Advanced Usage

This document covers some of Requests more advanced features.

Session Objects

The Session object allows you to persist certain parameters across requests. It also persists cookies across all requests made from the Session instance, and will use urllib3’s connection pooling. So if you’re making several requests to the same host, the underlying TCP connection will be reused, which can result in a significant performance increase (see HTTP persistent connection).

A Session object has all the methods of the main Requests API.

Let’s persist some cookies across requests:

s = requests.Session()

r = s.get('')

# '{"cookies": {"sessioncookie": "123456789"}}'

Sessions can also be used to provide default data to the request methods. This is done by providing data to the properties on a Session object:

s = requests.Session()
s.auth = ('user', 'pass')
s.headers.update({'x-test': 'true'})

# both 'x-test' and 'x-test2' are sent
s.get('', headers={'x-test2': 'true'})

Any dictionaries that you pass to a request method will be merged with the session-level values that are set. The method-level parameters override session parameters.

Note, however, that method-level parameters will not be persisted across requests, even if using a session. This example will only send the cookies with the first request, but not the second:

s = requests.Session()

r = s.get('', cookies={'from-my': 'browser'})
# '{"cookies": {"from-my": "browser"}}'

r = s.get('')
# '{"cookies": {}}'

If you want to manually add cookies to your session, use the Cookie utility functions to manipulate Session.cookies.

Sessions can also be used as context managers:

with requests.Session() as s:

This will make sure the session is closed as soon as the with block is exited, even if unhandled exceptions occurred.

Remove a Value From a Dict Parameter

Sometimes you’ll want to omit session-level keys from a dict parameter. To do this, you simply set that key’s value to None in the method-level parameter. It will automatically be omitted.

All values that are contained within a session are directly available to you. See the Session API Docs to learn more.

Request and Response Objects

Whenever a call is made to requests.get() and friends, you are doing two major things. First, you are constructing a Request object which will be sent off to a server to request or query some resource. Second, a Response object is generated once Requests gets a response back from the server. The Response object contains all of the information returned by the server and also contains the Request object you created originally. Here is a simple request to get some very important information from Wikipedia’s servers:

>>> r = requests.get('')

If we want to access the headers the server sent back to us, we do this:

>>> r.headers
{'content-length': '56170', 'x-content-type-options': 'nosniff', 'x-cache':
'HIT from cp1006.eqiad.wmnet, MISS from cp1010.eqiad.wmnet', 'content-encoding':
'gzip', 'age': '3080', 'content-language': 'en', 'vary': 'Accept-Encoding,Cookie',
'server': 'Apache', 'last-modified': 'Wed, 13 Jun 2012 01:33:50 GMT',
'connection': 'close', 'cache-control': 'private, s-maxage=0, max-age=0,
must-revalidate', 'date': 'Thu, 14 Jun 2012 12:59:39 GMT', 'content-type':
'text/html; charset=UTF-8', 'x-cache-lookup': 'HIT from cp1006.eqiad.wmnet:3128,
MISS from cp1010.eqiad.wmnet:80'}

However, if we want to get the headers we sent the server, we simply access the request, and then the request’s headers:

>>> r.request.headers
{'Accept-Encoding': 'identity, deflate, compress, gzip',
'Accept': '*/*', 'User-Agent': 'python-requests/1.2.0'}

Prepared Requests

Whenever you receive a Response object from an API call or a Session call, the request attribute is actually the PreparedRequest that was used. In some cases you may wish to do some extra work to the body or headers (or anything else really) before sending a request. The simple recipe for this is the following:

from requests import Request, Session

s = Session()

req = Request('POST', url, data=data, headers=headers)
prepped = req.prepare()

# do something with prepped.body
prepped.body = 'No, I want exactly this as the body.'

# do something with prepped.headers
del prepped.headers['Content-Type']

resp = s.send(prepped,


Since you are not doing anything special with the Request object, you prepare it immediately and modify the PreparedRequest object. You then send that with the other parameters you would have sent to requests.* or Session.*.

