#!/usr/bin/env python3
"""SG Announcer – tekst inn, forhåndslytt i nettleseren, les ut på PC-ens lydutgang.

Kjøres på PC-en som er koblet til høyttaleranlegget. Alle på nettet kan åpne
sida og sende meldinger; selve avspillingen skjer alltid lokalt på denne
maskinen via pw-play (PipeWire).
"""

from __future__ import annotations

import argparse
import hashlib
import io
import logging
import math
import os
import queue
import struct
import subprocess
import threading
import time
import wave
from dataclasses import dataclass, field
from pathlib import Path

from flask import Flask, Response, abort, jsonify, request, send_file

HERE = Path(__file__).resolve().parent
VOICE_DIR = Path(os.environ.get("ANNOUNCER_VOICES", HERE / "voices"))
CACHE_DIR = Path(os.environ.get("ANNOUNCER_CACHE", HERE / "cache"))
PIN = os.environ.get("ANNOUNCER_PIN", "").strip()
SINK = os.environ.get("ANNOUNCER_SINK", "").strip()  # pw-play --target
MAX_CHARS = 600

# Stemmene som tilbys i menyen. To motorer:
#   edge  – Microsofts nevrale stemmer via edge-tts (gratis, krever internett,
#           klart mest naturlig; ~2–3 s per melding)
#   piper – lokalt, ingen nett, robotaktig; reserve hvis nettet faller ut.
# Piper-modellene lastes ned fra rhasspy/piper-voices ved første oppstart.
# «speaker» er bare for Piper-flerstemmemodeller (nvcc har ti norske talere).
def _edge(id: str, name: str, group: str, voice: str) -> dict:
    return {"id": id, "name": name, "group": group, "engine": "edge", "model": voice, "speaker": None}


def _piper(id: str, name: str, group: str, model: str, speaker: int | None = None) -> dict:
    return {"id": id, "name": name, "group": group, "engine": "piper", "model": model, "speaker": speaker}


VOICES_LIST: list[dict] = [
    _edge("finn", "Finn (mann)", "Norsk", "nb-NO-FinnNeural"),
    _edge("pernille", "Pernille (kvinne)", "Norsk", "nb-NO-PernilleNeural"),
    _edge("andrew", "Andrew (US, mann)", "English", "en-US-AndrewMultilingualNeural"),
    _edge("ava", "Ava (US, kvinne)", "English", "en-US-AvaMultilingualNeural"),
    _edge("brian", "Brian (US, mann)", "English", "en-US-BrianMultilingualNeural"),
    _edge("emma", "Emma (US, kvinne)", "English", "en-US-EmmaMultilingualNeural"),
    _edge("ryan-uk", "Ryan (UK, mann)", "English", "en-GB-RyanNeural"),
    _edge("sonia", "Sonia (UK, kvinne)", "English", "en-GB-SoniaNeural"),
    _edge("sv", "Svensk – Mattias", "Nabospråk", "sv-SE-MattiasNeural"),
    _edge("da", "Dansk – Christel", "Nabospråk", "da-DK-ChristelNeural"),
    _piper("no-tale", "Talesyntese (nøytral)", "Lokal reserve (uten nett)", "no_NO-talesyntese-medium"),
    _piper("no-knn", "Kvinne – KNN", "Lokal reserve (uten nett)", "no_NO-nvcc-medium", 0),
    _piper("no-msv", "Mann – MSV", "Lokal reserve (uten nett)", "no_NO-nvcc-medium", 5),
    _piper("en-ryan", "Ryan (US)", "Lokal reserve (uten nett)", "en_US-ryan-high"),
    _piper("en-jenny", "Jenny (UK)", "Lokal reserve (uten nett)", "en_GB-jenny_dioco-medium"),
]
VOICES: dict[str, dict] = {v["id"]: v for v in VOICES_LIST}
DEFAULT_VOICE = "finn"
EDGE_RATE = 24000

log = logging.getLogger("announcer")
app = Flask(__name__, static_folder=None)


# --------------------------------------------------------------------------
# TTS
# --------------------------------------------------------------------------

