awacke1's picture
Update app.py
04a1711 verified
raw
history blame
23.6 kB
import gradio as gr
from pathlib import Path
import datetime
import re
import os
import shutil
import fitz # PyMuPDF
from PIL import Image
from collections import defaultdict
import io
from pypdf import PdfWriter
import random
from reportlab.platypus import SimpleDocTemplate, Paragraph, Spacer, Table, TableStyle, PageBreak, BaseDocTemplate, Frame, PageTemplate, Image as ReportLabImage
from reportlab.lib.styles import getSampleStyleSheet, ParagraphStyle
from reportlab.lib.pagesizes import letter, A4, legal, landscape
from reportlab.lib.units import inch
from reportlab.lib import colors
from reportlab.pdfbase import pdfmetrics
from reportlab.pdfbase.ttfonts import TTFont
from functools import partial
# --- Configuration & Setup ---
CWD = Path.cwd()
LAYOUTS = {
"A4 Portrait": {"size": A4},
"A4 Landscape": {"size": landscape(A4)},
"Letter Portrait": {"size": letter},
"Letter Landscape": {"size": landscape(letter)},
"Legal Portrait": {"size": legal},
"Legal Landscape": {"size": landscape(legal)},
}
OUTPUT_DIR = CWD / "generated_pdfs"
PREVIEW_DIR = CWD / "previews"
FONT_DIR = CWD # Assumes fonts are in the same directory as the script
# Create necessary directories
OUTPUT_DIR.mkdir(exist_ok=True)
PREVIEW_DIR.mkdir(exist_ok=True)
# --- Font & Emoji Handling ---
def register_local_fonts():
"""Finds and registers all .ttf files from the application's base directory."""
print("--- Font Registration Process Starting ---")
text_font_names = []
emoji_font_name = None
noto_emoji_path = FONT_DIR / "NotoColorEmoji-Regular.ttf"
if not noto_emoji_path.exists():
print(f"Warning: Emoji font not found at {noto_emoji_path}. Emojis may not render correctly.")
print("Please download 'NotoColorEmoji-Regular.ttf' and place it in the application directory.")
print(f"Scanning for fonts in: {FONT_DIR.absolute()}")
font_files = list(FONT_DIR.glob("*.ttf"))
print(f"Found {len(font_files)} .ttf files: {[f.name for f in font_files]}")
for font_path in font_files:
try:
font_name = font_path.stem
pdfmetrics.registerFont(TTFont(font_name, str(font_path)))
pdfmetrics.registerFont(TTFont(f"{font_name}-Bold", str(font_path)))
pdfmetrics.registerFont(TTFont(f"{font_name}-Italic", str(font_path)))
pdfmetrics.registerFont(TTFont(f"{font_name}-BoldItalic", str(font_path)))
pdfmetrics.registerFontFamily(font_name, normal=font_name, bold=f"{font_name}-Bold", italic=f"{font_name}-Italic", boldItalic=f"{font_name}-BoldItalic")
if "notocoloremoji-regular" in font_name.lower():
emoji_font_name = font_name
elif "notoemoji" not in font_name.lower():
text_font_names.append(font_name)
except Exception as e:
print(f"Could not register font {font_path.name}: {e}")
if not text_font_names:
print("WARNING: No text fonts found. Adding 'Helvetica' as a default.")
text_font_names.append('Helvetica')
print(f"Successfully registered user-selectable fonts: {text_font_names}")
print(f"Emoji font set to: {emoji_font_name}")
print("--- Font Registration Process Finished ---")
return sorted(text_font_names), emoji_font_name
def apply_emoji_font(text: str, emoji_font_name: str) -> str:
"""Intelligently wraps emoji characters in a <font> tag to ensure they render correctly."""
if not emoji_font_name:
return text
emoji_pattern = re.compile(f"([{re.escape(''.join(map(chr, range(0x1f600, 0x1f650))))}"
f"{re.escape(''.join(map(chr, range(0x1f300, 0x1f5ff))))}"
f"{re.escape(''.join(map(chr, range(0x1f900, 0x1f9ff))))}"
f"{re.escape(''.join(map(chr, range(0x2600, 0x26ff))))}"
f"{re.escape(''.join(map(chr, range(0x2700, 0x27bf))))}]+)")
return emoji_pattern.sub(fr'<font name="{emoji_font_name}">\1</font>', text)
# --- AI Content Generation (Simulation) ---
def generate_ai_content_api(prompt):
"""
Simulates a call to an LLM to generate markdown content.
