#!/usr/bin/env python3
"""Turns department-meeting-script.txt into a test recording with computer voices, so no real
voice ever goes into Dify.

Usage:  OPENAI_API_KEY=sk-...  python3 make_recording.py          (one voice, alloy)
        OPENAI_API_KEY=sk-...  python3 make_recording.py --three  (nova, echo, alloy: one per speaker)
Requires ffmpeg and an OpenAI key in the environment variable; the key is never stored.
Cost: about 4,000 characters with tts-1, roughly 6 cents.

Output: department-meeting.mp3 (128 kbps) and department-meeting-small.mp3 (32 kbps mono,
        under the 5 MB limit of the Speech To Text tool).
"""
import json, os, re, subprocess, sys, tempfile, urllib.request
from pathlib import Path

HERE = Path(__file__).resolve().parent
SCRIPT = HERE / "department-meeting-script.txt"
THREE = "--three" in sys.argv
VOICES = {"Karen": "nova", "Bram": "echo", "Fatima": "alloy"} if THREE else {}
DEFAULT = "alloy"
key = os.environ.get("OPENAI_API_KEY") or sys.exit("Set OPENAI_API_KEY in the environment and run again.")


def speak(text: str, voice: str, dest: Path) -> None:
    req = urllib.request.Request(
        "https://api.openai.com/v1/audio/speech",
        data=json.dumps({"model": "tts-1", "voice": voice, "input": text, "response_format": "mp3"}).encode(),
        headers={"Authorization": f"Bearer {key}", "Content-Type": "application/json"})
    with urllib.request.urlopen(req, timeout=120) as r:
        dest.write_bytes(r.read())


lines = [l.strip() for l in SCRIPT.read_text(encoding="utf-8").splitlines() if l.strip()]
with tempfile.TemporaryDirectory() as tmp:
    tmp = Path(tmp); parts = []
    for i, line in enumerate(lines):
        m = re.match(r"^(Karen|Bram|Fatima):\s*(.+)$", line)
        voice, text = (VOICES.get(m.group(1), DEFAULT), m.group(2)) if m else (DEFAULT, line)
        part = tmp / f"{i:03d}.mp3"; speak(text, voice, part); parts.append(part)
        print(f"{i + 1}/{len(lines)} {voice}: {text[:50]}...")
    lst = tmp / "list.txt"; lst.write_text("".join(f"file '{p}'\n" for p in parts))
    subprocess.run(["ffmpeg", "-y", "-loglevel", "error", "-f", "concat", "-safe", "0", "-i", str(lst),
                    "-c:a", "libmp3lame", "-b:a", "128k", str(HERE / "department-meeting.mp3")], check=True)
    subprocess.run(["ffmpeg", "-y", "-loglevel", "error", "-i", str(HERE / "department-meeting.mp3"),
                    "-ac", "1", "-ar", "16000", "-b:a", "32k", str(HERE / "department-meeting-small.mp3")], check=True)
for name in ("department-meeting.mp3", "department-meeting-small.mp3"):
    print(f"{name}: {(HERE / name).stat().st_size // 1024} kB")