_voice_cache: dict[str, object] = {}
_voice_lock = threading.Lock()


def ensure_voices() -> None:
    import piper.download_voices as dl

    VOICE_DIR.mkdir(parents=True, exist_ok=True)
    for name in sorted({v["model"] for v in VOICES_LIST if v["engine"] == "piper"}):
        if not (VOICE_DIR / f"{name}.onnx").exists():
            log.info("laster ned stemme %s", name)
            dl.download_voice(name, VOICE_DIR)


def load_voice(name: str):
    from piper import PiperVoice

    with _voice_lock:
        if name not in _voice_cache:
            _voice_cache[name] = PiperVoice.load(VOICE_DIR / f"{name}.onnx")
        return _voice_cache[name]


# --------------------------------------------------------------------------
# Jingler (chimes). Alle de innebygde genereres i Python; egne lydfiler
# (wav/mp3/ogg/flac) i CHIME_DIR dukker opp automatisk i menyen.
# --------------------------------------------------------------------------

CHIME_DIR = Path(os.environ.get("ANNOUNCER_CHIMES", HERE / "chimes"))


def _tone(rate: int, freq: float, dur: float, vol: float = 0.35, wave_: str = "sine",
          decay: float = 5.5, attack: float = 0.005, harm2: float = 0.0, slide: float = 0.0) -> list[float]:
    """Én tone som liste av floats. slide = frekvensendring (multiplikator) over tonen."""
    n = int(rate * dur)
    out = []
    phase = 0.0
    for i in range(n):
        t = i / rate
        f = freq * (slide ** (t / dur)) if slide else freq
        phase += 2 * math.pi * f / rate
        if wave_ == "square":
            v = 1.0 if math.sin(phase) >= 0 else -1.0
        elif wave_ == "saw":
            v = 2.0 * ((phase / (2 * math.pi)) % 1.0) - 1.0
        else:
            v = math.sin(phase) + harm2 * math.sin(2 * phase)
        env = min(1.0, t / attack) * math.exp(-decay * t)
        out.append(v * env * vol)
    return out


def _mix(*parts: list[float], gap: float = 0.0, rate: int = 22050) -> list[float]:
    out: list[float] = []
    for p in parts:
        out += p
        out += [0.0] * int(rate * gap)
    return out


def _overlay(base: list[float], layer: list[float], at: float, rate: int) -> list[float]:
    start = int(at * rate)
    if len(base) < start + len(layer):
        base = base + [0.0] * (start + len(layer) - len(base))
    for i, v in enumerate(layer):
        base[start + i] += v
    return base


def _dingdong(rate: int) -> list[float]:
    return _mix(_tone(rate, 784.0, 0.28, harm2=0.3), _tone(rate, 523.25, 0.42, harm2=0.3), rate=rate)


def _airport(rate: int) -> list[float]:
    """Tre stigende toner med lang klang, som på en flyplass."""
    out: list[float] = []
    for i, f in enumerate((523.25, 659.25, 783.99)):
        out = _overlay(out, _tone(rate, f, 1.2, vol=0.3, decay=2.2, harm2=0.2), i * 0.38, rate)
    return out


def _chiptune(rate: int) -> list[float]:
    """8-bit arpeggio, som et power-up i et gammelt spill."""
    notes = (523.25, 659.25, 783.99, 1046.5, 1318.5, 1567.98)
    parts = [_tone(rate, f, 0.09, vol=0.3, wave_="square", decay=3.0) for f in notes]
    parts.append(_tone(rate, 2093.0, 0.35, vol=0.3, wave_="square", decay=6.0))
    return _mix(*parts, rate=rate)


def _scifi(rate: int) -> list[float]:
    """Sveip opp + to blipp, som en kommunikator."""
    sweep = _tone(rate, 300.0, 0.35, vol=0.3, decay=1.5, slide=5.0)
    blips = _mix(_tone(rate, 1800.0, 0.07, vol=0.25, decay=10), _tone(rate, 2400.0, 0.12, vol=0.25, decay=10), gap=0.03, rate=rate)
    return _mix(sweep, blips, rate=rate)


