Go to file
Ashik K 134afba3f5 documentation fix 2023-04-10 08:31:53 +02:00
README.md documentation fix 2023-04-10 08:31:53 +02:00
smartsh.py add support for text-davinci-003 2023-04-10 08:20:38 +02:00

README.md

smartsh.py is a simple python script that I wrote which can take a task description (any string) as argument and query OpenAPI's API to tell you how to accomplish the given task using a shell command. Yes, chatgpt from command line (sort of). But let me tell you a simple trick that you can use to supercharge your bash with it!

bash (I think starting from 4.0) conveniently provides a handler for situations when a command that the user entered is invalid.

You just need to provide a function named command_not_found_handle () and point it to the action to be taken on, yes, when a command is not found :-)

In this case, you need to add the following to your .bashrc for the magic to work:

export OPENAI_API_KEY="your_api_key"

export SMARTSH_PATH="path_where_you_checked_out_smartsh"

command_not_found_handle () {
    echo "Let's get help from OpenAI API!"
    python3 $SMARTSH_PATH/smartsh.py "$@"
}

Some examples:

bash$ Show the most recent file in the present working directory
Let's get help from OpenAI API!
ls -t | head -n1
bash$ Kill all Google chrome renderer processes
Let's get help from OpenAI API!
The command to accomplish this task is:

pkill -f "chrome.*renderer"


This command uses the `pkill` command to send a signal to all processes whose name matches the regular expression `chrome.*renderer`. This will effectively kill all Google Chrome renderer processes.

You can set the following additional environment variables which smartsh can make use of:

OPENAI_MODEL_ID: The model to use: Supported models are "text-davinci-003", "gpt-3.5-turbo" If the variable is not set, we use "gpt-3.5-turbo" by default

SMARTSH_DEBUG : Will print additional debug info when running smartsh