<?php
declare(strict_types=1);

/**
 * Minimal TTS bridge (microgenerator client) + queue publish + DEBUG LOG
 * POST /tts/speak-wav.php
 * Body (JSON): {"text":"...","lang":"ja-JP","voice":"ja-JP-Chirp3-HD-Leda","audio_format":"MP3"}
 * Response: {"ok":true,"id":"...","url":"/tts/out/<id>.mp3"}
 */

use Google\Cloud\TextToSpeech\V1\Client\TextToSpeechClient;  // microgenerator client
use Google\Cloud\TextToSpeech\V1\AudioConfig;
use Google\Cloud\TextToSpeech\V1\AudioEncoding;
use Google\Cloud\TextToSpeech\V1\SynthesisInput;
use Google\Cloud\TextToSpeech\V1\VoiceSelectionParams;
use Google\Cloud\TextToSpeech\V1\SynthesizeSpeechRequest;

header('Content-Type: application/json; charset=utf-8');
set_time_limit(25);

/* ---- CONFIG ---- */
const GCP_CREDENTIALS    = 'E:\\gkey\\tts-sa.json';
const OUT_DIR            = __DIR__ . '/out';
const PUBLIC_URL_BASE    = '/tts/out';
const QUEUE_FILE         = __DIR__ . '/queue.jsonl';
const DEBUG_LOG_FILE     = __DIR__ . '/tts-debug.jsonl';

const ENFORCE_AUTH       = false;            // flip to true to require X-API-Key
const API_KEY            = 'passTTS123';

const DEFAULT_LANG       = 'ja-JP';
const DEFAULT_VOICE      = 'ja-JP-Chirp3-HD-Sulafat';
const DEFAULT_RATE       = 0.0;             // 0.25–2.0 (non-Chirp only)
const DEFAULT_PITCH      = 5.0;             // semitones, -20.0..+20.0
const DEFAULT_VOLUME_DB  = 0.0;              // dB, -96.0..+16.0
const DEFAULT_EFFECTS    = [];               // e.g., ['wearable-class-device']
const DEFAULT_FORMAT     = 'MP3';            // MP3 | OGG | LINEAR16

/* ---- DEBUG LOGGING ---- */
function logj(string $event, array $data = []): void {
  $rec = $data + [
    'event' => $event,
    'ts'    => date('c'),
  ];
  @file_put_contents(
    DEBUG_LOG_FILE,
    json_encode($rec, JSON_UNESCAPED_SLASHES) . "\n",
    FILE_APPEND | LOCK_EX
  );
}
set_error_handler(function($sev, $msg, $file, $line){
  logj('php_error', ['severity'=>$sev,'message'=>$msg,'file'=>$file,'line'=>$line]);
  return false; // also let PHP handle it normally
});
set_exception_handler(function($e){
  logj('php_exception', ['error'=>$e->getMessage(),'trace'=>$e->getTraceAsString()]);
  http_response_code(500);
  $payload = ['ok'=>false,'error'=>'Unhandled: '.$e->getMessage()];
  echo json_encode($payload);
  logj('response', $payload);
  exit;
});

/* ---- COMPOSER AUTOLOAD (local vendor only) ---- */
$autoload = __DIR__ . DIRECTORY_SEPARATOR . 'vendor' . DIRECTORY_SEPARATOR . 'autoload.php';
if (!is_file($autoload)) {
  http_response_code(500);
  $payload = ['ok'=>false,'error'=>'Composer autoload not found at '.$autoload];
  echo json_encode($payload);
  logj('response', $payload);
  exit;
}
require $autoload;
if (!class_exists(\Google\Cloud\TextToSpeech\V1\Client\TextToSpeechClient::class)) {
  http_response_code(500);
  $payload = ['ok'=>false,'error'=>'TextToSpeechClient class not found under V1\\Client.'];
  echo json_encode($payload);
  logj('response', $payload);
  exit;
}

/* ---- AUTH (optional) ---- */
if (ENFORCE_AUTH) {
  $hdrs = function_exists('getallheaders') ? (getallheaders() ?: []) : [];
  $clientKey = $hdrs['X-API-Key'] ?? $hdrs['x-api-key'] ?? '';
  if ($clientKey !== API_KEY) {
    http_response_code(401);
    $payload = ['ok'=>false,'error'=>'Unauthorized'];
    echo json_encode($payload);
    logj('response', $payload);
    exit;
  }
}

