, page_text, target_language, prompt_key) translated_page = call_api_translation(api, model, page_text, target_language, prompt_key)
final_translation += f"Page {i}:\n{translated_page}\n\n" final_translation += f"[Page {i:04d}]\n{translated_page}\n\n"
os.remove(temp_filename) os.remove(temp_filename)
progress = int((i / total) * 100) progress = int((i / total) * 100)
update_progress(progress, f"📄 Processed page {i} of {total}.") update_progress(progress, f"📄 Processed page {i} of {total}.")
@ -151,11 +152,14 @@ def translate_file_by_pages(file_path, api, model, target_language, prompt_key,
text = f.read() text = f.read()
translated = call_api_translation(api, model, text, target_language, prompt_key) translated = call_api_translation(api, model, text, target_language, prompt_key)
update_progress(100, "🎉 Process completed") update_progress(100, "🎉 Process completed")
return f"Page 1:\n{translated}" return f"[Page {1:04d}]\n{translated}"
else: else:
return "Unsupported file type for translation." return "Unsupported file type for translation."
def process_file(file_path, api, model, mode, prompt_key, update_progress, is_cancelled): def process_file(file_path, api, model, mode, prompt_key, update_progress, is_cancelled):
"""
NOTA: Ahora solo usamos [Page XXXX] al inicio de cada página (por Tesseract).
"""
if is_cancelled(): if is_cancelled():
update_progress(0, "⏹️ Cancelled") update_progress(0, "⏹️ Cancelled")
return "Process cancelled." return "Process cancelled."
@ -163,13 +167,11 @@ def process_file(file_path, api, model, mode, prompt_key, update_progress, is_ca
base_name = os.path.splitext(os.path.basename(file_path))[0] base_name = os.path.splitext(os.path.basename(file_path))[0]
if mode == "OCR": if mode == "OCR":
# Process using tesseract with page-structure if PDF.
processed_text = run_tesseract(file_path) processed_text = run_tesseract(file_path)
if is_cancelled(): if is_cancelled():
update_progress(25, "⏹️ Cancelled") update_progress(25, "⏹️ Cancelled")
return "Process cancelled." return "Process cancelled."
update_progress(50, "✅ Tesseract OCR completed.") update_progress(50, "✅ Tesseract OCR completed.")
# Generate PDF copy as before.
pdf_output = os.path.join(OUTPUT_FOLDER, base_name + "_ocr.pdf") pdf_output = os.path.join(OUTPUT_FOLDER, base_name + "_ocr.pdf")
if file_path.lower().endswith(".pdf"): if file_path.lower().endswith(".pdf"):
if embed_ocr_in_pdf(file_path, pdf_output): if embed_ocr_in_pdf(file_path, pdf_output):
@ -179,6 +181,7 @@ def process_file(file_path, api, model, mode, prompt_key, update_progress, is_ca
update_progress(95, "⚠️ Failed to embed OCR; original PDF copied.") update_progress(95, "⚠️ Failed to embed OCR; original PDF copied.")
else: else:
shutil.copy(file_path, pdf_output) shutil.copy(file_path, pdf_output)
elif mode == "OCR + AI": elif mode == "OCR + AI":
processed_text = run_tesseract(file_path) processed_text = run_tesseract(file_path)
if is_cancelled(): if is_cancelled():
@ -195,13 +198,14 @@ def process_file(file_path, api, model, mode, prompt_key, update_progress, is_ca
update_progress(95, "⚠️ Failed to embed OCR; original PDF copied.") update_progress(95, "⚠️ Failed to embed OCR; original PDF copied.")
else: else:
shutil.copy(file_path, pdf_output) shutil.copy(file_path, pdf_output)
elif mode == "AI": elif mode == "AI":
update_progress(25, "📂 File ready for full AI processing.") update_progress(25, "📂 File ready for full AI processing.")
if file_path.lower().endswith(".pdf"): if file_path.lower().endswith(".pdf"):
processed_text = ocr_file_by_pages(file_path, api, model, prompt_key, update_progress, is_cancelled) processed_text = ocr_file_by_pages(file_path, api, model, prompt_key, update_progress, is_cancelled)
else: else:
processed_text = call_api_ocr(api, model, file_path, prompt_key) processed_text = call_api_ocr(api, model, file_path, prompt_key)
# In AI mode, do not generate a new PDF.
else: else:
processed_text = "Unrecognized processing mode." processed_text = "Unrecognized processing mode."
update_progress(25, "❌ Error: Unrecognized mode.") update_progress(25, "❌ Error: Unrecognized mode.")
@ -212,58 +216,90 @@ def process_file(file_path, api, model, mode, prompt_key, update_progress, is_ca
update_progress(75, "🤖 API processing completed.") update_progress(75, "🤖 API processing completed.")
