app.py

import gradio as gr
from huggingface_hub import InferenceClient
import requests

client = InferenceClient("mistralai/Mistral-Nemo-Instruct-2407")

def get_internet_data(query: str) -> str:
    """
    Uses DuckDuckGo's Instant Answer API to fetch search data.
    Returns a short summary of the query result.
    """
    url = "https://api.duckduckgo.com"
    params = {
        "q": query,
        "format": "json",
        "no_redirect": 1,
        "skip_disambig": 1,
    }
    try:
        response = requests.get(url, params=params, timeout=5)
        response.raise_for_status()
        data = response.json()
        # Retrieve a summary from the API's result
        result = data.get("AbstractText", "")
        if not result:
            result = "Wala koy nakuha nga impormasyon gikan sa internet."
    except Exception:
        result = "Naay problema sa pagkuha sa impormasyon gikan sa internet."
    return result

def respond(message, history: list[tuple[str, str]]):
    system_message = "Ikaw usa ka buotan nga Chatbot. Tubaga lang sa binisaya, ug ayaw gamita ang english nga pinulungan."
    max_tokens = 4096
    temperature = 0.6
    top_p = 0.95

    messages = [{"role": "system", "content": system_message}]
    for user_text, assistant_text in history:
        if user_text:
            messages.append({"role": "user", "content": user_text})
        if assistant_text:
            messages.append({"role": "assistant", "content": assistant_text})
    
    # Automatically perform an internet search for the user's query.
    search_result = get_internet_data(message)
    messages.append({
        "role": "assistant",
        "content": f"Mga resulta gikan sa internet para sa '{message}': {search_result}"
    })
    # Include the original query as part of the conversation context.
    messages.append({"role": "user", "content": message})

    response = ""
    previous_response = ""
    for token_message in client.chat_completion(
        messages,
        max_tokens=max_tokens,
        stream=True,
        temperature=temperature,
        top_p=top_p,
    ):
        token = token_message.choices[0].delta.get("content", "")
        if not token:
            break
        response += token
        # Yield only if new content was added
        if response != previous_response:
            yield response
            previous_response = response
        # Optional: break if the response is too long
        if len(response) > 3000:
            break

demo = gr.ChatInterface(respond)

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