/* ---- INPUT ---- */
$raw = file_get_contents('php://input') ?: '';
logj('request_raw', ['bytes'=>strlen($raw), 'preview'=>substr($raw,0,256)]);

$body = json_decode($raw, true);
if (!is_array($body)) {
  // allow plain text as a fallback
  $raw = trim($raw);
  if ($raw === '') {
    http_response_code(400);
    $payload = ['ok'=>false,'error'=>'Missing body'];
    echo json_encode($payload);
    logj('response', $payload);
    exit;
  }
  $body = ['text' => $raw];
}
if (!isset($body['text']) || trim((string)$body['text']) === '') {
  http_response_code(400);
  $payload = ['ok'=>false,'error'=>'Missing "text"'];
  echo json_encode($payload);
  logj('response', $payload);
  exit;
}
$text        = trim((string)$body['text']);
$lang        = (string)($body['lang'] ?? DEFAULT_LANG);
$voice       = trim((string)($body['voice'] ?? DEFAULT_VOICE));
$format      = strtoupper((string)($body['audio_format'] ?? DEFAULT_FORMAT));
// Optional audio tuning from body
$rate        = isset($body['rate']) ? (float)$body['rate'] : DEFAULT_RATE;
$pitch       = isset($body['pitch']) ? (float)$body['pitch'] : DEFAULT_PITCH;
$volumeDb    = isset($body['volume_db']) ? (float)$body['volume_db'] : DEFAULT_VOLUME_DB;
$effectsBody = $body['effects_profile'] ?? DEFAULT_EFFECTS;
$effects     = is_array($effectsBody) ? $effectsBody : (($effectsBody!=='') ? [$effectsBody] : []);

logj('request_parsed', [
  'lang'=>$lang,
  'voice'=>$voice,
  'format'=>$format,
  'rate'=>$rate,
  'pitch'=>$pitch,
  'volume_db'=>$volumeDb,
  'effects_profile'=>$effects,
  'text_len'=>strlen($text),
  'text_preview'=>substr($text,0,160)
]);

/* ---- PREP OUTPUT ---- */
@mkdir(OUT_DIR, 0775, true);
$id      = date('Ymd_His') . '_' . bin2hex(random_bytes(3));
$ext     = map_ext($format);
$absPath = sprintf('%s/%s.%s', OUT_DIR, $id, $ext);
$url     = sprintf('%s/%s.%s', PUBLIC_URL_BASE, $id, $ext);

/* ---- HELPERS ---- */
function map_encoding(string $fmt): int {
  switch (strtoupper($fmt)) {
    case 'OGG':      return AudioEncoding::OGG_OPUS;
    case 'LINEAR16': return AudioEncoding::LINEAR16;
    case 'MP3':
    default:         return AudioEncoding::MP3;
  }
}
function map_ext(string $fmt): string {
  switch (strtoupper($fmt)) {
    case 'OGG':      return 'ogg';
    case 'LINEAR16': return 'ogg'; // (or 'oga' if you prefer)
    case 'LINEAR16': return 'wav';
    case 'MP3':
    default:         return 'mp3';
  }
}
function ssml_escape_text(string $s): string {
  $s = str_replace(['&','<','>'], ['&amp;','&lt;','&gt;'], $s);
  $s = preg_replace('/\s+/u', ' ', $s);
  return trim($s);
}
/** Simple JP container you already had; safe for non-Chirp voices. */
function build_basic_ssml(string $text): string {
  $t = ssml_escape_text($text);
  return <<<SSML
<speak xml:lang="ja-JP">
  <p>
    <prosody rate="medium">
      $t
    </prosody>
    <break time="250ms"/>
  </p>
</speak>
SSML;
}
/** Chirp detector: “chirp” anywhere in the name. */
function is_chirp_voice(string $name): bool {
  return (strpos(strtolower($name), 'chirp') !== false);
}
function clampf(float $v, float $lo, float $hi): float {
  return max($lo, min($hi, $v));
}

/* ---- BUILD REQUEST ---- */
$isChirp = is_chirp_voice($voice ?: DEFAULT_VOICE);

