MonkeyLearn API for Python

Official Python client for the MonkeyLearn API. Build and run machine learning models for language processing from your Python apps.

Installation

You can use pip to install the library:

$ pip install monkeylearn

Alternatively, you can just clone the repository and run the setup.py script:

$ python setup.py install

Usage

Before making requests to the API, you need to create an instance of the MonkeyLearn client. You will have to use your account API Key:

from monkeylearn import MonkeyLearn

# Instantiate the client Using your API key
ml = MonkeyLearn('<YOUR API TOKEN HERE>')

Requests

From the MonkeyLearn client instance, you can call any endpoint (check the available endpoints below). For example, you can classify a list of texts using the public Sentiment analysis classifier:

response = ml.classifiers.classify(
    model_id='cl_Jx8qzYJh',
    data=[
        'Great hotel with excellent location',
        'This is the worst hotel ever.'
    ]
)

Responses

The response object returned by every endpoint call is a MonkeyLearnResponse object. The body attribute has the parsed response from the API:

print(response.body)
# =>  [
# =>      {
# =>          'text': 'Great hotel with excellent location',
# =>          'external_id': null,
# =>          'error': false,
# =>          'classifications': [
# =>              {
# =>                  'tag_name': 'Positive',
# =>                  'tag_id': 1994,
# =>                  'confidence': 0.922,
# =>              }
# =>          ]
# =>      },
# =>      {
# =>          'text': 'This is the worst hotel ever.',
# =>          'external_id': null,
# =>          'error': false,
# =>          'classifications': [
# =>              {
# =>                  'tag_name': 'Negative',
# =>                  'tag_id': 1941,
# =>                  'confidence': 0.911,
# =>              }
# =>          ]
# =>      }
# =>  ]

You can also access other attributes in the response object to get information about the queries used or available:

print(response.plan_queries_allowed)
# =>  300

print(response.plan_queries_remaining)
# =>  240

print(response.request_queries_used)
# =>  2

Errors

Endpoint calls may raise exceptions. Here is an example on how to handle them:

from monkeylearn.exceptions import PlanQueryLimitError, MonkeyLearnException

try:
    response = ml.classifiers.classify('[MODEL_ID]', data=['My text'])
except PlanQueryLimitError as e:
    # No monthly queries left
    # e.response contains the MonkeyLearnResponse object
    print(e.error_code, e.detail)
except MonkeyLearnException:
    raise

Available exceptions:

class	Description
MonkeyLearnException	Base class for every exception below.
RequestParamsError	An invalid parameter was sent. Check the exception message or response object for more information.
AuthenticationError	Authentication failed, usually because an invalid token was provided. Check the exception message. More about Authentication.
ForbiddenError	You don't have permissions to perform the action on the given resource.
ModelLimitError	You have reached the custom model limit for your plan.
ModelNotFound	The model does not exist. Check the model_id.
TagNotFound	The tag does not exist. Check the tag_id parameter.
PlanQueryLimitError	You have reached the monthly query limit for your plan. Consider upgrading your plan. More about Plan query limits.
PlanRateLimitError	You have sent too many requests in the last minute. Check the exception detail. More about Plan rate limit.
ConcurrencyRateLimitError	You have sent too many requests in the last second. Check the exception detail. More about Concurrency rate limit.
ModelStateError	The state of the model is invalid. Check the exception detail.

Auto-batching

Classify and Extract endpoints might require more than one request to the MonkeyLearn API in order to process every text in the data parameter. If the auto_batch parameter is True (which is the default value), you won't have to keep the data length below the max allowed value (200). You can just pass the full list and the library will handle the batching and make the necessary requests. If the retry_if_throttled parameter is True (which is the default value), it will also wait and retry if the API throttled a request.

