unstructured

Implements a registry-based partitioning system that automatically detects document file types (PDF, DOCX, PPTX, XLSX, HTML, images, email, audio, plain text, XML) via FileType enum and routes to specialized format-specific processors through _PartitionerLoader. The partition() entry point in unstructured/partition/auto.py orchestrates this routing, dynamically loading only required dependencies for each format to minimize memory overhead and startup latency.

Implements a three-tier processing strategy pipeline for PDFs and images: FAST (PDFMiner text extraction only), HI_RES (layout detection + element extraction via unstructured-inference), and OCR_ONLY (Tesseract/Paddle OCR agents). The system automatically selects or allows explicit strategy specification, with intelligent fallback logic that escalates from text extraction to layout analysis to OCR when content is unreadable. Bounding box analysis and layout merging algorithms reconstruct document structure from spatial coordinates.

Implements table detection and extraction that preserves table structure (rows, columns, cell content) with cell-level metadata (coordinates, merged cells). Supports extraction from PDFs (via layout detection), images (via OCR), and Office documents (via native parsing). Handles complex tables (nested headers, merged cells, multi-line cells) with configurable extraction strategies.

Implements image detection and extraction from documents (PDFs, Office files, HTML) that preserves image metadata (dimensions, coordinates, alt text, captions). Supports image-to-text conversion via OCR for image content analysis. Extracts images as separate Element objects with links to source document location. Handles image preprocessing (rotation, deskewing) for improved OCR accuracy.

Implements serialization layer (unstructured/staging/base.py 103-229) that converts extracted Element objects to multiple output formats (JSON, CSV, Markdown, Parquet, XML) while preserving metadata. Supports custom serialization schemas, filtering by element type, and format-specific optimizations. Enables lossless round-trip conversion for certain formats.

Implements bounding box utilities for analyzing spatial relationships between document elements (coordinates, page numbers, relative positioning). Supports coordinate normalization across different page sizes and DPI settings. Enables spatial queries (e.g., find elements within a region) and layout reconstruction from coordinates. Used internally by layout detection and element merging algorithms.

Implements evaluation framework (unstructured/metrics/) that measures extraction quality through text metrics (precision, recall, F1 score) and table metrics (cell accuracy, structure preservation). Supports comparison against ground truth annotations and enables benchmarking across different strategies and document types. Collects processing metrics (time, memory, cost) for performance monitoring.

Provides API client abstraction (unstructured/api/) for integration with cloud document processing services and hosted Unstructured platform. Supports authentication, request batching, and result streaming. Enables seamless switching between local processing and cloud-hosted extraction for cost/performance optimization. Includes retry logic and error handling for production reliability.

Defines a typed element model (unstructured/documents/elements.py) with 20+ element types (Title, NarrativeText, Table, Image, Header, Footer, PageBreak, etc.) that represent document components. Each element carries rich metadata including bounding box coordinates, page numbers, language detection, table structure (rows/columns), image dimensions, and custom key-value pairs. The metadata system supports serialization to JSON, CSV, Markdown, and other formats while preserving structural information.

Provides chunking capabilities that split extracted elements into semantically coherent chunks optimized for embedding models and RAG retrieval. The chunking system respects element boundaries (e.g., keeps paragraphs together), supports configurable chunk size and overlap, and can leverage element metadata (type, coordinates) to make intelligent splitting decisions. Integration with LangChain enables seamless pipeline composition for vector database ingestion.

Implements specialized partitioners for Microsoft Office formats that extract text, tables, and images while preserving document structure (headings, lists, formatting). Uses python-docx, python-pptx, and openpyxl libraries to parse Office XML formats and reconstruct logical document hierarchy. Supports extraction of embedded images, hyperlinks, and table structure with cell-level granularity.

Implements HTML partitioner that parses HTML/XML documents using BeautifulSoup or lxml, extracts semantic content from tags (h1-h6 for headings, p for paragraphs, table for tables), and reconstructs document structure. Handles common web patterns (navigation, sidebars, footers) by filtering noise elements. Supports extraction of links, metadata (title, description), and image alt text.

Implements partitioners for email formats (EML, MSG, MBOX) and message protocols that extract message headers (from, to, subject, date), body text, and attachments. Reconstructs email threads by parsing In-Reply-To and References headers. Supports extraction of quoted text and signature detection to separate original content from replies.

Implements audio partitioner that transcribes speech to text using Whisper or other speech recognition models. Extracts speaker segments, timestamps, and confidence scores. Supports multiple audio formats (MP3, WAV, FLAC, OGG) and handles long-form audio by chunking into segments for processing. Integrates with language detection for multilingual support.

Implements language detection at document and element level using langdetect or textblob, enabling multilingual document processing. Detects language for each extracted element, supports language-specific text processing (e.g., CJK character handling), and enables filtering by language. Integrates with OCR agents to select language-specific models for improved accuracy.

Provides strategy configuration system (FAST, HI_RES, OCR_ONLY) that allows users to trade off speed vs accuracy based on use case. Supports per-document strategy selection, timeout configuration, and resource limits (memory, CPU). Includes metrics collection for performance monitoring and optimization. Enables fine-tuning of partitioner parameters (e.g., OCR language, layout detection thresholds).