However, the above code will lose some of the advantages of having a Requests Session object. In particular, Session-level state such as cookies will not get applied to your request. To get a PreparedRequest with that state applied, replace the call to Request.prepare() with a call to Session.prepare_request(), like this:

from requests import Request, Session

s = Session()
req = Request('GET',  url, data=data, headers=headers)

prepped = s.prepare_request(req)

# do something with prepped.body
prepped.body = 'Seriously, send exactly these bytes.'

# do something with prepped.headers
prepped.headers['Keep-Dead'] = 'parrot'

resp = s.send(prepped,


When you are using the prepared request flow, keep in mind that it does not take into account the environment. This can cause problems if you are using environment variables to change the behaviour of requests. For example: Self-signed SSL certificates specified in REQUESTS_CA_BUNDLE will not be taken into account. As a result an SSL: CERTIFICATE_VERIFY_FAILED is thrown. You can get around this behaviour by explicitly merging the environment settings into your session:

from requests import Request, Session

s = Session()
req = Request('GET', url)

prepped = s.prepare_request(req)

# Merge environment settings into session
settings = s.merge_environment_settings(prepped.url, {}, None, None, None)
resp = s.send(prepped, **settings)


SSL Cert Verification

Requests verifies SSL certificates for HTTPS requests, just like a web browser. By default, SSL verification is enabled, and Requests will throw a SSLError if it’s unable to verify the certificate:

>>> requests.get('')
requests.exceptions.SSLError: hostname '' doesn't match either of '*', ''

I don’t have SSL setup on this domain, so it throws an exception. Excellent. GitHub does though:

>>> requests.get('')
<Response [200]>

You can pass verify the path to a CA_BUNDLE file or directory with certificates of trusted CAs:

>>> requests.get('', verify='/path/to/certfile')

or persistent:

s = requests.Session()
s.verify = '/path/to/certfile'


If verify is set to a path to a directory, the directory must have been processed using the c_rehash utility supplied with OpenSSL.

This list of trusted CAs can also be specified through the REQUESTS_CA_BUNDLE environment variable. If REQUESTS_CA_BUNDLE is not set, CURL_CA_BUNDLE will be used as fallback.

Requests can also ignore verifying the SSL certificate if you set verify to False:

>>> requests.get('', verify=False)
<Response [200]>

Note that when verify is set to False, requests will accept any TLS certificate presented by the server, and will ignore hostname mismatches and/or expired certificates, which will make your application vulnerable to man-in-the-middle (MitM) attacks. Setting verify to False may be useful during local development or testing.

By default, verify is set to True. Option verify only applies to host certs.

Client Side Certificates

You can also specify a local cert to use as client side certificate, as a single file (containing the private key and the certificate) or as a tuple of both files’ paths:

>>> requests.get('', cert=('/path/client.cert', '/path/client.key'))
<Response [200]>

or persistent:

s = requests.Session()
s.cert = '/path/client.cert'

If you specify a wrong path or an invalid cert, you’ll get a SSLError:

>>> requests.get('', cert='/wrong_path/client.pem')
SSLError: [Errno 336265225] _ssl.c:347: error:140B0009:SSL routines:SSL_CTX_use_PrivateKey_file:PEM lib


The private key to your local certificate must be unencrypted. Currently, Requests does not support using encrypted keys.

CA Certificates

Requests uses certificates from the package certifi. This allows for users to update their trusted certificates without changing the version of Requests.

Before version 2.16, Requests bundled a set of root CAs that it trusted, sourced from the Mozilla trust store. The certificates were only updated once for each Requests version. When certifi was not installed, this led to extremely out-of-date certificate bundles when using significantly older versions of Requests.

For the sake of security we recommend upgrading certifi frequently!

Body Content Workflow

By default, when you make a request, the body of the response is downloaded immediately. You can override this behaviour and defer downloading the response body until you access the Response.content attribute with the stream parameter:

tarball_url = ''
r = requests.get(tarball_url, stream=True)

At this point only the response headers have been downloaded and the connection remains open, hence allowing us to make content retrieval conditional:

if int(r.headers['content-length']) < TOO_LONG:
  content = r.content

You can further control the workflow by use of the Response.iter_content() and Response.iter_lines() methods. Alternatively, you can read the undecoded body from the underlying urllib3 urllib3.HTTPResponse at Response.raw.