Let's say you send a data parameter with 300 texts and auto_batch is enabled. The list will be split internally and two requests will be sent to MonkeyLearn with 200 and 100 texts, respectively. If all requests respond with a 200 status code, the responses will be appended and you will get the 300 classifications as usual in the MonkeyLearnResponse.body attribute:

data = ['Text to classify'] * 300
response = ml.classifiers.classify('[MODEL_ID]', data)
assert len(response.body) == 300  # => True

Now, let's say you only had 200 queries left when trying the previous example, the second internal request would fail since you wouldn't have queries left after the first batch and a PlanQueryLimitError exception would be raised. The first 200 (successful) classifications will be in the exception object. However, if you don't manage this exception with an except clause, those first 200 successful classifications will be lost. Here's how you should handle that case:

from monkeylearn.exceptions import PlanQueryLimitError

data = ['Text to classify'] * 300
batch_size = 200

try:
    response = ml.classifiers.classify('[MODEL_ID]', data, batch_size=batch_size)
except PlanQueryLimitError as e:
    partial_predictions = e.response.body  # The body of the successful responses
    non_2xx_raw_responses = r.response.failed_raw_responses  # List of requests responses objects
else:
    predictions = response.body

This is very convenient and usually should be enough. If you need more flexibility, you can manage batching and rate limits yourself.

from time import sleep
from monkeylearn.exceptions import PlanQueryLimitError, ConcurrencyRateLimitError, PlanRateLimitError

data = ['Text to classify'] * 300
batch_size = 200
predictions = []

for i in range(0, len(data), batch_size):
    batch_data = data[i:i + batch_size]

    retry = True
    while retry:
        try:
            retry = True
            response = ml.classifiers.classify('[MODEL_ID]', batch_data, auto_batch=False,
                                               retry_if_throttled=False)
        except PlanRateLimitError as e:
            sleep(e.seconds_to_wait)
        except ConcurrencyRateLimitError:
            sleep(2)
        except PlanQueryLimitError:
            raise
        else:
            retry = False

    predictions.extend(response.body)

This way you'll be able to control every request that is sent to the MonkeyLearn API.

Available endpoints

These are all the endpoints of the API. For more information about each endpoint, check out the API documentation.

Classifiers

Classify

def MonkeyLearn.classifiers.classify(model_id, data, production_model=False, batch_size=200,
                                     auto_batch=True, retry_if_throttled=True)

Parameters:

Parameter	Type	Description
model_id	str	Classifier ID. It always starts with 'cl', for example, 'cl_oJNMkt2V'.
data	list[str or dict]	A list of up to 200 data elements to classify. Each element must be a string with the text or a dict with the required text key and the text as the value. You can provide an optional external_id key with a string that will be included in the response.
production_model	bool	Indicates if the classifications are performed by the production model. Only use this parameter with custom models (not with the public ones). Note that you first need to deploy your model to production either from the UI model settings or by using the Classifier deploy endpoint.
batch_size	int	Max number of texts each request will send to MonkeyLearn. A number from 1 to 200.
auto_batch	bool	Split the data list into smaller valid lists, send each one in separate request to MonkeyLearn, and merge the responses.
retry_if_throttled	bool	If a request is throttled, sleep and retry the request.

Example:

data = ['First text', {'text': 'Second text', 'external_id': '2'}]
response = ml.classifiers.classify('[MODEL_ID]', data)

Classifier detail

def MonkeyLearn.classifiers.detail(model_id, retry_if_throttled=True)

Parameters:

Parameter	Type	Description
model_id	str	Classifier ID. It always starts with 'cl', for example, 'cl_oJNMkt2V'.
retry_if_throttled	bool	If a request is throttled, sleep and retry the request.

