emma

WIP !!!

🎅 Intelligent mention bot for GitHub organizations

Install

$ pip3 install -r requirements.txt

Deploy

$ sh deploy.sh [provider]

Developement setup

Setup ngrok

Download and install ngrok

Run ngrok

$ ./ngork http [port]

Setup GitHub webhook

Goto Repository -> Settings -> Webhooks -> Add webhook

Add Payload URL from ngrok public URL. Click on Add webhook

Add configuration file

[Credentials]
username = *********
password = *********

[Repository]
org = *********
repo = *********

Run emma

$ bash ./init.sh

Enpoints

/payload

Programmatic API

from emma import Parser

# create parser object
parser = Parser()

# parse local diff file - returns a generator
parsed_diff = parser.parse_diff('/home/rajika/projects/sublime-vmd', 'master')
print(parsed_diff)

# parse raw diff to get following dict
#  {
#      'file_names': file_names,
#      'deleted_lines': deleted_lines,
#      'added_lines': added_lines
#  }

parsed_raw_diff = parser.parse_raw_diff(raw_diff)

from emma import Parser

# create parser object
parser = Parser()

# parse diff from GitHub diff
parsed_diff = parser.get_pr_diff('https://patch-diff.githubusercontent.com/raw/facebook/react/pull/3.diff')
print(parsed_diff)

Algorithm

Emma uses supervised learning in machine learning to learn about a given repository. With the increase of data set, emma becomes more intelligent. Therefore mature repositories can gain better results.

Dataset generation

Emma generates a data set for each and every repo in order to train the machine learning model. Dataset is structured as follows.

• Filename, Timestamp, Commit author, Previous author

eg - bin,321,2017-04-18T19:47:30+0518,inbox.rajika@gmail.com,inbox.rajika@gmail.com

Heuristics

Following are the heuristics used by emma to predict the best possible reviewer for a pull request.

• Deleted lines
• Added lines
• Modified lines