If you set stream to True when making a request, Requests cannot release the connection back to the pool unless you consume all the data or call Response.close. This can lead to inefficiency with connections. If you find yourself partially reading request bodies (or not reading them at all) while using stream=True, you should make the request within a with statement to ensure it’s always closed:

with requests.get('', stream=True) as r:
    # Do things with the response here.


Excellent news — thanks to urllib3, keep-alive is 100% automatic within a session! Any requests that you make within a session will automatically reuse the appropriate connection!

Note that connections are only released back to the pool for reuse once all body data has been read; be sure to either set stream to False or read the content property of the Response object.

Streaming Uploads

Requests supports streaming uploads, which allow you to send large streams or files without reading them into memory. To stream and upload, simply provide a file-like object for your body:

with open('massive-body', 'rb') as f:'http://some.url/streamed', data=f)


It is strongly recommended that you open files in binary mode. This is because Requests may attempt to provide the Content-Length header for you, and if it does this value will be set to the number of bytes in the file. Errors may occur if you open the file in text mode.

Chunk-Encoded Requests

Requests also supports Chunked transfer encoding for outgoing and incoming requests. To send a chunk-encoded request, simply provide a generator (or any iterator without a length) for your body:

def gen():
    yield 'hi'
    yield 'there''http://some.url/chunked', data=gen())

For chunked encoded responses, it’s best to iterate over the data using Response.iter_content(). In an ideal situation you’ll have set stream=True on the request, in which case you can iterate chunk-by-chunk by calling iter_content with a chunk_size parameter of None. If you want to set a maximum size of the chunk, you can set a chunk_size parameter to any integer.

POST Multiple Multipart-Encoded Files

You can send multiple files in one request. For example, suppose you want to upload image files to an HTML form with a multiple file field ‘images’:

<input type="file" name="images" multiple="true" required="true"/>

To do that, just set files to a list of tuples of (form_field_name, file_info):

>>> url = ''
>>> multiple_files = [
...     ('images', ('foo.png', open('foo.png', 'rb'), 'image/png')),
...     ('images', ('bar.png', open('bar.png', 'rb'), 'image/png'))]
>>> r =, files=multiple_files)
>>> r.text
  'files': {'images': 'data:image/png;base64,iVBORw ....'}
  'Content-Type': 'multipart/form-data; boundary=3131623adb2043caaeb5538cc7aa0b3a',


It is strongly recommended that you open files in binary mode. This is because Requests may attempt to provide the Content-Length header for you, and if it does this value will be set to the number of bytes in the file. Errors may occur if you open the file in text mode.

Event Hooks

Requests has a hook system that you can use to manipulate portions of the request process, or signal event handling.

Available hooks:


The response generated from a Request.

You can assign a hook function on a per-request basis by passing a {hook_name: callback_function} dictionary to the hooks request parameter:

hooks={'response': print_url}

That callback_function will receive a chunk of data as its first argument.

def print_url(r, *args, **kwargs):

Your callback function must handle its own exceptions. Any unhandled exception won’t be passed silently and thus should be handled by the code calling Requests.

If the callback function returns a value, it is assumed that it is to replace the data that was passed in. If the function doesn’t return anything, nothing else is affected.

def record_hook(r, *args, **kwargs):
    r.hook_called = True
    return r

Let’s print some request method arguments at runtime:

>>> requests.get('', hooks={'response': print_url})
<Response [200]>

You can add multiple hooks to a single request. Let’s call two hooks at once:

>>> r = requests.get('', hooks={'response': [print_url, record_hook]})
>>> r.hook_called

You can also add hooks to a Session instance. Any hooks you add will then be called on every request made to the session. For example:

>>> s = requests.Session()
>>> s.hooks['response'].append(print_url)
>>> s.get('')
 <Response [200]>

A Session can have multiple hooks, which will be called in the order they are added.

Custom Authentication

Requests allows you to specify your own authentication mechanism.

