backend/app.py

import asyncio
from fastapi import FastAPI, WebSocket
from fastapi.staticfiles import StaticFiles
from fastapi.responses import FileResponse
import uvicorn
import json
from transformers import pipeline
from collections import deque
from collections import defaultdict
import math
import sys
import random
import os

sys.path.append(".")

app = FastAPI()

app.mount("/static", StaticFiles(directory="../frontend"), name="static")


# 🔹 Load Emotion Detection Model
emotion_classifier = pipeline(
    "zero-shot-classification",
    model="MarfinF/marfin_emotion",
    framework="pt"
)

# 🔹 Emotion-to-Mood Mapping
emotion_to_mood = {
    "senang": "happy",
    "sedih": "sad",
    "marah": "excited",
    "takut": "relaxed",
    "cinta": "romantic"
}

# 🔹 WebSocket Clients
clients = {}
chat_history = deque(maxlen=4)

mood_to_genre = {
    "happy": "pop",
    "sad": "acoustic",
    "excited": "rock",
    "intense": "cinematic",
    "romantic": "rnb",
    "chill": "chill"
}

# 🔹 Detect Emotion
def detect_emotion(text):
    labels = ["takut", "marah", "sedih", "senang", "cinta"]
    result = emotion_classifier(text, candidate_labels=labels)
    top_emotion = result['labels'][0]
    top_scores = result['scores'][0]
    return top_emotion, top_scores

# 🔹 Get Music Recommendations
def get_recommendations_by_mood(mood):
    genre_folder = mood_to_genre.get(mood, "pop")
    folder_path = f"../music/{genre_folder}"
    print("folder path")
    print(folder_path)

    # Check if folder exists
    if not os.path.exists(folder_path):
        return []
    print("folder exist")

    # List and shuffle songs
    songs = [f"../music/{genre_folder}/{song}" for song in os.listdir(folder_path) if song.endswith(".mp3")]
    random.shuffle(songs)
    return songs[:3]  # Return top 3 shuffled songs

def softmax(scores):
    exp_scores = [math.exp(score) for score in scores]
    total = sum(exp_scores)
    return [exp_score / total for exp_score in exp_scores]

# 🔹 Broadcast User List
async def broadcast_user_list():
    user_list = list(clients.keys())
    message = json.dumps({
        "type": "user_list",
        "users": user_list
    })
    for client in clients.values():
        await client.send_text(message)

# 🔹 Periodic Music Recommender every 30 seconds
async def periodic_recommendation():
    while True:
        user_list = list(clients.keys())
        if len(user_list) >= 2:
            await asyncio.sleep(10)
            if clients:  # Only run if someone is connected
                if len(chat_history) > 0:
                    # 1. Detect emotion dan ambil (label, score)
                    print("chat history")
                    print(chat_history)
                    emotions = [detect_emotion(msg) for msg in chat_history]
                    print("Detected Emotions:", emotions)

                    # 2. Group by emotion + sum score
                    emotion_score_sum = defaultdict(float)
                    for label, score in emotions:
                        emotion_score_sum[label] += score

                    # 3. Softmax
                    labels = list(emotion_score_sum.keys())
                    scores = list(emotion_score_sum.values())
                    softmax_scores = softmax(scores)

                    # 4. Pair label + softmax_score
                    softmax_result = list(zip(labels, softmax_scores))
                    print("Softmax Result:", softmax_result)

                    # 5. Dominant emotion
                    most_common_emotion = max(softmax_result, key=lambda x: x[1])[0]
                    print("Dominant Emotion:", most_common_emotion)
                    mood = emotion_to_mood.get(most_common_emotion, "chill")
                    music_recommendations = get_recommendations_by_mood(mood)
                else:
                    music_recommendations = ["chill"]  # default if no chat

                recommendation_response = {
                    "recommendations": music_recommendations,
                    "genre": mood_to_genre.get(mood, "pop")
                }

                for client in clients.values():
                    await client.send_text(json.dumps(recommendation_response))
        else:
            await asyncio.sleep(2)
            await broadcast_user_list()

# 🔹 Start periodic task
@app.on_event("startup")
async def start_recommender():
    asyncio.create_task(periodic_recommendation())

# 🔹 WebSocket Endpoint
@app.websocket("/chat/{username}")
async def chat_endpoint(websocket: WebSocket, username: str):
    await websocket.accept()
    clients[username] = websocket
    print(f"{username} joined")

    await broadcast_user_list()

    try:
        while True:
            data = await websocket.receive_text()
            message_data = json.loads(data)

            chat_history.append(message_data["message"])
            response = {
                "username": message_data["username"],
                "message": message_data["message"]
            }

            # Broadcast message to all clients
            for client in clients.values():
                await client.send_text(json.dumps(response))

    except Exception as e:
        print(f"{username} disconnected: {e}")
        del clients[username]
        await broadcast_user_list()

@app.get("/")
def read_root():
    return FileResponse("../frontend/index.html")

if __name__ == "__main__":
    uvicorn.run(app, host="0.0.0.0", port=7860)