In a real application, this would contain `fetch` calls to a generative AI API.
"""
if not prompt:
return "# The Golem awaits your command!\n\nPlease enter a prompt in the box above and click '๐Ÿง  Animate Golem!' to get started. I can help you write reports, stories, poems, and more! โœจ"
import time
time.sleep(1.5)
sample_story = f"""
# The Quest for the Sunstone โ˜€๏ธ
Long ago, in the mystical land of Aerthos, a creeping shadow began to fall across the valleys. The once-vibrant flowers ๐ŸŒธ drooped, and the rivers ran slow. The elders knew the cause: the light of the Sunstone, hidden deep within Mount Cinder, was fading.
## The Prophecy ๐Ÿ“œ
An ancient prophecy spoke of a hero who would rekindle the stone. It read:
> When darkness drapes the verdant ground,
> A soul of courage shall be found.
> Through trials of fire ๐Ÿ”ฅ, wit, and might,
> They'll bring once more the sacred light.
## The Chosen One ๐Ÿฆธโ€โ™€๏ธ
A young woman named Elara, known for her kindness and bravery, was chosen. She accepted the quest, her only companions a loyal wolf ๐Ÿบ and a map gifted by the village shaman.
### The Journey Begins
Elara's journey was fraught with peril. She navigated enchanted forests and crossed treacherous chasms.
| Trial | Location | Challenge |
|------------------|-------------------|----------------------------------------|
| The Whispering | The Gloomwood | Resist maddening whispers of despair |
| The Riddle of | Sphinx Gate | Answer three impossible questions |
| The Fiery Path | The Magma Caverns | Walk barefoot across burning embers |
Finally, she reached the heart of the mountain. There, resting on a pedestal, was the dim Sunstone. Pouring her own hope and courage into it, the stone blazed with renewed life, banishing the shadows from Aerthos forever. The people rejoiced, and Elara became a legend.
"""
return f"# Golem's Vision for: '{prompt}'\n\n{sample_story}"
# --- PDF Generation & Handling ---
def _draw_header_footer(canvas, doc, header_text, footer_text, title):
"""Draws the header and footer on each page."""
canvas.saveState()
page_num = canvas.getPageNumber()
# Replace variables for header and footer. Note: [Total Pages] is only accurate at the end of the build.
final_footer_text = footer_text.replace("[Page #]", str(page_num)).replace("[Total Pages]", str(doc.page))
final_header_text = header_text.replace("[Page #]", str(page_num)).replace("[Title]", title)
# Header
if final_header_text:
canvas.setFont('Helvetica', 9)
canvas.setFillColor(colors.grey)
canvas.drawRightString(doc.width + doc.leftMargin, doc.height + doc.topMargin + 0.25*inch, final_header_text)
# Footer
if final_footer_text:
canvas.setFont('Helvetica', 9)
canvas.setFillColor(colors.grey)
canvas.drawString(doc.leftMargin, doc.bottomMargin - 0.25*inch, final_footer_text)
canvas.restoreState()
def markdown_to_story(markdown_text: str, font_name: str, emoji_font: str, font_size_body: int, font_size_h1: int, font_size_h2: int, font_size_h3: int):
"""Converts markdown to a ReportLab story with enhanced, user-configurable styling."""
