<?php

namespace App\Http\Controllers;

use App\Services\SpeechService;
use Illuminate\Http\JsonResponse;
use Illuminate\Http\Request;
use Illuminate\Support\Facades\Process;
use Illuminate\Support\Facades\Storage;
use Illuminate\Support\Str;

/**
 * Endpoint NUR fürs Transkribieren. Frontend bekommt den Text zurück und
 * routet ihn durch den normalen send-Flow (IntentDetector + Livewire).
 * NICHT mehr: handleUserMessage / speak() - das würde 2x getriggert + Doppel-Audio.
 */
class FoxVoiceController
{
    public function __invoke(Request $request, SpeechService $speech): JsonResponse
    {
        $upload = $request->file('audio');
        if (! $upload) {
            return response()->json(['error' => 'no_file', 'detail' => 'audio field missing'], 422);
        }
        if (! $upload->isValid()) {
            return response()->json([
                'error' => 'invalid_upload',
                'detail' => $upload->getErrorMessage(),
                'size' => $upload->getSize(),
            ], 422);
        }
        if ($upload->getSize() > 50_000_000) {
            return response()->json(['error' => 'too_large', 'size' => $upload->getSize()], 413);
        }

        // webm/ogg → wav konvertieren (whisper.cpp will 16kHz mono wav)
        $rawPath = $upload->getRealPath();
        $wavName = 'in_'.Str::random(10).'.wav';
        $wavPath = Storage::disk('audio')->path($wavName);

        Process::timeout(30)->run(sprintf(
            'ffmpeg -y -i %s -ar 16000 -ac 1 -f wav %s 2>&1',
            escapeshellarg($rawPath),
            escapeshellarg($wavPath),
        ));

        $text = $speech->transcribe($wavPath);

        if (! $text) {
            return response()->json(['text' => null, 'error' => 'Transkription fehlgeschlagen'], 422);
        }

        return response()->json(['text' => $text]);
    }
}