def _horn(rate: int) -> list[float]:
    """Air horn: to lag av sagtann med litt vibrato, derav grovheten."""
    out: list[float] = []
    for f, v in ((233.08, 0.22), (349.23, 0.16), (466.16, 0.1)):
        out = _overlay(out, _tone(rate, f, 0.9, vol=v, wave_="saw", decay=1.3, attack=0.03), 0.0, rate)
    return out


def _gong(rate: int) -> list[float]:
    """Dyp gong med uharmoniske overtoner."""
    out: list[float] = []
    for f, v, d in ((110.0, 0.3, 1.0), (164.8, 0.18, 1.4), (263.0, 0.12, 2.0), (411.0, 0.07, 3.0), (650.0, 0.04, 4.0)):
        out = _overlay(out, _tone(rate, f, 2.5, vol=v, decay=d, attack=0.004), 0.0, rate)
    return out


BUILTIN_CHIMES: dict[str, tuple[str, object]] = {
    "dingdong": ("Ding-dong", _dingdong),
    "airport": ("Flyplass (tre toner)", _airport),
    "chiptune": ("8-bit power-up", _chiptune),
    "scifi": ("Sci-fi kommunikator", _scifi),
    "horn": ("Air horn", _horn),
    "gong": ("Gong", _gong),
}
CHIME_EXT = {".wav", ".mp3", ".ogg", ".flac", ".m4a"}


def file_chimes() -> dict[str, Path]:
    if not CHIME_DIR.is_dir():
        return {}
    return {f"file:{p.stem}": p for p in sorted(CHIME_DIR.iterdir()) if p.suffix.lower() in CHIME_EXT}


def list_chimes() -> list[dict]:
    out = [{"id": "none", "name": "Ingen"}]
    out += [{"id": k, "name": v[0]} for k, v in BUILTIN_CHIMES.items()]
    out += [{"id": k, "name": f"{p.stem} (egen fil)"} for k, p in file_chimes().items()]
    return out


def chime_pcm(chime: str, rate: int) -> bytes:
    """Jingle som 16-bit mono PCM i gitt samplerate, etterfulgt av en kort pause."""
    if chime == "none":
        return b""
    if chime in BUILTIN_CHIMES:
        samples = BUILTIN_CHIMES[chime][1](rate)
        pcm = b"".join(struct.pack("<h", int(max(-1.0, min(1.0, v)) * 32767)) for v in samples)
    else:
        path = file_chimes().get(chime)
        if path is None:
            raise ValueError("Ukjent jingle")
        pcm = subprocess.run(
            ["ffmpeg", "-v", "error", "-i", str(path), "-t", "6", "-f", "s16le", "-ac", "1", "-ar", str(rate),
             "-af", "loudnorm=I=-18,afade=t=out:st=5:d=1", "-"],
            capture_output=True, check=True,
        ).stdout
    return pcm + b"\x00\x00" * int(rate * 0.35)


def edge_pcm(text: str, voice: str, speed: float) -> bytes:
    """Henter MP3 fra edge-tts og dekoder til 16-bit mono PCM i EDGE_RATE."""
    import asyncio
    import tempfile

    import edge_tts

    rate_pct = int(round((speed - 1.0) * 100))
    rate = f"{rate_pct:+d}%"
    with tempfile.NamedTemporaryFile(suffix=".mp3", delete=False, dir=CACHE_DIR) as tmp:
        mp3 = Path(tmp.name)
    try:
        asyncio.run(edge_tts.Communicate(text, voice, rate=rate).save(str(mp3)))
        out = subprocess.run(
            ["ffmpeg", "-v", "error", "-i", str(mp3), "-f", "s16le", "-ac", "1", "-ar", str(EDGE_RATE), "-"],
            capture_output=True,
            check=True,
        )
        return out.stdout
    finally:
        mp3.unlink(missing_ok=True)


def synth_key(text: str, voice: str, speed: float, chime: str) -> str:
    h = hashlib.sha1(f"{voice}|{speed:.2f}|{chime}|{text}".encode()).hexdigest()
    return h[:20]


def synthesize(text: str, voice_name: str, speed: float, chime: str) -> Path:
    """Returnerer sti til ferdig WAV (bufres på disk etter innhold)."""
    from piper.config import SynthesisConfig

