Skip to content
forked from ryszard/think

Think: AI-Assisted Command Line Mastery

License

Notifications You must be signed in to change notification settings

FaunoBarberini/think

 
 

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

23 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

🧠💻Think

⚡️Make AI your command line copilot⚡️

Overview

think is a command line tool that incorporates OpenAI's AI models to simplify and streamline your shell operations. No need to remember complex command line details, just describe what you want, and let think generate the shell code for you. Obviously, the heavy lifting here is done by the AI model, but think provides a convenient interface to interact with it.

Features

  • AI-powered command generation: Just describe in your own words what you want to achieve, and think will generate the necessary shell command(s) for you.

  • Iterative refinement: Not satisfied with the command generated? No problem. You can provide additional feedback and guide the AI to produce a more suitable result.

  • Code editing: Before executing, you have the chance to review and edit the proposed command to make sure it's exactly what you want.

Installation

If you are on a Mac, you can install think using Homebrew:

brew install ryszard/ryszard/think

If you have Go installed, you can also just do

go install github.com/ryszard/think/cmd/think

In order to be able to use think, you need to have an OpenAI API key. think will for it in the OPENAI_API_KEY environment variable:

export OPENAI_API_KEY="sk-..."

Usage

Upon running, think opens an interactive session where you can discuss your task with the AI, review and edit the proposed code, and then execute it when ready. It has two modes, think and run. In thinking mode, you can describe your task to the AI, and it will generate the necessary shell code for you. In running mode, you can review and edit the proposed code, and then execute it when ready. In running mode, you can hit C-c or C-d to go back to thinking mode. If the AI gets something wrong, simply tell it to fix it. It helps if --send-output is true (see bellow).

Configuration

You can either pass flags or set environment variables to configure think. The following flags and env variables are available:

  • THINK_SEND_OUTPUT/--send-output (bool): If true, think will send some additional data to the AI: the actual command that you executed (with your edits), and the first 1000 bytes of the stdout and stderr. This is helpful, as it allows the AI to debug its mistakes. However, it also means that if you run a command that contains sensitive data, that data will be sent to the AI. --send-output is false by default.

  • THINK_MODEL/--model (string): Name of the model to use. Defaults to gpt-4.

Examples

Task: Find the 10 largest files in ~/Downloads

Do you always remember the exact syntax for the find command? I don't. But I know what I want to do, so I can just describe it to "Think" and let it generate the command for me.

think> find the 10 largest files in ~/Downloads
Find the 10 largest files in the ~/Downloads directory, sorted by size.
find ~/Downloads -type f -exec du -sh {} \; 2> /dev/null | sort -rh | head -n 10

run> find ~/Downloads -type f -exec du -sh {} \; 2> /dev/null | sort -rh | head -n 10
564M	/Users/ric/Downloads/Anaconda3-2023.03-1-MacOSX-arm64.pkg
564M	/Users/ric/Downloads/Anaconda3-2023.03-1-MacOSX-arm64 (1).pkg
224M	/Users/ric/Downloads/Skype-8.97.0.204.dmg
209M	/Users/ric/Downloads/googlechrome.dmg
191M	/Users/ric/Downloads/VSCode-darwin-universal.zip
160M	/Users/ric/Downloads/Discord.dmg
155M	/Users/ric/Downloads/Obsidian-1.1.16-universal.dmg
116M	/Users/ric/Downloads/google-cloud-cli-430.0.0-darwin-arm.tar.gz
113M	/Users/ric/Downloads/files-WhatsApp.dmg
112M	/Users/ric/Downloads/Messenger.185.0.0.5.228.dmg

Task: Analyze a CSV file

How about some CSV wrangling?

(I was actually not looking forward to preparing example data for this one, but then I remembered that I could just ask the AI to do it for me... 😅)

think> Please generate a simple CSV file with a few columns. At least one column should be numeric.
Creating a simple CSV file with three columns: ID, Name, Age
echo -e "ID,Name,Age\n1,Alice,30\n2,Bob,25\n3,Carol,28" > data.csv

run> echo -e "ID,Name,Age\n1,Alice,30\n2,Bob,25\n3,Carol,28" > data.csv
think> get me the first column from data.csv
Extracting the first column from data.csv
cut -d"," -f1 data.csv
run> cut -d"," -f1 data.csv
ID
1
2
3
think> please sum the third column
Calculating the sum of the third column in data.csv
awk -F"," 'NR>1 {sum += $3} END {print sum}' data.csv

run> awk -F"," 'NR>1 {sum += $3} END {print sum}' data.csv
83

Task: Create a Go project

Now let's do something more fancy:

think> create the directory structure for a go project named thinktest, with a go module and git repo. Add files like README, LICENSE, .gitignore...
Creating Go project directory structure with module, Git repo, README, LICENSE, and .gitignore files.
mkdir -p thinktest/{cmd,pkg} && cd thinktest && go mod init thinktest && git init && touch README.md LICENSE .gitignore

