autocomplete adds intelligent command-line suggestions for you directly in the terminal. Just type and it calls an LLM (OpenAI by default) and returns the top suggestions for you.
wget -qO- https://autocomplete.sh/install.sh | bashIt's faster than copy-pasting from Stack Overflow and ChatGPT.
The suggestions are more accurate since we've engineered the prompts to contain limited information of your terminal's state including:
- What kind of machine you are using:
$USER, $PWD, $OLDPWD, $HOME, $OSTYPE, $BASH, $TERM, $HOSTNAME env- Which variables are defined (but just the names and not the values)history- Recently executed commandsls- Recently modified files in the current directory--help- any additional help information for the current command
If you are curious, you can see the full prompt using
autocomplete command --dry-run "anything you want here"
Pull Requests are welcome if you want to make it better!
By default we cache the last 20 requests to reduce latency and costs.
Although writing 1,049 lines of bash was its own "reward", if you tried it and love it please show your support here!
autocomplete config< TODO INSERT PICTURE OF DEFAULT CONFIG>
Configurations can be changed with
autocomplete config set <key> <value>For example autocomplete config set api_key sk-p... will update your API key.
autocomplete config reset will restore config to the default values
autocomplete usage< TODO INSERT PICTURE OF USAGE>
The average cost for me is about half a penny per request using the latest gpt-4-omni model.
The next section has instructions for the lower cost gpt-3.5-turbo model.
GPT 3.5 has a lower cost model that is still quite effective. It costs about 10-20 calls per $0.01
Run the following commands to switch to the lower cost model
autocomplete config set model gpt-3.5-turbo
autocomplete config set api_prompt_cost 0.0000005
autocomplete config set api_completion_cost 0.0000015As always --help will get you more
autocomplete --help< TODO INSERT PICTURE OF help screen>
< TODO Replace each persona with a GIF of completions and the before + after text completions. Use USAGE.md>
Quickly manipulate datasets in the terminal to efficiently complete data transformations and move on to analysis or further processing.
ls
awk -F',' '{print $1}' data.csv | grep 'keyword'
spark-submit --master yarn my_script.py --input data.csv --output results/
python -c "import pandas as pd; df1 = pd.read_csv('data1.csv'); df2 = pd.read_csv('data2.csv'); merged_df = pd.merge(df1, df2, on='key_column'); merged_df.to_csv('merged_data.csv', index=False)"Swiftly deploy updates and focus on improving your codebase.
git init
gcc -o output_file input_file.c
pytest tests/
docker build -t my_image .
docker run -d -p 8080:80 my_image- Effortlessly navigate and control your system for seamless administration.
top
sudo apt install package_name
chmod 755 script.sh
sudo systemctl start service_nameBuild confidence and proficiency with every command.
cd path/to/directory
touch new_file.txt
cp file.txt destination/
cat file.txtStreamline tasks and reclaim valuable time for what matters most.
sed -i 's/old_text/new_text/g' *.txt
tar -czvf archive.tar.gz directory/
tar -xzvf archive.tar.gz
cat file.txt | grep 'keyword' | wc -lResolve issues and understand commands with ease.
man command_name
apropos keyword
help
dpkg -l | grep package_nameInstall locally
git clone git@github.com:closedloop-technologies/autocomplete-sh.git
# Create a symlink to install the script in your path
ln -s $PWD/autocomplete.sh $HOME/.local/bin/autocomplete
# Run the install script
. autocomplete.sh installPre Commit Hooks for shellcheck
pip install pre-commit
pre-commit install
Tests via Bats
sudo apt install bats
bats tests
Test Installation
docker build -t autocomplete-sh .
docker run -it autocomplete-sh
. autocomplete.sh install
autocomplete --help
Currently maintained by Sean Kruzel @closedloop as a member of Closedloop.tech
More maintainers and bug fixers are quite welcome, and we hope to grow the community around this project. Governance will be re-evaluated as the project evolves.
See the LICENSE file for details.