alchemark-ai 0.1.6

AlcheMark AI - PDF to Markdown conversion with rich metadata and element analysis

AlcheMark.

Your files ready for Gen AI ✨🚀

AlcheMark is a lightweight PDF to Markdown, alchemical-inspired toolkit that transmutes PDF documents into structured Markdown pages—complete with rich metadata and markdown element annotations—empowering you to uncover insights page by page.

Installation

# Install from PyPI
pip install alchemark-ai

# Or install from source
git clone https://github.com/matthsena/AlcheMark-ai.git
cd AlcheMark-ai
pip install -e .

Usage

from alchemark_ai import pdf2md

# Convert PDF to markdown
# pdf_file_path: Path to the PDF document to be converted
# process_images: When True, extracts images from the PDF (default: False)
# keep_images_inline: When True, keeps base64 images inline in markdown; when False, 
#                     replaces with references to image hashes (default: False)
results = pdf2md("path/to/document.pdf", process_images=True, keep_images_inline=True)

# Each result is a FormattedResult object with the structure:
# {
#   "metadata": {
#     "file_path": str,       # Path to the PDF file
#     "page": int,            # Page number
#     "page_count": int,      # Total number of pages
#     "text_length": int,     # Length of the extracted text
#     "processed_timestamp": float  # Processing timestamp
#   },
#   "elements": {
#     "tables": List[Table],  # Tables extracted from the page
#     "images": List[Image],  # Images extracted from the page (with optional base64 data)
#     "titles": List[str],    # Titles/headers detected
#     "lists": List[str],     # List items detected
#     "links": List[Link]     # Links with text and URL
#   },
#   "text": str,              # Markdown text content
#   "tokens": int,            # Number of tokens in the text
#   "language": str           # Detected language
# }

# Access the markdown text of the first page
markdown_text = results[0].text

# Get metadata for the first page
page_number = results[0].metadata.page
total_pages = results[0].metadata.page_count

# Check elements detected in the page
tables_count = len(results[0].elements.tables)
images_count = len(results[0].elements.images)

Google Colab Example

Try AlcheMark AI directly in your browser with our interactive Google Colab notebook!

Overview

AlcheMark AI provides a seamless solution for converting PDF documents into well-structured Markdown format. The tool not only extracts the text content but also analyzes and catalogs various elements like tables, images, headings, lists, and links while tracking token counts for LLM compatibility.

Key Features

• PDF to Markdown Conversion: Transform PDF documents into clean, organized Markdown
• Rich Metadata Extraction: Preserve document metadata including title, author, creation date
• Element Analysis: Automatic detection and counting of markdown elements (headings, lists, links)
• Table & Image Support: Extract and format tables and images from PDFs
• Inline Image Handling: Option to keep images inline as base64 or replace with image references
• Token Counting: Built-in token counting using tiktoken for LLM integration
• Structured Output: Get page-by-page results with detailed metadata

Extracted Data Fields

Field	Type	Description
metadata.file_path	str	Path to the original PDF file
metadata.page	int	Current page number
metadata.page_count	int	Total number of pages in the document
metadata.text_length	int	Character count of the extracted text
metadata.processed_timestamp	float	Unix timestamp when the page was processed
elements.tables	List[Table]	Tables extracted from the page with their structure preserved
elements.images	List[Image]	Images extracted from the page with their metadata, including optional base64 content and hash
elements.titles	List[str]	Headings and titles detected in the page
elements.lists	List[str]	List items (ordered and unordered) found in the page
elements.links	List[Link]	Hyperlinks with their display text and target URLs
text	str	The complete markdown text content of the page
tokens	int	Token count for the page (useful for LLM context planning)
language	str	Detected language of the page content

Configuration Options

Option	Default	Description
process_images	False	Enable extraction and processing of images from the PDF
keep_images_inline	False	Keep images inline as base64 in the markdown text. When set to False, images are replaced with references ([IMAGE](hash))

Test Coverage

AlcheMark AI maintains a high test coverage to ensure reliability:

Name                                     Stmts   Miss  Cover
------------------------------------------------------------
alchemark_ai/configs/logger.py               2      0   100%
alchemark_ai/formatter/formatter_md.py      80      2    98%
alchemark_ai/models/FormattedResult.py      25      0   100%
alchemark_ai/models/PDFResult.py            56      0   100%
alchemark_ai/pdf2md/pdf2md.py               30      0   100%
------------------------------------------------------------
TOTAL                                      193      2    99%

Current test suite includes 37 tests covering all major functionality, with an overall coverage of 99%.

Contributing

Contributions are welcome! Please feel free to submit a Pull Request.