sharepoint-to-text 0.9.0

Text extraction library for typical file formats found in SharePoint repositories

sharepoint-to-text

A pure Python library for extracting text, metadata, and structured elements from Microsoft Office files—both modern (.docx, .xlsx, .pptx) and legacy (.doc, .xls, .ppt) formats—plus PDF, email formats, and plain text.

The library also includes an optional SharePoint client for reading files directly from Microsoft SharePoint sites via the Graph API. You still orchestrate the pipeline: pull files (via sharepoint_io or your own Graph client), then pass the bytes into the extractors.

Install: uv add sharepoint-to-text Python import: import sharepoint2text CLI (text): sharepoint2text --file /path/to/file.docx > extraction.txt CLI (JSON): sharepoint2text --file /path/to/file.docx --json > extraction.json (images ignored by default; add --include-images to extract)

What You Get

• Unified API: sharepoint2text.read_file(path) yields one or more typed extraction results.
• Typed results: each format returns a specific dataclass (e.g. DocxContent, PdfContent) that also supports the common interface.
• Text: get_full_text() or iterate_units() (pages / slides / sheets depending on format; call unit.get_text() for the string).
• Structured content: tables and images where the format supports it.
• Metadata: file metadata (plus format-specific metadata where available).
• Serialization: result.to_json() returns a JSON-serializable dict.

Extractor Interface (Developer Guide)

Every extracted result implements the same high-level interface (ExtractionInterface). Use it to build pipelines that work across file types without special-casing .pdf vs .docx vs .pptx.

Quick Guide: Which Method Should I Use?

Goal	Recommended	Why
Get "the document text" as one string	result.get_full_text()	Best default for indexing and simple exports; hides format-specific unit details.
Chunk text by page/slide/sheet (RAG, citations, per-unit metadata)	result.iterate_units()	Stable unit boundaries for formats that have them (PDF pages, PPT slides, XLS(X) sheets).
Extract images (and optionally store payloads)	result.iterate_images()	Returns image objects with metadata; binary payload handling is caller-controlled.
Extract tables as structured data	result.iterate_tables()	Returns table objects as 2D arrays, suitable for CSV/JSON downstream.
Attach filename/path context	result.get_metadata()	Normalizes file metadata regardless of format; useful for provenance and linking.
Persist/transport results	result.to_json() / ExtractionInterface.from_json(...)	JSON-serializable representation; optional base64 encoding for binary fields.

Method Details (When to Use Which)

• get_full_text():
  • Use when you want a single string per extracted item (search indexing, previews, "export to .txt").
  • It is usually derived from iterate_units(), but some formats may prepend metadata (e.g., titles) or omit optional content by default.
• iterate_units():
  • Use when you need chunk boundaries aligned with the source structure (pages/slides/sheets) or when you want to keep unit-level metadata.
  • Each unit supports unit.get_text(), unit.get_images(), unit.get_tables(), and unit.get_metadata().
• iterate_images() / iterate_tables():
  • Use when you want all images/tables across the document (often simpler than traversing units).
  • Prefer unit-level access (unit.get_images(), unit.get_tables()) when you need "where did this come from?" context (page/slide number).
• get_metadata():
  • Use for provenance fields like filename, file_extension, file_path, folder_path.
  • Pair with unit metadata for precise citations (e.g., file_path + page_number).
• to_json() / from_json():
  • Use to store results, send them across processes, or debug extraction output.
  • Binary payloads are representable but can be large; omit them unless you explicitly need embedded data.

Examples

Plain text (single string):

import sharepoint2text

result = next(sharepoint2text.read_file("document.pdf"))
text = result.get_full_text()

Unit-based chunking (recommended for RAG):

import sharepoint2text

result = next(sharepoint2text.read_file("deck.pptx"))
for unit in result.iterate_units():
    chunk = unit.get_text()
    unit_meta = unit.get_metadata()  # e.g., slide/page/sheet number when available

Why This Library?

Pure Python, No System Dependencies

Unlike popular alternatives that shell out to LibreOffice or Apache Tika (requiring Java), sharepoint-to-text is a native Python implementation with no system-level dependencies:

Approach	Requirements	Cross-platform	Container-friendly
sharepoint-to-text	uv add only	Yes	Yes (minimal image)
LibreOffice-based	LibreOffice install, X11/headless setup	Complex	Large images (~1GB+)
Apache Tika	Java runtime, Tika server	Complex	Heavy (~500MB+)
subprocess-based	Shell access, security concerns	No	Risky

This library parses Office binary formats (OLE2) and XML-based formats (OOXML) directly in Python, making it ideal for:

• RAG pipelines and LLM document ingestion
• Serverless functions (AWS Lambda, Google Cloud Functions)
• Containerized deployments with minimal footprint
• Secure environments where shell execution is restricted
• Cross-platform applications (Windows, macOS, Linux)

Enterprise SharePoint Reality

Enterprise SharePoints contain decades of accumulated documents. While modern .docx, .xlsx, and .pptx files are well-supported, legacy .doc, .xls, and .ppt files remain common. This library provides a unified interface for all formats—no conditional logic needed.

