A Python HTTP client for the Kinto API.
Clone or download
Latest commit f300dda Dec 5, 2018
Type Name Latest commit message Commit time
Failed to load latest commit information.
.coveragerc Rename kinto-client to kinto-http. Jun 9, 2016
.gitignore Add pytest_cache/ to .gitignore Oct 11, 2018
.travis.yml Update dependencies. Dec 3, 2018
CONTRIBUTING.md Minor nitpicks in CONTRIBUTING.md guide Nov 4, 2016
LICENSE Preparing release 1.0.0 Nov 9, 2015
MANIFEST.in Add CONTRIBUTING.md to Python package manifest May 25, 2017
Makefile Update dependencies. Dec 3, 2018
README.rst Solved typos on README Dec 5, 2018
dev-requirements.txt Get rid of mock external module. Oct 15, 2018
requirements.txt Upgrade requirements.txt. Dec 3, 2018
setup.cfg Don't build universal wheel since we don't support python 2.7 anymore. Oct 15, 2018
setup.py Back to development: 10.2.0 Nov 13, 2018
tox.ini Update tests environment. Dec 3, 2018


Kinto python client


Kinto is a service that allows users to store and synchronize arbitrary data, attached to a user account. Its primary interface is HTTP.

kinto-http is a Python library that eases the interactions with a Kinto server instance. A project with related goals is also available for JavaScript.


Use pip:

$ pip install kinto-http



Operations are always performed directly on the server, and no synchronisation features have been implemented yet.

  • The first version of this API doesn't cache any access nor provides any refresh mechanism. If you want to be sure you have the latest data available, issue another call.

Here is an overview of what the API provides:

from kinto_http import Client

client = Client(server_url="http://localhost:8888/v1",
                auth=('alexis', 'p4ssw0rd'))

records = client.get_records(bucket='default', collection='todos')
for i, record in enumerate(records):
    record['title'] = 'Todo {}'.format(i)

for record in records:

Creating a client

The passed auth parameter is a requests authentication policy, allowing authenticating using whatever scheme fits you best.

By default, Kinto supports Firefox Accounts and Basic authentication policies.

from kinto_http import Client
credentials = ('alexis', 'p4ssw0rd')

client = Client(server_url='http://localhost:8888/v1',

It is also possible to pass the bucket and the collection to the client at creation time, so that this value will be used by default.

auth = ("token", "secret")
client = Client(bucket="payments", collection="receipts", auth=auth)

After creating a client, you can also replicate an existing one and overwrite some key arguments.

client2 = client.clone(collection="orders")

Using a Bearer token to authenticate

import requests

class BearerTokenAuth(requests.auth.AuthBase):
    def __init__(self, token):
        self.token = token

    def __call__(self, r):
        r.headers['Authorization'] = 'Bearer ' + self.token
        return r

auth = BearerTokenAuth("a67fjnewgre5")
client = Client(bucket="main", collection="tippytop", auth=auth)

Using FxA from a script with the email/password

from fxa.plugins.requests import FxABearerTokenAuth

auth = FxABearerTokenAuth(
    email, passwd,
client = Client(bucket="payments", collection="receipts", auth=auth)

Getting server information

You can use the server_info method to get the server information:

from kinto_http import Client

client = Client(server_url='http://localhost:8888/v1')
info = client.server_info()
assert 'schema' in info['capabilities'], "Server doesn't support schema validation."

Handling buckets

All operations are rooted in a bucket. It makes little sense for one application to handle multiple buckets at once (but it is possible). If no specific bucket name is provided, the "default" bucket is used.

from kinto_http import Client
from kinto_http.patch_type import BasicPatch, MergePatch, JSONPatch
credentials = ('alexis', 'p4ssw0rd')

client = Client(server_url='http://localhost:8888/v1',

# To create a bucket.

# To get an existing bucket
bucket = client.get_bucket(id='payments')

# Or retrieve all readable buckets.
buckets = client.get_buckets()

# To create or replace an existing bucket.
client.update_bucket(id='payments', data={'description': 'My payments data.'})

# Or modify some fields in an existing bucket.
# The Kinto server supports different types of patches, which can be used from kinto_http.patch_type.
client.patch_bucket(id='payments', changes=BasicPatch({'status': 'updated'}))

# It is also possible to manipulate bucket permissions (see later)
client.patch_bucket(id='payments', changes=BasicPatch(permissions={}))

# Or delete a bucket and everything under.

# Or even every writable buckets.


A group associates a name to a list of principals. It is useful in order to handle permissions.

