ShellOracle
ShellOracle is an innovative terminal utility designed for intelligent shell command generation, bringing a new level of
efficiency to your command-line interactions. ShellOracle currently supports Ollama, llmman, OpenAI, Deepseek, LocalAI, and Grok!
Show your support for ShellOracle and keep an eye out for exciting new developments by clicking the ⭐ and a 👀!
Table of Contents
Features
Key features of ShellOracle include:
- Seamless shell command generation from written descriptions
- Command history for easy reference
- Unix pipe support for advanced command chaining
- Self-hosted for full control over your environment
- Highly configurable to adapt to your preferences
Installation
Installing ShellOracle is easy!
- pipx install the
shelloracle packagepipx install shelloracle
- Configure ShellOracle and follow the prompts
shor config init
- Refer to the providers section for specific details regarding your chosen provider.
Upgrading to the latest version of ShellOracle is just as simple!
- pipx upgrade the
shelloracle packagepipx upgrade shelloracle
Installation with pip is supported, however, pipx is preferred for its automatic environment isolation.
Usage
ShellOracle is designed to be used as a BASH/ZSH/fish widget activated by the CTRL+F keyboard shortcut.
- Press CTRL+F
- Describe your command
- Press Enter
The generated command will be inserted into your shell prompt after a brief processing period.
Other ways to run ShellOracle
ShellOracle can be run as a Python module with python3 -m shelloracle or using its entrypoint shor, however,
running ShellOracle with this method will not automatically insert the result into your shell prompt.
Tips
- If you press CTRL+F with text in your ZLE buffer, all text left of your cursor will carry over to your ShellOracle
prompt.
- ⬆️ arrow and ⬇️ arrow cycle through your prompt history.
- ShellOracle can be chained with other commands; try:
echo "find all the python files in my cwd" | shor
Providers
Ollama
Before using ShellOracle with Ollama, pull the model you chose in the configure step.
For example, if you chose gemma4:12b, run:
ollama pull gemma4:12b
Refer to the Ollama docs for installation, available models, and usage.
llmman
llmman serves the Ollama API on port 17434. Start it and pull the model you
chose in the configure step. For example, if you chose gemma4, run:
llmman serve
llmman pull gemma4
OpenAI
To use ShellOracle with OpenAI’s models, create an API key. Edit
your ~/.shelloracle/config.toml to change your provider and enter your API key.
You can use shor config edit as a shorthand to edit this file.
Deepseek
To use ShellOracle with Deepseek’s models, create an API key. Edit
your ~/.shelloracle/config.toml to change your provider and enter your API key.
You can use shor config edit as a shorthand to edit this file.
LocalAI
Refer to the LocalAI docs for installation, available models, and usage.
LM Studio
To use ShellOracle with LM Studio, you will first need to run a model and start the server.
For macOS users, it’s recommended to utillize a model that is using the MLX runtime.
Here is an example configuration:
[shelloracle]
provider = "OpenAICompat"
[provider.OpenAICompat]
base_url = "http://localhost:1234/v1"
api_key = "lm-studio"
model = "mistralai/devstral-small-2507"
XAI
To use ShellOracle with XAI’s models, create an API key.
Edit your ~/.shelloracle/config.toml to change your provider and enter your API key.
You can use shor config edit as a shorthand to edit this file.
Google
To use ShellOracle with Google’s models, create an API key.
Edit your ~/.shelloracle/config.toml to change your provider and enter your API key.
You can use shor config edit as a shorthand to edit this file.
Configuration
ShellOracle’s configuration is your gateway to tailoring the utility to match your preferences and requirements.
The ~/.shelloracle/config.toml file serves as the control center for customizing various aspects of ShellOracle’s
behavior. You can quickly access and edit this file using the command shor config edit.
System Requirements
Software
ShellOracle supports BASH, ZSH and fish on macOS and Linux.
Hardware
For cloud providers like OpenAI, there are no hardware requirements.
If running locally, refer to your model for hardware requirements.
Feedback
Encountered problems? File an issue. Feature requests are welcome,
and contributions can be made by opening a pull request.
License
This software is licensed under the GPLv3 license.
评论