0.14.1

Enhancements

• Refactor code related to embedded text extraction. The embedded text extraction code is moved from unstructured-inference to unstructured.

Features

• Large improvements to the ingest process:
  • Support for multiprocessing and async, with limits for both.
  • Streamlined to process when mapping CLI invocations to the underlying code
  • More granular steps introduced to give better control over process (i.e. dedicated step to uncompress files already in the local filesystem, new optional staging step before upload)
  • Use the python client when calling the unstructured api for partitioning or chunking
  • Saving the final content is now a dedicated destination connector (local) set as the default if none are provided. Avoids adding new files locally if uploading elsewhere.
  • Leverage last modified date when deciding if new files should be downloaded and reprocessed.
  • Add attribution to the pinecone connector
• Add support for Python 3.12. unstructured now works with Python 3.12!

0.14.0

0.14.0

BREAKING CHANGES

• Turn table extraction for PDFs and images off by default. Reverting the default behavior for table extraction to "off" for PDFs and images. A number of users didn't realize we made the change and were impacted by slower processing times due to the extra model call for table extraction.

Enhancements

• Skip unnecessary element sorting in partition_pdf(). Skip element sorting when determining whether embedded text can be extracted.
• Faster evaluation Support for concurrent processing of documents during evaluation
• Add strategy parameter to partition_docx(). Behavior of future enhancements may be sensitive the partitioning strategy. Add this parameter so partition_docx() is aware of the requested strategy.
• Add GLOBAL_WORKING_DIR and GLOBAL_WORKING_PROCESS_DIR configuration parameteres to control temporary storage.

Features

• Add form extraction basics (document elements and placeholder code in partition). This is to lay the ground work for the future. Form extraction models are not currently available in the library. An attempt to use this functionality will end in a NotImplementedError.

Fixes

• Add missing starting_page_num param to partition_image
• Make the filename and file params for partition_image and partition_pdf match the other partitioners
• Fix include_slide_notes and include_page_breaks params in partition_ppt
• Re-apply: skip accuracy calculation feature Overwritten by mistake
• Fix type hint for paragraph_grouper param paragraph_grouper can be set to False, but the type hint did not not reflect this previously.
• Remove links param from partition_pdf links is extracted during partitioning and is not needed as a paramter in partition_pdf.
• Improve CSV delimeter detection. partition_csv() would raise on CSV files with very long lines.
• Fix disk-space leak in partition_doc(). Remove temporary file created but not removed when file argument is passed to partition_doc().
• Fix possible SyntaxError or SyntaxWarning on regex patterns. Change regex patterns to raw strings to avoid these warnings/errors in Python 3.11+.
• Fix disk-space leak in partition_odt(). Remove temporary file created but not removed when file argument is passed to partition_odt().
• AstraDB: option to prevent indexing metadata

0.13.7

Enhancements

• Remove page_number metadata fields for HTML partition until we have a better strategy to decide page counting.
• Extract OCRAgent.get_agent(). Generalize access to the configured OCRAgent instance beyond its use for PDFs.
• Add calculation of table related metrics which take into account colspans and rowspans

Features

• add ability to get ratio of cid characters in embedded text extracted by pdfminer.

Fixes

• partition_docx() handles short table rows. The DOCX format allows a table row to start late and/or end early, meaning cells at the beginning or end of a row can be omitted. While there are legitimate uses for this capability, using it in practice is relatively rare. However, it can happen unintentionally when adjusting cell borders with the mouse. Accommodate this case and generate accurate .text and .metadata.text_as_html for these tables.
• Remedy macOS test failure not triggered by CI. Generalize temp-file detection beyond hard-coded Linux-specific prefix.
• Remove unnecessary warning log for using default layout model.
• Add chunking to partition_tsv Even though partition_tsv() produces a single Table element, chunking is made available because the Table element is often larger than the desired chunk size and must be divided into smaller chunks.

0.13.6

0.13.6

Enhancements

Features

Fixes

• ValueError: Invalid file (FileType.UNK) when parsing Content-Type header with charset directive URL response Content-Type headers are now parsed according to RFC 9110.

0.13.5

0.13.5

Enhancements

Features

Fixes

• KeyError raised when updating parent_id In the past, combining ListItem elements could result in reusing the same memory location which then led to unexpected side effects when updating element IDs.
• Bump unstructured-inference==0.7.29: table transformer predictions are now removed if confidence is below threshold

0.13.4

Enhancements

• Unique and deterministic hash IDs for elements Element IDs produced by any partitioning
  function are now deterministic and unique at the document level by default. Before, hashes were
  based only on text; however, they now also take into account the element's sequence number on a
  page, the page's number in the document, and the document's file name.
• Enable remote chunking via unstructured-ingest Chunking using unstructured-ingest was
  previously limited to local chunking using the strategies basic and by_title. Remote chunking
  options via the API are now accessible.
• Save table in cells format. UnstructuredTableTransformerModel is able to return predicted table in cells format

Features

• Add a PDF_ANNOTATION_THRESHOLD environment variable to control the capture of embedded links in partition_pdf() for fast strategy.
• Add integration with the Google Cloud Vision API. Adds a third OCR provider, alongside Tesseract and Paddle: the Google Cloud Vision API.

