API Reference

This page lists all of the interfaces exposed by the treq package.

Making Requests

The treq module provides several convenience functions for making requests. These functions all create a default treq.client.HTTPClient instance and pass their arguments to the appropriate HTTPClient method.

treq.request(method, url, **kwargs)[source]

Make an HTTP request.

  • method (str) – HTTP method. Example: 'GET', 'HEAD'. 'PUT', 'POST'.

  • url (hyperlink.DecodedURL, str, bytes, or hyperlink.EncodedURL) – http or https URL, which may include query arguments.

  • headers (Headers or None) – Optional HTTP Headers to send with this request.

  • params (dict w/ str or list/tuple of str values, list of 2-tuples, or None.) – Optional parameters to be append to the URL query string. Any query string parameters in the url will be preserved.

  • data (bytes, typing.BinaryIO, IBodyProducer, or None) –

    Arbitrary request body data.

    If files is also passed this must be a dict, a tuple or list of field tuples as accepted by MultiPartProducer. The request is assigned a Content-Type of multipart/form-data.

    If a dict, list, or tuple it is URL-encoded and the request assigned a Content-Type of application/x-www-form-urlencoded.

    Otherwise, any non-None value is passed to the client’s data_to_body_producer callable (by default, IBodyProducer), which accepts bytes and binary files like returned by open(..., "rb").

  • files

    Files to include in the request body, in any of the several formats:

    • [("fieldname", binary_file)]

    • [("fieldname", "filename", binary_file)]

    • [("fieldname, "filename', "content-type", binary_file)]

    Or a mapping:

    • {"fieldname": binary_file}

    • {"fieldname": ("filename", binary_file)}

    • {"fieldname": ("filename", "content-type", binary_file)}

    Each binary_file is a file-like object open in binary mode (like returned by open("filename", "rb")). The filename is taken from the file’s name attribute if not specified. The Content-Type is guessed based on the filename using mimetypes.guess_type() if not specified, falling back to application/octet-stream.

    While uploading Treq will measure the length of seekable files to populate the Content-Length header of the file part.

    If files is given the request is assigned a Content-Type of multipart/form-data. Additional fields may be given in the data argument.

  • json (dict, list, tuple, int, str, bool, or None) – Optional JSON-serializable content for the request body. Mutually exclusive with data and files.

  • auth (tuple of ('username', 'password')) – HTTP Basic Authentication information — see treq.auth.add_auth().

  • cookies (dict or cookielib.CookieJar) – Cookies to send with this request. The HTTP kind, not the tasty kind.

  • timeout (int) – Request timeout seconds. If a response is not received within this timeframe, a connection is aborted with CancelledError.

  • allow_redirects (bool) – Follow HTTP redirects. Default: True

  • browser_like_redirects (bool) – Follow redirects like a web browser: When a 301 or 302 redirect is received in response to a POST request convert the method to GET. See 7231 and BrowserLikeRedirectAgent). Default: False

  • unbuffered (bool) – Pass True to to disable response buffering. By default treq buffers the entire response body in memory.

  • reactor – Optional Twisted reactor.

  • persistent (bool) – Use persistent HTTP connections. Default: True

  • agent (twisted.web.iweb.IAgent) – Provide your own custom agent. Use this to override things like connectTimeout or BrowserLikePolicyForHTTPS. By default, treq will create its own Agent with reasonable defaults.

Return type:

Deferred that fires with an IResponse

Changed in version treq: 20.9.0

The url param now accepts hyperlink.DecodedURL and hyperlink.EncodedURL objects.

treq.get(url, headers=None, **kwargs)[source]

Make a GET request.

See treq.request()

treq.head(url, **kwargs)[source]

Make a HEAD request.

See treq.request()

treq.post(url, data=None, **kwargs)[source]

Make a POST request.

See treq.request()

treq.put(url, data=None, **kwargs)[source]

Make a PUT request.