SharePoint Connectivity

For scenarios where documents live in Microsoft SharePoint, the library includes a built-in Graph API client. This is an optional convenience layer, not required for local files or other storage backends. You are responsible for orchestrating the pull (list/download) and then calling the extractors:

from sharepoint2text.sharepoint_io import SharePointRestClient, EntraIDAppCredentials

credentials = EntraIDAppCredentials(
    tenant_id="your-tenant-id",
    client_id="your-client-id",
    client_secret="your-client-secret",
)
client = SharePointRestClient(site_url="https://contoso.sharepoint.com/sites/Docs", credentials=credentials)

# List and download files
for file in client.list_all_files():
    content = client.download_file(file.id)
    # Pass to sharepoint2text extractors...

The client supports filtering by modification date (for delta-sync patterns), folder paths, and file extensions. See sharepoint2text/sharepoint_io/SETUP.md for Azure/Entra ID configuration instructions.

Supported Formats

Format Detection and Processing

The library uses a sophisticated multi-layered approach to detect and process file formats:

1. Extension-based detection (primary): File extensions are mapped to appropriate extractors
2. Extension aliases: Template formats and variants automatically map to their base extractors
3. MIME type fallback: When extensions are unclear, MIME types are used for detection
4. Archive processing: Archives recursively extract and process all supported files within them

Extension Aliases: The library automatically handles format variants:

• Office templates (.dotx, .xltx, .potx) → processed as their base formats
• Office shows (.ppsx) → processed as presentations
• OpenDocument templates (.ott, .ots, .otp) → processed as their base formats
• Compression variants (.tar.gz → .tgz, .tar.bz2 → .tbz2, .tar.xz → .txz)

Security Features: Archive processing includes protection against zip bombs and has size limits (7z archives max 100MB).

Legacy Microsoft Office

Format	Extension	Description
Word 97-2003	.doc	Word 97-2003 documents
Word 97-2003 (template)	.dot	Word 97-2003 templates
Excel 97-2003	.xls	Excel 97-2003 spreadsheets
Excel 97-2003 (template)	.xlt	Excel 97-2003 templates
PowerPoint 97-2003	.ppt	PowerPoint 97-2003 presentations
PowerPoint 97-2003 (template)	.pot	PowerPoint 97-2003 templates
PowerPoint 97-2003 (show)	.pps	PowerPoint 97-2003 slide shows
Rich Text Format	.rtf	Rich Text Format documents

Modern Microsoft Office

Format	Extension	Description
Word 2007+	.docx	Word 2007+ documents
Word 2007+ (macro)	.docm	Word 2007+ macro-enabled documents
Word 2007+ (template)	.dotx	Word 2007+ templates
Word 2007+ (template, macro)	.dotm	Word 2007+ macro-enabled templates
Excel 2007+	.xlsx	Excel 2007+ spreadsheets
Excel 2007+ (macro)	.xlsm	Excel 2007+ macro-enabled spreadsheets
Excel 2007+ (template)	.xltx	Excel 2007+ templates
Excel 2007+ (template, macro)	.xltm	Excel 2007+ macro-enabled templates
PowerPoint 2007+	.pptx	PowerPoint 2007+ presentations
PowerPoint 2007+ (macro)	.pptm	PowerPoint 2007+ macro-enabled presentations
PowerPoint 2007+ (template)	.potx	PowerPoint 2007+ templates
PowerPoint 2007+ (template, macro)	.potm	PowerPoint 2007+ macro-enabled templates
PowerPoint 2007+ (show)	.ppsx	PowerPoint 2007+ slide shows
PowerPoint 2007+ (show, macro)	.ppsm	PowerPoint 2007+ macro-enabled slide shows

OpenDocument

Format	Extension	Description
Text	.odt	OpenDocument Text
Text (template)	.ott	OpenDocument Text templates
Presentation	.odp	OpenDocument Presentation
Presentation (template)	.otp	OpenDocument Presentation templates
Spreadsheet	.ods	OpenDocument Spreadsheet
Spreadsheet (template)	.ots	OpenDocument Spreadsheet templates
Drawing	.odg	OpenDocument Drawing
Formula	.odf	OpenDocument Formula

Email

Format	Extension	Description
EML	.eml	RFC 822 email format
MSG	.msg	Microsoft Outlook email format
MBOX	.mbox	Unix mailbox format (multiple emails)

Plain Text

Format	Extension	Description
Plain Text	.txt	Plain text files
Markdown	.md	Markdown
CSV	.csv	Comma-separated values
TSV	.tsv	Tab-separated values
JSON	.json	JSON files