styles = getSampleStyleSheet()
leading_body = font_size_body * 1.4
style_normal = ParagraphStyle('BodyText', fontName=font_name, fontSize=font_size_body, leading=leading_body, spaceAfter=6)
style_h1 = ParagraphStyle('h1', parent=styles['h1'], fontName=font_name, fontSize=font_size_h1, leading=font_size_h1*1.2, spaceBefore=12, textColor=colors.HexColor("#a855f7"))
style_h2 = ParagraphStyle('h2', parent=styles['h2'], fontName=font_name, fontSize=font_size_h2, leading=font_size_h2*1.2, spaceBefore=10, textColor=colors.HexColor("#6366f1"))
style_h3 = ParagraphStyle('h3', parent=styles['h3'], fontName=font_name, fontSize=font_size_h3, leading=font_size_h3*1.2, spaceBefore=8, textColor=colors.HexColor("#3b82f6"))
style_code = ParagraphStyle('Code', fontName='Courier', backColor=colors.HexColor("#333333"), textColor=colors.HexColor("#f472b6"), borderWidth=1, borderColor=colors.HexColor("#444444"), padding=8, leading=12, fontSize=9)
style_table_header = ParagraphStyle('TableHeader', parent=style_normal, fontName=f"{font_name}-Bold" if font_name != 'Helvetica' else 'Helvetica-Bold')
story = []
lines = markdown_text.split('\n')
in_code_block, in_table = False, False
code_block_text, table_data = "", []
first_heading = True
document_title = "Untitled Document"
for line in lines:
stripped_line = line.strip()
if stripped_line.startswith("```"):
if in_code_block:
escaped_code = code_block_text.replace('&', '&amp;').replace('<', '&lt;').replace('>', '&gt;')
story.append(Paragraph(escaped_code.replace('\n', '<br/>'), style_code))
story.append(Spacer(1, 0.1 * inch))
in_code_block, code_block_text = False, ""
else: in_code_block = True
continue
if in_code_block:
code_block_text += line + '\n'; continue
if stripped_line.startswith('|'):
if not in_table: in_table = True
if all(c in '-|: ' for c in stripped_line): continue
cells = [cell.strip() for cell in stripped_line.strip('|').split('|')]
table_data.append(cells)
continue
if in_table:
in_table = False
if table_data:
header_content = [apply_emoji_font(re.sub(r'\*\*(.*?)\*\*', r'<b>\1</b>', cell), emoji_font) for cell in table_data[0]]
header = [Paragraph(cell, style_table_header) for cell in header_content]
formatted_rows = [
[Paragraph(apply_emoji_font(re.sub(r'\*\*(.*?)\*\*', r'<b>\1</b>', cell), emoji_font), style_normal) for cell in row]
for row in table_data[1:]
]
table = Table([header] + formatted_rows, hAlign='LEFT', repeatRows=1)
table.setStyle(TableStyle([
('BACKGROUND', (0, 0), (-1, 0), colors.HexColor("#4a044e")),
('TEXTCOLOR', (0, 0), (-1, 0), colors.whitesmoke),
('GRID', (0, 0), (-1, -1), 1, colors.HexColor("#6b21a8")),
('VALIGN', (0,0), (-1,-1), 'MIDDLE'),
('TOPPADDING', (0,0), (-1,-1), 6),
('BOTTOMPADDING', (0,0), (-1,-1), 6),
]))
story.append(table); story.append(Spacer(1, 0.2 * inch))
table_data = []
if not stripped_line: continue
content, style, extra_args = stripped_line, style_normal, {}
if stripped_line.startswith("# "):
if not first_heading: story.append(PageBreak())
content, style = stripped_line.lstrip('# '), style_h1
if first_heading: document_title = content
first_heading = False
elif stripped_line.startswith("## "): content, style = stripped_line.lstrip('## '), style_h2
elif stripped_line.startswith("### "): content, style = stripped_line.lstrip('### '), style_h3
elif stripped_line.startswith(("- ", "* ")): content, extra_args['bulletText'] = stripped_line[2:], 'โ€ข'
formatted_content = re.sub(r'_(.*?)_', r'<i>\1</i>', re.sub(r'\*\*(.*?)\*\*', r'<b>\1</b>', content))
final_content = apply_emoji_font(formatted_content, emoji_font)
story.append(Paragraph(final_content, style, **extra_args))
return story, document_title
def create_pdf_preview(pdf_path: Path):
"""Generates a PNG preview of the first page of a PDF."""
