Ollama Powered Terminal/Shell AI Helper

This tool enhances your command-line experience by answering questions, proposing commands, analyzing command output, and helping correct errors—all powered by Ollama AI.

When you invoke it with a prompt, your question is sent to Ollama, and the AI's response is printed to your terminal. The tool automatically collects useful context and sends it along with your question to improve the quality of the answer.

Context sent to Ollama includes:

• The current operating system you use
• The current shell you are running in
• The content of the clipboard if you specify the --clip option

---

Usage

Simply run the command followed by your question:

ai How can I switch to my home directory?

A typical response might look like:

✨ You can switch to your home directory by using the `cd ~` command.

Command proposal: `cd ~`

---

Options

To see all available options, run:

ai --help

---

Configuration

You can specify some settings in a settings file in the file ~/.ai_config. If this file exists it will take precendence over the environment variables.

Ollama URL

This can be set in the config file via:

OLLAMA_API_BASE = <url>

alternatively set the OLLAMA_API_BASE environment variable to your Ollama server URL. If nothing is configured, the tool defaults to http://localhost:11434.

export OLLAMA_API_BASE=http://localhost:11434

Model Selection

Before starting, you must specify the Ollama model.

This can be set in the config file via:

OLLAMA_MODEL = <modelname>

or by setting the OLLAMA_MODEL environment variable:

export OLLAMA_MODEL=gemma3:4b

You can also set the model at runtime using the --model option:

ai --model gemma3:4b "Your prompt here"

Which model should I use?`This depends on your hardware but in general I get good answers by using small models like gemma3:4b or the small phi models. They produce quick responses which are most of of the time good enough. I do not recommend to use a reasoning model as they take longer to process your input with only little improvements. If you have better hardware you can try bigger models but often their output is way to verbose so it will not give you always a overall better experience.

Other Settings

You can also specify a prefix and suffix for the generated text. This is currently used to color the output of the AI. To do so specify in the config file: OUTPUT_START and OUTPUT_END

For example:

OUTPUT_START=\x1B[32m✨ 
OUTPUT_END=\x1B[0m

Changing the system prompt

To set your own system prompt you can do so by setting in the config file SYSTEM_PROMPT

---

Building

• Entrypoint: bin/
• Library code: lib/
• Example unit tests: test/

To build the executable, run:

dart compile exe bin/ai.dart

This will give you an bin/ai.exe on Linux you may want to rename that one.

---

Further Reading

• Ollama documentation