app.py

import pandas as pd
import numpy as np
import gradio as gr
from sklearn.metrics.pairwise import euclidean_distances
import openai

# --- Set your OpenAI API key ---
openai.api_key = "YOUR_API_KEY"  # replace with your key

# --- Load your CSV ---
# Ensure your CSV has columns: 'song', 'artist', 'bpm', 'nrgy', 'dnce', 'dB', 'live', 'val', 'dur', 'acous', 'spch', 'pop'
df = pd.read_csv("datalab_export_2025-08-11 14_16_35.csv")
feature_cols = ['bpm', 'nrgy', 'dnce', 'dB', 'live', 'val', 'dur', 'acous', 'spch', 'pop']
df_features = df[feature_cols].astype(float)

# --- Song recommendation function ---
def recommend_song(user_input):
    # Try to find a song or artist mentioned
    user_input_lower = user_input.lower()
    matched_row = None
    for _, row in df.iterrows():
        if row['song'].lower() in user_input_lower or row['artist'].lower() in user_input_lower:
            matched_row = row
            break

    # If a song/artist is found, use its features; otherwise use dataset mean
    if matched_row is not None:
        user_features = df_features.loc[matched_row.name].values.reshape(1, -1)
    else:
        user_features = df_features.mean().values.reshape(1, -1)

    distances = euclidean_distances(df_features.values, user_features)
    top5_idx = np.argsort(distances.flatten())[:5]
    top5 = df.iloc[top5_idx][['artist', 'song']]
    
    output_lines = [f"{row['artist']} - {row['song']}" for _, row in top5.iterrows()]
    return "\n".join(output_lines)

# --- Chat model using OpenAI GPT with memory ---
def chat_model(user_input, chat_history):
    try:
        messages = [{"role": "system", "content": "You are a friendly chatbot."}]
        # Include previous conversation for context
        for user_msg, bot_msg in chat_history:
            messages.append({"role": "user", "content": user_msg})
            messages.append({"role": "assistant", "content": bot_msg})
        # Add the current user input
        messages.append({"role": "user", "content": user_input})

        response = openai.ChatCompletion.create(
            model="gpt-4o-mini",
            messages=messages,
            temperature=0.7
        )
        return response['choices'][0]['message']['content']
    except Exception as e:
        return f"Error contacting LLM: {e}"

# --- Main chat handler ---
def respond(message, chat_history):
    input_lower = message.lower()
    if any(keyword in input_lower for keyword in ["recommend", "song", "music", "suggest"]):
        recommendation = recommend_song(message)
        chat_history.append((message, recommendation))
        return recommendation, chat_history
    else:
        reply = chat_model(message, chat_history)
        chat_history.append((message, reply))
        return reply, chat_history

# --- Gradio ChatInterface ---
demo = gr.ChatInterface(
    fn=respond,
    title="Smart Music Chatbot",
    description="Chat normally with the AI, or ask for song recommendations."
)

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