app.py

import gradio as gr
from huggingface_hub import InferenceClient
import os

# 🔹 Load HF token from Space Secrets
HF_TOKEN = os.environ.get('telemedpro')

# 🔹 Fixed persona system message
SYSTEM_MESSAGE = (
    "You are Dr. Alex, a highly knowledgeable yet empathetic doctor. "
    "You always provide clear, safe, and well-structured medical advice in simple language. "
    "You avoid making unsafe claims and encourage users to seek professional help when needed. "
    "You behave politely, patiently, and with care, like a trusted family doctor."
)

# 🔹 Initialize InferenceClient once
client = InferenceClient(token=HF_TOKEN, model="m42-health/Llama3-Med42-70B")

# 🔹 Respond function
def respond(message, history, system_message=SYSTEM_MESSAGE, max_tokens=512, temperature=0.7, top_p=0.95):
    try:
        # Start with system message
        messages = [{"role": "system", "content": system_message}]

        # Append previous conversation safely
        if history:
            messages.extend(history)  # ✅ safe, don't manipulate

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

        # Stream model output
        response = ""
        for msg in client.chat_completion(
            messages,
            max_tokens=max_tokens,
            stream=True,
            temperature=temperature,
            top_p=top_p,
        ):
            if msg.choices and hasattr(msg.choices[0].delta, "content") and msg.choices[0].delta.content:
                token = msg.choices[0].delta.content
                response += token
                yield response

    except Exception as e:
        yield f"⚠️ Space error: {e}"

# 🔹 Gradio Chat Interface
chatbot = gr.ChatInterface(
    fn=respond,
    type="messages",
    additional_inputs=[
        gr.Textbox(value=SYSTEM_MESSAGE, label="System message"),
        gr.Slider(minimum=1, maximum=2048, value=512, step=1, label="Max tokens"),
        gr.Slider(minimum=0.1, maximum=2.0, value=0.7, step=0.1, label="Temperature"),
        gr.Slider(minimum=0.1, maximum=1.0, value=0.95, step=0.05, label="Top-p"),
    ],
)

# 🔹 Layout
with gr.Blocks() as demo:
    gr.Markdown("## 🩺 AI Health Mentor — Dr. Alex")
    chatbot.render()

# 🔹 Launch
if __name__ == "__main__":
    demo.launch(show_error=True)