// INPUT: Chirp must be plain text; others can be SSML
if ($isChirp) {
  $input = (new SynthesisInput())->setText($text);
} else {
  $input = (new SynthesisInput())->setSsml(build_basic_ssml($text));
}

$voiceSel = (new VoiceSelectionParams())->setLanguageCode($lang);
if ($voice !== '') {
  $voiceSel->setName($voice);
} else {
  $voiceSel->setName(DEFAULT_VOICE);
}

$audioCfg = (new AudioConfig())->setAudioEncoding(map_encoding($format));
// Apply volume and effects for all voices
$audioCfg->setVolumeGainDb(clampf($volumeDb, -96.0, 16.0));
if (!empty($effects)) {
  $audioCfg->setEffectsProfileId($effects); // array OK
}
// Rate/Pitch for non-Chirp voices
if (!$isChirp) {
  $audioCfg->setSpeakingRate(clampf($rate, 0.25, 2.0));     // per docs
  $audioCfg->setPitch(clampf($pitch, -20.0, 20.0));         // semitones
}
logj('build_request', [
  'isChirp'=>$isChirp,
  'voice'=>$voiceSel->getName(),
  'lang'=>$voiceSel->getLanguageCode(),
  'format'=>$format,
  'out'=>$absPath
]);
logj('audio_tuning', [
  'isChirp'=>$isChirp,
  'rate'=>$rate,
  'pitch'=>$pitch,
  'volume_db'=>$volumeDb,
  'effects'=>$effects
]);

/* ---- CALL GOOGLE (microgenerator) ---- */
try {
  $opts = ['transport' => 'rest'];
  if (GCP_CREDENTIALS !== '' && is_file(GCP_CREDENTIALS)) {
    $opts['credentials'] = GCP_CREDENTIALS;
  }
  $client = new TextToSpeechClient($opts);

  $request = (new SynthesizeSpeechRequest())
               ->setInput($input)
               ->setVoice($voiceSel)
               ->setAudioConfig($audioCfg);

  try {
    $resp = $client->synthesizeSpeech($request, ['timeoutMillis' => 10000]);
    logj('call_primary_ok');
  } catch (\Throwable $e) {
    // Fallback: retry without specific voice name (language only)
    logj('call_primary_err', ['error'=>$e->getMessage()]);
    $request->setVoice((new VoiceSelectionParams())->setLanguageCode($lang));
    $resp = $client->synthesizeSpeech($request, ['timeoutMillis' => 10000]);
    logj('call_fallback_ok');
  }

  $bytes = $resp->getAudioContent();
  $blen  = is_string($bytes) ? strlen($bytes) : 0;
  logj('tts_result', ['bytes'=>$blen]);

  if (!$bytes) {
    http_response_code(502);
    $payload = ['ok'=>false,'error'=>'TTS returned empty audio'];
    echo json_encode($payload);
    logj('response', $payload);
    exit;
  }

  if (file_put_contents($absPath, $bytes) === false) {
    http_response_code(500);
    $payload = ['ok'=>false,'error'=>'Failed to write audio file','path'=>$absPath];
    echo json_encode($payload);
    logj('response', $payload);
    exit;
  }

  // Publish for listener/stream
  publishAudio($id, $url, time());

  $payload = ['ok'=>true,'id'=>$id,'url'=>$url];
  echo json_encode($payload, JSON_UNESCAPED_SLASHES);
  logj('response', $payload);

} catch (\Throwable $e) {
  http_response_code(500);
  $payload = ['ok'=>false,'error'=>$e->getMessage()];
  echo json_encode($payload);
  logj('response', $payload);
  exit;
}

/* ---- SSE QUEUE ---- */
function publishAudio(string $id, string $url, int $ts): void {
  $line = json_encode(['id'=>$id,'url'=>$url,'ts'=>$ts], JSON_UNESCAPED_SLASHES) . "\n";
  $f = fopen(QUEUE_FILE, 'ab');
  if ($f) {
    flock($f, LOCK_EX);
    fwrite($f, $line);
    fflush($f);
    flock($f, LOCK_UN);
    fclose($f);
    logj('queue_publish', ['id'=>$id,'url'=>$url]);
  } else {
    logj('queue_error', ['id'=>$id,'url'=>$url]);
  }
}