FaceWork Microwork Chatbot

This is the Facebook Messenger chatbot for microwork made by group 6 for the course IN4325 Information Retrieval.

Using the chatbot

Go to the Facebook page and click 'Message' to chat with our bot. You need to be added as developer or tester of the Facebook app, otherwise it will give you the silent treatment. If you post your Facebook account name on our slack channel, we will add you as soon as possible.

Project structure

The project directories and files are structured in the following tree:

• setup.py runs worker API, requester API and chatbot
• api
  • apiserver.py webserver that contains logic for the requester and worker API
  • database.py initializes database configuration
  • model.py datamodel used by both API's
  • tests integration tests for both API's
• chatbot
  • api_helper.py helper methods to call the worker API
  • chatbot.py entry point of chatbot, uses messaging and logger functionality
  • facebook_helper.py helper methods to call Facebook API and Facebook-specific functionality
  • local_test.py tests communication with chatbot without having to use the Facebook interface
  • logger.py wraps (heroku) logging functionality
  • messaging.py logic to handle the chatbot interaction flow by reading user messages and replying according to current chatbot state
• pipelines
  • pipeline_1.py pipeline for Albert Heijn Twitter WebCare evaluation task (Type 1: global content)
  • pipeline_2.py pipeline for Albert Heijn floor plan evaluation task (Type 2: local content)
  • pipeline_3.py pipeline for Albert Heijn store stocking evaluation task (Type 3: content generation)
  • reviewPipeline.py pipeline for generation of review tasks

Project set up

The following paragraph describe how to set up the project to run and test it locally.

First run

To run the chatbot (on Windows) execute the following steps:

• Make sure you have installed Python 3 (I used 3.6) and Heroku CLI. I also assume you have Git installed and know how it works.
• pip install virtualenvwrapper-win to install the virtualenv stuff we are going to use. For Unix you need to install virtualenvwrapper
• mkvirtualenv chatbot to enter our virtual environment. It doesn't matter how you call it, but it's useful to remember how you call it.
• pip install -r requirements.txt to locally install project dependencies in the virtual environment

If you run a Unix machine I assume you know how to adapt the above instructions to your own machine.

Useful for development

• heroku git:remote -a fathomless-cove-38602 adds the Heroku app to the local git to push the code to. You need to be added to the Heroku app for this to work.
• Execute workon chatbot to re-enter the virtual environment if you come back to develop.
• Execute git push heroku master to deploy the chatbot to Heroku. In order to push a branch different than master to Heroku, execute git push heroku your-branch-name:master.
• You can run the chatbot locally with heroku local. This will allow you to run the application locally, with the heroku database. If you have MySQL installed you can also run the app with your local database. In order to do this you need to enter your database settings in api/config/settings.json, and run the server app with python ./setup.py.

Testing locally

• Once you have the app running locally you can run the API tests using python -m unittest discover.

To test the communication between the chatbot and facebook do the following:

• Add a new Environment variable to your machine, with name PAGE_ACCESS_TOKEN and as value the token which can be found in Heroku at Settings -> Config vars
• Run the server locally as described above using the heroku local command, this will start the server at localhost:5000
• Start the chatbot/local_test.py script and follow the instructions