Unstructured

Enhancements

* **Support nested DOCX tables.** In DOCX, like HTML, a table cell can itself contain a table. In this case, create nested HTML tables to reflect that structure and create a plain-text table with captures all the text in nested tables, formatting it as a reasonable facsimile of a table.
* **Add connection check to ingest connectors** Each source and destination connector now support a `check_connection()` method which makes sure a valid connection can be established with the source/destination given any authentication credentials in a lightweight request.

Features

* **Add functionality to do a second OCR on cropped table images.** Changes to the values for scaling ENVs affect entire page OCR output(OCR regression) so we now do a second OCR for tables.
* **Adds ability to pass timeout for a request when partitioning via a `url`.** `partition` now accepts a new optional parameter `request_timeout` which if set will prevent any `requests.get` from hanging indefinitely and instead will raise a timeout error. This is useful when partitioning a url that may be slow to respond or may not respond at all.

Fixes

* **Fix logic that determines pdf auto strategy.** Previously, `_determine_pdf_auto_strategy` returned `hi_res` strategy only if `infer_table_structure` was true. It now returns the `hi_res` strategy if either `infer_table_structure` or `extract_images_in_pdf` is true.
* **Fix invalid coordinates when parsing tesseract ocr data.** Previously, when parsing tesseract ocr data, the ocr data had invalid bboxes if zoom was set to `0`. A logical check is now added to avoid such error.
* **Fix ingest partition parameters not being passed to the api.** When using the --partition-by-api flag via unstructured-ingest, none of the partition arguments are forwarded, meaning that these options are disregarded. With this change, we now pass through all of the relevant partition arguments to the api. This allows a user to specify all of the same partition arguments they would locally and have them respected when specifying --partition-by-api.
* **Support tables in section-less DOCX.** Generalize solution for MS Chat Transcripts exported as DOCX by including tables in the partitioned output when present.
* **Support tables that contain only numbers when partitioning via `ocr_only`** Tables that contain only numbers are returned as floats in a pandas.DataFrame when the image is converted from `.image_to_data()`. An AttributeError was raised downstream when trying to `.strip()` the floats.
* **Improve DOCX page-break detection.** DOCX page breaks are reliably indicated by `w:lastRenderedPageBreak` elements present in the document XML. Page breaks are NOT reliably indicated by "hard" page-breaks inserted by the author and when present are redundant to a `w:lastRenderedPageBreak` element so cause over-counting if used. Use rendered page-breaks only.

Enhancements

* **Adds include_header argument for partition_csv and partition_tsv** Now supports retaining header rows in CSV and TSV documents element partitioning.
* **Add retry logic for all source connectors** All http calls being made by the ingest source connectors have been isolated and wrapped by the `SourceConnectionNetworkError` custom error, which triggers the retry logic, if enabled, in the ingest pipeline.
* **Google Drive source connector supports credentials from memory** Originally, the connector expected a filepath to pull the credentials from when creating the client. This was expanded to support passing that information from memory as a dict if access to the file system might not be available.
* **Add support for generic partition configs in ingest cli** Along with the explicit partition options supported by the cli, an `additional_partition_args` arg was added to allow users to pass in any other arguments that should be added when calling partition(). This helps keep any changes to the input parameters of the partition() exposed in the CLI.
* **Map full output schema for table-based destination connectors** A full schema was introduced to map the type of all output content from the json partition output and mapped to a flattened table structure to leverage table-based destination connectors. The delta table destination connector was updated at the moment to take advantage of this.
* **Incorporate multiple embedding model options into ingest, add diff test embeddings** Problem: Ingest pipeline already supported embedding functionality, however users might want to use different types of embedding providers. Enhancement: Extend ingest pipeline so that users can specify and embed via a particular embedding provider from a range of options. Also adds a diff test to compare output from an embedding module with the expected output

Features

* **Allow setting table crop parameter** In certain circumstances, adjusting the table crop padding may improve table.

Fixes

* **Fixes `partition_text` to prevent empty elements** Adds a check to filter out empty bullets.
* **Handle empty string for `ocr_languages` with values for `languages`** Some API users ran into an issue with sending `languages` params because the API defaulted to also using an empty string for `ocr_languages`. This update handles situations where `languages` is defined and `ocr_languages` is an empty string.
* **Fix PDF tried to loop through None** Previously the PDF annotation extraction tried to loop through `annots` that resolved out as None. A logical check added to avoid such error.
* **Ingest session handler not being shared correctly** All ingest docs that leverage the session handler should only need to set it once per process. It was recreating it each time because the right values weren't being set nor available given how dataclasses work in python.
* **Ingest download-only fix.** Previously the download only flag was being checked after the doc factory pipeline step, which occurs before the files are actually downloaded by the source node. This check was moved after the source node to allow for the files to be downloaded first before exiting the pipeline.
* **Fix flaky chunk-metadata.** Prior implementation was sensitive to element order in the section resulting in metadata values sometimes being dropped. Also, not all metadata items can be consolidated across multiple elements (e.g. coordinates) and so are now dropped from consolidated metadata.
* **Fix tesseract error `Estimating resolution as X`** leaded by invalid language parameters input. Proceed with defalut language `eng` when `lang.py` fails to find valid language code for tesseract, so that we don't pass an empty string to tesseract CLI and raise an exception in downstream.

