- 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.
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
Initiate a discussion thread. You have three ways to do this.
- Select a topic from the list.
- Enter a topic.
- 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:
- Generate the next post.
- Enter your own post.
NOTE: Threads are automatically saved with the prefix thread_
in the current directory.
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.]
- 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