#!/usr/bin/env python3
"""
Generiert alle Voice-Line-MP3s via OpenAI TTS API.

Liest App/sims/logistik/assets/audio/voice-lines.json und legt pro Eintrag
zwei Dateien an: <id>.mp3 (std) und <id>_easy.mp3 (easy).

Aufruf:
    set -a; source App/.env.local; set +a
    python App/sims/logistik/scripts/generate-voice-lines.py

Vorhandene Dateien werden uebersprungen (FORCE=1 zum Ueberschreiben).
"""
import json
import os
import ssl
import sys
import time
import urllib.request
from pathlib import Path

SCRIPT_DIR = Path(__file__).resolve().parent
LINES_JSON = SCRIPT_DIR.parent / "assets" / "audio" / "voice-lines.json"
OUT_DIR    = SCRIPT_DIR.parent / "assets" / "audio" / "voice" / "de"

API_KEY = os.environ.get("OPENAI_API_KEY")
if not API_KEY:
    print("FEHLER: OPENAI_API_KEY muss als Env-Variable gesetzt sein.", file=sys.stderr)
    sys.exit(1)

FORCE = bool(os.environ.get("FORCE"))

SSL_CTX = ssl.create_default_context()
SSL_CTX.check_hostname = False
SSL_CTX.verify_mode = ssl.CERT_NONE


def tts(text, voice="nova", model="tts-1", speed=1.0):
    """Ruft OpenAI TTS auf, gibt mp3-bytes zurueck."""
    payload = json.dumps({
        "model": model,
        "voice": voice,
        "input": text,
        "speed": speed,
        "response_format": "mp3",
    }).encode("utf-8")
    req = urllib.request.Request(
        "https://api.openai.com/v1/audio/speech",
        data=payload,
        headers={
            "Authorization": "Bearer " + API_KEY,
            "Content-Type": "application/json",
        },
    )
    with urllib.request.urlopen(req, timeout=60, context=SSL_CTX) as resp:
        return resp.read()


def main():
    OUT_DIR.mkdir(parents=True, exist_ok=True)
    data = json.loads(LINES_JSON.read_text(encoding="utf-8"))
    voice = data.get("_voice", "nova")
    model = data.get("_model", "tts-1")
    lines = data["lines"]

    count_new = 0
    count_skip = 0
    total_chars = 0

    for entry in lines:
        lid = entry["id"]
        for variant, text_key, speed in (("", "std", 1.0), ("_easy", "easy", 0.88)):
            text = entry.get(text_key)
            if not text:
                continue
            fname = f"{lid}{variant}.mp3"
            out_path = OUT_DIR / fname
            if out_path.exists() and not FORCE:
                count_skip += 1
                print(f"ok {fname} (schon da)")
                continue
            total_chars += len(text)
            print(f"-> {fname}: {len(text)} Zeichen, speed={speed} ...", end="", flush=True)
            try:
                mp3 = tts(text, voice=voice, model=model, speed=speed)
            except Exception as e:
                print(f" FEHLER: {e}")
                continue
            out_path.write_bytes(mp3)
            size_kb = len(mp3) / 1024
            print(f" {size_kb:.0f} KB")
            count_new += 1
            time.sleep(0.35)   # sanftes Rate-Limit

    # tts-1 Kosten: $15 / 1M Zeichen (2024)
    cost_eur = total_chars / 1_000_000 * 15 * 0.92   # USD -> EUR approx
    print(f"\nFertig: {count_new} neu, {count_skip} uebersprungen.")
    print(f"Zeichen neu erzeugt: {total_chars} (~{cost_eur:.3f} EUR)")


if __name__ == "__main__":
    main()