Install AutoGPT on Windows

I don't know if you're familiar with AutoGPT, but it's a kind of God Mode for ChatGPT.

Basically, you give it a mission and the tool works through it using auto-prompts in ChatGPT.

What's more, it can interact with online and local applications and services, such as web browsers and document management (text files, csv).

Just so you know, there are already people making quite a lot of money with it.

The thing is, the other day I was trying to install it by following the instructions on its official website when quite a few things were failing, some because the instructions weren't clear enough and many others because of my limited Python knowledge.

So, if you're trying to install AutoGPT on your PC and the same thing is happening to you, here's the step-by-step process that worked for me.

Installation requirements

Python 3.8 or later: installing it is pretty straightforward. In the article where I explained how to install Stable Diffusion we already did it. In my case, I used version 10.

Git: in the same article we installed git. You can download the Windows version from this link.

OpenAI API Key: log in to your account and create it here.

It's not essential, but it is VERY useful to use version 4 of ChatGPT. If you don't have it yet, request it from here.

Pinecone API Key: so that your AutoGPT has long-term memory. Create an account here and go to the APIs section to get yours. You'll need the two values I point out later:

Updated Linux kernel: from here you can download it and set version 2 as the default, as indicated.

Docker Desktop for Windows: download it from here and install it normally.

Okay, with this you have everything you need if you don't want AutoGPT to talk to you (it could do so through the API from ElevenLabs)

Step-by-step installation

Let's start with installing the tool itself on a Windows PC.

Step 1: create the folder where it will be installed

Create a folder where the application's code will be stored.

In my case, I created it in the root of the hard drive:

Step 2: clone the repository

To do this, open the folder we just created, type "CMD" in the folder path field in the window and press Enter:

The console will open and we type this:

git clone https://github.com/Torantulino/Auto-GPT.git

Enter.

All the code will start downloading into the folder.

Step 3: install the dependencies

From the console, navigate to the "Auto-GPT" folder that will have been created inside your original folder. To do that, type:

cd Auto-GPT

Press Enter.

This command should start installing the dependencies:

pip install -r requirements.txt

I started getting errors here, so I replaced it with this one:

C:UserspcadminAppDataLocalProgramsPythonPython310python.exe -m pip install -r requirements.txt

Because when I installed Python I didn't tick the corresponding box to add the path to PATH, what I do is replace pip with the full path where Python is installed on my PC:

Pay attention to what I've highlighted in bold, because it will be different in your case:

• The first thing is your Windows username.
• The second is the Python folder, whose name will change depending on the version installed (in my case it's 3.10).

Save this path because we'll use it again.

Step 4: change parameters in the .env file

All the necessary files and dependencies will now have been installed. Among them, you will see a file called ".env.template".

Now make a copy of that file and rename it to ".env":

Open ".env" with a text editor, because we are going to change the parameters. Here are the ones you need to uncomment (removing the "#" from the beginning of the line) and the custom values to add (in bold):

• RESTRICT_TO_WORKSPACE=True
• USER_AGENT="Mozilla/5.0 (Macintosh; Intel Mac OS X 10_15_4) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/83.0.4103.97 Safari/537.36"
• EXIT_KEY=n
• OPENAI_API_KEY=enter your OpenAI key
• TEMPERATURE=0
• USE_AZURE=False
• SMART_LLM_MODEL=gpt-4
• FAST_LLM_MODEL=gpt-3.5-turbo
• FAST_TOKEN_LIMIT=4000
• SMART_TOKEN_LIMIT=4000
• MEMORY_BACKEND=local
• MEMORY_INDEX=auto-gpt
• PINECONE_API_KEY=enter your Pinecone key
• PINECONE_ENV=enter your Pinecone "Environment"
• IMAGE_PROVIDER=dalle
• IMAGE_SIZE=256
• HUGGINGFACE_IMAGE_MODEL=CompVis/stable-diffusion-v1-4
• HUGGINGFACE_API_TOKEN=your-huggingface-api-token
• SD_WEBUI_AUTH=
• SD_WEBUI_URL=your local webUI URL if you have one (not required)

These are the uncommented parameters and the values I have added to my project.

I haven't had much time to play around with them, just enough to check that the tool works and does what it's supposed to.

Save the file and we're now ready to use AutoGPT.

Launch AutoGPT

In theory, you would only need to type this in the console:

python scripts/main.py

But because of the PATH issue mentioned above, in my case I have to type this instead:

C:UserspcadminAppDataLocalProgramsPythonPython310python.exe -m autogpt

Remember that the values in bold are the ones you'll need to check in your case.

If everything goes as expected, the console will take a little while and you'll see something like this:

AutoGPT will ask what its goal is and a few more questions about how it will be used, theoretically to optimize its work.

Once you fill in the details for your use case, the tool will start carrying out tasks, asking you first whether you think what it proposes is appropriate. If so, you only have to press the "y" key and the application will start its task:

Exit AutoGPT

Very simple: just press "n" when AutoGPT asks whether we want to continue and the program will stop running.

Conclusions

As you can see, installing AutoGPT isn't very complicated, but it does require a few tricks that can leave you stuck if you don't know them.

Personally, when you start using it and see everything it does and how it suggests doing it, that is when you really "feel" that AI is going to start carrying out many, many manual tasks.

I hope this helps you start using it in your day-to-day work too.