Fixes

• Remove ElementMetadata.section field.. This field was unused, not populated by any partitioners.

0.13.3

Enhancements

• Remove duplicate image elements. Remove image elements identified by PDFMiner that have similar bounding boxes and the same text.
• Add support for start_index in html links extraction
• Add strategy arg value to _PptxPartitionerOptions. This makes this paritioning option available for sub-partitioners to come that may optionally use inference or other expensive operations to improve the partitioning.
• Support pluggable sub-partitioner for PPTX Picture shapes. Use a distinct sub-partitioner for partitioning PPTX Picture (image) shapes and allow the default picture sub-partitioner to be replaced at run-time by one of the user's choosing.
• Introduce starting_page_number parameter to partitioning functions It applies to those partitioners which support page_number in element's metadata: PDF, TIFF, XLSX, DOC, DOCX, PPT, PPTX.
• Redesign the internal mechanism of assigning element IDs This allows for further enhancements related to element IDs such as deterministic and document-unique hashes. The way partitioning functions operate hasn't changed, which means unique_element_ids continues to be False by default, utilizing text hashes.

Features

Fixes

• Add support for extracting text from tag tails in HTML. This fix adds ability to generate separate elements using tag tails.
• Add support for extracting text from <b> tags in HTML Now partition_html() can extract text from <b> tags inside container tags (like <div>, <pre>).
• Fix pip-compile make target Missing base.in dependency missing from requirments make file added

0.13.2

0.13.2

Enhancements

Features

Fixes

• Brings back missing word list files that caused partition failures in 0.13.1.

0.13.1

0.13.1

Enhancements

• Drop constraint on pydantic, supporting later versions All dependencies has pydantic pinned at an old version. This explicit pin was removed, allowing the latest version to be pulled in when requirements are compiled.

Features

• Add a set of new ElementTypes to extend future element types

Fixes

• Fix partition_html() swallowing some paragraphs. The partition_html() only considers elements with limited depth to avoid becoming the text representation of a giant div. This fix increases the limit value.
• Fix SFTP Adds flag options to SFTP connector on whether to use ssh keys / agent, with flag values defaulting to False. This is to prevent looking for ssh files when using username and password. Currently, username and password are required, making that always the case.

0.13.0

0.13.0

Enhancements

• Add .metadata.is_continuation to text-split chunks. .metadata.is_continuation=True is added to second-and-later chunks formed by text-splitting an oversized Table element but not to their counterpart Text element splits. Add this indicator for CompositeElement to allow text-split continuation chunks to be identified for downstream processes that may wish to skip intentionally redundant metadata values in continuation chunks.
• Add compound_structure_acc metric to table eval. Add a new property to unstructured.metrics.table_eval.TableEvaluation: composite_structure_acc, which is computed from the element level row and column index and content accuracy scores
• Add .metadata.orig_elements to chunks. .metadata.orig_elements: list[Element] is added to chunks during the chunking process (when requested) to allow access to information from the elements each chunk was formed from. This is useful for example to recover metadata fields that cannot be consolidated to a single value for a chunk, like page_number, coordinates, and image_base64.
• Add --include_orig_elements option to Ingest CLI. By default, when chunking, the original elements used to form each chunk are added to chunk.metadata.orig_elements for each chunk. * The include_orig_elements parameter allows the user to turn off this behavior to produce a smaller payload when they don't need this metadata.
• Add Google VertexAI embedder Adds VertexAI embeddings to support embedding via Google Vertex AI.

Features

• Chunking populates .metadata.orig_elements for each chunk. This behavior allows the text and metadata of the elements combined to make each chunk to be accessed. This can be important for example to recover metadata such as .coordinates that cannot be consolidated across elements and so is dropped from chunks. This option is controlled by the include_orig_elements parameter to partition_*() or to the chunking functions. This option defaults to True so original-elements are preserved by default. This behavior is not yet supported via the REST APIs or SDKs but will be in a closely subsequent PR to other unstructured repositories. The original elements will also not serialize or deserialize yet; this will also be added in a closely subsequent PR.
• Add Clarifai destination connector Adds support for writing partitioned and chunked documents into Clarifai.

Fixes

• Fix clean_pdfminer_inner_elements() to remove only pdfminer (embedded) elements merged with inferred elements. Previously, some embedded elements were removed even if they were not merged with inferred elements. Now, only embedded elements that are already merged with inferred elements are removed.
• Clarify IAM Role Requirement for GCS Platform Connectors. The GCS Source Connector requires Storage Object Viewer and GCS Destination Connector requires Storage Object Creator IAM roles.
• Change table extraction defaults Change table extraction defaults in favor of using skip_infer_table_types parameter and reflect these changes in documentation.
• Fix OneDrive dates with inconsistent formatting Adds logic to conditionally support dates returned by office365 that may vary in date formatting or may be a datetime rather than a string. See previous fix for SharePoint
• Adds tracking for AstraDB Adds tracking info so AstraDB can see what source called their api.
• Support AWS Bedrock Embeddings in ingest CLI The configs required to instantiate the bedrock embedding class are now exposed in the api and the version of boto being used meets the minimum requirement to introduce the bedrock runtime required to hit the service.
• Change MongoDB redacting Original redact secrets solution is causing issues in platform. This fix uses our standard logging redact solution.