pork-chop

An AI chat bot

Conversations

We use ChatterBot, a machine learning conversational dialog engine to power Pork Chop's conversations.

Training

Pork Chop has a few options for training:

• Train with csv data - The success and training time with this is dependent on the size and quality of the data you provide it. For reference, a 70,000 message file we trained pork chop with took around 10 minutes on a AWS EC2 t3.micro VM. See example_training.csv for reference.
• Train with english corpus - Train with the small english corpus provided by ChatterBot
• Train with Ubuntu corpus - This is a huge data set which will take a long time to download and train

Command Modules

Pork chop will respond to the following command modules:

• !example - Template module
• !usage - Display modules and their use
• !kanye - Display a random Kanye West quote.
• !stonks <symbol> - Get stock price information
• !turn <user> [list] - Turn a user or list number of turns for all users
• !helpmecook <food> - Get recipe help for cooking a tasty vegan meal with spoonacular api (try cooking manwich)
• !yt <query> - Get the best-match YouTube video url for your query
• !reddit <query> - Get the best-match subreddit for your query
• !word - Display a randomly generated fake word + definition (from the awesome thisworddoesnotexist.com)

Deployment

Pork Chop can be deployed on any server with python3.

Bot Registration

• Navigate to dev.groupme.com/bots
• Click the orange button labeled "create bot"
• Choose the groupchat you would like pork chop to live in
• For the callback URL, enter the ip or domain of the server Pork Chop will be run on with the extension /bot as this is where the flask handler is set to.
• If you are running pork chop on serverless mode set up the callback URL appropriate to you AWS Lambda function
• You can leave Avatar URL blank or set it to your desired image

Server

• git clone https://github.com/danerwilliams/pork-chop.git Clone pork-chop
• cd pork-chop enter pork-chop directory
• pip3 install -r requirements.txt install pork-chop's python3 dependencies
• Set environment variables for secrets:
  • $EDITOR .env record environment variables to a file with text editor of choice, minimum need to set BOT_ID (use .env.example for reference)
  • export $(grep -v '^#' .env) set environment variables from .env file
• OR use .secrets json formatted file for secrets:
  • $EDITOR .secrets record secrets to file with text editor of choice, minimum need to set BOT_ID (use .secrets.example for reference)
• sudo ./pork-chop.py -d deploy pork chop on port 80 (you can add &> /dev/null & to ignore output and run in background)

Flags

./pork-chop.py -h

usage: pork-chop [-h] [-c CORES] [-t TRAIN [TRAIN ...]] [-d] [-u] [-e]
                 [-n NAME]

optional arguments:
  -h, --help            show this help message and exit
  -c CORES, --cores CORES
                        Number of CPU cores to use
  -t TRAIN [TRAIN ...], --train TRAIN [TRAIN ...]
                        Train bot from csv data files
  -d, --deploy          Deploy bot with flask on port 80 (requires sudo)
  -u, --ubuntu          Train bot from ubuntu corpus
  -e, --english         Train bot from english corpus
  -n NAME, --name NAME  Change Pork Chop's name from default

References

Inspired by ginglis13/shortstop and pbui/bobbit

Useful Documentation:

• ChatterBot
• Flask
• GroupMe

Team

This project is maintained by the following person(s) and a bunch of awesome contributors.

Dane Williams	Conor Murphy	James Heneghan	Ben Green

License

MIT License