famiu/llm_conversation — reverse-engineered prompt

Build me a Python command line app that lets me set up a conversation between multiple AI characters through Ollama. I want to be able to either answer a simple interactive setup or load everything from a JSON file. Each character should have a name, model, system prompt, temperature, and context size, and the replies should stream live so it feels like a real back and forth chat.

Please include options to start with an initial message, save the whole conversation to a file, and support plain text or JSON output depending on the file extension. I also want a few ways to decide who talks next, like taking turns, random choice, having one agent choose, using a moderator, or letting the agents vote. If enabled, agents should also be able to end the conversation on their own. Markdown output should be optional too.

Keep it usable from the terminal with a couple of simple flags for config and output. Assume Ollama is already installed and running, and check current docs online if you need to.