More deterministic element ordering when using hi_res PDF parsing strategy (from unstructured-inference bump to 0.5.4)
Make large model available (from unstructured-inference bump to 0.5.3)
Combine inferred elements with extracted elements (from unstructured-inference bump to 0.5.2)
---------
Co-authored-by: Roman Isecke <roman@unstructured.io>
Co-authored-by: Crag Wolfe <crag@unstructured.io>
- Adds reusable validation scripts (check-x.sh) to minimize repeated (or near-repeated) code and create one source of truth
- Restructures the location of download and output folders such that they are nested in the test_unstructured_ingest directory
- Adds gitignore for output folders / files to avoid them accidentally getting checked into the repository
- Construct paths as reusable variables declared at top of scripts
- Sort order of flag for ingest calls, across all tests (this makes it easier to parse at a glance)
- OVERWRITE_FIXTURES removes all old fixtures for path to guarantee no stale results are left behind
- Bonus: don't check/exit on expected number of expected outputs when OVERWRITE_FIXTURES is true
- Bonus: exclude file_directory from Slack and Discord test scripts (match convention in all others)
* sets OVERWRITE_FIXTURES to default to false in test-ingest-local-single-file.sh
* fixes incorrect expected results
* update expected results to properly parse Korean text
* bonus: installs language pack for Korean in CI and ingest fixture workflows
Avoid setting metadata in constructor signature for elements because that can lead to unexpected object reuse (and modification).
Bonus refactor for PageBreak to have text values of "".
---------
Co-authored-by: Alan Bertl <alan@unstructured.io>
Co-authored-by: Crag Wolfe <crag@unstructuredai.io>
* add max partition size logic
* work splitting logic into split_by_paragraph
* pass through max_partition to other functions
* added test for splitting long document
* add type hint
* add documentation
* version and changelog
* ingest-test-fixtures-update
* Update ingest test fixtures (#819)
Co-authored-by: MthwRobinson <MthwRobinson@users.noreply.github.com>
* retrigger ci
* ingest-test-fixtures-update
* ingest-test-fixtures-update
* Update ingest test fixtures (#821)
Co-authored-by: MthwRobinson <MthwRobinson@users.noreply.github.com>
* update default for partition_xml
* update version for release
* update msg doc string
---------
Co-authored-by: MthwRobinson <MthwRobinson@users.noreply.github.com>
* feature(html partition): parse pre tag
* chore: update CHANGELOG.md
* style: black format xml.py
* Added tests dor html with pre tag
* remove skip test, update parse pre tag
* fix style
* chore: spell check
* chore: update changelog & version
* chore: update ingest test fixtures
* chore: add exception handling if `element.text` is `None` in `_read_xml`
* test: add more sanity testing on the `.text` content of the element(s)
* refactor: move the conditional logic for <pre> outside of the `try/except` block
---------
Co-authored-by: cragwolfe <crag@unstructured.io>
Co-authored-by: christinestraub <christinemstraub@gmail.com>
* initial pass on text in figures
* refactor text extraction
* update tests
* fix title test
* add test for docs that require recursive text grab
* version and changelog
* ingest-test-fixtures-update
* there are 8 pdf files now
* add support for page numbers in docx when present
* version and changelog
* add comment on page numbers
* add header and footer to doc elements list
* update integrations docs
* include_page_breaks kwarg for doc and docx
* merge element metadata for pagebreaks
* fix typo
* fix changelog typo
* change page number default to None
* add initial_page_number kwarg
* make page number tests in pdf more explicit
* revert test file
* update ingest tests
* update test fixture outputs
* updates to IRS forms fixtures
* ingest-test-fixtures-update
* Update ingest test fixtures (#759)
Co-authored-by: MthwRobinson <MthwRobinson@users.noreply.github.com>
---------
Co-authored-by: Unstructured-DevOps <111007769+Unstructured-DevOps@users.noreply.github.com>
Co-authored-by: MthwRobinson <MthwRobinson@users.noreply.github.com>
Unstructured-ingest biomed connector: Adds max retries, max request time with backoff and decay.
---------
Co-authored-by: Crag Wolfe <crag@unstructuredai.io>
Summary
* Updates "fast" PDF output element ordering to be consistent across Python versions by using the X,Y coordinates of elements extracted
* Added PDFs ingest tests with fast strategy with new script ./test_unstructured_ingest/test-ingest-pdf-fast-reprocess.sh
Updated ingest tests procedure:
* Processing files with hi_res strategy, and preserve downloads to repo files-ingest-download/<ingest_test_name>
* Reprocessing all PDFs with fast strategy from local file files-ingest-download, the partition outputs are stored at expected-structured-output/pdf-fast-reprocess/<ingest_test_name>
Test
* Reproduce tests with ./scripts/ingest-test-fixtures-update.sh , should expect no update. Also don't need any secret tokens since relevant tests won't produce PDFs.