preview_path = PREVIEW_DIR / f"{pdf_path.stem}.png"
try:
doc = fitz.open(pdf_path)
page = doc.load_page(0)
pix = page.get_pixmap(dpi=150)
pix.save(str(preview_path))
doc.close()
return str(preview_path)
except Exception as e:
print(f"Could not create preview for {pdf_path.name}: {e}")
try:
img = Image.new('RGB', (400, 500), color = '#111827')
img.save(str(preview_path))
return str(preview_path)
except: return None
# --- Main API Function ---
def generate_pdfs_api(files, ai_content, layouts, fonts, num_columns, header_text, footer_text, font_size_body, font_size_h1, font_size_h2, font_size_h3, margin_top, margin_bottom, margin_left, margin_right, progress=gr.Progress(track_tqdm=True)):
if not files and not ai_content.strip(): raise gr.Error("Please conjure some content or upload an image before alchemizing!")
if not layouts: raise gr.Error("You must select a scroll (page layout)!")
if not fonts: raise gr.Error("A scribe needs a font! Please choose one.")
shutil.rmtree(OUTPUT_DIR, ignore_errors=True); shutil.rmtree(PREVIEW_DIR, ignore_errors=True)
OUTPUT_DIR.mkdir(); PREVIEW_DIR.mkdir()
md_content = ai_content if ai_content and ai_content.strip() else ""
image_files = [Path(f.name) for f in files if Path(f.name).suffix.lower() in ['.png', '.jpg', '.jpeg']] if files else []
log_updates, generated_pdf_paths = "", []
for layout_name in progress.tqdm(layouts, desc=" brewing potions..."):
for font_name in progress.tqdm(fonts, desc=f" enchanting scrolls with {layout_name}..."):
merger = PdfWriter()
if md_content:
md_buffer = io.BytesIO()
story, title = markdown_to_story(md_content, font_name, EMOJI_FONT_NAME, font_size_body, font_size_h1, font_size_h2, font_size_h3)
pagesize = LAYOUTS[layout_name]["size"]
doc = BaseDocTemplate(md_buffer, pagesize=pagesize,
leftMargin=margin_left*inch, rightMargin=margin_right*inch,
topMargin=margin_top*inch, bottomMargin=margin_bottom*inch)
frame_width = (doc.width / num_columns) - (num_columns - 1) * 0.1*inch
frames = [Frame(doc.leftMargin + i * (frame_width + 0.2*inch), doc.bottomMargin, frame_width, doc.height, id=f'col_{i}') for i in range(num_columns)]
header_footer_callback = partial(_draw_header_footer, header_text=header_text, footer_text=footer_text, title=title)
page_template = PageTemplate(id='main_template', frames=frames, onPage=header_footer_callback)
doc.addPageTemplates([page_template])
doc.build(story)
md_buffer.seek(0)
merger.append(fileobj=md_buffer)
for img_path in image_files:
try:
with Image.open(img_path) as img: img_width_px, img_height_px = img.size
img_width_pt, img_height_pt = img_width_px * (inch / 72), img_height_px * (inch / 72)
img_buffer = io.BytesIO()
img_doc = SimpleDocTemplate(img_buffer, pagesize=(img_width_pt, img_height_pt), leftMargin=0, rightMargin=0, topMargin=0, bottomMargin=0)
img_doc.build([ReportLabImage(img_path, width=img_width_pt, height=img_height_pt)])
img_buffer.seek(0)
merger.append(fileobj=img_buffer)
except Exception as e: log_updates += f"โš ๏ธ Failed to process image {img_path.name}: {e}\n"
if len(merger.pages) > 0:
time_str = datetime.datetime.now().strftime('%H%M%S')
clean_layout = layout_name.replace(' ', '')
filename = f"Scroll_{clean_layout}_{font_name}_x{num_columns}_{time_str}.pdf"