Enhancements

* **Add table structure evaluation helpers** Adds functions to evaluate the similarity between predicted table structure and actual table structure.
* **Use `yolox` by default for table extraction when partitioning pdf/image** `yolox` model provides higher recall of the table regions than the quantized version and it is now the default element detection model when `infer_table_structure=True` for partitioning pdf/image files
* **Remove pdfminer elements from inside tables** Previously, when using `hi_res` some elements where extracted using pdfminer too, so we removed pdfminer from the tables pipeline to avoid duplicated elements.
* **Fsspec downstream connectors** New destination connector added to ingest CLI, users may now use `unstructured-ingest` to write to any of the following:
* Azure
* Box
* Dropbox
* Google Cloud Service

Features

* **Update `ocr_only` strategy in `partition_pdf()`** Adds the functionality to get accurate coordinate data when partitioning PDFs and Images with the `ocr_only` strategy.

Fixes

* **Fixed SharePoint permissions for the fetching to be opt-in** Problem: Sharepoint permissions were trying to be fetched even when no reletad cli params were provided, and this gave an error due to values for those keys not existing. Fix: Updated getting keys to be with .get() method and changed the "skip-check" to check individual cli params rather than checking the existance of a config object.
* **Fixes issue where tables from markdown documents were being treated as text** Problem: Tables from markdown documents were being treated as text, and not being extracted as tables. Solution: Enable the `tables` extension when instantiating the `python-markdown` object. Importance: This will allow users to extract structured data from tables in markdown documents.
* **Fix wrong logger for paddle info** Replace the logger from unstructured-inference with the logger from unstructured for paddle_ocr.py module.
* **Fix ingest pipeline to be able to use chunking and embedding together** Problem: When ingest pipeline was using chunking and embedding together, embedding outputs were empty and the outputs of chunking couldn't be re-read into memory and be forwarded to embeddings. Fix: Added CompositeElement type to TYPE_TO_TEXT_ELEMENT_MAP to be able to process CompositeElements with unstructured.staging.base.isd_to_elements
* **Fix unnecessary mid-text chunk-splitting.** The "pre-chunker" did not consider separator blank-line ("\n\n") length when grouping elements for a single chunk. As a result, sections were frequently over-populated producing a over-sized chunk that required mid-text splitting.
* **Fix frequent dissociation of title from chunk.** The sectioning algorithm included the title of the next section with the prior section whenever it would fit, frequently producing association of a section title with the prior section and dissociating it from its actual section. Fix this by performing combination of whole sections only.
* **Fix PDF attempt to get dict value from string.** Fixes a rare edge case that prevented some PDF's from being partitioned. The `get_uris_from_annots` function tried to access the dictionary value of a string instance variable. Assign `None` to the annotation variable if the instance type is not dictionary to avoid the erroneous attempt.

Enhancements

* **Leverage dict to share content across ingest pipeline** To share the ingest doc content across steps in the ingest pipeline, this was updated to use a multiprocessing-safe dictionary so changes get persisted and each step has the option to modify the ingest docs in place.

Features

Fixes

* **Removed `ebooklib` as a dependency** `ebooklib` is licensed under AGPL3, which is incompatible with the Apache 2.0 license. Thus it is being removed.
* **Caching fixes in ingest pipeline** Previously, steps like the source node were not leveraging parameters such as `re_download` to dictate if files should be forced to redownload rather than use what might already exist locally.

Enhancements

* **Add text CCT CI evaluation workflow** Adds cct text extraction evaluation metrics to the current ingest workflow to measure the performance of each file extracted as well as aggregated-level performance.

Features

* **Functionality to catch and classify overlapping/nested elements** Method to identify overlapping-bboxes cases within detected elements in a document. It returns two values: a boolean defining if there are overlapping elements present, and a list reporting them with relevant metadata. The output includes information about the `overlapping_elements`, `overlapping_case`, `overlapping_percentage`, `largest_ngram_percentage`, `overlap_percentage_total`, `max_area`, `min_area`, and `total_area`.
* **Add Local connector source metadata** python's os module used to pull stats from local file when processing via the local connector and populates fields such as last modified time, created time.

