skip to navigation
skip to content

Not Logged In

python-epo-ops-client 0.1.3

Python Client for the European Patent Office's Open Patent Services API

Latest Version: 1.0.0


python-epo-ops-client is an Apache2 Licensed client library for accessing the European Patent Office’s (“EPO”) Open Patent Services (“OPS”) v.3.1 (based on v 1.2.10 of the reference guide).

import epo_ops

anonymous_client = epo_ops.Client()  # Instantiate a default client
response = anonymous_client.published_data(  # Retrieve bibliography data
  reference_type = 'publication',  # publication, application, priority
  input = epo_ops.models.Docdb('1000000', 'EP', 'A1'),  # original, docdb, epodoc
  endpoint = 'biblio',  # optional, defaults to biblio in case of published_data
  constituents = []  # optional, list of constituents

registered_client = epo_ops.RegisteredClient(key='abc', secret='xyz')
registered_client.access_token  # To see the current token
response = registered_client.published_data()


python_epo_ops_client abstracts away the complexities of access EPO OPS:

  • Format the requests properly
  • Bubble up quota problems as proper HTTP errors
  • Handle token authentication and renewals automatically
  • Handle throttling properly
  • Add optional caching to minimize impact on the OPS servers

There are two main layers to python_epo_ops_client: Client and Middleware.


The Client contains all the formatting and token handling logic and is what you’ll interact with mostly.

When you issue a request, the response is a requests.Response object. If response.status_code != 200 then an exception will be raised, it’s your responsibility to handle those exceptions if you want to. The one case that’s handled by the RegisteredClient is when its access token has expired: in this case, the client will automatically handle the HTTP 400 status and renew the token.

Note that the Client does not attempt to interpret the data supplied by OPS, so it’s your responsibility to parse the XML or JSON payload for your own purpose.

The following custom exceptions are raised for cases when OPS quotas are exceeded, they are all subclasses of requests.HTTPError and offer the same behavior:

  • AnonymousQuotaPerMinuteExceeded
  • AnonymousQuotaPerDayExceeded
  • IndividualQuotaPerHourExceeded
  • RegisteredQuotaPerWeekExceeded

Again, it’s up to you to parse the response and decide what to do.

Currently the Client only knows how to issue request for the following services:

  • /published-data/search (search)
  • /published-data (retrieval)
  • /family (inpadoc)

Please submit pull requests for other services by enhancing the epo_ops.api.Client class.


All requests and responses are passed through each middleware object listed in client.middlewares. Requests are processed in the order listed, and responses are processed in the reverse order.

Each middleware should subclass middlewares.Middleware and implement the process_request and process_response methods.

There are two middleware classes out of the box: Throttler and Dogpile. Throttler is in charge of the OPS throttling rules and will delay requests accordingly. Dogpile is an optional cache which will cache all HTTP status 200, 404, 405, and 413 responses.

By default, only the Throttler middleware is enabled, if you want to enable caching:

import epo_ops

middlewares = [
registered_client = epo_ops.RegisteredClient(

Note that caching middleware should be first in most cases.


Dogpile is based on (surprise) dogpile.cache. By default it is instantiated with a DBMBackend region with timeout of 2 weeks.

Dogpile takes three optional instantiation parameters:

  • region: You can pass whatever valid dogpile.cache Region you want to backend the cache
  • kwargs_handlers: A list of keyword argument handlers, which it will use to process the kwargs passed to the request object in order to extract elements for generating the cache key. Currently one handler is implemented (and instantiated by default) to make sure that the X-OPS-Range request header is part of the cache key.
  • http_status_codes: A list of HTTP status codes that you would like to have cached. By default 200, 404, 405, and 413 responses are cached.

Note: dogpile.cache is not installed by default, if you want to use it, pip install dogpile.cache in your project.


Throttler contains all the logic for handling different throttling scenarios. Since OPS throttling is based on a one minute rolling window, we must persist historical (at least for the past minute) throtting data in order to know what the proper request frequency is. Each Throttler must be instantiated with a Storage object.


The Storage object is responsible for:

  1. Knowing how to update the historical record with each request (Storage.update()), making sure to observe the one minute rolling window rule.
  2. Calculating how long to wait before issuing the next request (Storage.delay_for()).

Currently the only Storage backend provided is SQLite, but you can easily write your own Storage backend (such as file, Redis, etc.). To use a custom Storage type, just pass the Storage object when you’re instantiating a Throttler object. See epo_ops.middlewares.throttle.storages.Storage for more implementation details.


Tests are written using pytest. To run the tests:

  1. Register a OPS user login with EPO
  2. Create an app
  3. Set the OPS_KEY and OPS_SECRET environment variables accordingly
  4. make test

The tests must be run with a working internet connection, since both OPS and the mock Apiary services are online.

Release History

0.1.3 (2014-05-21)

  • Python 3.4 compatibility
  • Updated requests dependency to 2.3.0

0.1.2 (2014-03-04)

  • Python 2.6 and 3.3 compatibility

0.1.1 (2014-03-01)

  • Allow configuration of which HTTP responses (based on status code) to cache

0.1.0 (2014-02-20)

  • Introduced dogpile.cache for caching http200 resopnses
  • Introduced the concept of middleware

0.0.1 (2014-01-21)

  • Initial release
File Type Py Version Uploaded on Size
python-epo-ops-client-0.1.3.tar.gz (md5) Source 2014-05-21 15KB
python_epo_ops_client-0.1.3-py2.py3-none-any.whl (md5) Python Wheel 2.7 2014-05-21 20KB
  • Downloads (All Versions):
  • 96 downloads in the last day
  • 474 downloads in the last week
  • 2038 downloads in the last month