Two AI assistants to scrape emails and teach code with Auto-GPT.
You need to clone the latest stable branch from Auto-GPT repository:
git clone -b stable https://github.com/Significant-Gravitas/Auto-GPT
Go inside the new folder:
cd Auto-GPT
Copy .env.template, rename it .env and add your OPENAI_API_KEY.
Launch your Docker daemon, and build the image:
docker-compose build auto-gpt
If you don't have Docker, check an alternative method. You can do:
pip install -r requirements.txt
Run:
docker-compose run --rm auto-gpt
You can also pass options.
To define a model:
docker-compose run --rm auto-gpt --gpt3only
To let the bot do its job without having to validate every step:
docker-compose run --rm auto-gpt --continuous
Answer --manual to the first question to be able to give a name to your AI bot.
Run:
./run.sh
Here is the first example:
Name: email-finder-gpt
Role: an AI designed to find the emails of the top 10 coding youtubers and save them to a file
Goals: ['find the emails of the top 10 coding youtubers', 'save the emails to a file']
The result will be stored in autogpt/auto_gpt_workspace.
Here is the second example:
Name: coding-exercise-gpt
Role: an AI designed to help me, a coding instructor, write exercises for students learning to code
Goals: ["Write a text description for an exercise testing students' ability to write Python classes", 'Write any starter code needed for the exercise and save it to a file', 'Write the solution code needed for the exercise and save it to a file', 'Write tests for the exercise solution and save them to a file']
The result is stored in coding-exercise-gpt.
To test it:
cd coding-exercise-gpt
python3 person_tests.py
- setting up Auto-GPT with Docker.
- creating an email finder AI.
- creating a teaching assistant.
Based on AutoGPT Tutorial - Create Your Own AI Agents! by Colt Steele (2023).