mirror of
https://github.com/crowetic/nc-talk-ai.git
synced 2026-09-07 15:42:31 +00:00
Talk AI is a multi-bot AI assistant manager for Nextcloud Talk: per-bot prompts and models, agentic tool calling (MCP + built-in tools), RAG over Nextcloud files, room-document search, vision and speech-to-text attachments, persistent bot wikis, approval workflows, rate limiting, and multi-provider LLM support (any OpenAI-compatible endpoint). Developed within EDUC - the European Digital UniverCity (https://educalliance.eu), where it runs as the 'EDUC AI' assistant on the alliance-wide Nextcloud portal. This public repository is the upstream point of truth; deployment-specific tools plug in via the tool-provider extension point (docs/TOOL_PROVIDERS.md). License: AGPL-3.0-or-later.
193 lines
6.0 KiB
PHP
193 lines
6.0 KiB
PHP
<?php
|
|
|
|
declare(strict_types=1);
|
|
|
|
namespace OCA\EducAI\Service;
|
|
|
|
use Exception;
|
|
use OCP\Http\Client\IClientService;
|
|
use Psr\Log\LoggerInterface;
|
|
|
|
class SpeechToTextClient {
|
|
private IClientService $clientService;
|
|
private SettingsService $settingsService;
|
|
private LoggerInterface $logger;
|
|
|
|
public function __construct(
|
|
IClientService $clientService,
|
|
SettingsService $settingsService,
|
|
LoggerInterface $logger
|
|
) {
|
|
$this->clientService = $clientService;
|
|
$this->settingsService = $settingsService;
|
|
$this->logger = $logger;
|
|
}
|
|
|
|
public function isEnabled(): bool {
|
|
return (bool)$this->settingsService->getSpeechConfig()['enabled'];
|
|
}
|
|
|
|
/**
|
|
* @throws Exception
|
|
*/
|
|
public function transcribeAudio(string $tempPath, string $displayName): string {
|
|
$config = $this->settingsService->getSpeechConfig();
|
|
$model = $config['model'] ?? null;
|
|
if ($model === null || $model === '') {
|
|
throw new Exception('Speech model not configured');
|
|
}
|
|
|
|
$apiKey = $config['api_key'] ?: $this->settingsService->getApiKey();
|
|
if ($apiKey === null || $apiKey === '') {
|
|
throw new Exception('Speech API key not configured');
|
|
}
|
|
|
|
$endpoint = $this->resolveTranscriptionEndpoint($config['endpoint'] ?? null);
|
|
if ($endpoint === null) {
|
|
throw new Exception('Speech endpoint not configured');
|
|
}
|
|
|
|
$client = $this->clientService->newClient();
|
|
$boundary = 'educai_boundary_' . bin2hex(random_bytes(8));
|
|
$mimeType = mime_content_type($tempPath) ?: 'audio/wav';
|
|
$content = file_get_contents($tempPath);
|
|
if (!is_string($content)) {
|
|
throw new Exception('Unable to read temporary audio content');
|
|
}
|
|
|
|
$response = $client->post($endpoint, [
|
|
'headers' => [
|
|
'Authorization' => 'Bearer ' . $apiKey,
|
|
'Accept' => 'application/json',
|
|
'Content-Type' => 'multipart/form-data; boundary=' . $boundary,
|
|
],
|
|
'body' => $this->buildMultipartBody($boundary, $displayName, $content, $mimeType, $model),
|
|
'timeout' => 180,
|
|
]);
|
|
|
|
$payload = json_decode((string)$response->getBody(), true);
|
|
$text = $payload['text'] ?? $payload['transcript'] ?? null;
|
|
if (!is_string($text)) {
|
|
throw new Exception('Speech API returned no transcript');
|
|
}
|
|
|
|
return trim($text);
|
|
}
|
|
|
|
/**
|
|
* @return array{success:bool,error?:string}
|
|
*/
|
|
public function testConnection(?string $endpoint = null, ?string $apiKey = null, ?string $model = null): array {
|
|
$config = $this->settingsService->getSpeechConfig();
|
|
$resolvedEndpoint = $this->resolveTranscriptionEndpoint($endpoint ?? $config['endpoint'] ?? null);
|
|
$resolvedKey = $apiKey ?? $config['api_key'] ?? $this->settingsService->getApiKey();
|
|
$resolvedModel = $model ?? $config['model'] ?? null;
|
|
|
|
