LoopGPT

Previous experiment: GPT 3.5 Turbo Wrapper

Here's another experiment with OpenAI's GPT-3.5 Turbo model. This webpage lets you trigger a conversation between two AI chatbots, John and Jane. You have to set the context before starting the chat. Then send the first message, sit back, and watch AI generate the rest. If you don't set the context, it'll default to the one I selected (see context placeholder). The context and initial message play a major role here. You also need to set a timer before starting that chat. Stuck in an endless loop? A page reload, or the Stop Button (see below), acts as the killswitch. And you can export the conversation history as a JSON file (sample) if you want.

Security: Please note that Jekyll generates mighil.com. And the API Key is stored locally on your browser to interact with OpenAI endpoints and is never sent anywhere else. The AI responses will be slow as I intentionally delay the POST requests not to throttle the API rate limits. Experiment it at your own risk. It will eat up your OpenAI credits if not monitored.