Smart PDF for Business: Extract Text, Headings, and Signatures from PDFs
Smart PDF for Business is a Python library for structured PDF content analysis, built on top of spaCy-layout. Using spaCy-layout’s page, block, and layout-aware text extraction capabilities, the library adds higher-level features for business workflows and document automation. Smart PDF for Business provides a unified interface to extract, search, and analyze PDF documents with structure-aware intelligence:
- Load PDFs from files, folders, or raw bytes.
- Extract headings, body text, sections, and layout-aware content.
- Search text using keywords or semantic similarity powered by SentenceTransformers.
- Detect handwritten or scanned signatures with bounding boxes, pages, and optional cropped images.
- Export results as plain text, Markdown, CSV, Excel, or JSON.
- Process multiple PDFs at once with batch utilities.
📝 Usage
⚠️ This package requires Python 3.10 or above.
pip install smart_pdf_for_business
After initialising a PDFDoc object using one of the factory methods, you can call its built-in functions. Most methods create a new PDFDoc object containing the result. This allows you to utilise the library's output capabilities.
from smart_pdf_for_business import PDFDoc
# Load a PDF
pdf = PDFDoc.from_file("example.pdf")
# Perform semantic search
clauses = pdf.search_by_meaning("termination clause", chunk_by="semantic")
# Export results
clauses.to_csv("output.csv")
clauses.to_excel("output.xlsx")
Alternatively, you can use as_tuple=True to return the result as a tuple.
from smart_pdf_for_business import PDFDoc
# Load a PDF
pdf = PDFDoc.from_file("example.pdf")
# Perform semantic search
clauses = pdf.search_by_meaning("termination clause", chunk_by="semantic", as_tuple=True)
# Print result as tuple
print(clauses)
📚 API
class PDFDoc
| Attribute | Description |
|---|---|
data: bytes |
Raw byte content of the PDF document. |
path: Optional[Path] |
File system path to the PDF file, if loaded from disk. |
name: Optional[str] |
Name of the PDF document (usually the filename). |
spacy_doc: Optional[Doc] |
Processed spaCy Doc object, containing text, layout, and markdown. |
sections: List[Tuple[str, str]] |
List of (title, body) tuples extracted from the document. |
weighted_sections: List[Tuple[Tuple[str, str], float]] |
List of sections paired with a semantic match weight. |
signatures: List[Signature] |
Detected signatures stored as Signature objects. |
| Methods | Description | Parameter |
|---|---|---|
from_file |
Creates a PDFDoc from a PDF file. Return: PDFDoc |
path: str | Path - Path to the PDF file |
from_folder |
Loads all PDFs from a folder (optionally recursively). Return: list[PDFDoc] |
folder_path: str | Path - Folder containing PDFs recursive: bool = True - Include subfolders |
from_bytes |
Creates a PDFDoc from raw PDF byte content. Return: PDFDoc |
data: bytes - PDF bytes name: Optional[str] = None - Name assigned to document |
from_byte_list |
Creates multiple PDFDoc objects from PDF byte streams. Return: list[PDFDoc] |
byte_list: List[bytes] - List of PDF byte streams names: Optional[List[str]] = None - Corresponding PDF names |
search_signature |
Detects signatures near keyword regions and optionally saves cropped images. Updates self.signatures. Return: PDFDoc |
keywords: list[str] - Signature-related keywords exact=False - Forbid substring matching max_distance=70 - Max pixel distance save_folder=None - Folder to save crops min_contrast=30 - Minimum image contrast filter_stroke_density=True - Stroke-density filtering enforce_text_type: bool = True - Filter signature text using nlp |
search_header |
Searches section headers for keyword matches. Return: PDFDoc or list[tuple[str, str]] |
keywords: list[str] - Header keywords exact: bool = True - Exact match as_tuple: bool = False - Return tuples |
search_body |
Searches section bodies for keyword matches. Return: PDFDoc or list[tuple[str, str]] |
keywords: list[str] - Body keywords exact: bool = True - Exact match as_tuple: bool = False - Return tuples |
search_by_meaning |
Performs semantic search across sections, sentences, paragraphs, or words. Return: PDFDoc or list[tuple[str, float]] |
query: str - Search query threshold: float = 0 - Minimum similarity chunk_by: str = "section" - Chunking strategy as_tuple: bool = False - Return raw tuples word_count: int = 5 - Words per chunk buffer_size: int = 1 - Chunk overlap chunk_text_semantically_threshold: float = 0.3 - Semantic chunk threshold max_results: int = 5 - Max results |
to_text |
Converts the PDF to plain text with optional signature annotations. Return: str |
annotate_signatures: bool = True - Annotate signature positions |
to_markdown |
Converts the PDF to Markdown with optional signature annotations. Return: str |
annotate_signatures: bool = True - Annotate signature positions |
to_dataframe |
Converts the document to a structured pandas DataFrame. Return: pandas.DataFrame |
None |
to_excel |
Exports the document DataFrame to an Excel (.xlsx) file. Return: None |