Any callable which is passed as the auth argument to a request method will have the opportunity to modify the request before it is dispatched.

Authentication implementations are subclasses of AuthBase, and are easy to define. Requests provides two common authentication scheme implementations in requests.auth: HTTPBasicAuth and HTTPDigestAuth.

Let’s pretend that we have a web service that will only respond if the X-Pizza header is set to a password value. Unlikely, but just go with it.

from requests.auth import AuthBase

class PizzaAuth(AuthBase):
    """Attaches HTTP Pizza Authentication to the given Request object."""
    def __init__(self, username):
        # setup any auth-related data here
        self.username = username

    def __call__(self, r):
        # modify and return the request
        r.headers['X-Pizza'] = self.username
        return r

Then, we can make a request using our Pizza Auth:

>>> requests.get('', auth=PizzaAuth('kenneth'))
<Response [200]>

Streaming Requests

With Response.iter_lines() you can easily iterate over streaming APIs such as the Twitter Streaming API. Simply set stream to True and iterate over the response with iter_lines:

import json
import requests

r = requests.get('', stream=True)

for line in r.iter_lines():

    # filter out keep-alive new lines
    if line:
        decoded_line = line.decode('utf-8')

When using decode_unicode=True with Response.iter_lines() or Response.iter_content(), you’ll want to provide a fallback encoding in the event the server doesn’t provide one:

r = requests.get('', stream=True)

if r.encoding is None:
    r.encoding = 'utf-8'

for line in r.iter_lines(decode_unicode=True):
    if line:


iter_lines is not reentrant safe. Calling this method multiple times causes some of the received data being lost. In case you need to call it from multiple places, use the resulting iterator object instead:

lines = r.iter_lines()
# Save the first line for later or just skip it

first_line = next(lines)

for line in lines:


If you need to use a proxy, you can configure individual requests with the proxies argument to any request method:

import requests

proxies = {
  'http': '',
  'https': '',

requests.get('', proxies=proxies)

Alternatively you can configure it once for an entire Session:

import requests

proxies = {
  'http': '',
  'https': '',
session = requests.Session()



Setting session.proxies may behave differently than expected. Values provided will be overwritten by environmental proxies (those returned by urllib.request.getproxies). To ensure the use of proxies in the presence of environmental proxies, explicitly specify the proxies argument on all individual requests as initially explained above.

See #2018 for details.

When the proxies configuration is not overridden per request as shown above, Requests relies on the proxy configuration defined by standard environment variables http_proxy, https_proxy, no_proxy, and all_proxy. Uppercase variants of these variables are also supported. You can therefore set them to configure Requests (only set the ones relevant to your needs):

$ export HTTP_PROXY=""
$ export HTTPS_PROXY=""
$ export ALL_PROXY="socks5://"

$ python
>>> import requests
>>> requests.get('')

To use HTTP Basic Auth with your proxy, use the http://user:password@host/ syntax in any of the above configuration entries:

$ export HTTPS_PROXY="http://user:pass@"

$ python
>>> proxies = {'http': 'http://user:pass@'}


Storing sensitive username and password information in an environment variable or a version-controlled file is a security risk and is highly discouraged.

To give a proxy for a specific scheme and host, use the scheme://hostname form for the key. This will match for any request to the given scheme and exact hostname.

proxies = {'': ''}

Note that proxy URLs must include the scheme.

Finally, note that using a proxy for https connections typically requires your local machine to trust the proxy’s root certificate. By default the list of certificates trusted by Requests can be found with:

from requests.utils import DEFAULT_CA_BUNDLE_PATH

You override this default certificate bundle by setting the REQUESTS_CA_BUNDLE (or CURL_CA_BUNDLE) environment variable to another file path:

$ export REQUESTS_CA_BUNDLE="/usr/local/myproxy_info/cacert.pem"
$ export https_proxy=""

$ python
>>> import requests
>>> requests.get('')


New in version 2.10.0.

In addition to basic HTTP proxies, Requests also supports proxies using the SOCKS protocol. This is an optional feature that requires that additional third-party libraries be installed before use.