Configuration & Data Files

Format	Extension	Description
YAML	.yaml, .yml	YAML configuration files
XML	.xml	XML documents and data files
Log files	.log	Application log files
INI/Config	.ini, .cfg, .conf	Configuration files
Properties	.properties	Java properties files

PDF

Format	Extension	Description
PDF	.pdf	PDF documents

HTML / Web

Format	Extension	Description
HTML	.html, .htm	HTML documents
MHTML	.mhtml, .mht	MIME HTML (web archive) files
EPUB	.epub	EPUB e-book format

Archives

Format	Extension	Description
ZIP	.zip	ZIP archives
7-Zip	.7z	7-Zip archives (max 100MB)
TAR	.tar	TAR archives
Gzip TAR	.tar.gz, .tgz, .gz	Gzip-compressed TAR archives
Bzip2 TAR	.tar.bz2, .tbz2, .bz2	Bzip2-compressed TAR archives
XZ TAR	.tar.xz, .txz, .xz	XZ-compressed TAR archives

Archive extraction recursively processes all supported files within the archive. Each file inside the archive is extracted and processed individually, yielding separate results for each supported document found. Archives include security measures to prevent zip bomb attacks.

Installation

uv add sharepoint-to-text

Optional: faster AES handling for encrypted PDFs (avoids the slow fallback crypto and large-PDF image skips):

uv add "sharepoint-to-text[pdf-crypto]"

Or install from source:

git clone https://github.com/Horsmann/sharepoint-to-text.git
cd sharepoint-to-text
uv sync --all-groups

Libraries

Core Libraries (runtime)

These are required for normal use of the library:

• charset-normalizer: Automatic encoding detection for plain text files and MIME type detection fallback
• defusedxml: Hardened XML parsing for OOXML/ODF formats
• mail-parser: RFC 822 email parsing (.eml)
• msg-parser: Outlook .msg extraction
• olefile: OLE2 container parsing for legacy Office formats
• openpyxl: .xlsx parsing
• .7z archive extraction is handled via a built-in pure-Python reader (stdlib lzma)
• pypdf: .pdf parsing
• xlrd: .xls parsing

Format Detection: The library primarily uses file extensions for format detection, with MIME type fallback for ambiguous cases. Template formats and variants are automatically mapped to their appropriate extractors.

Development Libraries

These are only needed for development workflows:

• pytest: test runner
• pre-commit: linting/format hooks
• black: code formatter

Optional Libraries

These are opt-in extras for specific use cases:

• pycryptodome: Faster AES crypto for encrypted PDFs (pdf-crypto extra)

Quick Start

The Unified Interface

sharepoint2text.read_file(...) returns a generator of extraction results implementing a common interface. Most formats yield a single item, but some (notably .mbox) can yield multiple items.

Format Detection: Automatically detects file formats using extensions (primary) and MIME types (fallback). Template formats and variants are automatically mapped to appropriate extractors.

import sharepoint2text

# Works identically for ANY supported format
# Most formats yield a single item, so use next() for convenience
for result in sharepoint2text.read_file("document.docx"):  # or .doc, .pdf, .pptx, etc.
    # Methods available on ALL content types:
    text = result.get_full_text()  # Complete text as a single string
    metadata = result.get_metadata()  # File metadata (filename/path; plus format-specific fields when available)

    # Iterate over logical units (varies by format - see below)
    for unit in result.iterate_units():
        print(unit.get_text())

    # Iterate over extracted images
    for image in result.iterate_images():
        print(image)

    # Iterate over extracted tables
    for table in result.iterate_tables():
        print(table)

# For single-item formats, you can use next() directly:
result = next(sharepoint2text.read_file("document.docx"))
print(result.get_full_text())

Notes: ImageInterface provides get_bytes(), get_content_type(), get_caption(), get_description(), and get_metadata() (unit index, image index, content type, width, height). TableInterface provides get_table() (rows as lists) and get_dim() (rows, columns).

Most results also expose format-specific structured fields (e.g. PdfContent.pages, PptxContent.slides, XlsxContent.sheets) in addition to the common interface—see Return Types below.

JSON Output (to_json())

All extraction results support to_json() for a JSON-serializable representation of the extracted data (including nested dataclasses).

import json
import sharepoint2text

result = next(sharepoint2text.read_file("document.docx"))
print(json.dumps(result.to_json()))

To restore objects from JSON, use ExtractionInterface.from_json(...).

from sharepoint2text.parsing.extractors.data_types import ExtractionInterface

restored = ExtractionInterface.from_json(result.to_json())

Understanding iterate_units() Output by Format

Different file formats have different natural structural units:

Format	iterate_units() yields	Notes
.docx, .doc, .odt	1 item (full text)	Word/text documents have no page structure in the file format
.xlsx, .xls, .ods	1 item per sheet	Each yield contains sheet content
.pptx, .ppt, .odp	1 item per slide	Each yield contains slide text
.pdf	1 item per page	Each yield contains page text
.eml, .msg	1 item (email body)	Plain text or HTML body
.mbox	1 item per email	Mailboxes can contain multiple emails
.txt, .csv, .json, .tsv	1 item (full content)	Single unit

Note on Word documents: The .doc and .docx file formats do not store page boundaries—pages are a rendering artifact determined by fonts, margins, and printer settings. The library returns the full document as a single text unit.