path: str | Path - Output .xlsx path sheet_name: str = "Sheet1" - Sheet name |
to_csv |
Exports the document DataFrame to a CSV file. Return: None |
path: str | Path - Output CSV file path |
to_json |
Serializes the document into JSON, optionally writing to a file. Return: str (JSON) or None (if saved to file) |
path: Optional[str | Path] = None - Optional save path indent: int = 2 - JSON formatting |
class PDFDocBatch
| Attribute | Description |
|---|---|
pdfdocs: list[PDFDoc] |
List containing all PDFDoc instances in the batch. |
| Methods | Description | Parameter |
|---|---|---|
from_folder |
Creates a batch from all PDFs in a folder. Return: PDFDocBatch |
path: str | Path - Path to the folder containing PDFs recursive: bool = True - Whether to include subfolders |
from_byte_list |
Creates a batch from a list of PDF byte streams. Return: PDFDocBatch |
byte_list: List[bytes] - List of PDF bytes names: Optional[List[str]] = None - Optional list of names corresponding to each PDF |
from_pdfdoc_list |
Creates a batch from an existing list of PDFDoc objects. Return: PDFDocBatch |
pdfdoc_list: List[PDFDoc] - List of PDFDoc instances |
extend |
Extends the batch with another batch or list of PDFs. Return: None |
other: Union[List[PDFDoc], PDFDocBatch] - Batch or list of PDFs to append |
append |
Appends a single PDFDoc to the batch. Return: None |
pdfdoc: PDFDoc - PDFDoc instance to append |
search_signature |
Detects handwritten or scanned signatures associated with keyword regions for multiple PDFs. Return: PDFDocBatch |
keywords: List[str] - Keywords used to locate signature-related regions exact: bool = False - Whether keyword matching must be exact max_distance: int = 70 - Maximum pixel distance between a keyword block and signature save_folder: Optional[str | Path] = None - Folder to save cropped signature images, if provided min_contrast: int = 30 - Minimum pixel contrast threshold filter_stroke_density: bool = True - Filter image regions based on stroke density enforce_text_type: bool = True - Filter signature text based on stroke density |
search_header |
Searches headers in every PDF for keywords. Return: PDFDocBatch |
keywords: list[str] - Header keywords exact: bool = True - Exact or partial matching |
search_body |
Searches body text in every PDF for keywords. Return: PDFDocBatch |
keywords: list[str] - Body keywords exact: bool = True - Exact or partial matching |
search_by_meaning |
Performs semantic search across all PDFs. Return: PDFDocBatch |
query: str - Natural language query threshold: float = 0 - Min similarity score chunk_by: str = "section" - Chunking method word_count: int = 5 - Words per chunk buffer_size: int = 1 - Overlap size chunk_text_semantically_threshold: float = 0.3 - Semantic similarity threshold max_results: int = 5 - Max results per PDF |
to_dataframe |
Combines all PDFs in the batch into a DataFrame. PDFs with the same name overwrite each other. Return: pandas.DataFrame |
None |
to_excel |
Exports batch data to an Excel file. Return: None |
path: str | Path - Output Excel path sheet_name: str = "Sheet1" - Sheet name |
to_csv |
Exports batch data to a CSV file. Return: None |
path: str | Path - Output CSV file path |
dataclass Signature
| Attribute | Description |
|---|---|
keyword |
Keyword used for search |
text |
OCR text of signature region |
bbox |
Bounding box coordinates (x1, y1, x2, y2) |
page |
Page number |
img |
Cropped image as numpy.ndarray |
distance |
Distance to keyword |
Contributing
Contributions are welcome! Please fork the repository and submit a pull request.
1 maintainer
Metadata
1 maintainer
Release files for smart-pdf-for-business 1.0.0
For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.
Source distribution (sdist)
| File | Size | Uploaded | |
|---|---|---|---|
| smart_pdf_for_business-1.0.0.tar.gz | 11.0 MB | Details |
Built distribution (wheel)
| File | Interpreter | ABI | Platform | Reset |
|---|---|---|---|---|
| smart_pdf_for_business-1.0.0-py3-none-any.whl | Python 3 | none | any | Details |
Total release size: 11.1 MB
Release files / smart_pdf_for_business-1.0.0.tar.gz
| Download URL | smart_pdf_for_business-1.0.0.tar.gz |
|---|---|
| Size | 11.0 MB |
| Tags | Source |
|
SHA-256 checksum How to use checksums |
52966d45d7777a4fccdc66b0ed85a14ba35dbe27b3d4b93182e50fb0066dbf1d
|
|
BLAKE2b-256 checksum How to use checksums |
350178c088c62cbfd935bd55f31f6be1b3e32f7282beb744e6b140c76ca57f82
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
No |
| Uploaded via |
twine/6.2.0 CPython/3.13.3
|
Release files / smart_pdf_for_business-1.0.0-py3-none-any.whl
| Download URL | smart_pdf_for_business-1.0.0-py3-none-any.whl |
|---|---|
| Size | 20.9 kB |
| Tags | Python 3 |
|
SHA-256 checksum How to use checksums |
c61c7079ab5d490a013a98a39a834b509b4a5bef82b3b7c523de87756c6ea49a
|
|
BLAKE2b-256 checksum How to use checksums |
5a6dab116909d0f70efa24c0099c2453cb982aa9b186d9849bb7f4412ca3b1de
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
No |
| Uploaded via |
twine/6.2.0 CPython/3.13.3
|
