shellama

This is a simple utility that allows you to communicate with Ollama running locally in your shell. This can be used for quick iteration while programming or trouble shooting and is a small binary that does... very little! Here are some examples of what you can do with shellama

Prerequisites

Before running shellama you need ollama installed on your machine and running. The default model is also set to llama3.1, this needs to be downloaded if you don't plan to overwrite the model. If you want to use a different model ensure that it is pulled locally using the following command.

ollama pull llama3.1

Installing

shellama can be installed through homebrew or through GitHub releases!

brew tap rossedman/tap
brew install rossedman/tap/shellama

Using

shellama is meant to be used as a simple shell utility that can speed up development cycles. I built this to prevent from having to look things up or even paste things into GUI forms for ChatGPT. Because this uses ollama it also runs everything locally allowing you lots of leverage over what gets run.

You can ask shellama for simple examples like below:

$ shellama "provide an example of a bash for loop"

for i in {1..5}; do
  echo "Loop iteration: $i"
done

You can also cat files into shellama to be analyzed.

$ shellama "what does this file do?" $(cat main.go)

This is a Go (Golang) source file that builds and runs the `shellama` command. The main function joins two strings (`version` and `commit`) with "+" in between using `strings.Join()`, then passes them to `cmd.Execute()` to run the shellama command.

Using Profiles

Profiles are a way to use different combinations of models and prompts so you can have different "actors" that can be called on. Below is the example of a profile file placed at $HOME/.shellama.yaml

profiles:
  - name: default
    model: llama3.1
    prompt: |
      you are a simple programming assistant who only returns code
  - name: advanced
    model: llama3.1
    prompt: |
      you are a programming assistant who can also write advanced code and explain it in depth
  - name: editor
    model: llama3.1
    prompt: |
      you are a copy editor who can help with writing and editing text

This would allow you to run commands and switch profiles for different purposes

$ shellama --profile advanced "provide an example of a bash for loop"