AutoGPT is aopen source application developed in Python which uses i GPT-4 and GPT-3.5 language models through the API on OpenAI to read and write files, search the Internet, and learn from your mistakes.
With AutoGPT it is possible to give life to agents based ongenerative artificial intelligence able to take a single user command as a cue, and autonomously generate an intricate set of intermediate targets to satisfy the original request without further guidance.
Therefore, unlike OpenAI’s ChatGPT, Microsoft’s Bing Chat AI, and Google Bard, AutoGPT is not limited to a series of questions and answers, but acts independently by iterating over its own prompts and modifying each iteration to improve and work on it.
The system it is still in its infancy, but there are already several versions of AutoGPT available for developers, including Microsoft’s BabyAGI and Jarvis. There is another variant called AgentGPT accessible via browser for the general public. However, advanced features are limited to those with an OpenAI API subscription.
In this article, we will see how to do the setup su Windows.
1. All requirements for AutoGPT Windows setup
To use AutoGPT, the following requirements must be met.
Download one Python version 3.8 or higher for Windows directly from the official site. Just click on the yellow button “Download Python 3.11.3” and follow the simple installation instructions on Windows. Make sure you enable “Add python.exe to PATH” before installing Python.
Then you need to install Git always from the official site. Git is a distributed version control system designed to handle projects of any size with efficiency and speed because it allows you to keep track of all changes made to the source code. In this way, Git allows you to work collaboratively and manage your code efficiently. But it offers many other features, including the ability to clone a repository to your own machine, create and manage branches, merge branches, and commit changes.
At this point, you need to get the chiavi API OpenAI e Pinecone. In the first case, just visit the dedicated site and create an account (if the user does not already have one). Then, click on the button “Create new secret key”, give the key a name and copy it for safekeeping, perhaps in a text editor.
In the second case, you can connect to the official site and create an account. Pinecone will enable the language model to retrieve relevant information from memory for AI applications. Once on the site, click on “API Keys” in the sidebar on the left and then on “Create API Key” in the right pane. Give it a name like “autogpt” and click “Create Key”. Now, copy and store somewhere the values marked as “Value” ed “Environment” because they will be needed later.
Optionally, it is recommended to proceed with the request for the API Keys also with ElevenLabs by connecting directly to the site, if you also want to use the function AutoGPT speech synthesis.
2. Download the AutoGPT source code from GitHub
Once obtained, you can proceed with the installation of AutoGPT by following these steps.
If you are familiar with Git, you can directly clone the AutoGPT repository by entering the command: git clone https://github.com/Significant-Gravitas/Auto-GPT.git in the computer terminal and go straight to step 2. If not, follow the step below for Windows setup.
Download as a ZIP file the source code of AutoGPT from repository di GitHub “https://github.com/Significant-Gravitas/Auto-GPT/releases/tag/v0.3.1” and unzip the downloaded folder by right clicking the mouse and selecting “Extract All”. Then, choose in which position to extract the contents of the folder and confirm by clicking on the “Extract” item.
3. Add OpenAI and Pinecone API Keys
Now you need to enter the API Keys obtained from OpenAI and Pinecone in the file .env.template. Then, right-click the .env.template file and open it with an editor like WordPad. Paste the APIs you copied earlier into the “OPENAI_API_KEY” e “PINECONE_API_KEY”. In the “PINECONE_ENV=” you can enter the region of the Pinecone server.
After doing this, change the name of the .env.template file to .env.
4. How to install AutoGPT on Windows
Once you’ve entered your API keys, it’s time to run AutoGPT. All that remains is to click the right mouse button on the “AutoGPT” folder and select the “Copy as path“.
Now, open your computer terminal by searching for “Command Prompt” o “cmd” in the search bar menu Start di Windows and type cdadd a space, then paste the copied path with the key combination Ctrl + V.
After that, press the “Enter” key and navigate to the Auto-GPT folder. Once here, install all the libraries required to run Auto-GPT, by running the command: pip install -r requirements.txt
If that doesn’t work, try: py -m pip install -r requirements.txt.
Next, to start Auto-GPT on your system, run the command: python -m autogpt.
5. Some tips to use AutoGPT on your PC
On first run, AutoGPT will ask you to assign a name to AI. If you don’t want to build an AI for a specific use case, you can leave this field blank and press “Enter”. The system will choose the name “Entrepreneur-GPT” by default.
Next, you need to describe the role of AI and establish up to five goals. For example, you can ask it to save the information to a text or PDF file or to shut down after all the information has been retrieved.
Now, AutoGPT will start thinking. During the course of the action, it will ask for your permission after each request, which allows for test and adapt the project going to our needs. Type “y” and “Enter” to confirm.
It might as well connect to websites and collect information. Meanwhile, the user can read what the AI is thinking, how he thinks and plans his next move. AutoGPT also displays negative self-criticism or prompts which helps him to improve his decision making. Finally, take action.
If you want to run the AI continuously without asking for user permission, you can input “y -n” and press “Enter”. Be sure to replace n with a number. If you enter “y -5”, it means that AutoGPT will not ask our permission for the next 5 actions. The “Ctrl + C” shortcut stops any action.
After the completion of the task, the user should find the final information in the Auto-GPT folder inside the file “auto_gpt_workspace“.
To know more: Artificial Intelligence: what it is and what it can do for us