See treq.request()

treq.patch(url, data=None, **kwargs)[source]

Make a PATCH request.

See treq.request()

treq.delete(url, **kwargs)[source]

Make a DELETE request.

See treq.request()

Accessing Content

treq.collect(response, collector)[source]

Incrementally collect the body of the response.

This function may only be called once for a given response.

If the collector raises an exception, it will be set as the error value on response Deferred returned from this function, and the underlying HTTP transport will be closed.

  • response (IResponse) – The HTTP response to collect the body from.

  • collector (single argument callable) – A callable to be called each time data is available from the response body.

Return type:

Deferred that fires with None when the entire body has been read.


Read the contents of an HTTP response.

This function may be called multiple times for a response, it uses a WeakKeyDictionary to cache the contents of the response.


response (IResponse) – The HTTP Response to get the contents of.

Return type:

Deferred that fires with the content as a str.

treq.text_content(response, encoding='ISO-8859-1')[source]

Read the contents of an HTTP response and decode it with an appropriate charset, which may be guessed from the Content-Type header.

  • response (IResponse) – The HTTP Response to get the contents of.

  • encoding (str) – A charset, such as UTF-8 or ISO-8859-1, used if the response does not specify an encoding.

Return type:

Deferred that fires with a unicode string.

treq.json_content(response, **kwargs)[source]

Read the contents of an HTTP response and attempt to decode it as JSON.

This function relies on content() and so may be called more than once for a given response.

  • response (IResponse) – The HTTP Response to get the contents of.

  • kwargs – Any keyword arguments accepted by json.loads()

Return type:

Deferred that fires with the decoded JSON.

The HTTP Client

treq.client.HTTPClient has methods that match the signatures of the convenience request functions in the treq module.

class treq.client.HTTPClient(agent, cookiejar=None, data_to_body_producer=IBodyProducer)[source]
request(method, url, *, params=None, headers=None, data=None, files=None, json=<object object>, auth=None, cookies=None, allow_redirects=True, browser_like_redirects=False, unbuffered=False, reactor=None, timeout=None, _stacklevel=2)[source]

See treq.request().

get(url, **kwargs)[source]

See treq.get().

head(url, **kwargs)[source]

See treq.head().

post(url, data=None, **kwargs)[source]

See treq.post().

put(url, data=None, **kwargs)[source]

See treq.put().

patch(url, data=None, **kwargs)[source]

See treq.patch().

delete(url, **kwargs)[source]

See treq.delete().

Augmented Response Objects

treq.request(), treq.get(), etc. return an object which provides twisted.web.iweb.IResponse, plus a few additional convenience methods:

class treq.response._Response[source]

Incrementally collect the body of the response, per treq.collect().


collector – A single argument callable that will be called with chunks of body data as it is received.


A Deferred that fires when the entire body has been received.


Read the entire body all at once, per treq.content().


A Deferred that fires with a bytes object when the entire body has been received.


Collect the response body as JSON per treq.json_content().


kwargs – Any keyword arguments accepted by json.loads()

Return type:

Deferred that fires with the decoded JSON when the entire body has been read.


Read the entire body all at once as text, per treq.text_content().

Return type:

A Deferred that fires with a unicode string when the entire body has been received.


Get a list of all responses that (such as intermediate redirects), that ultimately ended in the current response. The responses are ordered chronologically.


A list of _Response objects


Get a copy of this response’s cookies.

Return type:


Inherited from twisted.web.iweb.IResponse:


See IResponse.deliverBody()


See IResponse.setPreviousResponse()


treq.auth.add_auth(agent, auth_config)[source]

Wrap an agent to perform authentication

  • agent – Agent to wrap.

  • auth_config – A ('username', 'password') tuple — see add_basic_auth().