Example:

response = ml.classifiers.detail('[MODEL_ID]')

Create Classifier

def MonkeyLearn.classifiers.create(name, description='', algorithm='nb', language='en',
                                   max_features=10000, ngram_range=(1, 1), use_stemming=True,
                                   preprocess_numbers=True, preprocess_social_media=False,
                                   normalize_weights=True, stopwords=True, whitelist=None,
                                   retry_if_throttled=True)

Parameters:

Parameter	Type	Description
name	str	The name of the model.
description	str	The description of the model.
algorithm	str	The algorithm used when training the model. It can be either "nb" or "svm".
language	str	The language of the model. Full list of supported languages.
max_features	int	The maximum number of features used when training the model. Between 10 and 100000.
ngram_range	tuple(int,int)	Indicates which n-gram range used when training the model. A list of two numbers between 1 and 3. They indicate the minimum and the maximum n for the n-grams used.
use_stemming	bool	Indicates whether stemming is used when training the model.
preprocess_numbers	bool	Indicates whether number preprocessing is done when training the model.
preprocess_social_media	bool	Indicates whether preprocessing of social media is done when training the model.
normalize_weights	bool	Indicates whether weights will be normalized when training the model.
stopwords	bool or list	The list of stopwords used when training the model. Use False for no stopwords, True for the default stopwords, or a list of strings for custom stopwords.
whitelist	list	The whitelist of words used when training the model.
retry_if_throttled	bool	If a request is throttled, sleep and retry the request.

Example:

response = ml.classifiers.create(name='New classifier', stopwords=True)

Edit Classifier

def MonkeyLearn.classifiers.edit(model_id, name=None, description=None, algorithm=None,
                                 language=None, max_features=None, ngram_range=None,
                                 use_stemming=None, preprocess_numbers=None,
                                 preprocess_social_media=None, normalize_weights=None,
                                 stopwords=None, whitelist=None, retry_if_throttled=None)

Parameters:

Parameter	Type	Description
model_id	str	Classifier ID. It always starts with 'cl', for example, 'cl_oJNMkt2V'.
name	str	The name of the model.
description	str	The description of the model.
algorithm	str	The algorithm used when training the model. It can be either "nb" or "svm".
language	str	The language of the model. Full list of supported languages.
max_features	int	The maximum number of features used when training the model. Between 10 and 100000.
ngram_range	tuple(int,int)	Indicates which n-gram range used when training the model. A list of two numbers between 1 and 3. They indicate the minimum and the maximum n for the n-grams used.
use_stemming	bool	Indicates whether stemming is used when training the model.
preprocess_numbers	bool	Indicates whether number preprocessing is done when training the model.
preprocess_social_media	bool	Indicates whether preprocessing of social media is done when training the model.
normalize_weights	bool	Indicates whether weights will be normalized when training the model.
stopwords	bool or list	The list of stopwords used when training the model. Use False for no stopwords, True for the default stopwords, or a list of strings for custom stopwords.
whitelist	list	The whitelist of words used when training the model.
retry_if_throttled	bool	If a request is throttled, sleep and retry the request.

Example:

response = ml.classifiers.edit('[MODEL_ID]', description='The new description of the classifier')

Delete classifier

def MonkeyLearn.classifiers.delete(model_id, retry_if_throttled=True)

Parameters:

Parameter	Type	Description
model_id	str	Classifier ID. It always starts with 'cl', for example, 'cl_oJNMkt2V'.
retry_if_throttled	bool	If a request is throttled, sleep and retry the request.

Example:

response = ml.classifiers.delete('[MODEL_ID]')

List Classifiers

def MonkeyLearn.classifiers.list(page=1, per_page=20, order_by='-created', retry_if_throttled=True)

Parameters:

Parameter	Type	Description
page	int	Specifies which page to get.
per_page	int	Specifies how many items per page will be returned.
order_by	string or list	Specifies the ordering criteria. It can either be a string for single criteria ordering or a list of strings for more than one. Each string must be a valid field name; if you want inverse/descending order of the field prepend a - (dash) character. Some valid examples are: 'is_public', '-name' or ['-is_public', 'name'].
retry_if_throttled	bool	If a request is throttled, sleep and retry the request.

Example:

response = ml.classifiers.list(page=2, per_page=5, order_by=['-is_public', 'name'])

Deploy

def MonkeyLearn.classifiers.deploy(model_id, retry_if_throttled=True)

Parameters:

Parameter	Type	Description
model_id	str	Classifier ID. It always starts with 'cl', for example, 'cl_oJNMkt2V'.
retry_if_throttled	bool	If a request is throttled, sleep and retry the request.

