Unstructured

This release focuses on performance optimizations across several internal functions and resolves a bug related to partially extracted elements. A critical dependency upgrade to pdfminer-six addresses a performance regression.

Version 0.18.26 pins deltalake to resolve ARM64 build issues, while 0.18.25 includes a security update by relaxing constraints and bumping pdfminer.six and urllib3.

This release includes an optimization for OCR extraction and a critical security update by bumping dependencies.

This patch release primarily addresses a security vulnerability by updating the fonttools dependency.

This release primarily focuses on security by updating the "unstructured-inference" dependency to version 1.1.2 to address known CVEs.

This release improves the VoyageAI integration, adds support for voyage-context-3, and enhances metadata flagging for extracted elements.

This set of releases focuses heavily on security by addressing multiple CVEs through dependency bumps and fixing a critical path traversal vulnerability in MSG attachment handling. Additionally, performance was improved in hash ID assignment.

This release focuses on performance improvements for paragraph grouping and HTML element processing, alongside dependency updates to address security vulnerabilities.

This release focuses on performance improvements, notably speeding up the sentence_count function, and addresses several security vulnerabilities by updating dependencies.

This patch improves the robustness of email parsing by handling more diverse date formats within email headers.

This release improves error handling during encoding detection by replacing UnicodeDecodeError with UnprocessableEntityError to prevent logging large file contents.

This release introduces support for '|' as a CSV delimiter and adds mapping for <input> tags based on their type attribute. It also switches the dependency used for character set normalization.

This minor release introduces a new environment variable, OCR_AGENT_CACHE_SIZE, for better memory management in OCR agents.

Version 0.18.9 introduces a new feature to convert elements to markdown and includes fixes for language detection with empty text and handling password-protected XLSX files.

This release introduces language detection capabilities for PDF documents at both the document and element levels.

This release focuses on stability, improving EPUB partitioning error handling and correcting serialization types for TableChunks.

This minor release primarily addresses an issue by increasing the field limit when parsing CSV files.

This minor release bumps the pillow dependency to address a security vulnerability (CVE).

This release focuses on stability and correctness, addressing several bugs related to HTML parsing, XML escaping, Markdown encoding, and table processing.

This release introduces the DocumentData element type for better handling of large document data and fixes an issue related to the encoding property in _CsvPartitioningContext.

This release drops support for Python 3.9 due to dependency conflicts and includes numerous bug fixes related to PDF parsing, stability, and security vulnerabilities. It also deprecates the `stage_for_label_studio` utility.

This release introduces the extraction of image URLs from HTML partitions and significantly speeds up hOCR parsing by switching to lxml. It also bumps the minimum required numpy version to greater than 2.

Version 0.17.0 introduces image inclusion during HTML partitioning and allows passing OCR/table agents, while removing the deprecated PageLayout.elements reference.

This minor release primarily addresses a bug in filetype detection when handling JSON byte streams.

This release introduces dynamic file type registration for partitioners and enhances image block extraction for CamelCase element types. It also adds support for converting JSON elements to HTML.

This minor release primarily addresses a bug in file type detection when handling SpooledTemporaryFile objects.

This minor release focuses on security by addressing open CVES and updating underlying dependencies.

This release introduces password support for PDF loading and new configuration options for PDF Miner, alongside performance improvements in layout merging and a fix for NDJSON file detection.

This release addresses a critical security vulnerability related to file inclusion in rst and org files by implementing sandboxing for file partitioning.

This release primarily focuses on fixing critical bugs related to table extraction logic and HTML partitioning within table cells. It also updates internal tooling configurations like `make tidy`.

This minor release refactors the VoyageAI integration and fixes a bug related to text ordering in layout element construction.

This release introduces vectorized data structures for layout processing to improve performance and fixes an issue with NLTK auto-downloading and a patch in pdfminer that caused token splitting errors.

This release updates the versions of `unstructured-inference` and `pdfminer-six`, resulting in the removal of `layoutparser` related dependencies from `unstructured-inference`.

This release primarily addresses a bug related to redundant passing of the 'infer_table_structure' argument during email partitioning with image attachments.

This release introduces character-level filtering for Tesseract output and resolves an issue with NLTK asset usage in Docker images, while removing automatic NLTK package downloading.