UnknownAuthConfig – When the format auth_config isn’t supported.

treq.auth.add_basic_auth(agent: <InterfaceClass twisted.web.iweb.IAgent>, username: str | bytes, password: str | bytes) <InterfaceClass twisted.web.iweb.IAgent>[source]

Wrap an agent to add HTTP basic authentication

The returned agent sets the Authorization request header according to the basic authentication scheme described in RFC 7617. This header contains the given username and password in plaintext, and thus should only be used over an encrypted transport (HTTPS).

Note that the colon (:) is used as a delimiter between the username and password, so if either parameter includes a colon the interpretation of the Authorization header is server-defined.

  • agent – Agent to wrap.

  • username – The username.

  • password – The password.



exception treq.auth.UnknownAuthConfig(config)[source]

The authentication config provided couldn’t be interpreted.

Test Helpers

The treq.testing module contains tools for in-memory testing of HTTP clients and servers.

StubTreq Objects

class treq.testing.StubTreq(resource)

StubTreq implements the same interface as the treq module or the HTTPClient class, with the limitation that it does not support the files argument.


Flush all data between pending client/server pairs.

This is only necessary if a Resource under test returns NOT_DONE_YET from its render method, making a response asynchronous. In that case, after each write from the server, flush() must be called so the client can see it.

As the methods on treq.client.HTTPClient:


See treq.request().


See treq.get().


See treq.head().


See treq.post().


See treq.put().


See treq.patch().


See treq.delete().

RequestTraversalAgent Objects

class treq.testing.RequestTraversalAgent(rootResource)[source]

IAgent implementation that issues an in-memory request rather than going out to a real network socket.


Flush all data between pending client/server pairs.

This is only necessary if a Resource under test returns NOT_DONE_YET from its render method, making a response asynchronous. In that case, after each write from the server, flush() must be called so the client can see it.

request(method, uri, headers=None, bodyProducer=None)[source]

Implement IAgent.request.

RequestSequence Objects

class treq.testing.RequestSequence(sequence, async_failure_reporter=None)[source]

For an example usage, see RequestSequence.consume().

Takes a sequence of:

[((method, url, params, headers, data), (code, headers, body)),

Expects the requests to arrive in sequence order. If there are no more responses, or the request’s parameters do not match the next item’s expected request parameters, calls sync_failure_reporter or async_failure_reporter.

For the expected request tuples:

  • method should be bytes normalized to lowercase.

  • url should be a str normalized as per the transformations in that (usually) preserve semantics. A URL to http://something-that-looks-like-a-directory would be normalized to http://something-that-looks-like-a-directory/ and a URL to http://something-that-looks-like-a-page/page.html remains unchanged.

  • params is a dictionary mapping bytes to list of bytes.

  • headers is a dictionary mapping bytes to list of bytes – note that twisted.web.client.Agent may add its own headers which are not guaranteed to be present (for instance, user-agent or content-length), so it’s better to use some kind of matcher like HasHeaders.

  • data is a bytes.

For the response tuples:

  • code is an integer representing the HTTP status code to return.

  • headers is a dictionary mapping bytes to bytes or str. Note that the value is not a list.

  • body is a bytes.

  • sequence (list) – A sequence of (request tuple, response tuple) two-tuples, as described above.

  • async_failure_reporter – An optional callable that takes a str message indicating a failure. It’s asynchronous because it cannot just raise an exception—if it does, Resource.render will just convert that into a 500 response, and there will be no other failure reporting mechanism.

When the async_failure_reporter parameter is not passed, async failures will be reported via a twisted.logger.Logger instance, which Trial’s test case classes (twisted.trial.unittest.TestCase and SynchronousTestCase) will translate into a test failure.


Some versions of twisted.trial.unittest.SynchronousTestCase report logged errors on the wrong test: see Twisted #9267.