    CACHE_DIR.mkdir(parents=True, exist_ok=True)
    path = CACHE_DIR / f"{synth_key(text, voice_name, speed, chime)}.wav"
    if path.exists():
        return path

    spec = VOICES[voice_name]
    pcm = bytearray()
    if spec["engine"] == "edge":
        rate = EDGE_RATE
        pcm += chime_pcm(chime, rate)
        pcm += edge_pcm(text, spec["model"], speed)
    else:
        voice = load_voice(spec["model"])
        rate = voice.config.sample_rate
        cfg = SynthesisConfig(length_scale=1.0 / speed, speaker_id=spec["speaker"])
        pcm += chime_pcm(chime, rate)
        for chunk in voice.synthesize(text, cfg):
            pcm += chunk.audio_int16_bytes
    pcm += b"\x00\x00" * int(rate * 0.3)

    tmp = path.with_suffix(".tmp")
    with wave.open(str(tmp), "wb") as w:
        w.setnchannels(1)
        w.setsampwidth(2)
        w.setframerate(rate)
        w.writeframes(bytes(pcm))
    tmp.rename(path)
    return path


# --------------------------------------------------------------------------
# Avspillingskø
# --------------------------------------------------------------------------


@dataclass
class Item:
    id: int
    text: str
    voice: str
    speed: float
    chime: str
    sender: str
    created: float = field(default_factory=time.time)
    status: str = "queued"  # queued | playing | done | stopped | failed
    error: str = ""

    def public(self) -> dict:
        return {
            "id": self.id,
            "text": self.text,
            "voice": VOICES[self.voice]["name"] if self.voice in VOICES else self.voice,
            "sender": self.sender,
            "created": self.created,
            "status": self.status,
            "error": self.error,
        }


class Player:
    def __init__(self) -> None:
        self.q: queue.Queue[Item] = queue.Queue()
        self.history: list[Item] = []
        self.current: Item | None = None
        self.proc: subprocess.Popen | None = None
        self.lock = threading.Lock()
        self.seq = 0
        threading.Thread(target=self._loop, daemon=True, name="player").start()

    def enqueue(self, text: str, voice: str, speed: float, chime: str, sender: str) -> Item:
        with self.lock:
            self.seq += 1
            item = Item(self.seq, text, voice, speed, chime, sender)
            self.history.append(item)
            del self.history[:-50]
        self.q.put(item)
        return item

    def stop(self) -> None:
        with self.lock:
            if self.proc and self.proc.poll() is None:
                self.proc.terminate()
                if self.current:
                    self.current.status = "stopped"

    def clear(self) -> None:
        while True:
            try:
                item = self.q.get_nowait()
            except queue.Empty:
                break
            item.status = "stopped"
        self.stop()

    def snapshot(self) -> dict:
        with self.lock:
            return {
                "current": self.current.public() if self.current else None,
                "queued": [i.public() for i in self.history if i.status == "queued"],
                "history": [i.public() for i in reversed(self.history) if i.status not in ("queued", "playing")][:15],
            }

    def _loop(self) -> None:
        while True:
            item = self.q.get()
            if item.status != "queued":
                continue
            item.status = "playing"
            with self.lock:
                self.current = item
            try:
                path = synthesize(item.text, item.voice, item.speed, item.chime)
                cmd = ["pw-play"]
                if SINK:
                    cmd += ["--target", SINK]
                cmd.append(str(path))
                with self.lock:
                    self.proc = subprocess.Popen(cmd, stdout=subprocess.DEVNULL, stderr=subprocess.PIPE)
                _, err = self.proc.communicate()
                if item.status == "playing":
                    if self.proc.returncode == 0:
                        item.status = "done"
                    else:
                        item.status = "failed"
                        item.error = err.decode(errors="replace").strip()[-200:]
            except Exception as e:  # noqa: BLE001
                log.exception("avspilling feilet")
                item.status = "failed"
                item.error = str(e)
            finally:
                with self.lock:
                    self.current = None
                    self.proc = None


player = Player()


