A Python script to orchestrate a group of multiple ChatGPT instances.
- Python 3
- Libraries:
openai rich
Create a file called config.json
in the root directory.
The following example config contains all of the understood options,
but only the api key is required.
{
"api_key": "...",
"organization": "org-...",
"max_tokens": 1024,
"verbose": false
}
ChatGPT is good at conversation and reasoning, but it's very bad at planning. We can get around this by spawning multiple instances with different system prompts. For example, you might have one instance that converses with the user, and a second "supervisor" instance which reads the chat and writes notes, directions, and plans for the conversation instance.