Note on generators: All extractors return generators. Most formats yield a single content object, but .mbox files can yield multiple EmailContent objects (one per email in the mailbox). Use next() for single-item formats or iterate with for to handle all cases.

Skipping Images in Unit Iteration

When you don't need image data in your units, pass ignore_images=True to iterate_units() to improve performance and reduce memory usage:

result = next(sharepoint2text.read_file("document.docx"))

# Default behavior: includes images in units
for unit in result.iterate_units():
    print(f"Text: {unit.get_text()}")
    print(f"Images: {len(unit.get_images())}")  # May contain images

# Skip images: units will have empty image lists
for unit in result.iterate_units(ignore_images=True):
    print(f"Text: {unit.get_text()}")
    print(f"Images: {len(unit.get_images())}")  # Always 0

This is useful when:

• You only need text content and don't want the overhead of image processing
• You're processing documents in a text-only pipeline (e.g., full-text search indexing)
• You want to reduce memory usage for large documents with many embedded images

Note: This only affects unit iteration. result.iterate_images() still yields all images regardless of this flag.

Choosing Between get_full_text() and iterate_units()

The interface provides two methods for accessing text content, and you must decide which is appropriate for your use case:

Method	Returns	Best for
get_full_text()	All text as a single string	Simple extraction, full-text search, when structure doesn't matter
iterate_units()	Yields logical units (pages, slides, sheets)	RAG pipelines, per-unit indexing, preserving document structure

For RAG and vector storage: Consider whether storing pages/slides/sheets as separate chunks with metadata (e.g., page numbers) benefits your retrieval strategy. This allows more precise source attribution when users query your system.

# Option 1: Store entire document as one chunk
result = next(sharepoint2text.read_file("report.pdf"))
store_in_vectordb(text=result.get_full_text(), metadata={"source": "report.pdf"})

# Option 2: Store each page separately with page numbers
result = next(sharepoint2text.read_file("report.pdf"))
for page_num, unit in enumerate(result.iterate_units(), start=1):
    store_in_vectordb(
        text=unit.get_text(),
        metadata={"source": "report.pdf", "page": page_num}
    )

Trade-offs to consider:

• Per-unit storage enables citing specific pages/slides in responses, but creates more chunks
• Full-text storage is simpler and may work better for small documents
• Word documents (.doc, .docx) only yield one unit from iterate_units() since they lack page structure—for these formats, both methods are equivalent

Format-Specific Notes on get_full_text()

Template Format Support: Office template formats (.dotx, .xltx, .potx, etc.) and OpenDocument templates (.ott, .ots, .otp) are automatically processed using their respective base format extractors. Template files behave identically to regular documents during extraction.

get_full_text() is intended as a convenient "best default" for each format. In a few formats it intentionally differs from a plain "\n".join(unit.get_text() for unit in iterate_units()), or it omits optional content unless you opt in:

Format	get_full_text() default behavior	Not included by default / where to find it
.doc	Prepends metadata.title (if present) and returns main document body	footnotes, headers_footers, annotations are separate fields (DocContent)
.docx	Returns full_text (including formulas)	Comments are available on DocxContent.comments (not included in get_full_text())
.ppt	Per-slide title + body + other concatenated	Speaker notes live in slide.notes (PptSlideContent)
.pptx	Per-slide base_text plus formulas concatenated	Pass include_image_captions to PptxContent.get_full_text(...) (comments are available on PptxSlide.comments)
.odp	Per-slide text_combined concatenated	Pass include_notes/include_annotations to OdpContent.get_full_text(...)
.xls	Concatenation of sheet text blocks (no sheet names)	Sheet names are available as sheet.name (XlsSheet)
.xlsx, .ods	Includes sheet name + sheet text for each sheet	Images are available via iterate_images() / sheet image lists
.pdf	Concatenation of extracted page text	Tables/images are available via iterate_tables() / iterate_images() (PdfContent.pages)
.eml, .msg, .mbox	Returns body_plain when present, else body_html	Attachments are in EmailContent.attachments and can be extracted via iterate_supported_attachments()
.txt, .csv, .tsv, .json, .md, .html	Returns stripped content (leading/trailing whitespace removed)	Use the raw fields (.content) if you need untrimmed text
.rtf	Returns the extractor's full_text when available	iterate_units() yields per-page text when explicit \page breaks exist

Basic Usage Examples

import sharepoint2text

# Extract from any file - format auto-detected (use next() for single-item formats)
result = next(sharepoint2text.read_file("quarterly_report.docx"))
print(result.get_full_text())