# To create a group.
client.create_group(id='receipts', bucket='payments', data={'members': ['blah', 'foo']})

# Or get an existing one.
group = client.get_group(id='receipts', bucket='payments')

# Or retrieve all groups in the bucket.
groups = client.get_groups(bucket='payments')

# To create or replace an existing bucket.
client.update_group(id='receipts', bucket='payments', 'data'={'members':['foo']})

# Or modify some fields in an existing group.
# This uses the server's support for JSON patch, but any patch_type is accepted.
client.patch_group(id='receipts', bucket='payments',
    changes=JSONPatch([{'op': 'add', 'path': '/data/members/0', 'value': 'ldap:user@corp.com'}]))

# To delete an existing group.
client.delete_group(id='receipts', bucket='payments')

# Or all groups in a bucket.


A collection is where records are stored.

# To create a collection.
client.create_collection(id='receipts', bucket='payments')

# Or get an existing one.
collection = client.get_collection(id='receipts', bucket='payments')

# Or retrieve all of them inside a bucket.
collections = client.get_collections(bucket='payments')

# To create or replace an exiting collection.
client.update_collection(id='receipts', bucket='payments', data={'description':'bleeh'})

# Or modify some fields of an existing collection.
client.patch_collection(id='receipts', bucket='payments', changes=MergePatch({'status':'updated'}))

# To delete an existing collection.
client.delete_collection(id='receipts', bucket='payments')

# Or every collections in a bucket.


Records can be retrieved from and saved to collections.

A record is a dict with the "permissions" and "data" keys.

 # You can pass a python dictionary to create the record.
 client.create_record(data={'status': 'done', title: 'Todo #1'},
                      collection='todos', bucket='default')

 # You can use id to specify the record id when creating it.
 client.create_record(id='todo2', data={'status': 'doing', 'title': 'Todo #2'},
                      collection='todos', bucket='default')

 # Or get an existing one by its id.
 record = client.get_record(id='todo2', collection='todos', bucket='default')

 # Or retrieve all records.
 records = client.get_records(collection='todos', bucket='default')

# Or retrieve records page by page.
for page in client.get_paginated_records(collection='todos', bucket='default'):
   # Do something with each page

 # Or retrieve records timestamp.
 records_timestamp = client.get_records_timestamp(collection='todos', bucket='default')

 # To replace a record using a previously fetched record
 client.update_record(data=record, collection='todos', bucket='default')

 # Or create or replace it by its id.
 client.update_record(data={'status': 'unknown'}, id='todo2', collection='todos', bucket='default')

 # Or modify some fields in an existing record.
 client.patch_record(changes=MergePatch({'assignee': 'bob'}), id='todo2', collection='todos', bucket='default')

 # To delete an existing record.

 # Or every records of a collection.


By default, authors will get read and write access to the manipulated objects. It is possible to change this behavior by passing a dict to the permissions parameter.

    data={'foo': 'bar'},
    permissions={'read': ['group:groupid']},


Every creation or modification operation on a distant object can be given a permissions parameter.

Buckets, collections and records have permissions which can be edited. For instance to give access to "leplatrem" to a specific record, you would do:

record = client.get_record(1234, collection='todos', bucket='alexis')

# During creation, it is possible to pass the permissions dict.
client.create_record(data={'foo': 'bar'}, permissions={})

Get or create

In some cases, you might want to create a bucket, collection, group or record only if it doesn't exist already. To do so, you can pass the if_not_exists=True to the create_* methods:

client.create_bucket(id='bucket', if_not_exists=True)


In some cases, you might want to delete a bucket, collection, group or record only if it exists already. To do so, you can pass the if_exists=True to the delete_* methods:

client.delete_bucket(id='bucket', if_exists=True)

Overwriting existing objects

Most of the methods take a safe argument, which defaults to True. If set to True and a last_modified field is present in the passed data, or if the if_match parameter is specified then a check will be added to the requests to ensure the record wasn't modified on the server side in the meantime.