You can get the dependencies for this feature from pip:

$ python -m pip install requests[socks]

Once you’ve installed those dependencies, using a SOCKS proxy is just as easy as using a HTTP one:

proxies = {
    'http': 'socks5://user:pass@host:port',
    'https': 'socks5://user:pass@host:port'

Using the scheme socks5 causes the DNS resolution to happen on the client, rather than on the proxy server. This is in line with curl, which uses the scheme to decide whether to do the DNS resolution on the client or proxy. If you want to resolve the domains on the proxy server, use socks5h as the scheme.


Requests is intended to be compliant with all relevant specifications and RFCs where that compliance will not cause difficulties for users. This attention to the specification can lead to some behaviour that may seem unusual to those not familiar with the relevant specification.


When you receive a response, Requests makes a guess at the encoding to use for decoding the response when you access the Response.text attribute. Requests will first check for an encoding in the HTTP header, and if none is present, will use charset_normalizer or chardet to attempt to guess the encoding.

If chardet is installed, requests uses it, however for python3 chardet is no longer a mandatory dependency. The chardet library is an LGPL-licenced dependency and some users of requests cannot depend on mandatory LGPL-licensed dependencies.

When you install requests without specifying [use_chardet_on_py3] extra, and chardet is not already installed, requests uses charset-normalizer (MIT-licensed) to guess the encoding.

The only time Requests will not guess the encoding is if no explicit charset is present in the HTTP headers and the Content-Type header contains text. In this situation, RFC 2616 specifies that the default charset must be ISO-8859-1. Requests follows the specification in this case. If you require a different encoding, you can manually set the Response.encoding property, or use the raw Response.content.

HTTP Verbs

Requests provides access to almost the full range of HTTP verbs: GET, OPTIONS, HEAD, POST, PUT, PATCH and DELETE. The following provides detailed examples of using these various verbs in Requests, using the GitHub API.

We will begin with the verb most commonly used: GET. HTTP GET is an idempotent method that returns a resource from a given URL. As a result, it is the verb you ought to use when attempting to retrieve data from a web location. An example usage would be attempting to get information about a specific commit from GitHub. Suppose we wanted commit a050faf on Requests. We would get it like so:

>>> import requests
>>> r = requests.get('')

We should confirm that GitHub responded correctly. If it has, we want to work out what type of content it is. Do this like so:

>>> if r.status_code ==
...     print(r.headers['content-type'])
application/json; charset=utf-8

So, GitHub returns JSON. That’s great, we can use the r.json method to parse it into Python objects.

>>> commit_data = r.json()

>>> print(commit_data.keys())
['committer', 'author', 'url', 'tree', 'sha', 'parents', 'message']

>>> print(commit_data['committer'])
{'date': '2012-05-10T11:10:50-07:00', 'email': '', 'name': 'Kenneth Reitz'}

>>> print(commit_data['message'])
makin' history

So far, so simple. Well, let’s investigate the GitHub API a little bit. Now, we could look at the documentation, but we might have a little more fun if we use Requests instead. We can take advantage of the Requests OPTIONS verb to see what kinds of HTTP methods are supported on the url we just used.

>>> verbs = requests.options(r.url)
>>> verbs.status_code

Uh, what? That’s unhelpful! Turns out GitHub, like many API providers, don’t actually implement the OPTIONS method. This is an annoying oversight, but it’s OK, we can just use the boring documentation. If GitHub had correctly implemented OPTIONS, however, they should return the allowed methods in the headers, e.g.

>>> verbs = requests.options('')
>>> print(verbs.headers['allow'])

Turning to the documentation, we see that the only other method allowed for commits is POST, which creates a new commit. As we’re using the Requests repo, we should probably avoid making ham-handed POSTS to it. Instead, let’s play with the Issues feature of GitHub.

This documentation was added in response to Issue #482. Given that this issue already exists, we will use it as an example. Let’s start by getting it.

>>> r = requests.get('')
>>> r.status_code

>>> issue = json.loads(r.text)

>>> print(issue['title'])
Feature any http verb in docs

>>> print(issue['comments'])

Cool, we have three comments. Let’s take a look at the last of them.

