To implement inter-pre-chunk overlap, we need a context that sees every
pre-chunk both before and after it is accumulated (from elements).
- We need access to the pre-chunk when it is completed so we can extract
the "tail" overlap to be applied to the next chunk.
- We need access to the as-yet-unpopulated pre-chunk so we can add the
prior tail to it as a prefix.
This "visibility" is split between `PreChunkBuilder` and the pre-chunker
itself, which handles `TablePreChunk`s without the builder.
Move `Table` element and TablePreChunk` formation into `PreChunkBuilder`
such that _all_ element types (adding `Table` elements in particular)
pass through it. Then `PreChunkBuilder` becomes the context we require.
The actual overlap harvesting and application will come in a subsequent
commit.
Many of the classes defined in `unstructured.chunking.title` are
applicable to any chunking strategy and will shortly be used for the
"by-character" chunking strategy as well.
Move these and their tests to `unstructured.chunking.base`.
Along the way, rename `TextPreChunkBuilder` to `PreChunkBuilder` because
it will be generalized in a subsequent PR to also take `Table` elements
such that inter-pre-chunk overlap can be implemented.
Otherwise, no logic changes, just moves.
Chunking options for things like chunk-size are largely independent of
chunking strategy. Further, validating the args and applying defaults
based on call arguments is sophisticated to make its use easy for the
caller. These details distract from what the chunker is actually doing
and would need to be repeated for every chunking strategy if left where
they are.
Extract these settings and the rules governing chunking behavior based
on options into its own immutable object that can be passed to any
component that is subject to optional behavior (pretty much all of
them).
The chunking subpackage `unstructured.chunking` currently contains only
the `title` module and the `@add_chunking_strategy()` decorator is
located in that module even though it has no special relationship to the
`by_title` chunking strategy.
Move it to the `__init__.py` module such that it is exported from
`unstructured.chunking`. Adjust all references, pretty much one per
partitioner, to import it from there.
This prepares the way for further separation of the chunking package
into modules, including a new `character` module for the `by_character`
chunking strategy.
This PR addresses
[CORE-2969](https://unstructured-ai.atlassian.net/browse/CORE-2969)
- pdfminer sometimes fail to decode text in an pdf file and returns cid
codes as text
- now those text will be considered invalid and be replaced with ocr
results in `hi_res` mode
## test
This PR adds unit test for the utility functions. In addition the file
below would return elements with text in cid code on main but proper
ascii text with this PR:
[005-CISA-AA22-076-Strengthening-Cybersecurity-p1-p4.pdf](https://github.com/Unstructured-IO/unstructured/files/13662984/005-CISA-AA22-076-Strengthening-Cybersecurity-p1-p4.pdf)
This change improves both cct accuracy and %missing scores:
**before:**
```
metric average sample_sd population_sd count
--------------------------------------------------
cct-accuracy 0.681 0.267 0.266 105
cct-%missing 0.086 0.159 0.159 105
```
**after:**
```
metric average sample_sd population_sd count
--------------------------------------------------
cct-accuracy 0.697 0.251 0.250 105
cct-%missing 0.071 0.123 0.122 105
```
[CORE-2969]:
https://unstructured-ai.atlassian.net/browse/CORE-2969?atlOrigin=eyJpIjoiNWRkNTljNzYxNjVmNDY3MDlhMDU5Y2ZhYzA5YTRkZjUiLCJwIjoiZ2l0aHViLWNvbS1KU1cifQ
---------
Co-authored-by: ryannikolaidis <1208590+ryannikolaidis@users.noreply.github.com>
Co-authored-by: badGarnet <badGarnet@users.noreply.github.com>
Co-authored-by: christinestraub <christinemstraub@gmail.com>
Closes#2218. When a csv has commas in its content, and the delimiter is
something else, Pandas may throw an error. We can sniff the csv and get
the correct delimiter to pass to Pandas. To verify, try partitioning the
file in the linked bug.
The original naming for the pre-cursor to a chunk in `chunk_by_title()`
was conflated with the idea of how these element subsequences were
bounded (by document-section) for that strategy. I mistakenly picked
that up as a universal concept but in fact no notion of section arises
in the `by_character` or other chunking strategies.
Fix this misconception by using the name `pre-chunk` for this concept
throughout.