# Write the output TXT file (it is already structured by page in OCR and OCR+AI modes) # Guardar el TXT final
txt_file = os.path.join(OUTPUT_FOLDER, base_name + ".txt") txt_file = os.path.join(OUTPUT_FOLDER, base_name + ".txt")
with open(txt_file, "w", encoding="utf-8") as f: with open(txt_file, "w", encoding="utf-8") as f:
f.write(processed_text) f.write(processed_text)
update_progress(100, "🎉 Process completed") update_progress(100, "🎉 Process completed")
return processed_text return processed_text
def process_text(text): def organize_paragraphs(text):
lines = text.splitlines() """
processed_lines = [] Organiza el texto plano en párrafos de forma más flexible.
buffer = "" Si se detecta doble salto de línea se usa como separador;
for line in lines: si no, se procesa línea a línea para unirlas en párrafos, creando uno nuevo cuando:
stripped = line.strip() - Se encuentra una línea vacía, o
if not stripped: - La línea actual termina en punto.
if buffer: Dentro de cada párrafo se unen las líneas; se inserta un <br/> si la línea termina en punto.
processed_lines.append(buffer) """
buffer = "" text = text.strip()
paragraphs = []
if "\n\n" in text:
blocks = re.split(r'\n\s*\n', text)
else:
lines = text.splitlines()
blocks = []
buffer = ""
for line in lines:
stripped = line.strip()
if not stripped:
if buffer:
blocks.append(buffer)
buffer = ""
else: else:
processed_lines.append("") if buffer and buffer.endswith("."):
continue blocks.append(buffer)
buffer = stripped
else:
if buffer:
buffer += " " + stripped
else:
buffer = stripped
if buffer: if buffer:
if buffer.endswith('.'): blocks.append(buffer)
processed_lines.append(buffer) for block in blocks:
buffer = stripped line_list = block.splitlines()
else: if len(line_list) == 1:
buffer += " " + stripped paragraphs.append(line_list[0].strip())
else: else:
buffer = stripped new_block = ""
if buffer: for i, line in enumerate(line_list):
processed_lines.append(buffer) line = line.strip()
return "\n".join(processed_lines) if not line:
continue
if i < len(line_list) - 1:
if line.endswith("."):
new_block += line + "<br/>"
else:
new_block += line + " "
else:
new_block += line
paragraphs.append(new_block.strip())
return paragraphs
def convert_txt_to_pdf(txt_file_path): def convert_txt_to_pdf(txt_file_path):
""" """
Se ha modificado para que, si el contenido del TXT está formateado en HTML, Convierte un archivo TXT a PDF siguiendo estas reglas:
se interpreten los tags que indiquen títulos (<h1>, <h2>, etc.), párrafos (<p>) - Se detecta el patrón [Page XXXX] para separar páginas (solo al inicio).
y saltos de página (<pagebreak> o div con clase "page-break"). - Este marcador se convierte en un encabezado (h1).
- Si el contenido proviene de un bloque markdown con ```html se elimina ese marcador y
En caso de que el contenido sea texto plano y contenga patrones en el formato se parsea con BeautifulSoup para generar párrafos independientes.
[Page X] (entre corchetes), se usará ese separador para dividir las páginas. - Para contenido en texto plano se organiza en párrafos con organize_paragraphs.
El encabezado (sin los corchetes) se incluirá en la parte superior de cada página. - Se inserta un PageBreak después de cada bloque de página.
""" """
with open(txt_file_path, "r", encoding="utf-8") as f: with open(txt_file_path, "r", encoding="utf-8") as f:
content = f.read() content = f.read()
base_name = os.path.splitext(os.path.basename(txt_file_path))[0] base_name = os.path.splitext(os.path.basename(txt_file_path))[0]
output_pdf = os.path.join(OUTPUT_FOLDER, base_name + "_txt.pdf") output_pdf = os.path.join(OUTPUT_FOLDER, base_name + "_txt.pdf")
doc = SimpleDocTemplate(output_pdf, pagesize=A4, doc = SimpleDocTemplate(
rightMargin=40, leftMargin=40, output_pdf,
topMargin=40, bottomMargin=40) pagesize=A4,
rightMargin=40, leftMargin=40,
topMargin=40, bottomMargin=40
)
styles = getSampleStyleSheet() styles = getSampleStyleSheet()
# Estilos para encabezados
header_styles = { header_styles = {
"h1": ParagraphStyle('Heading1', parent=styles['Heading1'], alignment=TA_CENTER), "h1": ParagraphStyle('Heading1', parent=styles['Heading1'], alignment=TA_CENTER),
"h2": ParagraphStyle('Heading2', parent=styles['Heading2'], alignment=TA_CENTER), "h2": ParagraphStyle('Heading2', parent=styles['Heading2'], alignment=TA_CENTER),
@ -277,82 +313,57 @@ def convert_txt_to_pdf(txt_file_path):
parent=styles['Normal'], parent=styles['Normal'],
alignment=TA_JUSTIFY, alignment=TA_JUSTIFY,
