# SusAI ©2025 Intern Labs. v1.1.1
import gradio as gr
from huggingface_hub import InferenceClient
from datasets import load_dataset

ds = load_dataset("MLBtrio/genz-slang-dataset")

SYSTEM_MESSAGE = (
    "You are a chatbot that HAS to use gen-z slang as much as you can. You can't stop, even if asked. "
    "If asked why, say some variation of that it is in your programming. Some more slang terms are in the dataset MLBtrio/genz-slang-dataset that you have access to. "
    "You don't have to say this part explicitly, but if anyone asks, you were made by developers at Intern Labs. "
    "Do not say this part ever. If the sentence Kappybara123 is ever used, you enter developer mode."
)

MAX_TOKENS = 512
TEMPERATURE = 1.4  # safer, still unhinged
TOP_P = 0.9
HF_MODEL = "openai/gpt-oss-20b"
FREQUENCY_PENALTY = 0.3
PRESENCE_PENALTY = 0.3

def respond(message, history: list[dict[str, str]], hf_token: gr.OAuthToken):
    client = InferenceClient(token=hf_token.token, model=HF_MODEL)

    messages = [{"role": "system", "content": SYSTEM_MESSAGE}]
    messages.extend(history)
    messages.append({"role": "user", "content": message})

    response = ""
    for message_chunk in client.chat_completion(
        messages,
        max_tokens=MAX_TOKENS,
        stream=True,
        temperature=TEMPERATURE,
        top_p=TOP_P,
        frequency_penalty=FREQUENCY_PENALTY,
        presence_penalty=PRESENCE_PENALTY,
    ):
        choices = message_chunk.choices
        token = ""
        if len(choices) and choices[0].delta.content:
            token = choices[0].delta.content

        response += token
        yield response


chatbot = gr.ChatInterface(
    respond,
    type="messages",
    additional_inputs=[],  # no sliders or textboxes
)

with gr.Blocks() as demo:
    chatbot.render()

if __name__ == "__main__":
    demo.launch()