`CheckBox` elements get special treatment during chunking. `CheckBox`
does not derive from `Text` and can contribute no text to a chunk. It is
considered "non-combinable" and so is emitted as-is as a chunk of its
own. A consequence of this is it breaks an otherwise contiguous chunk
into two wherever it occurs.
This is problematic, but becomes much more so when overlap is
introduced. Each chunk accepts a "tail" text fragment from its preceding
element and contributes its own tail fragment to the next chunk. These
tails represent the "overlap" between chunks. However, a non-text chunk
can neither accept nor provide a tail-fragment and so interrupts the
overlap. None of the possible solutions are terrific.
Give `Element` a `.text` attribute such that _all_ elements have a
`.text` attribute, even though its value is the empty-string for
element-types such as CheckBox and PageBreak which inherently have no
text. As a consequence, several `cast()` wrappers are no longer required
to satisfy strict type-checking.
This also allows a `CheckBox` element to be combined with `Text`
subtypes during chunking, essentially the same way `PageBreak` is,
contributing no text to the chunk.
Also, remove the `_NonTextSection` object which previously wrapped a
`CheckBox` element during pre-chunking as it is no longer required.
In preparation for work on generalized chunking including
`chunk_by_character()` and overlap, get `elements` module and tests
passing strict type-checking.
### Description
Given all the shell files that now exist in the repo, would be nice to
have linting/formatting around them (in addition to the existing
shellcheck which doesn't do anything to format the shell code). This PR
introduces `shfmt` to both check for changes and apply formatting when
the associated make targets are called.
closes#2222.
### Summary
The "table" elements are saved as `table-<pageN>-<tableN>.jpg`. This
filename is presented in the `image_path` metadata field for the Table
element. The default would be to not do this.
### Testing
PDF: [124_PDFsam_Basel III - Finalising post-crisis
reforms.pdf](https://github.com/Unstructured-IO/unstructured/files/13591714/124_PDFsam_Basel.III.-.Finalising.post-crisis.reforms.pdf)
```
elements = partition_pdf(
filename="124_PDFsam_Basel III - Finalising post-crisis reforms.pdf",
strategy="hi_res",
infer_table_structure=True,
extract_element_types=['Table'],
)
```
### Description
All other connectors that were not included in
https://github.com/Unstructured-IO/unstructured/pull/2194 are now
updated to follow the new pattern and mark any variables as sensitive
where it makes sense.
Core changes:
* All connectors now support an `AccessConfig` to mark data that's
needed for auth (i.e. username, password) and those that are sensitive
are designated appropriately using the new enhanced field.
* All cli configs on the cli definition now inherit from the base config
in the connector file to reuse the variables set on that dataclass
* The base writer class was updated to better generalize the new
approach given better use of dataclasses
* The base cli classes were refactored to also take into account the
need for a connector and write config when creating the respective
runner/writer classes.
* Any mismatch between the cli field name and the dataclass field name
were updated on the dataclass side to not impact the user but maintain
consistency
* Add custom redaction logic for mongodb URIs since the password is
expected to be a part of it. Now this:
`"mongodb+srv://ingest-test-user:r4hK3BD07b@ingest-test.hgaig.mongodb.net/"`
->
`"mongodb+srv://ingest-test-user:***REDACTED***@ingest-test.hgaig.mongodb.net/"`
in the logs
* Bundle all fsspec based files into their own packages.
* Refactor custom `_decode_dataclass` used for enhanced json mixin by
using a monkey-patch approach. The original approach was breaking on
optional nested dataclasses when serializing since the other methods in
`dataclasses_json_core` weren't using the new method. By monkey-patching
the original method with a new one, all other methods in that library
would use the new one.
---------
Co-authored-by: ryannikolaidis <1208590+ryannikolaidis@users.noreply.github.com>
Co-authored-by: rbiseck3 <rbiseck3@users.noreply.github.com>
### Description
The local source connector already supports filters on a list of glob
filters optionally provided. This same code was leveraged in the base
fsspec connector to filter on the full file paths returned from the
underlying fs library.
When no connector is provided in the folder structure, the code get
filename as connector instead. Fix the code so that if folder structure
has no connector subfolder, it leaves blank or None for connector field.
Adds source connector for SFTP which uses fsspec and paramiko via
fsspec. Paramiko is the standard sftp package for python used in pysftp
etc...
```
--username foo \
--password bar \
--remote-url sftp://localhost:47474/upload/
```
Will only download a specifically requested file if it has an extension.