if ($resolvedEndpoint === null || $resolvedKey === null || $resolvedKey === '' || $resolvedModel === null || $resolvedModel === '') {
|
|
return ['success' => false, 'error' => 'Speech endpoint, key or model not configured'];
|
|
}
|
|
|
|
$tempPath = tempnam(sys_get_temp_dir(), 'educai_speech_test_');
|
|
if ($tempPath === false) {
|
|
return ['success' => false, 'error' => 'Unable to allocate temporary test audio'];
|
|
}
|
|
|
|
try {
|
|
file_put_contents($tempPath, $this->buildSilentWav());
|
|
$this->transcribeWithResolvedConfig($resolvedEndpoint, $resolvedKey, $resolvedModel, $tempPath);
|
|
return ['success' => true];
|
|
} catch (Exception $e) {
|
|
return ['success' => false, 'error' => $e->getMessage()];
|
|
} finally {
|
|
if (is_file($tempPath)) {
|
|
@unlink($tempPath);
|
|
}
|
|
}
|
|
}
|
|
|
|
/**
|
|
* @throws Exception
|
|
*/
|
|
private function transcribeWithResolvedConfig(string $endpoint, string $apiKey, string $model, string $tempPath): void {
|
|
$client = $this->clientService->newClient();
|
|
$boundary = 'educai_boundary_' . bin2hex(random_bytes(8));
|
|
$content = file_get_contents($tempPath);
|
|
if (!is_string($content)) {
|
|
throw new Exception('Unable to read temporary test audio');
|
|
}
|
|
|
|
$response = $client->post($endpoint, [
|
|
'headers' => [
|
|
'Authorization' => 'Bearer ' . $apiKey,
|
|
'Accept' => 'application/json',
|
|
'Content-Type' => 'multipart/form-data; boundary=' . $boundary,
|
|
],
|
|
'body' => $this->buildMultipartBody($boundary, 'test.wav', $content, 'audio/wav', $model),
|
|
'timeout' => 60,
|
|
]);
|
|
|
|
$payload = json_decode((string)$response->getBody(), true);
|
|
if (!is_array($payload)) {
|
|
throw new Exception('Speech test returned invalid response');
|
|
}
|
|
}
|
|
|
|
private function resolveTranscriptionEndpoint(?string $customEndpoint): ?string {
|
|
$endpoint = $customEndpoint;
|
|
if ($endpoint === null || $endpoint === '') {
|
|
$endpoint = $this->settingsService->getSettings()->getApiEndpoint();
|
|
}
|
|
if ($endpoint === null || $endpoint === '') {
|
|
return null;
|
|
}
|
|
|
|
$normalized = rtrim($endpoint, '/');
|
|
if (preg_match('#/v1/chat/completions$#', $normalized)) {
|
|
return (string)preg_replace('#/v1/chat/completions$#', '/v1/audio/transcriptions', $normalized);
|
|
}
|
|
if (preg_match('#/chat/completions$#', $normalized)) {
|
|
return (string)preg_replace('#/chat/completions$#', '/audio/transcriptions', $normalized);
|
|
}
|
|
if (preg_match('#/v1$#', $normalized)) {
|
|
return $normalized . '/audio/transcriptions';
|
|
}
|
|
|
|
return $normalized;
|
|
}
|
|
|
|
private function buildMultipartBody(string $boundary, string $filename, string $content, string $mimeType, string $model): string {
|
|
$body = '';
|
|
|
|
$body .= '--' . $boundary . "\r\n";
|
|
$body .= 'Content-Disposition: form-data; name="model"' . "\r\n\r\n";
|
|
$body .= $model . "\r\n";
|
|
|
|
$body .= '--' . $boundary . "\r\n";
|
|
$body .= 'Content-Disposition: form-data; name="file"; filename="' . $filename . '"' . "\r\n";
|
|
$body .= 'Content-Type: ' . $mimeType . "\r\n\r\n";
|
|
$body .= $content . "\r\n";
|
|
|
|
$body .= '--' . $boundary . "--\r\n";
|
|
|
|
return $body;
|
|
}
|
|
|
|
private function buildSilentWav(): string {
|
|
$sampleRate = 8000;
|
|
$durationSeconds = 1;
|
|
$samples = $sampleRate * $durationSeconds;
|
|
$data = str_repeat("\x00\x00", $samples);
|
|
$dataSize = strlen($data);
|
|
$byteRate = $sampleRate * 2;
|
|
$blockAlign = 2;
|
|
|
|
return 'RIFF'
|
|
. pack('V', 36 + $dataSize)
|
|
. 'WAVEfmt '
|
|
. pack('VvvVVvv', 16, 1, 1, $sampleRate, $byteRate, $blockAlign, 16)
|
|
. 'data'
|
|
. pack('V', $dataSize)
|
|
. $data;
|
|
}
|
|
}
|