Adding a New Service

This page offers an overview about the (detailed) steps of how to integrate a service into the SIC Framework. Become familiar with what a service could (not) be by checking out https://socialrobotics.atlassian.net/wiki/spaces/CBSR/pages/2070183939 . The code for all the existing services can be found at https://bitbucket.org/socialroboticshub/processing/src/master/.

Summary How to Add a Service to the SIC Framework

create a new folder in https://bitbucket.org/socialroboticshub/processing/src/master/ with the name of the service
copy the certificate file from any of the other services' folders into the service’s folder
copy any additional files that the services may need into the service’s folder
create a factory file inheriting from the SICfactory in the service’s folder, and override the superclass’s methods
create a service file inheriting from SICservice in the service’s folder, and override the superclass’s methods
update the file in the root folder with the new service files
deploy the new service to the folder by running the deploy_to_docker.sh file
update the file in the docker folder with the new service
update the file in the docker folder with the new service’s dependencies
update the topics in the constructor of the Abstract Connector from the folder with the name of the new service
update the device listeners in enable_service in the Abstract Connector with the service
update the listened to channels in __listen in the Abstract Connector with the service
create the corresponding event handler method for the service in the Abstract Connector
create the corresponding event handler method for the service in the Basic Connector
use the service in a new file

The detailed explanation of these steps with a sentiment analysis example can be found below:

Detailed How to Add a (Sentiment) Service to the SIC Framework

There are two shared libraries that handle a lot of the common logic that is needed to interact with our framework: one for Python-based integrations and one for Java-based integrations (using Maven). In order to allow users to run services without worrying about compatibility and installations, Docker Compose is used.

If a service is not simply an alternative to an existing service, adding a new service will also require updates to the connectors (EIS and Python) in order to be fully integrated.

create a new folder in with the name of the service. sentiment_analysis ( ) will be used as the example folder and service in this case
copy the certificate file from any of the other services' folders into the sentimentfolder
copy the into the sentiment folder
create a file in the sentiment folder

from os import getcwd

from sic.factory import SICfactory
from nltk import download
from nltk.data import path

from sentiment_service import SentimentAnalysisService


class SentimentAnalysisFactory(SICfactory):
    def __init__(self):
        super(SentimentAnalysisFactory, self).__init__()

    def get_connection_channel(self):
        return 'sentiment_analysis'

    def create_service(self, connect, identifier, disconnect):
        return SentimentAnalysisService(connect, identifier, disconnect)


if __name__ == '__main__':
    cwd = getcwd()
    download('punkt', download_dir=cwd)
    download('omw-1.4', download_dir=cwd)
    download('averaged_perceptron_tagger', download_dir=cwd)
    download('wordnet', download_dir=cwd)
    path.append(cwd)

    sentiment_analysis_factory = SentimentAnalysisFactory()
    sentiment_analysis_factory.run()

5. create a file in the sentiment folder

""" This file shows an example of a sentiment service that uses the text_transcript channel result
    from the Dialogflow speech-to-text to get the the type of sentiment
"""

from pickle import load
from re import sub
from string import punctuation

from sic.service import SICservice
from nltk.stem.wordnet import WordNetLemmatizer
from nltk.tag import pos_tag
from nltk.tokenize import word_tokenize


class SentimentAnalysisService(SICservice):
    def __init__(self, connect, identifier, disconnect):
        super(SentimentAnalysisService, self).__init__(connect, identifier, disconnect)
        with open('classifier.pickle', 'rb') as pickle:
            self.classifier = load(pickle)
        self.lemmatizer = WordNetLemmatizer()

    def get_device_types(self):
        return ['mic']

    def get_channel_action_mapping(self):
        return {self.get_full_channel('text_transcript'): self.execute}

    def execute(self, message):
        sentence = message['data'].decode()
        tokens = self.remove_noise(word_tokenize(sentence))
        sentiment = self.classifier.classify(dict([token, True] for token in tokens))
        print(sentiment)
        self.publish('text_sentiment', sentiment)
...

6. update the file in the root folder with the new service files

...
echo Deploying sentiment_analysis...
cd ../audio_sentiment
cp -f {cert.pem,classifier.pickle,*.py} ../../docker/sic/sentiment    # extra files should also be inclduded here
...

7. deploy the new service to the folder by running the deploy_to_docker.sh file

8. update the file in the docker folder with the new service

9. update the file in the docker folder with the new service’s dependencies

10. update the topics in the constructor of the Abstract Connector abstract_connector.py from the folder with the name of the new service

topics = [..., 'text_sentiment']

11. update the corresponding list of devices listeners in enable_service in abstract_connector.py with my_service

12. update the channels in __listen in abstract_connector.py with my_service

13. create the corresponding event handler method on_my_service in abstract_connector.py

14. create the corresponding inherited event handler method on_my_service in basic_connector.py

15. use the new service by creating a new file sentiment_example.py, overriding the on_text_sentiment method and running it