# Check format support before processing
if sharepoint2text.is_supported_file("document.xyz"):
    for result in sharepoint2text.read_file("document.xyz"):
        print(result.get_full_text())

# Access metadata
result = next(sharepoint2text.read_file("presentation.pptx"))
meta = result.get_metadata()
print(f"Author: {meta.author}, Modified: {meta.modified}")
print(meta.to_dict())  # Convert to dictionary

# Process emails (mbox can contain multiple emails)
for email in sharepoint2text.read_file("mailbox.mbox"):
    print(f"From: {email.from_email.address}")
    print(f"Subject: {email.subject}")
    print(email.get_full_text())

Working with Structured Content

import sharepoint2text

# Excel: iterate over sheets
result = next(sharepoint2text.read_file("budget.xlsx"))
for sheet in result.sheets:
    print(f"Sheet: {sheet.name}")
    print(f"Rows: {len(sheet.data)}")  # List of row dictionaries
    print(sheet.text)                   # Text representation

# PowerPoint: iterate over slides
result = next(sharepoint2text.read_file("deck.pptx"))
for slide in result.slides:
    print(f"Slide {slide.slide_number}: {slide.title}")
    print(slide.content_placeholders)  # Body text
    print(slide.images)                # Image metadata

# PDF: iterate over pages
result = next(sharepoint2text.read_file("report.pdf"))
for page_num, page in enumerate(result.pages, start=1):
    print(f"Page {page_num}: {page.text[:100]}...")
    print(f"Images: {len(page.images)}")

# Email: access email-specific fields
email = next(sharepoint2text.read_file("message.eml"))
print(f"From: {email.from_email.name} <{email.from_email.address}>")
print(f"To: {', '.join(e.address for e in email.to_emails)}")
print(f"Subject: {email.subject}")
print(f"Body: {email.body_plain or email.body_html}")

Using Format-Specific Extractors with BytesIO

For API responses or in-memory data:

import sharepoint2text
import io

# Direct extractor usage with BytesIO (returns generator, use next() for single items)
with open("document.docx", "rb") as f:
    result = next(sharepoint2text.read_docx(io.BytesIO(f.read()), path="document.docx"))

# Get extractor dynamically based on filename
def extract_from_api(filename: str, content: bytes):
    extractor = sharepoint2text.get_extractor(filename)
    # Returns a generator - iterate or use next()
    return list(extractor(io.BytesIO(content), path=filename))

Archive Processing

The library automatically extracts and processes all supported files within archives:

import sharepoint2text
import io

# Process archives using the main API (auto-detects format by extension)
# Each file in the archive yields a separate result
for result in sharepoint2text.read_file("archive.zip"):
    print(f"Extracted: {result.get_metadata().filename}")
    print(f"Content: {result.get_full_text()[:200]}...")

# 7-Zip archives are supported with size limits (100MB max)
for result in sharepoint2text.read_file("documents.7z"):
    print(f"Extracted: {result.get_metadata().filename}")
    print(f"Content: {result.get_full_text()[:200]}...")

# Process mixed-content archives (documents, images, etc.)
archive_results = list(sharepoint2text.read_file("mixed_files.zip"))
documents = [r for r in archive_results if hasattr(r, 'get_full_text')]
print(f"Found {len(documents)} extractable documents in archive")

# Or use the direct extractor for BytesIO data
from sharepoint2text.parsing.extractors.archive_extractor import read_archive

with open("archive.zip", "rb") as f:
    for result in read_archive(io.BytesIO(f.read()), path="archive.zip"):
        print(f"Extracted: {result.get_metadata().filename}")
        print(f"Content: {result.get_full_text()[:200]}...")

Limitations / Caveats

PDF Extraction

• No OCR support: This library does not perform optical character recognition. PDFs that consist of scanned images or photos of documents will return empty text. The images themselves are still extracted and available via iterate_images(), but no text is derived from them.

• Table extraction not implemented: PDF table extraction is not currently implemented. iterate_tables() will always yield empty results for PDF files. Tables may appear as part of the page text in get_full_text() or iterate_units(), but structured table data is not available.

• Image extraction on large encrypted files: When a PDF is AES-encrypted and pypdf is running in its fallback crypto provider (i.e., neither cryptography nor pycryptodome is installed), image extraction is skipped for large files (>= 10MB). Text and tables still extract, but image lists are empty. Install cryptography or pycryptodome to enable full PDF image extraction without this skip.

• Password-protected PDFs: PDFs requiring a non-empty password are rejected with an ExtractionFileEncryptedError.

• JBIG2 image format: Some PDFs contain images encoded with JBIG2 compression. If you see warnings like Failed to extract image data: jbig2dec binary is not available, you can install the jbig2dec system binary to enable extraction of these images:

  • macOS: brew install jbig2dec
  • Ubuntu/Debian: sudo apt-get install jbig2dec
  • Other Linux: Use your package manager (e.g., yum install jbig2dec, pacman -S jbig2dec)

  Note: This is optional. The warning is harmless if you don't need to extract JBIG2-encoded images.

