Third Party Features

Clarifai ImageAPI

Note: Head over to https://developer.clarifai.com/signup/ and create a free account, once you're logged in go to https://developer.clarifai.com/account/applications/ and create a new application. You can find the client ID and Secret there. You get 5000 API-calls free/month.

If you want the script to get your CLARIFAI_API_KEY for your environment, you can do:

export CLARIFAI_API_KEY="<API KEY>"

Example with Imagecontent handling

session.set_do_comment(True, percentage=10)
session.set_comments(['Cool!', 'Awesome!', 'Nice!'])
session.set_use_clarifai(enabled=True)
session.clarifai_check_img_for(['nsfw'])
session.clarifai_check_img_for(['food', 'lunch', 'dinner'], comment=True, comments=['Tasty!', 'Nice!', 'Yum!'])
session.end()

Enabling Imagechecking

# default enabled=False , enables the checking with the Clarifai API (image
# tagging) if secret and proj_id are not set, it will get the environment
# variables 'CLARIFAI_API_KEY'.
session.set_use_clarifai(enabled=True, api_key='xxx')

Using Clarifai Public Models and Custom Models

If not specified by setting the models=['model_name1'] in session.set_use_clarifai, models will be set to general by default.

If you wish to check against a specific model or multiple models (see Support for Compound Model Queries below), you can specify the models to be checked as shown below.

To get a better understanding of the models and their associated concepts, see the Clarifai Model Gallery and Developer Guide

NOTE ON MODEL SUPPORT: At this time, the support for theFocus, Face Detection, Face Embedding, and General Embedding has not been added.

# Check image using the NSFW model
session.set_use_clarifai(enabled=True, api_key='xxx', models=['nsfw'])
# Check image using the Apparel model
session.set_use_clarifai(enabled=True, api_key='xxx', models=['apparel'])
# Check image using the Celebrity model
session.set_use_clarifai(enabled=True, api_key='xxx', models=['celebrity'])
# Check image using the Color model
session.set_use_clarifai(enabled=True, api_key=‘xxx’, models=[‘model’])
# Check image using the Demographics model
session.set_use_clarifai(enabled=True, api_key=‘xxx’, models=[‘demographics’])
# Check image using the Food model
session.set_use_clarifai(enabled=True, api_key='xxx', models=['food'])
# Check image using the Landscape Quality model
session.set_use_clarifai(enabled=True, api_key='xxx', models=['landscape quality'])
# Check image using the Logo model
session.set_use_clarifai(enabled=True, api_key='xxx', models=['logo'])
# Check image using the Moderation model
session.set_use_clarifai(enabled=True, api_key='xxx', models=['moderation'])
# Check image using the Portrait Quality model
session.set_use_clarifai(enabled=True, api_key='xxx', models=['portrait quality'])
# Check image using the Textures and Patterns model
session.set_use_clarifai(enabled=True, api_key='xxx', models=['textures'])
# Check image using the Travel model
session.set_use_clarifai(enabled=True, api_key='xxx', models=['travel'])
# Chaeck image using the Weddings model
session.set_use_clarifai(enabled=True, api_key='xxx', models=['weddings'])
# Check image using a custom model where model_name is name of your choosing (see Clarifai documentation for using custom models)
session.set_use_clarifai(enabled=True, api_key='xxx', models=['your-model-name'])

Filtering Inappropriate Images

# uses the clarifai api to check if the image contains nsfw content
# by checking against Clarifai's NSFW model
# -> won't comment if image is nsfw
session.set_use_clarifai(enabled=True, api_key='xxx', models=['nsfw'])
session.clarifai_check_img_for(['nsfw'])
# uses the clarifai api to check if the image contains inappropriate content
# by checking against Clarifai's Moderation model
# -> won't comment if image is suggestive or explicit
session.set_use_clarifai(enabled=True, api_key='xxx', models=['moderation'])
session.clarifai_check_img_for(['suggestive', 'explicit'])
# To adjust the threshold for accepted concept predictions and their
# respective score (degree of confidence) you can set the default probability
# parameter for Clarifai (default 50%). For example, you could set probability to 15%.
# -> any image with a nsfw score of 0.15 of higher will not be commented on
session.set_use_clarifai(enabled=True, api_key='xxx', probability= 0.15, models=['nsfw'])
session.clarifai_check_img_for(['nsfw'])