>>> r = requests.get(r.url + '/comments')
>>> r.status_code

>>> comments = r.json()

>>> print(comments[0].keys())
['body', 'url', 'created_at', 'updated_at', 'user', 'id']

>>> print(comments[2]['body'])
Probably in the "advanced" section

Well, that seems like a silly place. Let’s post a comment telling the poster that he’s silly. Who is the poster, anyway?

>>> print(comments[2]['user']['login'])

OK, so let’s tell this Kenneth guy that we think this example should go in the quickstart guide instead. According to the GitHub API doc, the way to do this is to POST to the thread. Let’s do it.

>>> body = json.dumps({u"body": u"Sounds great! I'll get right on it!"})
>>> url = u""

>>> r =, data=body)
>>> r.status_code

Huh, that’s weird. We probably need to authenticate. That’ll be a pain, right? Wrong. Requests makes it easy to use many forms of authentication, including the very common Basic Auth.

>>> from requests.auth import HTTPBasicAuth
>>> auth = HTTPBasicAuth('', 'not_a_real_password')

>>> r =, data=body, auth=auth)
>>> r.status_code

>>> content = r.json()
>>> print(content['body'])
Sounds great! I'll get right on it.

Brilliant. Oh, wait, no! I meant to add that it would take me a while, because I had to go feed my cat. If only I could edit this comment! Happily, GitHub allows us to use another HTTP verb, PATCH, to edit this comment. Let’s do that.

>>> print(content[u"id"])

>>> body = json.dumps({u"body": u"Sounds great! I'll get right on it once I feed my cat."})
>>> url = u""

>>> r = requests.patch(url=url, data=body, auth=auth)
>>> r.status_code

Excellent. Now, just to torture this Kenneth guy, I’ve decided to let him sweat and not tell him that I’m working on this. That means I want to delete this comment. GitHub lets us delete comments using the incredibly aptly named DELETE method. Let’s get rid of it.

>>> r = requests.delete(url=url, auth=auth)
>>> r.status_code
>>> r.headers['status']
'204 No Content'

Excellent. All gone. The last thing I want to know is how much of my ratelimit I’ve used. Let’s find out. GitHub sends that information in the headers, so rather than download the whole page I’ll send a HEAD request to get the headers.

>>> r = requests.head(url=url, auth=auth)
>>> print(r.headers)
'x-ratelimit-remaining': '4995'
'x-ratelimit-limit': '5000'

Excellent. Time to write a Python program that abuses the GitHub API in all kinds of exciting ways, 4995 more times.

Custom Verbs

From time to time you may be working with a server that, for whatever reason, allows use or even requires use of HTTP verbs not covered above. One example of this would be the MKCOL method some WEBDAV servers use. Do not fret, these can still be used with Requests. These make use of the built-in .request method. For example:

>>> r = requests.request('MKCOL', url, data=data)
>>> r.status_code
200 # Assuming your call was correct

Utilising this, you can make use of any method verb that your server allows.

Transport Adapters

As of v1.0.0, Requests has moved to a modular internal design. Part of the reason this was done was to implement Transport Adapters, originally described here. Transport Adapters provide a mechanism to define interaction methods for an HTTP service. In particular, they allow you to apply per-service configuration.

Requests ships with a single Transport Adapter, the HTTPAdapter. This adapter provides the default Requests interaction with HTTP and HTTPS using the powerful urllib3 library. Whenever a Requests Session is initialized, one of these is attached to the Session object for HTTP, and one for HTTPS.

Requests enables users to create and use their own Transport Adapters that provide specific functionality. Once created, a Transport Adapter can be mounted to a Session object, along with an indication of which web services it should apply to.

>>> s = requests.Session()
>>> s.mount('', MyAdapter())

The mount call registers a specific instance of a Transport Adapter to a prefix. Once mounted, any HTTP request made using that session whose URL starts with the given prefix will use the given Transport Adapter.


The adapter will be chosen based on a longest prefix match. Be mindful prefixes such as http://localhost will also match or It’s recommended to terminate full hostnames with a /.

