Need help with requests-respectful?
Click the “chat” button below for chat support from the developer who created it, or find similar developers for support.

About the developer

422 Stars 27 Forks Other 18 Commits 5 Opened issues


Minimalist Requests wrapper to work within rate limits of any amount of services simultaneously. Parallel processing friendly.

Services available


Need anything else?

Contributors list


If you know Python, you know Requests. Requests is love. Requests is life. Depending on your use cases, you may come across scenarios where you need to use Requests a lot. Services you consume may have rate-limiting policies in place or you may just happen to be in a good mood and feel like being a good Netizen. This is where requests-respectful can come in handy.


  • Is a minimalist wrapper on top of Requests to work within rate limits of any amount of services simultaneously
  • Can scale out of a single thread, single process or even a single machine
  • Enables maximizing your allowed requests without ever going over set limits and having to handle the fallout
  • Proxies Requests HTTP verb methods (for minimal code changes)
  • Works with both Python 2 and 3 and is fully tested
  • Is cool (hopefully?)

Typical requests call

import requests
response = requests.get("", params={"foo": "bar"})

Magic requests-respectful call - requests verb methods are proxied!

from requests_respectful import RespectfulRequester

rr = RespectfulRequester()

This can be done elsewhere but the realm needs to be registered!

rr.register_realm("Github", max_requests=100, timespan=60)

response = rr.get("", params={"foo": "bar"}, realms=["Github"], wait=True)

Conservative requests-respectful call - pass a lambda with a requests method call

import requests
from requests_respectful import RespectfulRequester

rr = RespectfulRequester()

This can be done elsewhere but the realm needs to be registered!

rr.register_realm("Github", max_requests=100, timespan=60)

request_func = lambda: requests.get("", params={"foo": "bar"}) response = rr.request(request_func, realms=["Github"], wait=True)


  • Redis > 2.8.0 (See FAQ if you are rolling your eyes)


pip install requests-respectful


Default Configuration Values

    "redis": {
        "host": "localhost",
        "port": 6379,
        "database": 0
    "safety_threshold": 10,
    "requests_module_name": "requests"

Configuration Keys

  • redis: Provides the
    of the Redis instance
  • safety_threshold: A rate-limited exception will be raised at (realmmaxrequests - safety_threshold). Prevents going over the limit of services in scenarios where a large amount of requests are issued in parallel
  • requestsmodulename: Provides the name of the Requests module used in the request lambdas. Should not need to be changed unless you import Requests as another name.

Overriding Configuration Values

With requests-respectful.config.yml

The library auto-detects the presence of a YAML file named requests-respectful.config.yml at the root of your project and will attempt to load configuration values from it.


requests-respectful.config.yml ```yaml redis: host: port: 6379 database: 5

safety_threshold: 25 ```

With the configure() class method

If you don't like having an extra file lying around, the library can also be configured at runtime using the configure() class method.

    redis={"host": "", "port": 6379, "database": 5},

In both cases, the resulting active configuration would be: ```python RespectfulRequester._config()

Out[1]: { "redis": { "host": "", "port": 6379, "database": 5 }, "safetythreshold": 25, "requestsmodule_name": "requests" } ```


In your quest to use requests-respectful, you should only ever have to bother with one class: RespectfulRequester. Instance this class and you can perform all important operations.

Before each example, it is assumed that the following code has already been executed.

from requests_respectful import RespectfulRequester
rr = RespectfulRequester()


Realms are simply named containers that are provided with a maximum requesting rate. You are responsible of the management (i.e. CRUD) of your realms.

Realms track the HTTP requests that are performed under them and will raise a catchable rate limit exception if you are over their allowed requesting rate.

Fetching the list of Realms


This returns a list of currently registered realm names.

Registering a Realm

rr.register_realm("Google", max_requests=10, timespan=1)
rr.register_realm("Github", max_requests=100, timespan=60)
rr.register_realm("Twitter", max_requests=150, timespan=300)


realm_tuples = [ ["Google", 10, 1], ["Github", 100, 60], ["Twitter", 150, 300] ]


Either of these registers 3 realms: * Google at a maximum requesting rate of 10 requests per second * Github at a maximum requesting rate of 100 requests per minute * Twitter at a maximum requesting rate of 150 requests per 5 minutes

Updating a Realm

rr.update_realm("Google", max_requests=25, timespan=5)

This updates the maximum requesting rate of Google to 25 requests per 5 seconds.