Example:

response = ml.classifiers.deploy('[MODEL_ID]')

Train

def MonkeyLearn.classifiers.train(model_id, retry_if_throttled=True)

Parameters:

Parameter	Type	Description
model_id	str	Classifier ID. It always starts with 'cl', for example, 'cl_oJNMkt2V'.
retry_if_throttled	bool	If a request is throttled, sleep and retry the request.

Example:

response = ml.classifiers.train('[MODEL_ID]')

Tag detail

def MonkeyLearn.classifiers.tags.detail(model_id, tag_id, retry_if_throttled=True)

Parameters:

Parameter	Type	Description
model_id	str	Classifier ID. It always starts with 'cl', for example, 'cl_oJNMkt2V'.
tag_id	int	Tag ID.
retry_if_throttled	bool	If a request is throttled, sleep and retry the request.

Example:

response = ml.classifiers.tags.detail('[MODEL_ID]', TAG_ID)

Create tag

def MonkeyLearn.classifiers.tags.create(model_id, name, retry_if_throttled=True)

Parameters:

Parameter	Type	Description
model_id	str	Classifier ID. It always starts with 'cl', for example, 'cl_oJNMkt2V'.
name	str	The name of the new tag.
retry_if_throttled	bool	If a request is throttled, sleep and retry the request.

Example:

response = ml.classifiers.tags.create('[MODEL_ID]', 'Positive')

Edit tag

def MonkeyLearn.classifiers.tags.edit(model_id, tag_id, name=None,
                                      retry_if_throttled=True)

Parameters:

Parameter	Type	Description
model_id	str	Classifier ID. It always starts with 'cl', for example, 'cl_oJNMkt2V'.
tag_id	int	Tag ID.
name	str	The new name of the tag.
retry_if_throttled	bool	If a request is throttled, sleep and retry the request.

Example:

response = ml.classifiers.tags.edit('[MODEL_ID]', TAG_ID, 'New name')

Delete tag

def MonkeyLearn.classifiers.tags.delete(model_id, tag_id, move_data_to=None,
                                        retry_if_throttled=True)

Parameters:

Parameter	Type	Description
model_id	str	Classifier ID. It always starts with 'cl', for example, 'cl_oJNMkt2V'.
tag_id	int	Tag ID.
move_data_to	int	An optional tag ID. If provided, training data associated with the tag to be deleted will be moved to the specified tag before deletion.
retry_if_throttled	bool	If a request is throttled, sleep and retry the request.

Example:

response = ml.classifiers.tags.delete('[MODEL_ID]', TAG_ID)

Upload data

def MonkeyLearn.classifiers.upload_data(model_id, data, retry_if_throttled=True)

Parameters:

Parameter	Type	Description
model_id	str	Classifier ID. It always starts with 'cl', for example, 'cl_oJNMkt2V'.
data	list[dict]	A list of dicts with the keys described below.
input_duplicates_strategy	str	Indicates what to do with duplicate texts in this request. Must be one of merge, keep_first or keep_last.
existing_duplicates_strategy	str	Indicates what to do with texts of this request that already exist in the model. Must be one of overwrite or ignore.
retry_if_throttled	bool	If a request is throttled, sleep and retry the request.

data dict keys:

Key	Description
text	A string of the text to upload.
tags	A list of tags that can be refered to by their numeric ID or their name. The text will be tagged with each tag in the list when created (in case it doesn't already exist on the model). Otherwise, its tags will be updated to the new ones. New tags will be created if they don't already exist.
markers	An optional list of string. Each one represents a marker that will be associated with the text. New markers will be created if they don't already exist.

