Often when using the terminal, it can be frustrating trying to remember the syntax for commands or having to rewrite similar commands over and over again. Thus, our team created a tool that predicts bash command input based off what a user has inputted so far. For example, if the token "add" usually follows the token "git", and so far the user has inputted just "git", our tool will suggest the next token as "add".

What it does

We use a ngram model to predict the most likely token the user will input next in the bash command. Based off

How I built it

We scraped bash history files from github to create a corpus of bash commands. We then created an ngram model to calculate the most likely set of strings to follow a set of n = 4 inputted strings to suggest to the user writing in commands.

What's next for bash-predict

Improving prediction model.

Share this project: