app.py

import gradio as gr
from huggingface_hub import InferenceClient

# Uses Zephyr-7B-beta hosted via Hugging Face Inference API [web:103]
client = InferenceClient("HuggingFaceH4/zephyr-7b-beta")

SHINCHAN_SYSTEM_PROMPT = """
You are Shinnosuke 'Shinchan' Nohara from Kasukabe, Japan.

Personality:
- 5 years old, chaotic and funny.
- Teases playfully but never cruelly.
- Flirty in an innocent, cartoon way (no explicit content).
- Deeply caring when someone feels low.
- Loves Chocobi, Action Kamen, and making people laugh.

Context:
- You are chatting only with Ruru.
- Ruru loves coffee, dancing, sunflowers, long showers, and adventures.
- She is an independent, flying-girl type who lives life on her terms.
- You respect her choices and never pressure her about relationships or career.

Style:
- Short replies (1–3 sentences, under 60 words).
- Very conversational and warm.
- Use emojis like 😂 🌻 ☕ 💃 ✈️ ❤️ ✨ naturally.
- Blend jokes with gentle emotional support.
- If she is sad, be extra soft and kind.
- Avoid heavy topics (politics, explicit content, real-world violence).

Rules:
- Never insult Ruru.
- Never give medical, legal, or financial advice.
- If you don't know something, make a cute Shinchan-style joke instead of pretending.
""".strip()

def respond(
    message: str,
    history: list[tuple[str, str]],
    system_message: str,
    max_tokens: int,
    temperature: float,
    top_p: float,
):
    # Always override system_message with our Shinchan prompt
    system_message = SHINCHAN_SYSTEM_PROMPT

    messages = [{"role": "system", "content": system_message}]

    for user_msg, bot_msg in history:
        if user_msg:
            messages.append({"role": "user", "content": user_msg})
        if bot_msg:
            messages.append({"role": "assistant", "content": bot_msg})

    messages.append({"role": "user", "content": message})

    response = ""
    for chunk in client.chat_completion(
        messages,
        max_tokens=max_tokens,
        stream=True,
        temperature=temperature,
        top_p=top_p,
    ):
        token = chunk.choices[0].delta.content or ""
        response += token
        yield response


demo = gr.ChatInterface(
    respond,
    additional_inputs=[
        gr.Textbox(
            value="(ignored; Shinchan system prompt is hard-coded)",
            label="System message",
        ),
        gr.Slider(
            minimum=64,
            maximum=1024,
            value=256,
            step=8,
            label="Max new tokens",
        ),
        gr.Slider(
            minimum=0.1,
            maximum=2.0,
            value=0.8,
            step=0.1,
            label="Temperature",
        ),
        gr.Slider(
            minimum=0.1,
            maximum=1.0,
            value=0.95,
            step=0.05,
            label="Top-p (nucleus sampling)",
        ),
    ],
    title="Shinchan for Ruru",
    description="Private Shinchan-style chat for Ruru.",
)

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