Filtering by Keyword

# uses the clarifai api to check if the image concepts contain the keyword(s)
# -> won't comment if image contains the keyword
session.clarifai_check_img_for(['building'])

Specialized Comments for Images with Specific Content

# checks the image for keywords food and lunch. To check for both, set full_match in
# in session.set_use_clarifia to True, and if both keywords are found,
# InstaPy will comment with the given comments. If full_match is False (default), it only
# requires a single tag to match Clarifai results.
session.set_use_clarifai(enabled=True, api_key='xxx', full_match=True)
session.clarifai_check_img_for(['food', 'lunch'], comment=True, comments=['Tasty!', 'Yum!'])
# If you only want to accept results with a high degree of confidence, you could
# set a probability to a higher value, like 90%.
session.set_use_clarifai(enabled=True, api_key='xxx', probability=0.90, full_match=True)
session.clarifai_check_img_for(['food', 'lunch'], comment=True, comments=['Tasty!', 'Yum!'])

Querying Multiple Models with Workflow (Single API Call)

You can query multiple Clarifai models with a single API call by setting up a custom workflow. Using a workflow is the recommended way to query multiple models. Alternatively, it is possible to query multiple models separately (see Querying Multiple Models (Multiple API Calls) below).

To setup a workflow, see the Workflow Documentation.

NOTE :As mentioned above, the Focus, Face Detection, Face Embedding, and General Embedding models are not current supported.

Once you have a workflow setup, you can use InstaPy to check images with the Clarifai Image API by setting the workflow parameter in session.set_use_clarifai to the name of your custom workflow.

Let's say you want to comment 'Great shot!' on images of men or women with the hashtag #selfie, but you want to make sure not to comment on images which might contain inappropriate content. To get general concepts, e.g. woman, you would setup your workflow using General and to check the image for the concepts nsfw and explicit you would also want to add NSFW and Moderation models to your workflow.

For example:

session.set_use_clarifai(enabled=True, api_key='xxx', workflow=['your-workflow'], proxy='123.123.123.123:5555')
session.clarifai_check_img_for(['woman', 'man'], ['nsfw', 'explicit', 'suggestive'], comment=True, comments=['Great shot!'])

If Clarifai's response includes the concepts of either woman or man but also includes at least nsfw, explicit, or suggestive, InstaPy will not comment. On the other hand, if Clarifai's response includes the concepts of either woman or man but does not include any of the concepts nsfw, explicit, or suggestive, InstaPy will add the comment Great shot!

Querying Multiple Models (Multiple API Calls)

In the event that you do not want to set up a workflow, you can also query multiple models using multiple API calls.

WARNING: If you are using a free account with Clarifiai, be mindful that the using compound API queries could greatly increase your chances of exceeding your allotment of free 5000 operations per month. The number of Clarifai billable operations per image check equals the number of models selected. For example, if you check 100 images against models=['general', 'nsfw', 'moderation'], the total number of billable operations will be 300.

Following the example above, to get general concepts, e.g. woman, you would use the model general and to check the image for the concepts nsfw and explicit you would also want to check the image against the NSFW and Moderation models.

For example:

session.set_use_clarifai(enabled=True, api_key='xxx', models=['general', 'nsfw', 'moderation'], proxy=None)
session.clarifai_check_img_for(['woman', 'man'], ['nsfw', 'explicit', 'suggestive'], comment=True, comments=['Great shot!'])

Using proxy to access clarifai: We have 3 options:

  1. ip:port
  2. user:pass@ip:port
  3. None

Checking Video

WARNING: Clarifai checks one frame of video for content for every second of video. That is, in a 60 second video, 60 billable operations would be run for every model that the video is being checked against. Running checks on video should only be used if you have special needs and are prepared to use a large number of billable operations.