run> mkdir -p thinktest/{cmd,pkg} && cd thinktest && go mod init thinktest && git init && touch README.md LICENSE .gitignore^C
think> the repo should live on github. My username is ryszard. The license is MIT.
Creating Go project with GitHub remote, MIT license, and adding necessary files.
mkdir -p thinktest/{cmd,pkg} && cd thinktest && go mod init thinktest && git init && echo -e "MIT License\n\nCopyright (c) $(date +%Y) $(git config user.name)\n\nPermission is hereby granted, free of charge, to any person obtaining a copy\nof this software and associated documentation files (the \"Software\"), to deal\nin the Software without restriction, including without limitation the rights\nto use, copy, modify, merge, publish, distribute, sublicense, and/or sell\ncopies of the Software, and to permit persons to whom the Software is\nfurnished to do so, subject to the following conditions:\n\nThe above copyright notice and this permission notice shall be included in all\ncopies or substantial portions of the Software.\n\nTHE SOFTWARE IS PROVIDED \"AS IS\", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR\nIMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,\nFITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE\nAUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER\nLIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,\nOUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE\nSOFTWARE." > LICENSE && touch README.md .gitignore && git add . && git commit -m "Initial commit" && git remote add origin [email protected]:ryszard/thinktest.git

run> mkdir -p thinktest/{cmd,pkg} && cd thinktest && go mod init thinktest && git init && echo -e "MIT License\n\nCopyright (c) $(date +%Y) $(git config user.name)\n\nPermission is hereby granted, free of charge, to any person obtaining a copy\nof this software and associated documentation files (the \"Software\"), to deal\nin the Software without restriction, including without limitation the rights\nto use, copy, modify, merge, publish, distribute, sublicense, and/or sell\ncopies of the Software, and to permit persons to whom the Software is\nfurnished to do so, subject to the following conditions:\n\nThe above copyright notice and this permission notice shall be included in all\ncopies or substantial portions of the Software.\n\nTHE SOFTWARE IS PROVIDED \"AS IS\", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR\nIMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,\nFITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE\nAUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER\nLIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,\nOUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE\nSOFTWARE." > LICENSE && touch README.md .gitignore && git add . && git commit -m "Initial commit" && git remote add origin [email protected]:ryszard/thinktest.git
go: creating new go.mod: module thinktest
go: to add module requirements and sums:
	go mod tidy
Initialized empty Git repository in /Users/ric/thinktest/.git/
[main (root-commit) 085e2a9] Initial commit
 4 files changed, 24 insertions(+)
 create mode 100644 .gitignore
 create mode 100644 LICENSE
 create mode 100644 README.md
 create mode 100644 go.mod

Note that it got the LICENSE text right (I diffed it), which is pretty cool. In earlier versions it was trying to download the license from the Internet, but that was obviously problematic for an AI without access to the Internet and a knowledge cutoff. I reminded it in the prompt about its limitations.

think> find all the image files in ~/Downloads. copy them to ./images (create it). Anonymize the names so that I don't leak information (this is for a demo)
Finding image files in ~/Downloads, creating ./images directory, copying them with anonymized names to the directory
mkdir -p ./images && find ~/Downloads -type f \( -iname "*.jpg" -o -iname "*.jpeg" -o -iname "*.png" \) -exec sh -c 'cp -t ./images -- "$0" "$(xxd -l 8 -ps -c 8 random)"' '{}' \;

run> mkdir -p ./images && find ~/Downloads -type f \( -iname "*.jpg" -o -iname "*.jpeg" -o -iname "*.png" \) -exec sh -c 'cp -t ./images -- "$0" "$(xxd -l 8 -ps -c 8 random)"' '{}' \;
xxd: random: No such file or directory
cp: illegal option -- t
usage: cp [-R [-H | -L | -P]] [-fi | -n] [-aclpsvXx] source_file target_file
       cp [-R [-H | -L | -P]] [-fi | -n] [-aclpsvXx] source_file ... target_directory
xxd: random: No such file or directory

   (plenty more errors here)

cp: illegal option -- t
usage: cp [-R [-H | -L | -P]] [-fi | -n] [-aclpsvXx] source_file target_file
       cp [-R [-H | -L | -P]] [-fi | -n] [-aclpsvXx] source_file ... target_directory

So, think quite often gets something wrong (especially that GPT-4 tends to assume it's using the GNU versions of various command line tools). However, that's not a big deal, as you can just tell it to fix it.

think> fix it
Finding image files in ~/Downloads, creating ./images directory, copying them with anonymized names to the directory
mkdir -p ./images && find ~/Downloads -type f \( -iname "*.jpg" -o -iname "*.jpeg" -o -iname "*.png" \) -exec sh -c 'cp "$0" "./images/$(xxd -l 8 -p -u /dev/urandom).${0##*.}"' '{}' \;

run> mkdir -p ./images && find ~/Downloads -type f \( -iname "*.jpg" -o -iname "*.jpeg" -o -iname "*.png" \) -exec sh -c 'cp "$0" "./images/$(xxd -l 8 -p -u /dev/urandom).${0##*.}"' '{}' \;
think> now please give me the average dimensions (height and width) of the files in images
Calculating the average dimensions (width and height) of image files in the ./images directory using sips command
find ./images -type f \( -iname "*.jpg" -o -iname "*.jpeg" -o -iname "*.png" \) -exec sips -g pixelWidth -g pixelHeight '{}' \; | awk '/Width/{w+=$NF;wc++} /Height/{h+=$NF;hc++} END {print "Width: " w/wc ", Height: " h/hc}'

