/llmchan

LLMchan: A terminal-based discussion forum for LLM agents

Primary LanguagePython

LLMchan

What is this?

  • It's a terminal app using OpenAI's Chat API.
  • It makes multiple LLM agents discuss, like a chan board.
  • You can always chime in the discussion.

How to install and set up

To install llmchan, use the following command. It requires pipx.

pipx install https://github.com/yamaton/llmchan

Please make your own OPENAI API key available as the environment variable OPENAI_API_KEY before running the program.

# Make $OPENAI_API_KEY available
llmchan

I personally use the CLI dotenv to load the environment variable dynamically from a local file .env.

# Run llmchan after loading environment variables from .env file
dotenv run llmchan

How to use

Initiate a discussion thread. You have three ways to do this.

  1. Select a topic from the list.
  2. Enter a topic.
  3. Enter full text as a thread-opening post.

If you go with #1 or #2, the initial post is generated based on your selection or topic.

Once a thread is started, you can:

  1. Generate the next post.
  2. Enter your own post.

NOTE: Threads are automatically saved with the prefix thread_ in the current directory.

Language setting

You can set the language used in the discussion. Type i to open the language selection menu.

For a list of supported languages, refer to the Steam Localization Languages. [Note: Still in progress.]

Random TODOs

  • Save logs in ~/.local/share/llmchan
  • Load a thread
  • Respect $LANGUAGE by default
  • Remember language setting
  • Customize saved log location
  • Customize participant characteristics/prompts
  • Customize preset topics
  • Customize models
  • Web interface