To have Clarifai run a predict on video posts, you can set the check_video argument in session.set_use_clarifai to True. By default, this argument is set to False. Even if you do not choose to check the entire video, Clarifai will still check the video's keyframe for content.

For example:

session.set_use_clarifai(enabled=True, api_key='xxx', check_video=True)

With video inputs, Clarifai's Predict API response will return a list of concepts at a rate of one frame for every second of a video.

Be aware that you cannot check video using a workflow and that only a select number of public models are currently supported. Models currently supported are: Apparel, Food, General, NSFW, Travel, and Wedding. In the event that the models being used do not support video inputs or you are using a workflow, the video's keyframe will still be checked for content.

Check out https://clarifai.com/demo to see some of the available tags.

Text Analytics

Yandex Translate API

Offers excellent language detection and synchronized translation for over 95 languages 😎 worldwide

This service currently is supported only by the Interact by Comments feature.

Usage

Go sign up on translate.yandex.com and get a free API_key; Then configure its usage at your quickstart script,

session.set_use_yandex(enabled=True,
API_key='',
match_language=True,
language_code="en")
Parameters

enabled : Put True to activate or False to deactivate the service usage;

API_key : The key which is required to authenticate HTTP requests to the API;

match_language : Enable if you would like to match the language of the text;

language_code : Set your desired language's code to match language (if it's enabled);

You can get the list of all supported languages and their codes at tech.yandex.com.

Rate Limits

In its free plan, the daily request limit is 1,000,000 characters and the monthly limit is 10,000,000 characters.

To increase the request limit, you can switch to the fee-based version of the service ($15/million chars)..

Examples

1-) Matching language;

session.set_use_yandex(enabled=True, API_key='', match_language=True, language_code="az")

Target text : "your technique encourages📸 me"

Now that text is gonna be labeled inappropriate COS its language is english rather than the desired azerbaijani..

2-) Enabling the Yandex service but NOT matching language; Since Yandex Translate is being used [internally] by the MeaningCloud service, you can just provide the API key of Yandex and enable it without enabling the match_language parameter what will be sufficient for the MeaningCloud to work..

session.set_use_yandex(enabled=True, API_key='', match_language=False)

And yes, you can enable Yandex service to make it be available for MeaningCloud and then also match language if you like, in the same setup just by turning the match_language parameter on..

Legal Notice

Powered by Yandex.Translate

MeaningCloud Sentiment Analysis API

Offers a detailed, multilingual analysis of all kind of unstructured content determining its sentiment ⚖

This service currently is supported only by the Interact by Comments feature.

Determines if text displays positive, negative, or neutral sentiment - or is not possible to detect. Phrases are identified with the relationship between them evaluated which identifies a global polarity value of the text.

Usage

1-) Go sign up (offers sign in with 😎 Github) on meaningcloud.com and get a free license_key; Then configure its usage at your quickstart script,

session.set_use_meaningcloud(enabled=True,
license_key='',
polarity="P",
agreement="AGREEMENT",
subjectivity="SUBJECTIVE",
confidence=94)

2-) Install its package for python by pip;

pip install MeaningCloud-python

3-) Turn on Yandex Translate service which is a requirement for the language detection & translation at request; To have it configured, read its documentation.

Parameters

enabled : Put True to activate or False to deactivate the service usage;

license_key : The license key is required to do calls to the API;

polarity : It indicates the polarity found (or not found) in the text and applies to the global polarity of the text; It's a graduated polarity - rates from very negative to very positive.

score_tagdefinition
"P+"match if text is strong positive
"P"match if text is positive or above
"NEU"match if text is neutral or above
"N"match if text is negative or above
"N+"match if text is strong negative or above
Nonedo not match per polarity found, at all

By "or above" it means- e.g., if you set polarity to "P", and text is "P+" then it'll also be appropriate (as it always leans towards positivity) ..

agreement : Identifies opposing opinions - contradictory, ambiguous; It marks the agreement between the sentiments detected in the text, the sentence or the segment it refers to.

agreementdefinition
"AGREEMENT"match if the different elements have the same polarity
"DISAGREEMENT"match if there is disagreement between the different elements' polarity
Nonedo not match per agreement found, at all

subjectivity : Identification of opinions and facts - distinguishes between objective and subjective; It marks the subjectivity of the text.

subjectivitydefinition
"SUBJECTIVE"match if text that has subjective marks
"OBJECTIVE"match if text that does not have any subjectivity marks
Nonedo not match per subjectivity found, at all

confidence : It represents the confidence associated with the sentiment analysis performed on the text and takes an integer number in the range of (0, 100];

If you don't want to match per confidence found, at all, use the value of None.