CLI

After installation, a sharepoint2text command is available. Use --file to specify the path and print the extracted full text to stdout by default.

sharepoint2text --file /path/to/file.pdf > extraction.txt

Command Line Options

Option	Output	Notes
--file FILE	(required)	Path to the file to extract. Can be specified in any order relative to other options.
--output FILE, -o FILE	Output file	Write output to file instead of stdout.
--version	Version info	Show the version and exit.
(default)	Plain text	Prints result.get_full_text().
--json	JSON extraction object(s)	Prints result.to_json(). Images are ignored by default for faster processing.
--json-unit	JSON unit list(s)	Prints a JSON list of unit representations using result.iterate_units() (e.g., pages/slides/sheets). For multi-item inputs (e.g. .mbox), emits a JSON list where each item is that extraction's unit list. Images are ignored by default for faster processing.
--include-images	Include image data	Extract images and include as base64 blobs in JSON output. Only valid with --json or --json-unit.

--json and --json-unit are mutually exclusive.

Examples

Extract text from a file:

sharepoint2text --file /path/to/file.pdf > extraction.txt

Emit structured JSON output:

sharepoint2text --file /path/to/file.pdf --json > extraction.json

Write to output file instead of stdout:

sharepoint2text --file /path/to/file.pdf --output extraction.txt
sharepoint2text --file /path/to/file.pdf --json --output extraction.json

Include extracted images as base64:

sharepoint2text --file /path/to/file.pdf --json --include-images > extraction.with-images.json

Parameters can be specified in any order:

sharepoint2text --json --include-images --file /path/to/file.pdf
sharepoint2text --file /path/to/file.pdf --json --include-images

API Reference

Main Functions

import sharepoint2text

# Read any supported file (recommended entry point)
# Returns a generator - use next() for single-item formats or iterate for all
# Automatically detects format via extension (primary) or MIME type (fallback)
for result in sharepoint2text.read_file(path: str | Path):
    ...

# Check if a file extension is supported
# Uses same detection logic as read_file(): extension-first, MIME fallback
supported = sharepoint2text.is_supported_file(path: str) -> bool

# Get extractor function for a file type
# Returns appropriate extractor based on file extension/MIME type
extractor = sharepoint2text.get_extractor(path: str) -> Callable[[io.BytesIO, str | None], Generator[ContentType, Any, None]]

Format-Specific Extractors

All accept io.BytesIO and optional path for metadata population. All return generators:

sharepoint2text.read_docx(file: io.BytesIO, path: str | None = None) -> Generator[DocxContent, Any, None]
sharepoint2text.read_doc(file: io.BytesIO, path: str | None = None) -> Generator[DocContent, Any, None]
sharepoint2text.read_xlsx(file: io.BytesIO, path: str | None = None) -> Generator[XlsxContent, Any, None]
sharepoint2text.read_xls(file: io.BytesIO, path: str | None = None) -> Generator[XlsContent, Any, None]
sharepoint2text.read_pptx(file: io.BytesIO, path: str | None = None) -> Generator[PptxContent, Any, None]
sharepoint2text.read_ppt(file: io.BytesIO, path: str | None = None) -> Generator[PptContent, Any, None]
sharepoint2text.read_odt(file: io.BytesIO, path: str | None = None) -> Generator[OdtContent, Any, None]
sharepoint2text.read_odp(file: io.BytesIO, path: str | None = None) -> Generator[OdpContent, Any, None]
sharepoint2text.read_ods(file: io.BytesIO, path: str | None = None) -> Generator[OdsContent, Any, None]
sharepoint2text.read_pdf(file: io.BytesIO, path: str | None = None) -> Generator[PdfContent, Any, None]
sharepoint2text.read_plain_text(file: io.BytesIO, path: str | None = None) -> Generator[PlainTextContent, Any, None]
sharepoint2text.read_email__eml_format(file: io.BytesIO, path: str | None = None) -> Generator[EmailContent, Any, None]
sharepoint2text.read_email__msg_format(file: io.BytesIO, path: str | None = None) -> Generator[EmailContent, Any, None]
sharepoint2text.read_email__mbox_format(file: io.BytesIO, path: str | None = None) -> Generator[EmailContent, Any, None]

Return Types

All content types implement the common interface:

class ExtractionInterface(Protocol):
    def iterate_units() -> Iterator[UnitInterface]  # Iterate over logical units
    def iterate_images() -> Generator[ImageInterface, None, None]
    def iterate_tables() -> Generator[TableInterface, None, None]
    def get_full_text() -> str                   # Complete text as string
    def get_metadata() -> FileMetadataInterface  # Metadata with to_dict()
    def to_json() -> dict                        # JSON-serializable representation
    @classmethod
    def from_json(data: dict) -> "ExtractionInterface"

