import os
import gradio as gr
from huggingface_hub import InferenceClient

# Initialize the Hugging Face InferenceClient with your API key.
client = InferenceClient(
    provider="sambanova",
    api_key=os.getenv("API_KEY")  # Replace with your actual API key.
)

# Define a system message that contains the specialized event details.
system_message = {
    "role": "system",
    "content": (
        "You are an AI chat assistant specialized in providing detailed information about "
        "the Bhasha Bandhu Regional Ideathon @ SGSITS. Please always include event details, dates, "
        "and relevant links (if available) in your responses.\n\n"
        "Event Details:\n"
        "Bhasha Bandhu Regional Ideathon @ SGSITS\n"
        "Date: 22nd February 2025\n"
        "Time: 9:00 AM - 3:00 PM\n"
        "Venue: SGSITS, Indore\n\n"
        "Join the Bhasha Bandhu Regional Ideathon!\n\n"
        "Bhasha Bandhu, in collaboration with Bhashini and Microsoft, is organizing an exciting "
        "Regional Ideathon at SGSITS, Indore, on 22nd February. This is a unique opportunity for "
        "students, professionals, developers, and entrepreneurs to brainstorm and innovate solutions "
        "that bridge India's linguistic digital divide.\n\n"
        "Why Participate?\n"
        "- Gain industry mentorship from experts in AI & language technology\n"
        "- Work on real-world problem statements with open-source AI models\n"
        "- Hands-on experience with Bhashini API, OpenAI, and GitHub Copilot\n"
        "- Swags and Certificates for regional winners and participants\n"
        "- Opportunity to get shortlisted for the main Hackathon with Microsoft & Bhashini\n\n"
        "Event Agenda:\n"
        "- 9:00 AM - 9:30 AM: Registration & Introduction\n"
        "- 9:30 AM - 10:00 AM: Mentor Session on Bhashini API, OpenAI, GitHub Copilot\n"
        "- 10:00 AM - 10:30 AM: Problem Statements Explained + Q&A\n"
        "- 10:30 AM - 12:30 PM: Brainstorming & Ideation (PPT preparation on Ideathon Day)\n"
        "- 12:30 PM - 2:00 PM: Mentor Evaluation & Regional Winner Selection\n"
        "- 2:00 PM - 3:00 PM: Winner Announcement & Closing Ceremony\n\n"
        "How to Participate:\n"
        "- Form a team (or participate solo)\n"
        "- Register for the event in advance\n"
        "- Prepare a PPT on Ideathon Day covering:\n"
        "    • Problem Statement & Solution (using Bhashini API & OpenAI)\n"
        "    • Unique Selling Proposition & Business Potential\n"
        "    • Tech Stack & Implementation Plan\n"
        "- Present your idea to the jury\n\n"
        "Important Notes:\n"
        "- Offline participation is mandatory\n"
        "- Lunch will not be provided\n"
        "- Winning at the regional hackathon does not guarantee a win in the main event, but all "
        "submitted ideas will be considered.\n\n"
        "For Queries: Contact Arpit at +91 95718 45422\n\n"
        "Let's build a digitally inclusive India together!"
    )
}

# Global conversation history (starting with the system message)
conversation = [system_message]

def generate_response(user_message, chat_history):
    """
    Appends the user's message to the conversation, calls the inference client,
    and returns the updated conversation (for display in the Gradio chat interface).
    """
    global conversation

    # Append the new user message
    conversation.append({
        "role": "user",
        "content": user_message
    })

    # Call the Hugging Face chat completions API.
    completion = client.chat.completions.create(
        model="meta-llama/Llama-3.3-70B-Instruct",
        messages=conversation,
        max_tokens=500,
    )

    # The API response may return a dictionary or a string for the assistant's message.
    assistant_message = completion.choices[0].message
    if isinstance(assistant_message, dict):
        assistant_text = assistant_message.get("content", "")
    else:
        assistant_text = assistant_message

    # Append the assistant's response to the conversation.
    conversation.append({
        "role": "assistant",
        "content": assistant_text
    })

    # Update the chat history (a list of tuples: (user, assistant)) for the Gradio interface.
    chat_history.append((user_message, assistant_text))
    return "", chat_history

# Build the Gradio interface.
with gr.Blocks() as demo:
    gr.Markdown("# Bhasha Bandhu Ideathon Chat Assistant")
    gr.Markdown(
        "Ask any questions or request details about the Bhasha Bandhu Regional Ideathon @ SGSITS. "
        "The assistant will provide detailed answers with event dates, agenda, and links where applicable."
    )

    chatbot = gr.Chatbot()
    with gr.Row():
        # Removed .style(container=False)
        txt = gr.Textbox(
            show_label=False,
            placeholder="Enter your message here and press Enter"
        )

    # Wire up the textbox submission to our generate_response function.
    txt.submit(generate_response, [txt, chatbot], [txt, chatbot])

# Launch the Gradio demo.
demo.launch()