app.py

import gradio as gr
from openai import OpenAI
import os
import json

css = '''
.gradio-container{max-width: 1000px !important}
h1{text-align:center}
footer {
    visibility: hidden
}
'''

# Access token for Hugging Face
ACCESS_TOKEN = os.getenv("HF_TOKEN")

# Initialize the client for the OpenAI model
client = OpenAI(
    base_url="https://api-inference.huggingface.co/v1/",
    api_key=ACCESS_TOKEN,
)

# File path for storing user preferences
USER_DATA_PATH = "user_data.json"

# Load user preferences if they exist
def load_user_preferences():
    if os.path.exists(USER_DATA_PATH):
        with open(USER_DATA_PATH, "r") as file:
            return json.load(file)
    return {}

# Save user preferences
def save_user_preferences(data):
    with open(USER_DATA_PATH, "w") as file:
        json.dump(data, file)

# Respond function that generates the assistant's reply
def respond(
    message,
    history: list[tuple[str, str]],
    system_message,
    max_tokens,
    temperature,
    top_p,
):
    # Load user preferences
    user_data = load_user_preferences()

    # Custom welcome message or save user input
    if message.lower().startswith("my name is"):
        user_data["name"] = message.split("is")[-1].strip()
        save_user_preferences(user_data)
        response = f"Nice to meet you, {user_data['name']}! How can I assist you with your travel plans today?"
        yield response
        return

    if message.lower().startswith("i like to travel to"):
        user_data["favorite_destination"] = message.split("to")[-1].strip()
        save_user_preferences(user_data)
        response = f"Got it! I noted that you enjoy traveling to {user_data['favorite_destination']}."
        yield response
        return

    if message.lower().startswith("my budget is"):
        user_data["budget"] = message.split("is")[-1].strip()
        save_user_preferences(user_data)
        response = f"Understood! I'll keep your budget of {user_data['budget']} in mind when suggesting travel options."
        yield response
        return

    # Use user's name and preferences in responses if available
    name = user_data.get("name", "Traveler")
    favorite_destination = user_data.get("favorite_destination", "various places")
    budget = user_data.get("budget", "not specified")

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

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

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

    response = f"Hello {name}! You mentioned you like traveling to {favorite_destination}. Let's plan something exciting within your budget of {budget}.\n"

    # Generate a response using the OpenAI client
    for message in client.chat.completions.create(
        model="meta-llama/Meta-Llama-3.1-8B-Instruct",
        max_tokens=max_tokens,
        stream=True,
        temperature=temperature,
        top_p=top_p,
        messages=messages,
    ):
        token = message.choices[0].delta.content
        response += token
        yield response

# Gradio interface
demo = gr.ChatInterface(
    respond,
    additional_inputs=[
        gr.Textbox(
            value="You are a friendly travel assistant. Offer personalized travel tips and remember user preferences.",
            label="System message"
        ),
        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",
        ),
    ],
    css=css,
    theme="allenai/gradio-theme",
)

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