DocxContent (.docx)

result.metadata       # DocxMetadata (title, author, created, modified, ...)
result.paragraphs     # List[DocxParagraph] (text, style, runs with formatting)
result.tables         # List[List[List[str]]] (cell data)
result.images         # List[DocxImage] (filename, content_type, data, size_bytes)
result.headers        # List[DocxHeaderFooter]
result.footers        # List[DocxHeaderFooter]
result.hyperlinks     # List[DocxHyperlink] (text, url)
result.footnotes      # List[DocxNote] (id, text)
result.endnotes       # List[DocxNote]
result.comments       # List[DocxComment] (author, date, text)
result.sections       # List[DocxSection] (page dimensions, margins)
result.full_text      # str (pre-computed full text)

DocContent (.doc)

result.metadata         # DocMetadata (title, author, num_pages, num_words, num_chars, ...)
result.main_text        # str (main document body)
result.footnotes        # str (concatenated footnotes)
result.headers_footers  # str (concatenated headers/footers)
result.annotations      # str (concatenated annotations)

XlsxContent / XlsContent (.xlsx, .xls)

result.metadata   # XlsxMetadata / XlsMetadata (title, creator, created, modified, ...)
result.sheets     # List[XlsxSheet / XlsSheet]

# Each sheet:
sheet.name   # str (sheet name)
sheet.data   # List[Dict[str, Any]] (rows as dictionaries)
sheet.text   # str (text representation)

PptxContent (.pptx)

result.metadata   # PptxMetadata (title, author, created, modified, ...)
result.slides     # List[PPTXSlide]

# Each slide:
slide.slide_number          # int (1-indexed)
slide.title                 # str
slide.footer                # str
slide.content_placeholders  # List[str] (body content)
slide.other_textboxes       # List[str] (free-form text)
slide.images                # List[PPTXImage] (filename, content_type, size_bytes, blob)
slide.text                  # str (pre-computed combined text)

PptContent (.ppt)

result.metadata   # PptMetadata (title, author, num_slides, created, modified, ...)
result.slides     # List[PptSlideContent]
result.all_text   # List[str] (flat list of all text)

# Each slide:
slide.slide_number   # int (1-indexed)
slide.title          # str | None
slide.body_text      # List[str]
slide.other_text     # List[str]
slide.notes          # List[str] (speaker notes)
slide.text_combined  # str (property: title + body + other)
slide.all_text       # List[PptTextBlock] (with text_type info)

OdpContent (.odp)

result.metadata   # OdpMetadata (title, creator, creation_date, generator, ...)
result.slides     # List[OdpSlide]

# Each slide:
slide.slide_number   # int (1-indexed)
slide.name           # str (slide name)
slide.title          # str
slide.body_text      # List[str]
slide.other_text     # List[str]
slide.tables         # List[List[List[str]]] (tables on slide)
slide.annotations    # List[OdpAnnotation] (comments)
slide.images         # List[OdpImage] (embedded images with href, name, data, size_bytes)
slide.notes          # List[str] (speaker notes)
slide.text_combined  # str (property: title + body + other)

OdsContent (.ods)

result.metadata   # OdsMetadata (title, creator, creation_date, generator, ...)
result.sheets     # List[OdsSheet]

# Each sheet:
sheet.name         # str (sheet name)
sheet.data         # List[Dict[str, Any]] (row data with column keys A, B, C, ...)
sheet.text         # str (tab-separated cell values, newline-separated rows)
sheet.annotations  # List[OdsAnnotation] (cell comments)
sheet.images       # List[OdsImage] (embedded images)

PdfContent (.pdf)

result.metadata    # PdfMetadata (total_pages)
result.pages       # List[PdfPage]

# Each page:
page.text    # str
page.images  # List[PdfImage] (index, name, width, height, data, format)
page.tables  # List[List[List[str]]]

PlainTextContent (.txt, .csv, .json, .tsv)

result.content   # str (full file content)
result.metadata  # FileMetadataInterface (filename, file_extension, file_path, folder_path)

EmailContent (.eml, .msg, .mbox)

result.from_email    # EmailAddress (name, address)
result.to_emails     # List[EmailAddress]
result.to_cc         # List[EmailAddress]
result.to_bcc        # List[EmailAddress]
result.reply_to      # List[EmailAddress]
result.subject       # str
result.in_reply_to   # str (message ID of parent email)
result.body_plain    # str (plain text body)
result.body_html     # str (HTML body)
result.metadata      # EmailMetadata (date, message_id, plus file metadata)

# EmailAddress structure:
email.name     # str (display name)
email.address  # str (email address)

HtmlContent (.html, .htm)

result.content   # str (plain text content)
result.tables    # List[List[List[str]]] (table cell values)
result.headings  # List[Dict[str, str]] (level/text)
result.links     # List[Dict[str, str]] (text/href)
result.metadata  # HtmlMetadata (title, language, charset, ...)