(i.e. `--remote-url sftp://localhost:47474/upload/bob.zip`) It will
treat any other remote_url as a folder path. This is intentional.
---------
Co-authored-by: potter-potter <david.potter@gmail.com>
Update to API key link per Slack convo.
There may be more that's needed, @ron-unstructured please run with this
PR if it's the right solution, close it if not, or make changes if
needed. I haven't looked for other places where the link might need to
be changed. The link appears to work, but any further investigation or
testing is appreciated.
Co-authored-by: Ronny H <138828701+ron-unstructured@users.noreply.github.com>
### Description
Building off of PR
https://github.com/Unstructured-IO/unstructured/pull/2179, updating
fsspec based connectors to use better authentication field handling.
This PR adds in the following changes:
* Update the base classes to inherit from the enhanced json mixin
* Add in a new access config dataclass that should be used as a nest
dataclass in the connector configs
* Update the code extracting configs out of the cli options dictionary
to support the nested access config if it exists on the parent config
* Update all fsspec connectors with explicit access configs given what
each one's SDKs support
* Update the json mixin and enhanced field to support a name override
when serializing/deserializing from json/dicts. This allows a different
name to be used for the CLI option than what the name of the field is on
the dataclass.
* Update all the writes to use class-based approach and share the same
structure of the runner classes
* Above update allowed for better code to be used in the base source and
destination CLI commands
* Add in utility code around paring a flat dictionary (coming from the
click based options) into dataclass-based configs with potentially
nested dataclasses.
**Slightly unrelated changes:**
* session handle removed from pinecone connector as this was breaking
the serialization of the write config and didn't have any benefit as a
connection was never being shared, the index used simply makes a new
http call each time it's invoked.
* Dedicated write configs were created for all destination connectors to
better support serialization
* Refactor of Elasticsearch connector included, with update to ingest
test to use auth
**TODOs**
* Left a `#TODO` in the code but the way session handler is implemented
right now, it breaks serialization since it adds a generic variable
based on the library being used for a connector (i.e.
`googleapiclient.discovery.Resource`) which is not serializable. This
will need to be updated to omit that from serialization but still
support the current workflow.
---------
Co-authored-by: ryannikolaidis <1208590+ryannikolaidis@users.noreply.github.com>
Co-authored-by: rbiseck3 <rbiseck3@users.noreply.github.com>
### Summary
This PR is the second part of the "image extraction" refactor to move it
from unstructured-inference repo to unstructured repo, the first part is
done in
https://github.com/Unstructured-IO/unstructured-inference/pull/299. This
PR adds logic to support extracting images.
### Testing
`git clone -b refactor/remove_image_extraction_code --single-branch
https://github.com/Unstructured-IO/unstructured-inference.git && cd
unstructured-inference && pip install -e . && cd ../`
```
elements = partition_pdf(
filename="example-docs/embedded-images.pdf",
strategy="hi_res",
extract_images_in_pdf=True,
)
print("\n\n".join([str(el) for el in elements]))
```
### Description
Given the filtering in the ingest logger, anything going to console
should go through that. This adds a linter that only checks for
`print()` statements in the ingest code and ignored it elsewhere for
now.
Follow-up PR to
[https://github.com/Unstructured-IO/unstructured/pull/2195](https://github.com/Unstructured-IO/unstructured/pull/2195).
Removes unnecessary calls to `get_api_key()`. That helper function is
supposed to only be used for tests decorated by
@pytest.mark.skipif(skip_outside_ci, reason="Skipping test run outside
of CI") (which are skipped because those tests are partitioning pdf/jpg
files).
These tests are partitioning emails and rely on the MockResponse at the
top of the file, so they don't need to call `get_api_key()` and it can
simply be removed from them.
Closes#1781.
- Adds a Weaviate destination connector
- The connector receives a host for the weaviate instance and a weaviate
class name.
- Defines a weaviate schema for json elements.
- Defines the pre-processing to conform unstructured's schema to the
proposed weaviate schema.
### Summary
This PR is the second part of `pdfminer` refactor to move it from
`unstructured-inference` repo to `unstructured` repo, the first part is
done in
https://github.com/Unstructured-IO/unstructured-inference/pull/294. This
PR adds logic to merge the extracted layout with the inferred layout.