Getting the maximum requests value of a Realm


This would return 25.

Getting the timespan value of a Realm


This would return 5.

Unregistering a Realm


This would unregister the Google realm, preventing further queries from executing on it.

Unregistering multiple Realms

rr.unregister_realms(["Google", "Github", "Twitter"])

This would unregister all 3 realms in one operation, preventing further queries from executing on them.


Using Requests HTTP verb methods

The library supports proxying calls to the 7 Requests HTTP verb methods (DELETE, GET, HEAD, OPTIONS, PATCH, POST, PUT). This is literally a Requests method so go crazy with your params, body, headers, auth etc. kwargs. The only major difference is that a realm kwarg is expected. A wait boolean kwargs can also be provided (the behavior is explained later).

These are all valid calls:

rr.get("", realms=["HTTPBin"])'', data = {'key':'value'}, realms=["HTTPBin"], wait=True)
rr.put('', data = {'key':'value'}, realms=["HTTPBin"])
rr.delete('', realms=["HTTPBin"])

If not rate-limited, these would return your usual requests.Response object.

Using a request lamba

If you are a purist and prefer not using fancy proxying, you are also allowed to create a lambda of your Requests call and pass it to the request() instance method.

request_func = lambda:'', data = {'key':'value'})
rr.request(request_func, realms=["HTTPBin"], wait=True)

If not rate-limited, this would return your usual requests.Response object.

Multiple realms per request

Starting in 0.2.0, you can have a single request count against multiple realms. The kwarg has been changed from

and works as you would expect it to.
rr.get("", realms=["HTTPBin", "HTTPBinUser123", "HTTPBinServer3"])

The kwarg

has been deprecated on requesting instance methods. It will still work with a warning until 0.3.0

Handling exceptions

Executing these calls will either return a requests.Response object with the results of the HTTP call or raise a RequestsRespectfulRateLimitedError exception. This means that you'll likely want to catch and handle that exception.

from requests_respectful import RequestsRespectfulRateLimitedError

try: response = rr.get("", realm="HTTPBin") except RequestsRespectfulRateLimitedError: pass # Possibly requeue that call or wait.

The wait kwarg

Both ways of requesting accept a wait kwarg that defaults to False. If switched on and the realm is currently rate-limited, the process will block, wait until it is safe to send requests again and perform the requests then. Waiting is perfectly fine for scripts or smaller operations but is discouraged for large, multi-realm, parallel tasks (i.e. Background Tasks like Celery workers).


  • Exist?
  • Exhaustive?
  • Facepalm tactics?
    Yes -  Redis calls aren't mocked and gets a few friendly calls

Run them with

python -m pytest tests --spec


Whoa, whoa, whoa! Redis?!

Yes. The use of Redis allows for requests-respectful to go multi-thread, multi-process and even multi-machine while still respecting the maximum requesting rates of registered realms. Operations like Redis' SETEX are key in designing and working with rate-limiting systems. If you are doing Python development, there is a decent chance you already work with Redis as it is one of the two options to use as Celery's backend and one of the 2 major caching options in Web development. If not, you can always keep things clean and use a Docker Container or even build it from source. Redis has kept a consistent record over the years of being lightweight, solid software.

How is this different than other throttling libraries?

  • Most other libraries will ask you to specify an interval at which to send requests and will literally loop over
    . This one will allow to send as many as you want, as fast as you want, as long as you are under the maximum requesting rate of your realm.
  • Other libraries don't have the concept of realms and separate requesting rate rules.
  • Other libraries don't scale outside of the process.
  • Most other libraries don't integrate this neatly with Requests

Roadmap / Contribution Ideas

  • Provide some introspection methods to get live realm stats
  • Create a curses realm stats monitor
  • Provide real-life use cases
  • Read the Docs RST Documentation
  • Mock out the Redis calls in the tests
  • Mock out the Requests calls in the tests

We use cookies. If you continue to browse the site, you agree to the use of cookies. For more information on our use of cookies please see our Privacy Policy.