Impira enables you to get everything you need from your PDFs, scanned documents, images, and more — with the help of machine learning. This API allows you to access Impira programatically through Python and the command line.
NOTE: This SDK is currently under active development and is likely to break backwards compatibility between point releases. We will update this disclaimer when this changes.
Below is an abbreviated set of documentation to help you get started. You can visit the full documentation by visiting the Impira SDK docs.
This SDK is tested with Python 3.7.4+ on Mac OS X and Linux systems. We have users using Windows as well; however, this scenario is not tested automatically. Please reach out if you run into any issues on Windows or another platform.
You can install the Impira Python SDK, CLI, and its dependencies directly through pip:
$ pip install 'impira[cli]'To install just the SDK and its dependencies use:
$ pip install impiraIf you would like to install the SDK and CLI to develop locally, you can run the following:
$ git clone git@github.com:impira/impira-python.git
$ cd impira-python
$ make developThis will create a virtualenv locally and install the library to it in a manner that automatically updates as you change the source code.
See CLI Commands in the docs.
The Impira Python SDK includes utilities to upload and label files, insert and update data, and query data. The core abstraction is the Impira object which represents an authenticated connection to your organization. The SDK makes heavy use of the pydantic library to automatically build up and validate function arugments.
To connect to an org, you simply instantiate the Impira object with your organization's name and API token. You can find your organization's name in the URL you visit to access Impira. For example, when you login, if the URL is https://app.impira.com/o/acme-corp-1a23/collections, then your organization's name is acme-corp-1a23. For instructions on obtaining an API token, please visit the Impira docs. For security reasons, we highly recommend storing both the organization name and API key in configuration or environment variables, not directly in the code itself. For the purpose of these examples, we will use the environment variables IMPIRA_ORG_NAME and IMPIRA_API_KEY.
from impira import Impira
import os
impira_api = Impira(os.environ["IMPIRA_ORG_NAME"], os.environ["IMPIRA_API_KEY"])When you instantiate the Impira API, it will automatically issue a ping request to validate your credentials and raise an InvalidRequest exception if it fails. You can disable this behavior by passing ping=False to the constructor.
Many function calls in the API reference a collection_id parameter. This id can be found by navigating to a collection in the application, and copying the identifier after fc. For example, for a collection at a URL like https://app.impira.com/o/acme-corp-1a23/fc/07b71143a26b7163, the collection_id is 07b71143a26b7163.
To upload one or more files, you must provide at a minimum a name and path (either local or a URL) for each file. The specification is defined in the FilePath type. You can also optionally specify a collection_id or None to upload the file globally (to "All files").
# Upload a file on your local machine
uids = impira_api.upload_files(collection_id, [
{"name": "foo.pdf", "path": "/Users/me/Desktop/foo.pdf"}
])
# Upload multiple files by specifying their URLs
uids = impira_api.upload_files(collection_id, [
{"name": "foo.pdf", "path": "http://website.com/foo.pdf"},
{"name": "bar.pdf", "path": "http://website.com/bar.pdf"},
])The uids variable is a list with a uid for each file. A file's uid is its unique identifier throughout the system. A file that belongs to more than one collection will have the same uid in each. You can also optionally specify your own uid while uploading a file. If two files have the same uid, the system will automatically replace the former with the latter, effectively versioning the file. For more information on uploading files, visit the Upload API docs.
Impira's API is asynchronous, meaning that uploading files and retrieving predictions from them occur in two separate API requests. While there are many advanced ways to query for data using IQL, the SDK offers a simple poll_for_results() method that allows you to wait for results to be available for your uploads. Using the uids returned from upload_files() (as demonstrated above), you can simply run something like
for row in impira_api.poll_for_results(collection_id, uids):
print(row)to retrieve each prediction. Note that poll_for_results() returns a generator, so you must iterate through its output to retrieve each result.
You can run arbitrary IQL queries through the API by simply invoking the query() method. The response is exactly the same format as the read API and the SDK also supports poll mode (by passing the mode="poll" argument). For example,
response = impira_api.query(
"@`file_collections::%s`[uid] highest:Uploaded limit:10" % (collection_id)
)
last_10_uids = [row["uid"] for row in response["data"]]will retrieve the uid of each of the last 10 files uploaded to the collection. For more information on how to construct IQL queries, see the IQL docs.
The examples directory contains end-to-end working examples for how to use the SDK.
upload_files.pywalks through uploading a file, either locally or through a URL, and then waiting for its results.upload_and_split_files.pywalks through uploading a file, either locally or through a URL, remove all pages after page 5 and splitting the remaining pages, and then waiting for its results.download_all_collections_query.pycreates an IQL query that queries all data across all collections
MIT License. Copyright 2021 Impira Inc.