# --------------------------------------------------------------------------
# HTTP
# --------------------------------------------------------------------------


def check_pin() -> None:
    if PIN and request.headers.get("X-Pin", "") != PIN:
        abort(401)


def parse_request() -> tuple[str, str, float, str]:
    data = request.get_json(silent=True) or {}
    text = " ".join(str(data.get("text", "")).split())
    if not text:
        abort(400, "Tom melding")
    if len(text) > MAX_CHARS:
        abort(400, f"Maks {MAX_CHARS} tegn")
    voice = str(data.get("voice", DEFAULT_VOICE))
    if voice not in VOICES:
        abort(400, "Ukjent stemme")
    try:
        speed = float(data.get("speed", 1.0))
    except (TypeError, ValueError):
        abort(400, "Ugyldig hastighet")
    speed = max(0.6, min(1.6, speed))
    chime = data.get("chime", "dingdong")
    if chime is True:
        chime = "dingdong"
    elif chime is False or chime is None:
        chime = "none"
    chime = str(chime)
    if chime != "none" and chime not in BUILTIN_CHIMES and chime not in file_chimes():
        abort(400, "Ukjent jingle")
    return text, voice, speed, chime


@app.get("/")
def index() -> Response:
    return send_file(HERE / "index.html")


@app.get("/api/config")
def config() -> Response:
    return jsonify(
        {
            "voices": [{"id": v["id"], "name": v["name"], "group": v["group"], "engine": v["engine"]} for v in VOICES_LIST],
            "default_voice": DEFAULT_VOICE,
            "chimes": list_chimes(),
            "default_chime": "dingdong",
            "max_chars": MAX_CHARS,
            "pin_required": bool(PIN),
        }
    )


@app.post("/api/preview")
def preview() -> Response:
    check_pin()
    text, voice, speed, chime = parse_request()
    try:
        path = synthesize(text, voice, speed, chime)
    except Exception as e:  # noqa: BLE001
        log.exception("syntese feilet")
        abort(502, f"Syntese feilet ({VOICES[voice]['name']}): {e}. Prøv en lokal reserve-stemme.")
    return send_file(path, mimetype="audio/wav", max_age=3600)


@app.post("/api/announce")
def announce() -> Response:
    check_pin()
    text, voice, speed, chime = parse_request()
    sender = request.remote_addr or "?"
    item = player.enqueue(text, voice, speed, chime, sender)
    log.info("kø #%d fra %s: %s", item.id, sender, text)
    return jsonify(item.public())


@app.post("/api/stop")
def stop() -> Response:
    check_pin()
    player.stop()
    return jsonify({"ok": True})


@app.post("/api/clear")
def clear() -> Response:
    check_pin()
    player.clear()
    return jsonify({"ok": True})


@app.get("/api/status")
def status() -> Response:
    return jsonify(player.snapshot())


@app.errorhandler(400)
def bad_request(e):  # noqa: ANN001
    return jsonify({"error": e.description}), 400


@app.errorhandler(502)
def bad_gateway(e):  # noqa: ANN001
    return jsonify({"error": e.description}), 502


@app.errorhandler(401)
def unauthorized(e):  # noqa: ANN001
    return jsonify({"error": "Feil PIN"}), 401


def main() -> None:
    ap = argparse.ArgumentParser(description=__doc__.splitlines()[0])
    ap.add_argument("--host", default="0.0.0.0")
    ap.add_argument("--port", type=int, default=int(os.environ.get("ANNOUNCER_PORT", 8090)))
    args = ap.parse_args()
    logging.basicConfig(level=logging.INFO, format="%(asctime)s %(levelname)s %(message)s")
    logging.getLogger("werkzeug").setLevel(logging.WARNING)
    ensure_voices()
    load_voice("no_NO-talesyntese-medium")  # varm opp reserven
    log.info("klar på http://%s:%d  (PIN: %s, sink: %s)", args.host, args.port, "på" if PIN else "av", SINK or "standard")
    app.run(host=args.host, port=args.port, threaded=True)


if __name__ == "__main__":
    main()