https://github.com/ngpepin/aido
Bash scripts for in-line CLI review, editing and execution of Bash commands suggested by an LLM in response to a user's requests for assistance.
https://github.com/ngpepin/aido
assistant-chat-bots bash bash-completion bash-prompt bashrc command-line-tool linux llama llm llm-agent mixtral-8x7b
Last synced: about 1 month ago
JSON representation
Bash scripts for in-line CLI review, editing and execution of Bash commands suggested by an LLM in response to a user's requests for assistance.
- Host: GitHub
- URL: https://github.com/ngpepin/aido
- Owner: ngpepin
- License: mit
- Created: 2024-03-06T17:30:56.000Z (over 1 year ago)
- Default Branch: main
- Last Pushed: 2024-07-19T18:35:38.000Z (11 months ago)
- Last Synced: 2025-05-11T00:01:43.112Z (about 2 months ago)
- Topics: assistant-chat-bots, bash, bash-completion, bash-prompt, bashrc, command-line-tool, linux, llama, llm, llm-agent, mixtral-8x7b
- Language: Shell
- Homepage:
- Size: 16.6 KB
- Stars: 0
- Watchers: 1
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
- License: LICENSE
Awesome Lists containing this project
README
# "AI do!" (aido) & "sudo AI do!" (saido)
Very simple/simplistic Bash scripts for in-line CLI review, editing and execution of Bash commands suggested by an LLM in response to a user's requests for assistance.
- `aido` and `saido` are Bash aliases leveraging a backing function designed to execute Bash commands that have been suggested by an LLM (Language Model) based on natural language user prompts.
- These tools integrate seamlessly into your `.bashrc` file, providing an in-line method to request Bash command guidance.
## Purpose
`aido` and `saido` are designed to provide users with a convenient interface for executing Bash commands as either a non-root user or as root, respectively, as suggested by a Language Learning Model (LLM). They serve as handy tools for those who may not remember every Bash command or who seek a targeted and actionable reminder without any tl;dr.
## Safety & Security
Users must manually approve and can modify any LLM-suggested command before execution. This measure ensures control and conscious user consent before any command execution, particularly for those prefixed by `sudo`. However, please note that this is a proof-of-concept and that users are wholly responsible for commands they execute, especially since `eval` is used for execution.
## Usage
Add the following aliases to your `.bashrc` to use (these are provided in the aido-bashrc-snippet.sh):
```bash
alias aido='aido_func aido -u'
alias saido='aido_func saido -a'
```
## Setup
Ensure the `aichat` package and any required LLM servers (like `ollama`) or cloud API services are properly configured on your machine. Add the provided function and aliases to your `.bashrc` file to enable the functionality.
## Dependencies
This script relies on the `aichat` package for LLM communication and might require local or remote LLM servers depending on your setup.
Refer to:
- `aichat` installation: [aichat GitHub page](https://github.com/sigoden/aichat)
- `ollama` server setup: [Ollama Homepage](https://ollama.com)
## Shortcomings and Constraints
The script may require prompt optimization for specific LLM models and does not support multiple command recommendations or maintain context between interactions.
## Test Coverage and Limitations
Currently tested with the `ollama` server and Eric Hartford's `dolphin-mixtral-8x7b` model. Other models and extensive prompt engineering have not been tested.
## Possible Alternatives
Explore other integrations like [Shell GPT](https://github.com/TheR1D/shell_gpt).
## License
This project is open-sourced under the MIT License. See the LICENSE file for details.
## Contributing
Contributions are welcome!
## Disclaimer
This project is for educational and proof-of-concept purposes only. No guarantees are provided regarding the safety, security, or suitability of the suggested commands. Users are responsible for reviewing and executing commands at their own risk.