Example:

response = ml.classifiers.upload_data(
    model_id='[MODEL_ID]',
    data=[{'text': 'text 1', 'tags': [TAG_ID_1, '[tag_name]']},
          {'text': 'text 2', 'tags': [TAG_ID_1, TAG_ID_2]}]
)

Extractors

Extract

def MonkeyLearn.extractors.extract(model_id, data, production_model=False, batch_size=200,
                                   retry_if_throttled=True, extra_args=None)

Parameters:

Parameter	Type	Description
model_id	str	Extractor ID. It always starts with 'ex', for example, 'ex_oJNMkt2V'.
data	list[str or dict]	A list of up to 200 data elements to extract from. Each element must be a string with the text or a dict with the required text key and the text as the value. You can also provide an optional external_id key with a string that will be included in the response.
production_model	bool	Indicates if the extractions are performed by the production model. Only use this parameter with custom models (not with the public ones). Note that you first need to deploy your model to production from the UI model settings.
batch_size	int	Max number of texts each request will send to MonkeyLearn. A number from 1 to 200.
retry_if_throttled	bool	If a request is throttled, sleep and retry the request.

Example:

data = ['First text', {'text': 'Second text', 'external_id': '2'}]
response = ml.extractors.extract('[MODEL_ID]', data=data)

Extractor detail

def MonkeyLearn.extractors.detail(model_id, retry_if_throttled=True)

Parameters:

Parameter	Type	Description
model_id	str	Extractor ID. It always starts with 'ex', for example, 'ex_oJNMkt2V'.
retry_if_throttled	bool	If a request is throttled, sleep and retry the request.

Example:

response = ml.extractors.detail('[MODEL_ID]')

List extractors

def MonkeyLearn.extractors.list(page=1, per_page=20, order_by='-created', retry_if_throttled=True)

Parameters:

Parameter	Type	Description
page	int	Specifies which page to get.
per_page	int	Specifies how many items per page will be returned.
order_by	string or list	Specifies the ordering criteria. It can either be a string for single criteria ordering or a list of strings for more than one. Each string must be a valid field name; if you want inverse/descending order of the field prepend a - (dash) character. Some valid examples are: 'is_public', '-name' or ['-is_public', 'name'].
retry_if_throttled	bool	If a request is throttled, sleep and retry the request.

Example:

response = ml.extractors.list(page=2, per_page=5, order_by=['-is_public', 'name'])

Workflows

Workflow detail

def MonkeyLearn.workflows.detail(model_id, step_id, retry_if_throttled=True)

Parameters:

Parameter	Type	Description
model_id	str	Workflow ID. It always starts with 'wf', for example, 'wf_oJNMkt2V'.
step_id	int	Step ID.
retry_if_throttled	bool	If a request is throttled, sleep and retry the request.

Example:

response = ml.workflows.detail('[MODEL_ID]', '[STEP_ID]')

Create workflow

def MonkeyLearn.workflows.create(name, db_name, steps, description='', webhook_url=None,
                                 custom_fields=None, sources=None, retry_if_throttled=True)

Parameters:

Parameter	Type	Description
name	str	The name of the model.
db_name	str	The name of the database where the data will be stored. The name must not already be in use by another database.
steps	list[dict]	A list of step dicts.
description	str	The description of the model.
webhook_url	str	An URL that will be called when an action is triggered.
custom_fields	[]	A list of custom_field dicts that represent user defined fields that come with the input data and that will be saved. It does not include the mandatory text field.
sources	{}	An object that represents the data sources of the workflow.

Example:

response = ml.workflows.create(
    name='Example Workflow',
    db_name='example_workflow',
    steps=[{
        name: 'sentiment',
        model_id: 'cl_pi3C7JiL'
    }, {
        name: 'keywords',
        model_id: 'ex_YCya9nrn'
    }])

Delete workflow

def MonkeyLearn.workflows.delete(model_id, retry_if_throttled=True)

Parameters:

Parameter	Type	Description
model_id	str	Workflow ID. It always starts with 'wf', for example, 'wf_oJNMkt2V'.
retry_if_throttled	bool	If a request is throttled, sleep and retry the request.

