AI-powered command helper for your terminal. Built for Bash, customizable and ready to extend.
Terminal‑integrated assistant that suggests shell commands with help from a Large Language Model (LLM).
"Describe what you want → pick a suggestion → optionally run it."
- LLM‑powered suggestions (OpenAI, Ollama)
- Interactive picker – always asks before running
- Context‑aware – optionally include cwd,
git status
, and recent history - Customizable via
commands.md
to let LLM know what you want - Custom context script – use
.aih_context.sh
to provide additional system context - Single Bash function (
aih
) for quick access - Configurable through
.env
or CLI flags - MIT‑licensed and Python 3.11+ compatible
# 1. Clone or unzip
git clone https://github.com/alan-mroczek/ai-cli-help.git
cd ai-cli-help
# 2. Run installer (installs deps with uv, sets up .env & shell hook)
python install.py
# 3. Reload shell
source ~/.bashrc
Note: The script uses uv for fast, isolated deps.
Ifuv
isn’t installed, the installer will fetch it.
aih --help
usage: aih [-h] [--context] [--model MODEL] [--max MAX_SUGGESTIONS]
[--no-confirm] [--history]
[prompt ...]
Suggest shell commands with LLM assistance.
positional arguments:
prompt Describe what you want to do
options:
-h, --help show this help message and exit
--context Include directory & git context in the LLM prompt
(default from INCLUDE_CONTEXT in .env)
--model MODEL Override model name (default from MODEL in .env)
--max MAX_SUGGESTIONS
Max suggestions (default from MAX_SUGGESTIONS in .env)
--no-confirm Skip command confirmation prompt (default from
REQUIRE_CONFIRMATION in .env)
--history Display command history and exit
Example session:
aih large files here
Suggestions:
1. du -ah . | sort -rh | head -n 10
Options:
Enter a number to select a command
r - Regenerate suggestions
c - Add a comment or clarification
q, 0 or empty - Quit
Your choice:
Key | Purpose | Example |
---|---|---|
OPENAI_API_KEY |
API key for OpenAI models | sk-… |
MODEL |
Default model (openai/gpt-4o-mini , ollama/codellama ) |
openai/gpt-4o-mini |
OLLAMA_API_URL |
Ollama chat endpoint | http://local…::11434/api/… |
MAX_SUGGESTIONS |
Limit shown suggestions | 3 |
Flag | Description |
---|---|
--context |
Include cwd, git info, & history |
--model |
Override model for a single call |
--max |
Override max suggestions |
This file is a free-form cheat-sheet for the LLM.
There’s no rigid schema—the model simply reads the text and tries to imitate or reuse whatever it finds—so write it in whatever style feels natural.
This optional script allows you to add custom context gathering commands.
- An empty executable script is created during installation
- Copy commands from
.aih_context.example.sh
or write your own - The script is executed each time you run
aih --context
- Output from each command will be included in the LLM prompt
- Perfect for adding project-specific context or system information
- Empty scripts are safely ignored and won't affect the context
ai-cli-help/
├── main.py # Entry point
├── model.py # LLM abstraction
├── utils.py # Helpers (spinner, context, env)
├── install.py # One‑shot installer
├── commands.sh # Bash wrapper (sources aih)
├── commands.md # Docs (git-ignored)
├── commands.example.md # Docs example
├── .aih_context.sh # Custom context script (git-ignored)
├── .aih_context.example.sh # Context script example
├── .env # Configuration (git-ignored)
├── .env.example # Configuration example, used as a template to create .env
└── LICENSE
Licensed under the MIT License – see LICENSE
for details.