This project demonstrates a closed-loop LLM dialogue, where two language models alternate turns and generate conversation without any human participation after initialization.
The system uses a shared base prompt with persona substitution and feeds each model turn back into the context, creating an autonomous conversational loop.
Designed as a single-file, local-first experiment using LM Studio’s OpenAI-compatible server.
This project is an experimental demo. It does not include safeguards, content filtering, or guarantees about model output. Use at your own risk.