Many of the details of implementing a Transport Adapter are beyond the scope of this documentation, but take a look at the next example for a simple SSL use- case. For more than that, you might look at subclassing the BaseAdapter.

Example: Specific SSL Version

The Requests team has made a specific choice to use whatever SSL version is default in the underlying library (urllib3). Normally this is fine, but from time to time, you might find yourself needing to connect to a service-endpoint that uses a version that isn’t compatible with the default.

You can use Transport Adapters for this by taking most of the existing implementation of HTTPAdapter, and adding a parameter ssl_version that gets passed-through to urllib3. We’ll make a Transport Adapter that instructs the library to use SSLv3:

import ssl
from urllib3.poolmanager import PoolManager

from requests.adapters import HTTPAdapter

class Ssl3HttpAdapter(HTTPAdapter):
    """"Transport adapter" that allows us to use SSLv3."""

    def init_poolmanager(self, connections, maxsize, block=False):
        self.poolmanager = PoolManager(
            num_pools=connections, maxsize=maxsize,
            block=block, ssl_version=ssl.PROTOCOL_SSLv3)

Example: Automatic Retries

By default, Requests does not retry failed connections. However, it is possible to implement automatic retries with a powerful array of features, including backoff, within a Requests Session using the urllib3.util.Retry class:

from urllib3.util import Retry
from requests import Session
from requests.adapters import HTTPAdapter

s = Session()
retries = Retry(
    status_forcelist=[502, 503, 504],
s.mount('https://', HTTPAdapter(max_retries=retries))

Blocking Or Non-Blocking?

With the default Transport Adapter in place, Requests does not provide any kind of non-blocking IO. The Response.content property will block until the entire response has been downloaded. If you require more granularity, the streaming features of the library (see Streaming Requests) allow you to retrieve smaller quantities of the response at a time. However, these calls will still block.

If you are concerned about the use of blocking IO, there are lots of projects out there that combine Requests with one of Python’s asynchronicity frameworks. Some excellent examples are requests-threads, grequests, requests-futures, and httpx.

Header Ordering

In unusual circumstances you may want to provide headers in an ordered manner. If you pass an OrderedDict to the headers keyword argument, that will provide the headers with an ordering. However, the ordering of the default headers used by Requests will be preferred, which means that if you override default headers in the headers keyword argument, they may appear out of order compared to other headers in that keyword argument.

If this is problematic, users should consider setting the default headers on a Session object, by setting Session.headers to a custom OrderedDict. That ordering will always be preferred.


Most requests to external servers should have a timeout attached, in case the server is not responding in a timely manner. By default, requests do not time out unless a timeout value is set explicitly. Without a timeout, your code may hang for minutes or more.

The connect timeout is the number of seconds Requests will wait for your client to establish a connection to a remote machine (corresponding to the connect()) call on the socket. It’s a good practice to set connect timeouts to slightly larger than a multiple of 3, which is the default TCP packet retransmission window.

Once your client has connected to the server and sent the HTTP request, the read timeout is the number of seconds the client will wait for the server to send a response. (Specifically, it’s the number of seconds that the client will wait between bytes sent from the server. In 99.9% of cases, this is the time before the server sends the first byte).

If you specify a single value for the timeout, like this:

r = requests.get('', timeout=5)

The timeout value will be applied to both the connect and the read timeouts. Specify a tuple if you would like to set the values separately:

r = requests.get('', timeout=(3.05, 27))

If the remote server is very slow, you can tell Requests to wait forever for a response, by passing None as a timeout value and then retrieving a cup of coffee.

r = requests.get('', timeout=None)


The connect timeout applies to each connection attempt to an IP address. If multiple addresses exist for a domain name, the underlying urllib3 will try each address sequentially until one successfully connects. This may lead to an effective total connection timeout multiple times longer than the specified time, e.g. an unresponsive server having both IPv4 and IPv6 addresses will have its perceived timeout doubled, so take that into account when setting the connection timeout.


Neither the connect nor read timeouts are wall clock. This means that if you start a request, and look at the time, and then look at the time when the request finishes or times out, the real-world time may be greater than what you specified.