I’ll explore what chatbot offerings from OpenAI will run on a retro computer and explain exactly how you can do it too! Since I recorded this video in February 2023, OpenAI has actually improved upon the chat bot models that are available via API. So expect even better results now than those shown in the video!
1. Setup a RaspberryPi (or similar Linux machine) on your LAN to act as your RetroBridge. Then connect to it over ssh from a modern computer to configure it.
2. Install the openai Python library on your RetroBridge:
pip3 install openai
3. Copy my custom Python3 ChatGPT 3.5 Turbo script to your RetroBridge user folder:
wget https://levimaaia.com/content/chatbot.py
4. Create an OpenAI account and obtain a new API key. Paste that key into the ~/chatbot.py file on your RetroBridge:
https://platform.openai.com/account/api-keys
5. Install telnetd on your RetroBridge (follow the steps at this URL):
https://adamtheautomator.com/linux-to-install-telnet/
6. Permanently enable ANSI terminal emulation on the RetroBridge by adding the following line to the ~/.profile file:
TERM=ansi
7. Ensure that the terminal emulation software on your Apple II (Proterm or Spectrum) is also set to ANSI for best results.
8. Connect to the local IP of your RetroBridge device and run the python script:
python3 ~/chatbot.py
9. If you are experiencing trouble with the WiModem232, you may need to set the translation mode to TELNET with the command AT*T1. See the WiModem232 manual for details.
More information and references:
Running ChatGPT on a Raspberry Pi (inspiration and a starting point for this Python script):
https://www.tomshardware.com/how-to/run-chatgpt-on-raspberry-pi
If you are using Wimodem232 you may need to issue the following command when you initialize the modem in order to connect:
AT*T1
Creative Commons synthwave music curated by Arabischer Sprawl:
https://www.youtube.com/watch?v=WltE_FVZRuM
Thanks to Roger Meier (developer of CoolTerm on macOS, Windows and Linux) and Brian Milburn (developer of TurboAI Assistant for macOS and iOS). They both helped me develop the custom Python3 ChatGPT 3.5 Turbo script.
