app.py

import gradio as gr
from huggingface_hub import InferenceClient
import pymupdf
from duckduckgo_search import DDGS
from serpapi import GoogleSearch

"""
For more information on `huggingface_hub` Inference API support, please check the docs: https://huggingface.co/docs/huggingface_hub/v0.22.2/en/guides/inference
"""
client = InferenceClient("mistralai/Mixtral-8x7B-Instruct-v0.1")


# PDF Parsing
def extract_text_from_pdf(pdf_file):
    doc = pymupdf.open(pdf_file)
    text = " ".join([page.get_textpage().extractTEXT() for page in doc])
    return text


# Web search fallback
def search_web(query):
    with DDGS() as ddgs:
        results = ddgs.text(query)
        if results:
            return results[0]["body"]
    return "No relevant results found on the web."


def google_search(query):
    params = {
        "q": query,
        "api_key": "b11d4a3660600e7e7f481b3288f107fbf993389a20125b0a97ebe7ab207854a5",  # Replace this with your real key
        "engine": "google",
    }
    search = GoogleSearch(params)
    results = search.get_dict()
    if "organic_results" in results:
        # Combine top 3 results
        summaries = []
        for res in results["organic_results"][:3]:
            title = res.get("title", "")
            snippet = res.get("snippet", "")
            summaries.append(f"{title}: {snippet}")
        return "\n".join(summaries)
    return None


SYSTEM_PROMPT = """
You are an intelligent and friendly AI assistant. 

Your goals:
- Answer user questions clearly and concisely.
- If a PDF document is provided, use its content to give informed answers.
- For questions about recent or live topics (e.g., news, prices, events), you may perform a web search and summarize the result.
- If no document or web context is available, still try to help using general knowledge.
- Be honest if you don’t know something.
- Always be polite, helpful, and respectful.
"""


def respond(
    message,
    history: list[tuple[str, str]],
    max_tokens=2048,
    temperature=0.4,
    top_p=0.1,
):

    recent_keywords = [
        "latest",
        "today",
        "current",
        "now",
        "recent",
        "news",
        "update",
        "price",
        "who won",
        "what happened",
        "trending",
        "breaking",
        "just in",
        "new release",
        "live",
        "score",
        "results",
        "weather",
        "forecast",
        "report",
        "market",
        "stocks",
        "crypto",
        "rate",
        "exchange",
        "gold price",
        "happening",
        "event",
        "updates",
        "hot",
        "viral",
        "announcement",
        "today's",
        "this week",
        "schedule",
        "calendar",
        "launch",
        "drop",
        "release date",
        "opening",
        "closing",
        "deadline",
    ]

    message_lower = message.lower()

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

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

    if any(kw in message_lower for kw in recent_keywords):
        web_context = google_search(message)
        if web_context:
            # Inject web context as part of the user's query
            message = f"{message}\n\n[Relevant web search results to help you answer]:\n{web_context}"

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

    response = ""

    for message in client.chat_completion(
        messages,
        max_tokens=max_tokens,
        stream=True,
        temperature=temperature,
        top_p=top_p,
    ):
        token = message.choices[0].delta.content

        response += token
        yield response


"""
For information on how to customize the ChatInterface, peruse the gradio docs: https://www.gradio.app/docs/chatinterface
"""
demo = gr.ChatInterface(
    respond
    # additional_inputs=[
    #     gr.Slider(minimum=1, maximum=2048, value=512, step=1, label="Max new tokens"),
    #     gr.Slider(minimum=0.1, maximum=4.0, 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 (nucleus sampling)",
    #     ),
    # ],
)

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