run> find ./images -type f \( -iname "*.jpg" -o -iname "*.jpeg" -o -iname "*.png" \) -exec sips -g pixelWidth -g pixelHeight '{}' \; | awk '/Width/{w+=$NF;wc++} /Height/{h+=$NF;hc++} END {print "Width: " w/wc ", Height: " h/hc}'
Width: 2644.09, Height: 1414.73
think> now please resize all of them so that they are 1/4 of the size of the average dimensions.
Resizing all image files in ./images to 1/4 of the average dimensions (width and height) calculated previously
find ./images -type f \( -iname "*.jpg" -o -iname "*.jpeg" -o -iname "*.png" \) -exec sips -Z $((2644/4))x$((1415/4)) '{}' \;

run> find ./images -type f \( -iname "*.jpg" -o -iname "*.jpeg" -o -iname "*.png" \) -exec sips -Z $((2644/4))x$((1415/4)) '{}' \;
images/690EC1F13FB62A19.jpeg
  ./images/690EC1F13FB62A19.jpeg
./images/E0E26311A30C6376.png
  ./images/E0E26311A30C6376.png
./images/BCB68ED864C9F4F2.jpg
  ./images/BCB68ED864C9F4F2.jpg
./images/D87454DB4AA998EC.jpg
  ./images/D87454DB4AA998EC.jpg
./images/CCDE085970025EA2.png
  ./images/CCDE085970025EA2.png
./images/26DD89AF3D7B67EE.png
  ./images/26DD89AF3D7B67EE.png
./images/8A2406199B92A538.png
  ./images/8A2406199B92A538.png
./images/D705AA10B2A43BB9.jpg
  ./images/D705AA10B2A43BB9.jpg
./images/B18EB23B54B98725.png
  ./images/B18EB23B54B98725.png
./images/26D909E004CAC1C1.png
  ./images/26D909E004CAC1C1.png
./images/B7590602F4CEFE9B.jpg
  ./images/B7590602F4CEFE9B.jpg
think> show me their dimensions
Showing the dimensions of the resized image files in the ./images directory using sips command
find ./images -type f \( -iname "*.jpg" -o -iname "*.jpeg" -o -iname "*.png" \) -exec sips -g pixelWidth -g pixelHeight '{}' \;

run> find ./images -type f \( -iname "*.jpg" -o -iname "*.jpeg" -o -iname "*.png" \) -exec sips -g pixelWidth -g pixelHeight '{}' \;
./images/690EC1F13FB62A19.jpeg
  pixelWidth: 484
  pixelHeight: 661
./images/E0E26311A30C6376.png
  pixelWidth: 661
  pixelHeight: 452
./images/BCB68ED864C9F4F2.jpg
  pixelWidth: 440
  pixelHeight: 661
./images/D87454DB4AA998EC.jpg
  pixelWidth: 661
  pixelHeight: 459
./images/CCDE085970025EA2.png
  pixelWidth: 661
  pixelHeight: 661
./images/26DD89AF3D7B67EE.png
  pixelWidth: 661
  pixelHeight: 184
./images/8A2406199B92A538.png
  pixelWidth: 661
  pixelHeight: 661
./images/D705AA10B2A43BB9.jpg
  pixelWidth: 661
  pixelHeight: 372
./images/B18EB23B54B98725.png
  pixelWidth: 661
  pixelHeight: 661
./images/26D909E004CAC1C1.png
  pixelWidth: 661
  pixelHeight: 184
./images/B7590602F4CEFE9B.jpg
  pixelWidth: 495
  pixelHeight: 661

Limitations

This uses AI, which is known to make mistakes. You should double check if the tools isn't proposing to do something stupid. Do not run commands blindly. think is not a replacement for your brain, it's a tool to help you be more productive.

Please note that think will send your shell and OS to the AI. With the --send-output flag, think this will send some additional data through the API: the actual command that you executed (with your edits), and the first 1000 bytes of the stdout and stderr. This is helpful, as it allows the AI to debug its mistakes. However, it also means that if you run a command that contains sensitive data, that data will be sent to the AI. --send-output is false by default, but please be mindful of this.

You should never type any passwords into think (or into your shell, for that matter). Typing passwords directly in the command line poses security risks. The shell's command history can inadvertently store these passwords in plain text. Additionally, they may be visible in the process list to others on the system. If commands are sent over a network, passwords could be exposed during transmission. Hence, it's advised to use safer alternatives like password prompts or secure password files.

Contributing

Just send me a pull request.

Status

Experimental, use at your own risk.

Acknowledgements

I am grateful to OpenAI for making their models available to the public ❤️

License

MIT. Please see the LICENSE file for details.

About

Think: AI-Assisted Command Line Mastery

Resources

License

Stars

Watchers

Forks

Packages

No packages published

Languages

  • Go 100.0%