Running Amico

The amico cli is a standalone Amico agent runtime. Everyone can run an Amico autonomous AI agent and control its tasks and workflows just in a configuration file, without modifying the source code.

Installation

Installing from Cargo

You have to set up rust toolchain first. The recommended way is to use rustup.

After installing rustup on your system, run the following command to set up the rust stable tool chain:

Terminal

rustup default stable

Now you can install the amico cli using the following command:

Terminal

cargo install amico

You may have to add ~/.cargo/bin to your $PATH for the amico command to work. If you are using bash or zsh, you can do it like this:

Terminal

export PATH="$PATH:$HOME/.cargo/bin"

To check if the amico command is installed, run the following command:

Terminal

amico --version

Configuration

To run the Amico standalone runtime, you will need to set up a configuration file. The configuration file is a TOML file that specifies the agent's behavior and tasks.

For Linux systems, amico searches for the Amico.toml configuration file in the following order:

amico.toml in the current working directory (cwd);
$XDG_CONFIG_HOME/amico/Amico.toml;
$HOME/.config/amico/Amico.toml;
$HOME/.amico/Amico.toml;
/etc/amico/Amico.toml;

You can also specify the configuration file path explicitly using the --config or -c option:

Terminal

amico --config <path to config file>

Example Configuration

Important: Amico is not stable yet. The configuration file format may change.

# The example configuration file of a single-agent runtime
 
# Configuration file standard version
version = 0
 
# The name of the agent
name = "Amico"
 
# Standard runtime using `tokio::main`
runtime = "standard"
 
# Model Providers
[providers.openai]
base_url = "https://api.openai.com/v1"
api_key = "sk-XXX"
 
# Custom provider
[[providers.custom]]
name = "my-custom-provider"
base_url = "http://localhost:8000"
schema = "openai"
api_key = "XXX"
 
# Plugins configuration
[plugins]
# Use `std` plugins for low-level infrastracture
knowledge_base = "std"
task_executor = "std"
action_selector = "std"
 
[plugins.effectors.twitter]
enabled = true
consumer_key = "XXX"
consumer_secret = "XXX"
access_token = "XXX"
access_token_secret = "XXX"
 
[plugins.sensors.twitter_reader]
enabled = true
api_key = "XXX"
 
# Agent Runtime Infrastructure (Low-Level Plugins) Configuration
 
# RAG system
[plugins.config.knowledge_base]
database = "in-memory"
embedding_model = "text-embedding-ada-002"
 
# Action Selector Event Settings
[plugins.config.action_selector]
event_expiry_time = 60 # Event default expiry time (in seconds)