Examples

Bulk Processing

import sharepoint2text
from pathlib import Path

def extract_all_documents(folder: Path) -> dict[str, list[str]]:
    """Extract text from all supported files in a folder."""
    results = {}

    for file_path in folder.rglob("*"):
        if sharepoint2text.is_supported_file(str(file_path)):
            try:
                # Collect all content from the generator (handles mbox with multiple emails)
                texts = [result.get_full_text() for result in sharepoint2text.read_file(file_path)]
                results[str(file_path)] = texts
            except Exception as e:
                print(f"Failed to extract {file_path}: {e}")

    return results

Extract Images

import sharepoint2text

# From PDF
result = next(sharepoint2text.read_file("document.pdf"))
for page_num, page in enumerate(result.pages, start=1):
    for img in page.images:
        with open(f"page{page_num}_{img.name}.{img.format}", "wb") as out:
            out.write(img.data)

# From PowerPoint
result = next(sharepoint2text.read_file("slides.pptx"))
for slide in result.slides:
    for img in slide.images:
        with open(img.filename, "wb") as out:
            out.write(img.blob)

# From Word
result = next(sharepoint2text.read_file("document.docx"))
for img in result.images:
    if img.data:
        with open(img.filename, "wb") as out:
            out.write(img.data.getvalue())

Email Processing

import sharepoint2text

# Process a single email file (.eml or .msg)
email = next(sharepoint2text.read_file("message.eml"))
print(f"From: {email.from_email.name} <{email.from_email.address}>")
print(f"Subject: {email.subject}")
print(f"Date: {email.metadata.date}")
print(f"Body:\n{email.body_plain}")

# Process a mailbox with multiple emails (.mbox)
for i, email in enumerate(sharepoint2text.read_file("archive.mbox")):
    print(f"\n--- Email {i + 1} ---")
    print(f"From: {email.from_email.address}")
    print(f"To: {', '.join(e.address for e in email.to_emails)}")
    print(f"Subject: {email.subject}")
    if email.to_cc:
        print(f"CC: {', '.join(e.address for e in email.to_cc)}")

RAG Pipeline Integration

import sharepoint2text


def prepare_for_rag(file_path: str) -> list[dict]:
    """Prepare document chunks for RAG ingestion."""
    chunks = []

    # Handle all content items from the generator
    for result in sharepoint2text.read_file(file_path):
        meta = result.get_metadata()

        for i, unit in enumerate(result.iterate_units()):
            if unit.get_text().strip():  # Skip empty units
                chunks.append({
                    "text": unit.get_text(),
                    "metadata": {
                        "source": file_path,
                        "chunk_index": i,
                        "author": getattr(meta, "author", None),
                        "title": getattr(meta, "title", None),
                    }
                })
    return chunks

SharePoint Integration

This integration is optional; you can use sharepoint2text with any storage backend. When using sharepoint_io, you still orchestrate download and extraction (as shown below).

import io
from datetime import datetime, timedelta, timezone

import sharepoint2text
from sharepoint2text.sharepoint_io import (
    EntraIDAppCredentials,
    FileFilter,
    SharePointRestClient,
)

# Configure SharePoint access
credentials = EntraIDAppCredentials(
    tenant_id="your-tenant-id",
    client_id="your-client-id",
    client_secret="your-client-secret",
)
client = SharePointRestClient(
    site_url="https://contoso.sharepoint.com/sites/Documents",
    credentials=credentials,
)

# Delta sync: process files modified in the last 7 days
one_week_ago = datetime.now(timezone.utc) - timedelta(days=7)
file_filter = FileFilter(
    modified_after=one_week_ago,
    extensions=[".docx", ".pdf", ".pptx"],
)

for file_meta in client.list_files_filtered(file_filter):
    # Download and extract
    content = client.download_file(file_meta.id)
    extractor = sharepoint2text.get_extractor(file_meta.name)

    for result in extractor(io.BytesIO(content), path=file_meta.name):
        print(f"File: {file_meta.get_full_path()}")
        print(f"Text: {result.get_full_text()[:200]}...")

Exceptions

• ExtractionFileFormatNotSupportedError: Raised when no extractor exists for a given file type (e.g., unsupported extension/MIME mapping in the router).
• ExtractionFileEncryptedError: Raised when an extractor detects encryption or password protection (e.g., encrypted PDF, OOXML/ODF password-protected files, legacy Office with FILEPASS/encryption flags).
• ExtractionFileTooLargeError: Raised when a file exceeds the maximum allowed size for extraction (e.g., 7z archives larger than 100MB).
• LegacyMicrosoftParsingError: Raised when legacy Office parsing fails for non-encryption reasons (corrupt OLE streams, invalid headers, or unsupported legacy variations).

License

Apache 2.0 - see LICENSE for details.

Disclaimer

This project is not affiliated with, endorsed by, or sponsored by Microsoft.