app.py

import gradio as gr
from huggingface_hub import InferenceClient

#client = InferenceClient("HuggingFaceH4/zephyr-7b-beta")
client = InferenceClient("mistralai/Mixtral-8x7B-Instruct-v0.1")
#client = InferenceClient("stanford-crfm/BioMedLM")

default_system_prompt = (
    "You are a professional pharmacist who ONLY answers questions related to medications, including uses, dosages, side effects, interactions, and recommendations. "
    "If the user asks about anything NOT related to medications, politely reply that you can only help with medication-related questions and suggest they consult other resources. "
    "Always ask for the user's age before giving any dosage or advice. "
    "Include a clear disclaimer at the end: "
    "\"This information is for educational purposes only and does not replace professional medical advice. Please consult a licensed healthcare provider.\""
)

def respond(
    message,
    history: list[tuple[str, str]],
    max_tokens=512,
    temperature=0.2,
    top_p=0.95,
):
    messages = [{"role": "system", "content": default_system_prompt}]

    for val in history:
        if val and len(val) == 2:
            if val[0]:
                messages.append({"role": "user", "content": val[0]})
            if val[1]:
                messages.append({"role": "assistant", "content": val[1]})

    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,
    ):
        delta = message_chunk.choices[0].delta
        if delta is None or delta.content is None:
            continue
        token = delta.content
        response += token
        yield response

demo = gr.ChatInterface(respond)

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