output_path = OUTPUT_DIR / filename
with open(output_path, "wb") as f: merger.write(f)
generated_pdf_paths.append(output_path)
log_updates += f"โœ… Successfully alchemized: {filename}\n"
gallery_previews = [create_pdf_preview(p) for p in generated_pdf_paths]
final_gallery = [g for g in gallery_previews if g is not None]
return final_gallery, log_updates if log_updates else "โœจ All scrolls alchemized successfully! โœจ", [str(p) for p in generated_pdf_paths]
# --- Gradio UI Definition ---
AVAILABLE_FONTS, EMOJI_FONT_NAME = register_local_fonts()
# A custom, magical theme
theme = gr.themes.Base(
primary_hue=gr.themes.colors.purple,
secondary_hue=gr.themes.colors.indigo,
neutral_hue=gr.themes.colors.gray,
font=(gr.themes.GoogleFont("MedievalSharp"), "ui-sans-serif", "system-ui", "sans-serif"),
).set(
body_background_fill="#111827",
body_background_fill_dark="#111827",
body_text_color="#d1d5db",
body_text_color_dark="#d1d5db",
button_primary_background_fill="#a855f7",
button_primary_background_fill_dark="#a855f7",
button_primary_text_color="#ffffff",
button_primary_text_color_dark="#ffffff",
button_secondary_background_fill="#6366f1",
button_secondary_background_fill_dark="#6366f1",
button_secondary_text_color="#ffffff",
button_secondary_text_color_dark="#ffffff",
block_background_fill="#1f2937",
block_background_fill_dark="#1f2937",
block_label_background_fill="#1f2937",
block_label_background_fill_dark="#1f2937",
block_title_text_color="#a855f7",
block_title_text_color_dark="#a855f7",
input_background_fill="#374151",
input_background_fill_dark="#374151",
)
with gr.Blocks(theme=theme, title="The PDF Alchemist") as demo:
gr.Markdown("# โœจ The PDF Alchemist โœจ")
gr.Markdown("Turn your ideas into beautifully crafted PDF scrolls. Use the power of AI or upload your own treasures.")
with gr.Row(equal_height=False):
with gr.Column(scale=2):
with gr.Tabs():
with gr.TabItem("โ‘  Conjure Your Content ๐Ÿ“œ"):
gr.Markdown("### ๐Ÿค– The Idea Golem")
ai_prompt = gr.Textbox(label="Command the Golem", placeholder="e.g., 'A recipe for a dragon's breath chili...'")
generate_ai_btn = gr.Button("๐Ÿง  Animate Golem!")
ai_content_output = gr.Textbox(label="Golem's Manuscript (Editable)", lines=15, interactive=True)
gr.Markdown("<hr style='border-color: #374151;'>")
gr.Markdown("### ๐Ÿ“ค Or, Upload Your Treasures")
uploaded_files = gr.File(label="Upload Images (.png, .jpg)", file_count="multiple", file_types=[".png", ".jpg", ".jpeg"])
gr.Markdown("*<p style='font-size:0.8rem; color: #9ca3af;'>Note: The Golem's manuscript takes precedence. Images are always added.</p>*")
with gr.TabItem("โ‘ก Define Your Canvas ๐ŸŽจ"):
gr.Markdown("### ๐Ÿ“ Choose Your Scroll")
selected_layouts = gr.CheckboxGroup(choices=list(LAYOUTS.keys()), label="Page Layouts", value=["A4 Portrait"])
num_columns_slider = gr.Slider(label="Number of Text Columns", minimum=1, maximum=4, step=1, value=1)
gr.Markdown("### โœ๏ธ Scribe Your Margins")
gr.Markdown("<p style='font-size:0.8rem; color: #9ca3af;'>Use `[Page #]` and `[Title]`. `[Total Pages]` works best in the footer.</p>")
header_input = gr.Textbox(label="Header Inscription", value="[Title]", placeholder="e.g., Arcane Folio - [Page #]")