When not subclassing Trial’s classes you must pass async_failure_reporter and implement equivalent behavior or errors will pass silently. For example:

async_failures = []
sequence_stubs = RequestSequence([...], async_failures.append)
stub_treq = StubTreq(StringStubbingResource(sequence_stubs))
with sequence_stubs.consume(self.fail):  # self = unittest.TestCase

self.assertEqual([], async_failures)


sequence_stubs = RequestSequence([...])
stub_treq = StubTreq(StringStubbingResource(sequence_stubs))
# self = twisted.trial.unittest.SynchronousTestCase
with sequence_stubs.consume(self.fail):

If there are still remaining expected requests to be made in the sequence, fails the provided test case.


sync_failure_reporter – A callable that takes a single message reporting failures. This can just raise an exception - it does not need to be asynchronous, since the exception would not get raised within a Resource.


a context manager that can be used to ensure all expected requests have been made.


bool representing whether the entire sequence has been consumed. This is useful in tests to assert that the expected requests have all been made.

StringStubbingResource Objects

class treq.testing.StringStubbingResource(get_response_for)[source]

A resource that takes a callable with 5 parameters (method, url, params, headers, data) and returns (code, headers, body).

The resource uses the callable to return a real response as a result of a request.

The parameters for the callable are:

  • method, the HTTP method as bytes.

  • url, the full URL of the request as text.

  • params, a dictionary of query parameters mapping query keys lists of values (sorted alphabetically).

  • headers, a dictionary of headers mapping header keys to a list of header values (sorted alphabetically).

  • data, the request body as bytes.

The callable must return a tuple of (code, headers, body) where the code is the HTTP status code, the headers is a dictionary of bytes (unlike the headers parameter, which is a dictionary of lists), and body is a string that will be returned as the response body.

If there is a stubbing error, the return value is undefined (if an exception is raised, Resource will just eat it and return 500 in its place). The callable, or whomever creates the callable, should have a way to handle error reporting.


Produce a response according to the stubs provided.

HasHeaders Objects

class treq.testing.HasHeaders(headers)[source]

Since Twisted adds headers to a request, such as the host and the content length, it’s necessary to test whether request headers CONTAIN the expected headers (the ones that are not automatically added by Twisted).

This wraps a set of headers, and can be used in an equality test against a superset if the provided headers. The headers keys are lowercased, and keys and values are compared in their bytes-encoded forms.

Headers should be provided as a mapping from strings or bytes to a list of strings or bytes.

MultiPartProducer Objects

treq.multipart.MultiPartProducer is used internally when making requests which involve files.

class treq.multipart.MultiPartProducer(fields, boundary=None, cooperator=<module 'twisted.internet.task' from '/home/docs/checkouts/readthedocs.org/user_builds/treq/envs/release-23.11.0/lib/python3.11/site-packages/twisted/internet/task.py'>)[source]

MultiPartProducer takes parameters for a HTTP request and produces bytes in multipart/form-data format defined in RFC 2388 and RFC 2046.

The encoded request is produced incrementally and the bytes are written to a consumer.

Fields should have form: [(parameter name, value), ...]

Accepted values:

  • Unicode strings (in this case parameter will be encoded with utf-8)

  • Tuples with (file name, content-type, IBodyProducer objects)

Since MultiPartProducer can accept objects like IBodyProducer which cannot be read from in an event-driven manner it uses uses a Cooperator instance to schedule reads from the underlying producers. Reading is also paused and resumed based on notifications from the IConsumer provider being written to.

  • _fields – Sorted parameters, where all strings are enforced to be unicode and file objects stacked on bottom (to produce a human readable form-data request)

  • _cooperate – A method like Cooperator.cooperate which is used to schedule all reads.

  • boundary – The generated boundary used in form-data encoding


Temporarily suspend copying bytes from the input file to the consumer by pausing the CooperativeTask which drives that activity.


Undo the effects of a previous pauseProducing and resume copying bytes to the consumer by resuming the CooperativeTask which drives the write activity.


Start a cooperative task which will read bytes from the input file and write them to consumer. Return a Deferred which fires after all bytes have been written.


consumer – Any IConsumer provider


Permanently stop writing bytes from the file to the consumer by stopping the underlying CooperativeTask.