GPT-4 CLI with persistence in 10 lines of code.

November 03, 2023 • Written by Jakob Serlier

Tags: LLM, short, gpt-4

A short one: I needed a GPT-4 CLI interface (that's a RAS Syndrome), but most options seemed quite cluttered, and I like code-golf. Here an implementation of GPT-4 including 10-message persistence (OS agnostic) in 10 lines of code:

def main():
    import openai, os, pickle
    from pathlib import Path
    c, h = (lambda m: openai.ChatCompletion.create(model='gpt-4', messages=m)), (lambda r, c: {"role": r, "content": c})
    m = pickle.load(open(Path(os.getenv('APPDATA', os.path.expanduser('~')),'ai_chat.pkl'), 'rb')) \
        if Path(os.getenv('APPDATA', os.path.expanduser('~')),'ai_chat.pkl').exists() else [h('system', 'You are an AI assitant')]

    while (i := input('You: ').strip()) != 'exit':
        print(f'AI: {(r:=c((m:=[*m,h("user",i)]))["choices"][0]["message"]["content"])}')
        m = [*m, h("assistant", r)][-10:] # Truncate to 10 messages
        pickle.dump(m, open(Path(os.getenv('APPDATA', os.path.expanduser('~')),'ai_chat.pkl'), 'wb'))


PS C:\Projects\Personal\xsgpt> xsgpt
You: Hello! My name is jakob
AI: Hello, Jakob! How can I assist you today?
You: exit
PS C:\Projects\Personal\xsgpt> xsgpt
You: What is my name?
AI: Your name is Jakob.

(Installable) packaged version and source found here.


Back to Home