Rate Limits

It gives you 20 000 single API calls per each month (starting from the date you have signed up). It has no daily limit but if you hit the limit set for number of requests can be carried out concurrently (per second) it'll return with error code of 104 rather than the result 😉

Language Support

MeaningCloud currently supports a generic sentiment model (called general) in these languages: english, spanish, french, italian, catalan, and portuguese.

You can define your own sentiment models using the user sentiment models console and work with them in the same way as with the sentiment models it provides.

But no need to worry IF your language or target audience's language is NONE of those officially supported. Cos, to increase the coverage and support all other languages, as well, Yandex Translate service comes to rescue! It detects the text's langugage before passing it to MeaningCloud, and, if its language is not supported by MeaningCloud, it translates it into english and only then passes it to MeaningCloud Sentiment Analysis..

Examples

a -) Match ONLY per polarity and agreement

session.set_use_meaningcloud(enabled=True, license_key='', polarity="P", agreement="AGREEMENT")

Target text : "I appreciate your innovative thinking that results, brilliant images"

Sentiment Analysis results for the text:

score_tagagreementsubjectivityconfidence
"P+""AGREEMENT""SUBJECTIVE"100

Now that text is gonna be labeled appropriate COS its polarity is "P+" which is more positive than "P" and agreement values also do match..

b -) Match FULLY

session.set_use_meaningcloud(enabled=True, license_key='', polarity="P+", agreement="AGREEMENT", subjectivity="SUBJECTIVE", confidence=98)

Target text : "truly fantastic but it looks sad!"

Sentiment Analysis results for the text:

score_tagagreementsubjectivityconfidence
"P""DISAGREEMENT""SUBJECTIVE"92

Now that text is gonna be labeled inappropriate COS its polarity is "P" which is less positive than "P+" and also, agreement values also do NOT match, and lastly, confidence is below user-defined 98..

Legal Notice

This project uses MeaningCloud™ (http://www.meaningcloud.com) for Text Analytics.


Telegram Integration

This feature allows to connect your InstaPy session with a Telegram bot and send commands to the InstaPy session

Prerequisites

You will need to create a token, for this go into your Telegram App and talk with @fatherbot. You will also need to set your username as it is checked to ensure that you are authorized to access the InstaPy session, to do so go to Settings -> Profile -> Username.

Supported actions

There are 3 supported actions:

  • /start : will start the interaction between the bot and instapy. Please note: that the telegram bot cannot send you messages until you first send it a /start message. The bot will store the chat_id in the logs folder file telegram_chat_id.txt to be reused in further sessions (so you have to actually do /start just one time)
  • /report : will gather and show the current session statistics
  • /stop: will set the aborting flag to True

Examples

from instapy.plugins import InstaPyTelegramBot
session = InstaPy(username=insta_username,
password=insta_password,
bypass_with_mobile=True)
telegram = InstaPyTelegramBot(token='insert_real_token_here', telegram_username='my_username', instapy_session=session)
# if you want to receive the information when the session ends
# just add the following before your session.end()
telegram.end()
session.end()

Additional parameters:

  • debug=True if you want low level telegram debug information
  • proxy if you need one, here is the structure that needs to be passed
example_proxy = {
'proxy_url': 'http://PROXY_HOST:PROXY_PORT/',
# Optional, if you need authentication:
'username': 'PROXY_USER',
'password': 'PROXY_PASS',
}
telegram = InstaPytelegramBot(... , proxy=example_proxy)

Additional functionality

you can use

telegram.send_message(text="this is a message")

So you are able to send additional message inside your script if needed. Remember that the telegram bot
is not able to send messages as long as you haven't done at least one /start.