The updated workflow for the `hi_res` strategy:
* pass the document (as data/filename) to the `inference` repo to get
`inferred_layout` (DocumentLayout)
* pass the `inferred_layout` returned from the `inference` repo and the
document (as data/filename) to the `pdfminer_processing` module, which
first opens the document (create temp file/dir as needed), and splits
the document by pages
* if is_image is `True`, return the passed
inferred_layout(DocumentLayout)
* if is_image is `False`:
* get extracted_layout (TextRegions) from the passed
document(data/filename) by pdfminer
* merge `extracted_layout` (TextRegions) with the passed
`inferred_layout` (DocumentLayout)
* return the `inferred_layout `(DocumentLayout) with updated elements
(all merged LayoutElements) as merged_layout (DocumentLayout)
* pass merged_layout and the document (as data/filename) to the `OCR`
module, which first opens the document (create temp file/dir as needed),
and splits the document by pages (convert PDF pages to image pages for
PDF file)
### Note
This PR also fixes issue #2164 by using functionality similar to the one
implemented in the `fast` strategy workflow when extracting elements by
`pdfminer`.
### TODO
* image extraction refactor to move it from `unstructured-inference`
repo to `unstructured` repo
* improving natural reading order by applying the current default
`xycut` sorting to the elements extracted by `pdfminer`
### Summary
Closes#2033
Updates `partition_via_api` to use `UnstructuredClient` for api calls
instead of `requests`.
Updates associated tests.
Note: This PR does **not** update `partition_multiple_via_api` as
documentation in `unstructured-python-client` indicates it does not
support multiple files. A new issue should be opened to add that
functionality to `unstructured-python-client`.
---------
Co-authored-by: Klaijan <klaijan@unstructured.io>
Co-authored-by: Roman Isecke <136338424+rbiseck3@users.noreply.github.com>
Co-authored-by: ryannikolaidis <1208590+ryannikolaidis@users.noreply.github.com>
Co-authored-by: rbiseck3 <rbiseck3@users.noreply.github.com>
### Description
Currently the ingest unit tests are running in both the source and
destination ingest steps. This PR moved it out into it's own step that
can run on a more basic runner since it doesn't require multiple
cores/cpus for parallelization.
Further optimizations:
* Added the changelog step as a dependency for the ingest tests to avoid
running them (fail fast) of the pipeline has already failed due to the
changelog not being updated.
* The cache was never actually being saved if it needed to be recreated,
a save job was added at the end of each custom action
To test:
> cd docs && make html
Change logs:
* Examples are reorganized to have its own page
* Removed two old examples, ie. "file-utils" & "sentiment analysis".
* Added two examples: "RAG with Unstructured, LangChain, and ChromaDB" &
"Multi-Files Processing with S3 Connector and API"
* Reorganized and added detailed API documentation: (i) usage, (ii)
SDKs, (iii) Azure Marketplace, (iv) AWS Marketplace, (v) parameters and
validation errors
### Summary
Add a procedure to repair PDF when the PDF structure is invalid for
`PDFminer` to process.
This PR handles two cases of `PSSyntaxError Invalid dictionary
construct: ...`:
* PDFminer open entire document and create pages generator on
`PDFPage.get_pages(fp)`: [sentry log
example](https://unstructuredio.sentry.io/issues/4655715023/?alert_rule_id=14681339&alert_type=issue¬ification_uuid=d8db4cf4-686f-4504-8a22-74a79a8e966f&project=4505909127086080&referrer=slack)
* PDFminer's interpreter process a single page on
`interpreter.process_page(page)`: [sentry log
example](https://unstructuredio.sentry.io/issues/4655898781/?referrer=slack¬ification_uuid=0d929d48-f490-4db8-8dad-5d431c8460bc&alert_rule_id=14681339&alert_type=issue)
**Additional tech details:**
* Add new dependency `pikepdf` in `requirements/extra-pdf-image.in`,
which is used for repairing PDF.
* Add new denpendenct `pypdf` in `requirements/extra-pdf-image.in`,
which is used to find the error page from entire document by reading the
PDF file again (can't find a way to split pdf in PDFminer).
* Refactor the `is null` check for `get_uris_from_annots`, since the
root cause is that `get_uris` passed a None `annots` to
`get_uris_from_annots`, so the Null check should happen in `get_uris`.
