ayo

ayo (pronounced "hey yo!") is a simple command line AI assistant. It can run fully offline using ollama and its bank of pre-trained models.

Quick Start

Pre-requisites

You will need ollama installed and running on your machine, or an accessible server.

Install

Visit the latest release page and download the appropriate binary for your system. Put the binary somewhere accessible in your PATH.

Add Tools

Put a list of tools you want ayo to be able to use in the toolbox directory. By default ayo will look into ~/.config/ayo/toolbox. More about tools in the tools section.

Run

ayo explain my ~/.profile file

Usage

🧰 Tools

Ayo can use all tools from its toolbox to take actions on your local system.

Toolbox

The toolbox is a directory where ayo looks for tools to use. By default, ayo will look into ~/.config/ayo/toolbox. You can change the toolbox location by setting the AYO_TOOLBOX environment variable.

Tool

A tool is a JSON file that describes how ayo can use a command on your system.

{
  "type": "function",
  "function": {
    "name": "cat",
    "description": "Get the content of a file",
    "parameters": {
      "type": "object",
      "properties": {
        "file": {
          "type": "string",
          "description": "The file to read"
        }
      },
      "required": [
        "file"
      ]
    }
  },
  "cmd": "cat",
  "args": [
    "{{.file}}"
  ]
}

Breaking down the JSON file:

• type is the type of tool. Currently, only function is supported.
• function describes the function which will be sent to the model.
  • name is the name of the function. It must be unique in ayo's toolbox.
  • description is a short description of the function.
  • parameters describes the parameters of the function.
    • type is the type of the parameters. Currently, only object is supported.
    • properties describes the properties of the object.
    • required is an array of required properties.
• cmd is the command to run. It will be executed using the os/exec package.
• args is an array of arguments to pass to the command. The arguments can be templated using Go's text/template package. The template will be executed with the function's parameters values returned by the model.