QUICKSTART.md

Quickstart Guide

For the complete getting started tutorial series <- click here

Welcome to the Quickstart Guide! This guide will walk you through the process of setting up and running your own AutoGPT agent. Whether you're a seasoned AI developer or just starting out, this guide will provide you with the necessary steps to jumpstart your journey in the world of AI development with AutoGPT.

System Requirements

This project supports Linux (Debian based), Mac, and Windows Subsystem for Linux (WSL). If you are using a Windows system, you will need to install WSL. You can find the installation instructions for WSL here.

Getting Setup

1. Fork the Repository To fork the repository, follow these steps:

   • Navigate to the main page of the repository.

   

   • In the top-right corner of the page, click Fork.

   

   • On the next page, select your GitHub account to create the fork under.
   • Wait for the forking process to complete. You now have a copy of the repository in your GitHub account.

2. Clone the Repository To clone the repository, you need to have Git installed on your system. If you don't have Git installed, you can download it from here. Once you have Git installed, follow these steps:

   • Open your terminal.
   • Navigate to the directory where you want to clone the repository.
   • Run the git clone command for the fork you just created

   

   • Then open your project in your ide

   

3. Setup the Project Next we need to setup the required dependencies. We have a tool for helping you do all the tasks you need to on the repo. It can be accessed by running the run command by typing ./run in the terminal.

   The first command you need to use is ./run setup This will guide you through the process of setting up your system. Initially you will get instructions for installing flutter, chrome and setting up your github access token like the following image:

   Note: for advanced users. The github access token is only needed for the ./run arena enter command so the system can automatically create a PR

   

   You can keep running the command to get feedback on where you are up to with your setup. When setup has been completed, the command will return an output like this:

   

Creating Your Agent

Now setup has been completed its time to create your agent template. 
Do so by running the `./run agent create YOUR_AGENT_NAME` replacing YOUR_AGENT_NAME with a name of your choice. Examples of valid names: swiftyosgpt or SwiftyosAgent or swiftyos_agent

Upon creating your agent its time to offically enter the Arena!
Do so by running `./run arena enter YOUR_AGENT_NAME`

Note: for adavanced yours, create a new branch and create a file called YOUR_AGENT_NAME.json in the arena directory. Then commit this and create a PR to merge into the main repo. Only single file entries will be permitted. The json file needs the following format.

{
 "github_repo_url": "https://github.com/Swiftyos/YourAgentName",
 "timestamp": "2023-09-18T10:03:38.051498",
 "commit_hash_to_benchmark": "ac36f7bfc7f23ad8800339fa55943c1405d80d5e",
 "branch_to_benchmark": "master"
}

• github_repo_url: the url to your fork
• timestamp: timestamp of the last update of this file
• commit_hash_to_benchmark: the commit hash of your entry. You update each time you have an something ready to be officially entered into the hackathon
• branch_to_benchmark: the branch you are using to develop your agent on, default is master.

Running your Agent

Your agent can started using the ./run agent start YOUR_AGENT_NAME

This start the agent on http://localhost:8000/

The frontend can be accessed from http://localhost:8000/, you will first need to login using either a google account or your github account.

Upon logging in you will get a page that looks something like this. With your task history down the left hand side of the page and the 'chat' window to send tasks to your agent.

When you have finished with your agent, or if you just need to restart it, use Ctl-C to end the session then you can re-run the start command.

If you are having issues and want to ensure the agent has been stopped there is a ./run agent stop command which will kill the process using port 8000, which should be the agent.

Benchmarking your Agent

The benchmarking system can also be accessed using the cli too:

agpt % ./run benchmark
Usage: cli.py benchmark [OPTIONS] COMMAND [ARGS]...

  Commands to start the benchmark and list tests and categories

Options:
  --help  Show this message and exit.

Commands:
  categories  Benchmark categories group command
  start       Starts the benchmark command
  tests       Benchmark tests group command
agpt % ./run benchmark categories     
Usage: cli.py benchmark categories [OPTIONS] COMMAND [ARGS]...

  Benchmark categories group command

Options:
  --help  Show this message and exit.

Commands:
  list  List benchmark categories command
agpt % ./run benchmark tests      
Usage: cli.py benchmark tests [OPTIONS] COMMAND [ARGS]...

  Benchmark tests group command

Options:
  --help  Show this message and exit.

Commands:
  details  Benchmark test details command
  list     List benchmark tests command

The benchmark has been split into different categories of skills you and test your agent on. You can see what categories are available with

./run benchmark categories list
# And what tests are available with
./run benchmark tests list

Finally you can run the benchmark with

./run benchmark start YOUR_AGENT_NAME