Implement Gemini audio voice note replies

This commit is contained in:
Hamza-Ayed
2026-05-22 15:10:13 +03:00
parent 395c8ee8eb
commit 2e7cd11f85
5 changed files with 264 additions and 13 deletions

View File

@@ -0,0 +1,88 @@
<?php
// Secure token check to prevent unauthorized execution on production
if (($_GET['token'] ?? '') !== 'nabeh_test_audio_1298') {
http_response_code(403);
die('Unauthorized access');
}
require_once dirname(__DIR__) . '/app/bootstrap.php';
use App\Services\GeminiService;
use App\Models\ChatbotRule;
use App\Core\Database;
header('Content-Type: text/plain; charset=utf-8');
echo "=== Starting Chatbot Audio Reply Diagnostics ===\n\n";
// 1. Fetch Chatbot Rules and check for Gemini Key
$rules = Database::select("SELECT * FROM chatbot_rules WHERE trigger_type = 'gemini_ai' AND is_active = 1 LIMIT 1");
if (empty($rules)) {
echo "❌ [Database] No active chatbot rule found with 'gemini_ai' trigger type.\n";
exit(1);
}
$rule = $rules[0];
$apiKey = $rule['gemini_api_key'] ?: getenv('GEMINI_API_KEY');
if (empty($apiKey)) {
echo "❌ [Config] Gemini API Key is empty. Please set gemini_api_key in the active chatbot rule or GEMINI_API_KEY in the .env file.\n";
exit(1);
}
echo "✅ [Database] Found active Gemini rule for Company ID: {$rule['company_id']}\n";
echo " [Config] Gemini API Key starts with: " . substr($apiKey, 0, 6) . "...\n";
// 2. Test Gemini API Audio Output (Puck Voice, gemini-flash-lite-latest)
echo "\n--- Testing Gemini Audio Response ---\n";
$systemPrompt = "You are a friendly customer service assistant. Speak in a warm, welcoming tone.";
$testMsg = "مرحبا، هل متجركم مفتوح اليوم؟";
echo "Sending text prompt: '$testMsg'\n";
$startTime = microtime(true);
$audioResponse = GeminiService::generateAudioResponse($apiKey, $systemPrompt, $testMsg, 'Puck');
$elapsedTime = round(microtime(true) - $startTime, 2);
if ($audioResponse && !empty($audioResponse['audio'])) {
echo "✅ [Gemini] Successfully generated voice note in {$elapsedTime} seconds!\n";
echo " [Gemini] Audio MimeType: " . $audioResponse['mimeType'] . "\n";
echo " [Gemini] Audio Size: " . strlen($audioResponse['audio']) . " base64 chars\n";
// 3. Test Audio-to-Audio conversion
echo "\n--- Testing Audio-to-Audio (Speech-to-Speech) ---\n";
$startTime = microtime(true);
$audioResponse2 = GeminiService::generateAudioResponseFromAudio($apiKey, $systemPrompt, $audioResponse['audio'], $audioResponse['mimeType'], 'Puck');
$elapsedTime2 = round(microtime(true) - $startTime, 2);
if ($audioResponse2 && !empty($audioResponse2['audio'])) {
echo "✅ [Gemini] Successfully generated Audio-to-Audio response in {$elapsedTime2} seconds!\n";
echo " [Gemini] Audio MimeType: " . $audioResponse2['mimeType'] . "\n";
echo " [Gemini] Audio Size: " . strlen($audioResponse2['audio']) . " base64 chars\n";
} else {
echo "❌ [Gemini] Audio-to-Audio generation failed. Check errors below:\n";
}
} else {
echo "❌ [Gemini] Audio response generation failed. Please check the logs.\n";
}
// 4. Check Node.js WhatsApp Gateway Status
echo "\n--- Checking Node.js Gateway Status ---\n";
$gatewayUrl = rtrim(getenv('WHATSAPP_GATEWAY_URL') ?: 'http://localhost:3722', '/');
echo "Gateway URL: {$gatewayUrl}\n";
$ch = curl_init($gatewayUrl . '/health');
curl_setopt($ch, CURLOPT_RETURNTRANSFER, true);
curl_setopt($ch, CURLOPT_TIMEOUT, 5);
$healthResponse = curl_exec($ch);
$healthHttpCode = curl_getinfo($ch, CURLINFO_HTTP_CODE);
$healthError = curl_error($ch);
curl_close($ch);
if ($healthHttpCode === 200) {
echo "✅ [Gateway] Gateway is ONLINE and healthy.\n";
echo " [Gateway] Response: {$healthResponse}\n";
} else {
echo "❌ [Gateway] Gateway is OFFLINE or returning error. HTTP Code: {$healthHttpCode}. Error: {$healthError}\n";
}
echo "\n=== Diagnostics Complete ===\n";