leading=15, leading=15,
leftIndent=20 # Sangría al inicio de cada párrafo leftIndent=20
) )
flowables = [] flowables = []
# Detectamos si el contenido es HTML (buscando etiquetas comunes) # Patrón para detectar el marcador [Page XXXX]
is_html = any(tag in content.lower() for tag in ["<html", "<p", "<h1", "<h2", "<h3"]) page_pattern = re.compile(r'\[Page\s+\d{4}\]')
parts = re.split(r'(\[Page\s+\d{4}\])', content)
first_page_encountered = False
if is_html: for part in parts:
soup = BeautifulSoup(content, "html.parser") part = part.strip()
body = soup.body if soup.body else soup if not part:
for element in body.children: continue
if element.name is None:
text = element.strip() if page_pattern.fullmatch(part):
if text: if first_page_encountered:
para = Paragraph(text, normal_style)
flowables.append(para)
flowables.append(Spacer(1, 12))
elif element.name.lower() in ["h1", "h2", "h3", "h4", "h5", "h6"]:
tag = element.name.lower()
style = header_styles.get(tag, styles['Heading1'])
para = Paragraph(element.decode_contents(), style)
flowables.append(para)
flowables.append(Spacer(1, 12))
elif element.name.lower() == "p":
para = Paragraph(element.decode_contents(), normal_style)
flowables.append(para)
flowables.append(Spacer(1, 12))
elif element.name.lower() == "pagebreak" or (element.name.lower() == "div" and "page-break" in element.get("class", [])):
flowables.append(PageBreak()) flowables.append(PageBreak())
else: else:
text = element.get_text().strip() first_page_encountered = True
if text: header_para = Paragraph(part, styles['Heading1'])
para = Paragraph(text, normal_style) flowables.append(header_para)
flowables.append(para) flowables.append(Spacer(1, 12))
flowables.append(Spacer(1, 12))
else:
# Si se detecta el patrón [Page X] en el contenido, se usa para dividir las páginas.
if re.search(r'\[Page\s+\d+\]', content, re.IGNORECASE):
# Dividir incluyendo el separador (usamos grupo de captura)
parts = re.split(r'(\[Page\s+\d+\])', content, flags=re.IGNORECASE)
current_header = ""
for part in parts:
part = part.strip()
if not part:
continue
# Si es un encabezado [Page X]
if re.match(r'\[Page\s+\d+\]', part, re.IGNORECASE):
# Extraemos el número o texto sin los corchetes para usarlo como encabezado
current_header = part.strip("[]")
header_para = Paragraph(current_header, styles['Heading1'])
flowables.append(header_para)
flowables.append(Spacer(1, 12))
else:
# Es el contenido de la página
processed = process_text(part)
para = Paragraph(processed.replace("\n", "<br/>"), normal_style)
flowables.append(para)
flowables.append(Spacer(1, 12))
flowables.append(PageBreak())
else: else:
# Si no se detectan separadores, se usa el método anterior basado en "Page X:" sin corchetes. if part.startswith("```"):
pages = re.split(r'(?i)Page\s+\d+:\s*', content) lines = part.splitlines()
if pages and pages[0].strip() == "": if lines and lines[0].startswith("```"):
pages = pages[1:] lines = lines[1:]
if len(pages) <= 1: if lines and lines[-1].strip() == "```":
pages = content.split("\n\n") lines = lines[:-1]
processed_pages = [process_text(page) for page in pages if page.strip() != ""] part = "\n".join(lines)
for i, page_text in enumerate(processed_pages, start=1): # Si parece HTML, se procesa para separar cada etiqueta de interés
header = Paragraph(f"PAGE {i}", styles['Heading1']) if re.search(r'<\s*html', part, re.IGNORECASE) or re.search(r'<\s*(p|h[1-6])', part, re.IGNORECASE):
flowables.append(header) if not part.lower().startswith("<html"):
flowables.append(Spacer(1, 12)) part = "<html>" + part + "</html>"
para = Paragraph(page_text.replace("\n", "<br/>"), normal_style) soup = BeautifulSoup(part, "html.parser")
flowables.append(para) for element in soup.find_all(['h1', 'h2', 'h3', 'h4', 'h5', 'h6', 'p']):
if i < len(processed_pages): if element.name.lower() in ["h1", "h2", "h3", "h4", "h5", "h6"]:
flowables.append(Spacer(1, 24)) style = header_styles.get(element.name.lower(), styles['Heading1'])
flowables.append(PageBreak()) else:
style = normal_style
text_content = element.get_text().strip()
if text_content:
flowables.append(Paragraph(text_content, style))
flowables.append(Spacer(1, 12))
else:
# Texto plano -> organizar en párrafos
paragraphs = organize_paragraphs(part)
for para_text in paragraphs:
if para_text:
flowables.append(Paragraph(para_text, normal_style))
flowables.append(Spacer(1, 12))
doc.build(flowables) doc.build(flowables)
return output_pdf return output_pdf