Add functionality to try other common encodings for html, xml files if an error related to the encoding is raised and the user has not specified an encoding.
Change auto.py to have a None default for encoding
Remove the unused parameter encoding from partition_pdf
Add functionality to the read_txt_file utility function to handle file-like object from URL
Updated to the the latest version of unstructured-inference. detectron2 now gets implemented with onnxruntime, yay!
---------
Co-authored-by: Matt Robinson <mrobinson@unstructured.io>
* Initial commit of discord connector
based off of initial work by @tnachen with modifications
https://github.com/tnachen/unstructured/tree/tnachen/discord_connector
* Add test file
change format of imports
* working version of the connector
More work to be done to tidy it up and add any additional options
* add to test fixtures update
* fix spacing
* tests working, switching to bot testing channel
* add additional channel
add reprocess to tests
* add try clause to allow for exit on error
Update changelog and bump version
* add updated expected output filtes
* add logic to check if —discord-period is an integer
Add more to option description
* fix lint error
* Update discord reqs
* PR feedback
* add newline
* another newline
---------
Co-authored-by: Justin Bossert <packerbacker21@hotmail.com>
* change strategy arg defalut to auto in partition
* passing --partition-strategy down
* add strategy="hi_res" to test (default changed)
* made an error on param name, added note
Adds filetype to metadata. I've created a decorator that adds metadata to a list of elements. This replaces some existing boilerplate, but also adds a nice layered approach to determining the filetype. Since in some cases several partition_ functions handle a file in various formats, the partition function that first touches a file will be the last one to alter its metadata, resulting in the correct filetype metadata.
Tests are added to make sure:
* When partition is used, any content type or auto file type detection will override file-specific partition function metadata
* Both auto and file-specific partitioning gives the desired filetype metadata
Won't work with image files currently... the plumbing is there to use the image format inferred by PIL, but we need to pull in the fix from this PR to unstructured-inference .
* added functions for determining auto stratgy
* change default strategy to auto
* tests for auto strategy
* update docs
* changelog and version
* bump version
* remove ingest file in wrong location
* update jpg output
* typo fix
* added method for extracting datetime
* change filename metadata to the base filename
* fix filename metadata for msg
* changelog and bump version
* fix expected structured output
* newline back in file
* reset outpout file
* update filename output
* update test fixtures
* update fixture
* switch to using PDF objects
* linting, linting, linting
* couple more tweaks
* added test for chevron-page
* version and changelog
* linting, linting, linting
* now processing 4 files
* function to check if pdf is extractable
* add fallback logic for unextractable pdfs
* tests for docs with copy protection
* add test for unprocessable pdf
* update docs
* changelog and version
* update logic for images; reset file before proceeding
* 3 files for api tests
* docs update
* pip-compile new reqs
* bump inference version
* add language to pdf and image calls
* tests for passing in language
* version bump and changelog
* update docs
* pass ocr_languages in auto
* updated test fixtures
* typo in doc string
* group broken paragraphs with fast strategy
* changelog and version
* fix broken tests for text.py
* formatting for paragraph pattern re
* fix test
* fix whitespace substitution
* one more test tweak
* blurb to account for short lines
* fix for shorter paragraphs
* update changelog
* remove extra line break from auto
* retrigger ci
* trying skipping azure
* skip azure (test)
* updated github and azure fixtures
* update slack fixture
Previously, if there was an error (non-zero exit code) in an ingest test script,
the script would still complete and echo a warning about mismatched outputs
and how to regenerate the fixtures. However, this statement is irrelevant and
misleading: if the ingest failed with a non-zero exit code in the first place,
that is the failure that should be debugged -- don't confuse the user with
a comment about outputs.
This connector takes a slack channel id, token and other options to
pull conversation history for a channel and store it as a text file that
is then processed by unstructured into expected output.
* Update test fixtures that should have been updated in prior commit
* Disable biomed ingest tests for now, the fail more often than not
* Bonus: echo `tesseract --version` in the update script, since that is a key thing that influences fixture outputs.
* Add --partition-by-api and --partition-host args to ingest
* Fix error in make check
* Bump changelog
* Add a test ingest script
Also add a workaround for the test causing 400s from our api. Seems we need to make sure
unstructured-api can handle getting a file.content_type of None.
* Remove the content type workaround
- Updates CI to install tesseract version 5.3.0 (better than 4.x in various ways incl. perf.).
- Adds azure expected output fixtures for more useful reference points and as a repro for Some PDF's with scanned images return empty elements #346 .
- Adds a script to regenerate ingest test fixtures that is run in an ubuntu docker container (like CI), with the same version of tesseract. See the comments in scripts/ingest-test-fixtures-update.sh for details.
- Updates expected outputs with above script.
- Updates individual test-ingest scripts to update expected .json output if OVERWRITE_FIXTURES=true.