Fixes

* **Fixes elements partitioned from an image file missing certain metadata** Metadata for image files, like file type, was being handled differently from other file types. This caused a bug where other metadata, like the file name, was being missed. This change brought metadata handling for image files to be more in line with the handling for other file types so that file name and other metadata fields are being captured.
* **Adds `typing-extensions` as an explicit dependency** This package is an implicit dependency, but the module is being imported directly in `unstructured.documents.elements` so the dependency should be explicit in case changes in other dependencies lead to `typing-extensions` being dropped as a dependency.
* **Stop passing `extract_tables` to `unstructured-inference` since it is now supported in `unstructured` instead** Table extraction previously occurred in `unstructured-inference`, but that logic, except for the table model itself, is now a part of the `unstructured` library. Thus the parameter triggering table extraction is no longer passed to the `unstructured-inference` package. Also noted the table output regression for PDF files.
* **Fix a bug in Table partitioning** Previously the `skip_infer_table_types` variable used in `partition` was not being passed down to specific file partitioners. Now you can utilize the `skip_infer_table_types` list variable when calling `partition` to specify the filetypes for which you want to skip table extraction, or the `infer_table_structure` boolean variable on the file specific partitioning function.
* **Fix partition docx without sections** Some docx files, like those from teams output, do not contain sections and it would produce no results because the code assumes all components are in sections. Now if no sections is detected from a document we iterate through the paragraphs and return contents found in the paragraphs.
* **Fix out-of-order sequencing of split chunks.** Fixes behavior where "split" chunks were inserted at the beginning of the chunk sequence. This would produce a chunk sequence like [5a, 5b, 3a, 3b, 1, 2, 4] when sections 3 and 5 exceeded `max_characters`.
* **Deserialization of ingest docs fixed** When ingest docs are being deserialized as part of the ingest pipeline process (cli), there were certain fields that weren't getting persisted (metadata and date processed). The from_dict method was updated to take these into account and a unit test added to check.
* **Map source cli command configs when destination set** Due to how the source connector is dynamically called when the destination connector is set via the CLI, the configs were being set incorrectoy, causing the source connector to break. The configs were fixed and updated to take into account Fsspec-specific connectors.

Enhancements

* **Duplicate CLI param check** Given that many of the options associated with the `Click` based cli ingest commands are added dynamically from a number of configs, a check was incorporated to make sure there were no duplicate entries to prevent new configs from overwriting already added options.
* **Ingest CLI refactor for better code reuse** Much of the ingest cli code can be templated and was a copy-paste across files, adding potential risk. Code was refactored to use a base class which had much of the shared code templated.

Features

* **Table OCR refactor** support Table OCR with pre-computed OCR data to ensure we only do one OCR for entrie document. User can specify
ocr agent tesseract/paddle in environment variable `OCR_AGENT` for OCRing the entire document.
* **Adds accuracy function** The accuracy scoring was originally an option under `calculate_edit_distance`. For easy function call, it is now a wrapper around the original function that calls edit_distance and return as "score".
* **Adds HuggingFaceEmbeddingEncoder** The HuggingFace Embedding Encoder uses a local embedding model as opposed to using an API.
* **Add AWS bedrock embedding connector** `unstructured.embed.bedrock` now provides a connector to use AWS bedrock's `titan-embed-text` model to generate embeddings for elements. This features requires valid AWS bedrock setup and an internet connectionto run.

Fixes

* **Import PDFResourceManager more directly** We were importing `PDFResourceManager` from `pdfminer.converter` which was causing an error for some users. We changed to import from the actual location of `PDFResourceManager`, which is `pdfminer.pdfinterp`.
* **Fix language detection of elements with empty strings** This resolves a warning message that was raised by `langdetect` if the language was attempted to be detected on an empty string. Language detection is now skipped for empty strings.
* **Fix chunks breaking on regex-metadata matches.** Fixes "over-chunking" when `regex_metadata` was used, where every element that contained a regex-match would start a new chunk.
* **Fix regex-metadata match offsets not adjusted within chunk.** Fixes incorrect regex-metadata match start/stop offset in chunks where multiple elements are combined.
* **Map source cli command configs when destination set** Due to how the source connector is dynamically called when the destination connector is set via the CLI, the configs were being set incorrectoy, causing the source connector to break. The configs were fixed and updated to take into account Fsspec-specific connectors.
* **Fix metrics folder not discoverable** Fixes issue where unstructured/metrics folder is not discoverable on PyPI by adding an `__init__.py` file under the folder.
* **Fix a bug when `parition_pdf` get `model_name=None`** In API usage the `model_name` value is `None` and the `cast` function in `partition_pdf` would return `None` and lead to attribution error. Now we use `str` function to explicit convert the content to string so it is garanteed to have `starts_with` and other string functions as attributes
* **Fix html partition fail on tables without `tbody` tag** HTML tables may sometimes just contain headers without body (`tbody` tag)