* Add more type protection in `get_uris_from_annots` when using any
`PDFObjRef.resolve()` as `dict` (it could still be a `PDFObjRef`). This
should fix :
* https://github.com/Unstructured-IO/unstructured/issues/1922 where
`annotation_dict` is a `PDFObjRef`
* https://github.com/Unstructured-IO/unstructured/issues/1921 where
`rect` is a `PDFObjRef`
### Test
Added three test files (both are larger than 500 KB) for unittests to
test:
* Repair entire doc
* Repair one page
* Reprocess failure after repairing one page (just return the elements
before error page in this case).
* Also seems like splitting the document into smaller pages could fix
this problem, but not sure why. For example, I saw error from reprocess
in the whole
[cancer.pdf](https://github.com/Unstructured-IO/unstructured/files/13461616/cancer.pdf)
doc, but no error when i split the pdf by error page....
* tested if i can repair the entire doc again in this case, saw other
error which means repairing is not helping imo
* PDFminer can process the whole doc after pikepdf only repaired the
entire doc in the first place, but we can't repair by pages in this way
---------
Co-authored-by: cragwolfe <crag@unstructured.io>
### Description
Modify the logger being used by ingest to leverage a new class
implemented inheriting `logging.Formatter` which adds in some middleware
to update the message being logged to omit any sensitive content. It
does this by dynamically pulled out any valid json from the string being
logged and runs that through a `hide_sensitive_fields` method which
updated any values that are considered sensitive. Replaces the original
json strings with the `json.dumps` version of the new dictionary.
### Description
Often times there are tests being skipped either due to missing env vars
or explicitly defined in the base script but these get lost in the logs.
This PR updates the scripts to leverage a custom error code if being
skipped due to missing env vars and this custom error code is being
caught by the base script and logs all files being skipped to a file. At
the end of the script, this file gets logged in the CI output.
Closes#1843
Ingest connector for HubSpot. Supports:
- Calls: Logs from calls related to contacts, companies and tickets
- Communications: Logs from SMS/Whatsapp related to contacts, companies
and tickets
- Notes: Notes related to CRM notes
- Products: CRM products
- Emails: Logs from emails sent to CRM objects.
- Tasks: CRM tasks
From each record, `body/`description`information is grabbed. When a
title property is available, this is registered at the beggining of the
output file. The CLI receives three params:
- `api-token`: [Private
app](https://developers.hubspot.com/docs/api/private-apps) token.
- `object-types: One of the noted supported objects in the form of a
comma separated list: `calls,products,tasks`
- `custom-properties`: Custom properties to grab information from. Must
be in the form
`<object_type>:<custom_property_id>,<object_type>:<custom_property_id>`
---------
Co-authored-by: ryannikolaidis <1208590+ryannikolaidis@users.noreply.github.com>
Co-authored-by: rvztz <rvztz@users.noreply.github.com>
### Description
When passing in a remote path for fsspec-based source connectors, the
base directory was always being included in the output path itself. This
was updated to exclude the base directory any only include any child
directories relative to the base one.
---------
Co-authored-by: ryannikolaidis <1208590+ryannikolaidis@users.noreply.github.com>
Co-authored-by: rbiseck3 <rbiseck3@users.noreply.github.com>
### Description
There are some source ingest connectors that would be more efficient to
read the content in batches rather than use an entire process per
document. For example, reading from ElasticSearch. Given an index with
possible hundreds of documents, reading each one individually is not as
optimal as reading in batches. To try and maintain as much of the ingest
doc paradigm already being supported, a new class `BaseIngestDocBatch`
was added to handle reading in batches. It produces a list of
`BaseSingleIngestDoc` which is what all current implementations were
renamed to. This list is generated after it runs its `get_files` method.
Past the source node, all other steps in the pipeline should not be
affected, this is just an optimization for the read step.
**Additional Changes:**
* Removed use of jq and instead converted this into a fields filter on
the content to let the database handle the filtering and limit the
amount of data being pulled in.
---------
Co-authored-by: ryannikolaidis <1208590+ryannikolaidis@users.noreply.github.com>
Co-authored-by: rbiseck3 <rbiseck3@users.noreply.github.com>
Currently the `check-diff-evaluation-metrics` only runs when there is
file to perform evaluation on. Add the checking condition to skip the
action when there is none. Additionally, more refactoring and
`visualize` option for both evaluation calculation functions is also
added.
### Summary
This should fix the broken unit test on main CI
* change the strategy in
`test_partition_multiple_via_api_valid_request_data_kwargs` from `fast`
to `auto`, since the test was using `fast` for images, and we don't
support it.