Batching operations

Rather than issuing a request for each and every operation, it is possible to batch the requests. The client will then issue as little requests as possible.

It is possible to do batch requests using a Python context manager (with):

with client.batch() as batch:
    for idx in range(0, 100):
        batch.update_record(data={'id': idx})

Reading data from batch operations is achieved by using the results() method available after a batch context is closed.

with client.batch() as batch:

r1, r2, r3 = batch.results()

Besides the results() method, a batch object shares all the same methods as another client.

Retry on error

When the server is throttled (under heavy load or maintenance) it can return error responses.

The client can hence retry to send the same request until it succeeds. To enable this, specify the number of retries on the client:

client = Client(server_url='http://localhost:8888/v1',

The Kinto protocol lets the server define the duration in seconds between retries. It is possible (but not recommended) to force this value in the clients:

client = Client(server_url='http://localhost:8888/v1',


When the server responses are paginated, the client will download every pages and merge them transparently.

However, it is possible to specify a limit for the number of items to be retrieved in one page:

records = client.get_records(_limit=10)

In order to retrieve every available pages with a limited number of items in each of them, you can specify the number of pages:

records = client.get_records(_limit=10, pages=float('inf'))  # Infinity

Generating endpoint paths

You may want to generate some endpoint paths, you can use the get_endpoint utility to do so:

client = Client(server_url='http://localhost:8888/v1',
                auth=('token', 'your-token'),

# '/buckets/payments/collections/receipts/records/c6894b2c-1856-11e6-9415-3c970ede22b0'

Handling datetime and date objects

In addition to the data types supported by JSON, kinto-http.py also supports native Python date and datetime objects.

In case a payload contain a date or a datetime object, kinto-http.py will encode it as an ISO formatted string.

Please note that this transformation is only one-way. While reading a record, if a string contains a ISO formated string, kinto-http.py will not convert it to a native Python date or datetime object.

If you know that a field will be a datetime, you might consider encoding it yourself to be more explicit about it being a string for Kinto.

Command-line scripts

In order to have common arguments and options for scripts, some utilities are provided to ease configuration and initialization of client from command-line arguments.

import argparse
import logging

from kinto_http import cli_utils

logger = logging.getLogger(__name__)

if __name__ == "__main__":
    parser = argparse.ArgumentParser(description="Download records")

    args = parser.parse_args()

    cli_utils.setup_logger(logger, args)

    logger.debug("Instantiate Kinto client.")
    client = cli_utils.create_client_from_args(args)

    logger.info("Fetch records.")
    records = client.get_records()
    logger.warn("{} records.".format(len(records)))

The script now accepts basic options:

$ python example.py --help

usage: example.py [-h] [-s SERVER] [-a AUTH] [-b BUCKET] [-c COLLECTION] [-v]
                  [-q] [-D]

Download records

optional arguments:
  -h, --help            show this help message and exit
  -s SERVER, --server SERVER
                        The location of the remote server (with prefix)
  -a AUTH, --auth AUTH  BasicAuth token:my-secret
  -b BUCKET, --bucket BUCKET
                        Bucket name.
  -c COLLECTION, --collection COLLECTION
                        Collection name.
  --retry RETRY         Number of retries when a request fails
  --retry-after RETRY_AFTER
                        Delay in seconds between retries when requests fail
                        (default: provided by server)
  -v, --verbose         Show all messages.
  -q, --quiet           Show only critical errors.
  -D, --debug           Show all messages, including debug messages.

Run tests

In one terminal, run a Kinto server:

$ make runkinto

In another, run the tests against it:

$ make tests