Example:

response = ml.workflows.delete('[MODEL_ID]')

Step detail

def MonkeyLearn.workflows.steps.detail(model_id, step_id, retry_if_throttled=True)

Parameters:

Parameter	Type	Description
model_id	str	Workflow ID. It always starts with 'wf', for example, 'wf_oJNMkt2V'.
step_id	int	Step ID.
retry_if_throttled	bool	If a request is throttled, sleep and retry the request.

Example:

response = ml.workflows.steps.detail('[MODEL_ID]', STEP_ID)

Create step

def MonkeyLearn.workflows.steps.create(model_id, name, step_model_id, input=None,
                                         conditions=None, retry_if_throttled=True)

Parameters:

Parameter	Type	Description
model_id	str	Workflow ID. It always starts with 'wf', for example, 'wf_oJNMkt2V'.
name	str	The name of the new step.
step_model_id	str	The ID of the MonkeyLearn model that will run in this step. Must be an existing classifier or extractor.
input	str	Where the input text to use in this step comes from. It can be either the name of a step or input_data (the default), which means that the input will be the original text.
conditions	list[dict]	A list of condition dicts that indicate whether this step should execute or not. All the conditions in the list must be true for the step to execute.
retry_if_throttled	bool	If a request is throttled, sleep and retry the request.

Example:

response = ml.workflows.steps.create(model_id='[MODEL_ID]',  name='sentiment',
                                     step_model_id='cl_pi3C7JiL')

Delete step

def MonkeyLearn.workflows.steps.delete(model_id, step_id, retry_if_throttled=True)

Parameters:

Parameter	Type	Description
model_id	str	Workflow ID. It always starts with 'wf', for example, 'wf_oJNMkt2V'.
step_id	int	Step ID.
retry_if_throttled	bool	If a request is throttled, sleep and retry the request.

Example:

response = ml.workflows.steps.delete('[MODEL_ID]', STEP_ID)

Upload workflow data

def MonkeyLearn.workflows.data.create(model_id, data, retry_if_throttled=True)

Parameters:

Parameter	Type	Description
model_id	str	Workflow ID. It always starts with 'wf', for example, 'wf_oJNMkt2V'.
data	list[dict]	A list of dicts with the keys described below.
retry_if_throttled	bool	If a request is throttled, sleep and retry the request.

data dict keys:

Key	Description
text	A string of the text to upload.
[custom field name]	The value for a custom field for this text. The type of the value must be the one specified when the field was created.

Example:

response = ml.workflows.data.create(
    model_id='[MODEL_ID]',
    data=[{'text': 'text 1', 'rating': 3},
          {'text': 'text 2', 'rating': 4}]
)

List workflow data

def MonkeyLearn.workflows.data.list(model_id, batch_id=None, is_processed=None,
                                    sent_to_process_date_from=None, sent_to_process_date_to=None,
                                    page=None, per_page=None, retry_if_throttled=True)

Parameters:

Parameter	Type	Description
page	int	The page number to be retrieved.
per_page	int	The maximum number of items the page should have. The maximum allowed value is 50.
batch_id	int	The ID of the batch to retrieve. If unspecified, data from all batches is shown.
is_processed	bool	Whether to return data that has been processed or data that has not been processed yet. If unspecified, both are shown indistinctly.
sent_to_process_date_from	str	An ISO formatted date which specifies the oldest sent_date of the data to be retrieved.
sent_to_process_date_to	str	An ISO formatted date which specifies the most recent sent_date of the data to be retrieved.

Example:

response = ml.workflows.data.list('[MODEL_ID]', batch_id=1839, page=1)

Create custom field

def MonkeyLearn.workflows.custom_fields.create(model_id, name, data_type, retry_if_throttled=True)

Parameters:

Parameter	Type	Description
model_id	str	Workflow ID. It always starts with 'wf', for example, 'wf_oJNMkt2V'.
name	str	The name of the new custom field.
data_type	str	The type of the data of the field. It must be one of string, date, text, integer, float, bool.
retry_if_throttled	bool	If a request is throttled, sleep and retry the request.

Example:

response = ml.workflows.custom_fields.create(model_id='[MODEL_ID]',  name='rating',
                                             data_type='integer')