import os, openai, requests, json, subprocess

# 1️⃣  Generate lyrics
lyrics = openai.ChatCompletion.create(
    model="gpt-4-turbo",
    messages=[{"role":"user","content":"Write a 3‑verse pop lyric about a neon sunrise"}]
).choices[0].message.content

# 2️⃣  Create a melody (10 seconds demo)
music_resp = requests.post(
    "https://api.google.com/musiclm/generate",
    headers={"Authorization": f"Bearer {os.getenv('GOOGLE_API_KEY')}"},
    json={"prompt":"upbeat synth pop, 120 BPM", "duration_sec":10}
)
melody_path = "melody.wav"
with open(melody_path, "wb") as f:
    f.write(music_resp.content)

# 3️⃣  Synthesize vocals
vocal_resp = openai.Audio.synthesize(
    model="vocalizer-1",
    text=lyrics,
    voice="female_pop"
)
vocal_path = "vocals.wav"
with open(vocal_path, "wb") as f:
    f.write(vocal_resp)

# 4️⃣  Mix & master with Ozone AI (CLI must be installed)
subprocess.run([
    "ozone-cli", "mix",
    "--input", melody_path,
    "--input", vocal_path,
    "--output", "final_mix.wav",
    "--target-lufs", "-14"
])

print("✅ AI‑generated demo ready: final_mix.wav")

import pandas as pd, requests, time, os
from datetime import datetime

SLACK_WEBHOOK = os.getenv("SLACK_WEBHOOK")
GOOGLE_TRENDS_API = "https://api.trends.google.com/v1/queries"

def fetch_trends():
    r = requests.get(GOOGLE_TRENDS_API, params={"q":"AI music","geo":"global","hl":"en"})
    return r.json()["topQueries"]   # list of dicts {keyword, score}

def load_history():
    if os.path.exists("trends_history.csv"):
        return pd.read_csv("trends_history.csv")
    return pd.DataFrame(columns=["timestamp","keyword","score"])

def alert(keyword, pct):
    payload = {"text":f":musical_note: *{keyword}* spiked **{pct:.1f}%** in the last hour!"}
    requests.post(SLACK_WEBHOOK, json=payload)

history = load_history()

while True:
    now = datetime.utcnow()
    trends = fetch_trends()
    for q in trends[:10]:
        row = {"timestamp":now, "keyword":q["keyword"], "score":q["score"]}
        history = pd.concat([history, pd.DataFrame([row])], ignore_index=True)

        # compare with previous hour
        prev = history[(history["keyword"]==q["keyword"]) &
                       (history["timestamp"] < now) &
                       (history["timestamp"] >= now - pd.Timedelta(hours=1))]
        if not prev.empty:
            pct_change = (q["score"] - prev["score"].iloc[-1]) / prev["score"].iloc[-1] * 100
            if pct_change > 30:
                alert(q["keyword"], pct_change)

    history.to_csv("trends_history.csv", index=False)
    time.sleep(3600)   # wait an hour