DOCX Document Stamping¶

PrimeStamp supports Microsoft Word DOCX files with full content hashing, metadata extraction, and structural fingerprinting. Because DOCX is a ZIP-based format containing XML, PrimeStamp can provide both whole-document and per-section authentication.

Quick Start¶

from primestamp.integrations.docx_extract import DocxExtractor, PrimeStampDocx

extractor = DocxExtractor()

# Extract from file path
result = extractor.extract("report.docx")

# Or from bytes
result = extractor.extract_from_bytes(docx_bytes)

print(f"Paragraphs: {result.paragraph_count}")
print(f"Words: {result.word_count}")
print(f"Headings: {result.heading_count}")
print(f"Tables: {result.table_count}")
print(f"Content hash: {result.content_hash}")

integration = PrimeStampDocx(extractor)
stamped = await integration.stamp_docx(docx_bytes, "report.docx", preset="standard")

Dependencies¶

pip install python-docx
# or install with the integrations extra:
pip install "primestamp[integrations]"

Features¶

Paragraph-level extraction with style detection
Heading hierarchy reconstruction (H1-H9)
Table extraction with row/column hashing
Image reference cataloging
Header/footer content capture
Comment and revision tracking
Footnote and endnote extraction
Section-level Merkle tree for partial proofs
Metadata extraction (author, dates, custom properties)

Structural Elements¶

Element	Description
`heading`	Headings with level (1-9) from Word styles
`paragraph`	Body text paragraphs
`list_item`	Bulleted or numbered list items
`table`	Tables with cell-level content
`image`	Embedded image references
`header`	Page header content
`footer`	Page footer content
`footnote`	Footnotes
`endnote`	Endnotes
`comment`	Review comments with author and date

for element in result.elements:
    print(f"[{element.element_type}] {element.content[:80]}")
    if element.heading_level:
        print(f"  Level: {element.heading_level}")
    if element.style_name:
        print(f"  Style: {element.style_name}")

Heading Hierarchy¶

PrimeStamp reconstructs the full heading hierarchy from Word's named styles:

for element in result.elements:
    if element.element_type == "heading":
        indent = "  " * (element.heading_level - 1)
        print(f"{indent}H{element.heading_level}: {element.content}")

Section-Level Merkle Tree¶

Each structural element gets its own hash, combined into a Merkle tree for partial proofs. This enables proving a specific clause exists in a contract without revealing the rest:

proof = result.generate_section_proof(section_index=5)

Metadata Extraction¶

metadata = result.metadata
print(f"Title: {metadata.get('title')}")
print(f"Author: {metadata.get('author')}")
print(f"Last modified by: {metadata.get('last_modified_by')}")
print(f"Created: {metadata.get('created')}")
print(f"Modified: {metadata.get('modified')}")
print(f"Revision: {metadata.get('revision')}")

Comments and Revisions¶

for comment in result.comments:
    print(f"Comment by {comment.author} at {comment.date}: {comment.text}")

for revision in result.revisions:
    print(f"Revision by {revision.author}: {revision.type}")

Table Extraction¶

for element in result.elements:
    if element.element_type == "table":
        print(f"Table with {element.metadata.get('rows')} rows:")
        for row in element.metadata.get('row_data', []):
            print(f"  | {' | '.join(row)} |")

Content-Based Hashing¶

DOCX stamping hashes extracted text content rather than raw zip bytes. Style-only changes (font, color, spacing) do not affect the content hash. Structural changes (reordering paragraphs, adding headings) do.

CLI Usage¶

# Stamp a DOCX file
primestamp stamp contract.docx --preset standard

# Extract and display structure
primestamp fingerprint contract.docx --show-structure

# Verify a stamped DOCX
primestamp verify contract.docx

# Compare two versions
primestamp fingerprint --compare contract_v1.docx contract_v2.docx

Stamping Pipeline¶

When PrimeStamp stamps a DOCX file, it:

Unzips the DOCX and parses the XML structure
Extracts paragraphs, headings, tables, and other elements
Reads metadata from core properties
Captures comments, revisions, footnotes, and endnotes
Computes a content hash from all extracted text
Builds a Merkle tree over section hashes
Optionally generates a Prime-Space fingerprint
Creates a signed stamp with structural metadata
Submits the hash to configured blockchain anchors

Comparison with PDF¶

Feature	PDF	DOCX
Extraction granularity	Page-level	Paragraph-level
Heading detection	Heuristic (font-based)	Explicit (Word styles)
Comments/revisions	Annotations only	Full revision tracking
Merkle tree	Per page	Per structural element
Dependencies	`pypdf`	`python-docx`