footer_input = gr.Textbox(label="Footer Inscription", value="Page [Page #] of [Total Pages]", placeholder="e.g., Top Secret - Page [Page #]")
with gr.TabItem("โ‘ข Perfect Your Style ๐Ÿ’…"):
gr.Markdown("### ๐Ÿ”ค Master the Glyphs")
selected_fonts = gr.CheckboxGroup(choices=AVAILABLE_FONTS, label="Fonts", value=[AVAILABLE_FONTS[0]] if AVAILABLE_FONTS else [])
font_size_body_slider = gr.Slider(label="Body Text Size (pt)", minimum=8, maximum=16, step=1, value=10)
with gr.Row():
font_size_h1_slider = gr.Slider(label="H1 Size", minimum=16, maximum=32, step=1, value=24)
font_size_h2_slider = gr.Slider(label="H2 Size", minimum=14, maximum=28, step=1, value=18)
font_size_h3_slider = gr.Slider(label="H3 Size", minimum=12, maximum=24, step=1, value=14)
gr.Markdown("### ๐Ÿ“ Set Your Boundaries (inches)")
with gr.Row():
margin_top_slider = gr.Slider(label="Top", minimum=0.25, maximum=1.5, step=0.05, value=0.75)
margin_bottom_slider = gr.Slider(label="Bottom", minimum=0.25, maximum=1.5, step=0.05, value=0.75)
with gr.Row():
margin_left_slider = gr.Slider(label="Left", minimum=0.25, maximum=1.5, step=0.05, value=0.75)
margin_right_slider = gr.Slider(label="Right", minimum=0.25, maximum=1.5, step=0.05, value=0.75)
generate_pdfs_btn = gr.Button("๐Ÿ”ฎ Alchemize PDF!", variant="primary", size="lg")
with gr.Column(scale=3):
gr.Markdown("### ๐Ÿ–ผ๏ธ The Scrying Pool (Previews)")
gallery_output = gr.Gallery(label="Generated PDF Previews", show_label=False, elem_id="gallery", columns=2, height=700, object_fit="contain")
log_output = gr.Markdown(label="Alchemist's Log", value="Your log of successful transmutations will appear here...")
downloadable_files_output = gr.Files(label="Collect Your Scrolls")
# --- API Calls & Examples ---
generate_ai_btn.click(fn=generate_ai_content_api, inputs=[ai_prompt], outputs=[ai_content_output])
inputs_list = [uploaded_files, ai_content_output, selected_layouts, selected_fonts, num_columns_slider, header_input, footer_input, font_size_body_slider, font_size_h1_slider, font_size_h2_slider, font_size_h3_slider, margin_top_slider, margin_bottom_slider, margin_left_slider, margin_right_slider]
outputs_list = [gallery_output, log_output, downloadable_files_output]
generate_pdfs_btn.click(fn=generate_pdfs_api, inputs=inputs_list, outputs=outputs_list)
gr.Examples(
examples=[["A technical summary of how alchemy works"], ["A short poem about a grumpy gnome"], ["A sample agenda for a wizard's council meeting"]],
inputs=[ai_prompt], outputs=[ai_content_output], fn=generate_ai_content_api, cache_examples=False
)
if __name__ == "__main__":
if not (FONT_DIR / "NotoColorEmoji-Regular.ttf").exists():
print("\n" + "="*80)
print("WARNING: 'NotoColorEmoji-Regular.ttf' not found.")
print("Please download it from Google Fonts and place it in the script's directory for emojis to render correctly.")
print("="*80 + "\n")
if not any("MedievalSharp" in s for s in AVAILABLE_FONTS):
print("\n" + "="*80)
print("WARNING: 'MedievalSharp' font not found for the default theme.")
print("Please download it from Google Fonts to get the intended 'alchemist' look.")
print("The app will fall back to another font.")
print("="*80 + "\n")
demo.launch(debug=True)