更新

2026-08-03 10:00:05 +08:00
parent 0fb03d0bca
commit 01729b1e0b
109 changed files with 7577 additions and 1153 deletions
@@ -9,6 +9,7 @@ use app\model\UploadFile;
 use app\service\AgentCatalog;
 use app\service\ComfyJobDeferredException;
 use app\service\ComfyUIService;
+use app\service\CosyVoiceService;
 use app\service\DifyService;
 use app\service\DocumentTextService;
 use app\service\OpenAIService;
@@ -18,6 +19,133 @@ use think\facade\Log;

 class Chat extends BaseApi
 {
+    public function speech()
+    {
+        $this->authUser();
+        $input = $this->request->post();
+        $text = trim((string) ($input['text'] ?? ''));
+
+        if ($text === '') {
+            return $this->error('语音内容不能为空', 422);
+        }
+        if (mb_strlen($text) > 600) {
+            return $this->error('单次语音内容不能超过 600 个字符', 422);
+        }
+
+        $modelId = isset($input['model_id']) && $input['model_id'] !== ''
+            ? (int) $input['model_id']
+            : null;
+        $persona = CosyVoiceService::getPersona();
+        $voice = trim((string) ($persona['fallback_voice'] ?? $input['voice'] ?? 'marin'));
+        $allowedVoices = [
+            'alloy', 'ash', 'ballad', 'coral', 'echo', 'fable', 'nova',
+            'onyx', 'sage', 'shimmer', 'verse', 'marin', 'cedar',
+        ];
+        if (!in_array($voice, $allowedVoices, true)) {
+            $voice = 'marin';
+        }
+
+        $speech = null;
+        if (CosyVoiceService::canAttempt()) {
+            try {
+                $speech = CosyVoiceService::speech($text);
+            } catch (\Throwable $error) {
+                Log::warning('CosyVoice speech fallback: ' . $error->getMessage());
+            }
+        }
+
+        if (!$speech) {
+            $model = OpenAIService::getSpeechModel($modelId);
+            $speech = OpenAIService::speech($model, $text, $voice);
+            $speech['provider'] = 'openai';
+        }
+
+        return response($speech['audio'], 200, [
+            'Content-Type'           => $speech['content_type'],
+            'Content-Length'         => (string) strlen($speech['audio']),
+            'Cache-Control'          => 'no-store',
+            'X-Content-Type-Options' => 'nosniff',
+            'X-TTS-Provider'         => $speech['provider'] ?? 'unknown',
+        ]);
+    }
+
+    public function speechStream(): never
+    {
+        $this->authUser();
+        $input = $this->request->post();
+        $text = trim((string) ($input['text'] ?? ''));
+        $requestId = trim((string) ($input['request_id'] ?? ''));
+        if (!preg_match('/^[A-Za-z0-9._:-]{8,128}$/', $requestId)) {
+            $requestId = bin2hex(random_bytes(16));
+        }
+
+        if ($text === '' || mb_strlen($text) > 600 || !CosyVoiceService::canAttempt()) {
+            http_response_code($text === '' || mb_strlen($text) > 600 ? 422 : 503);
+            header('Content-Type: application/json; charset=utf-8');
+            echo json_encode([
+                'code' => 1,
+                'message' => $text === ''
+                    ? '语音内容不能为空'
+                    : (mb_strlen($text) > 600 ? '单次语音内容不能超过 600 个字符' : 'CosyVoice 暂时不可用'),
+                'data' => null,
+            ], JSON_UNESCAPED_UNICODE);
+            exit;
+        }
+
+        while (ob_get_level() > 0) {
+            ob_end_clean();
+        }
+        @ini_set('zlib.output_compression', '0');
+        ignore_user_abort(false);
+        OpenAIService::sseHeaders();
+
+        $persona = CosyVoiceService::getPersona();
+        OpenAIService::sseEvent('meta', [
+            'provider' => 'cosyvoice',
+            'request_id' => $requestId,
+            'cancel_url' => rtrim((string) ($persona['base_url'] ?? ''), '/')
+                . '/cancel/' . rawurlencode($requestId),
+            'format' => 'pcm_s16le',
+            'sample_rate' => (int) ($persona['sample_rate'] ?? 24000),
+            'channels' => 1,
+        ]);
+
+        try {
+            $result = CosyVoiceService::streamSpeech($text, static function (string $pcm): void {
+                OpenAIService::sseEvent('audio', [
+                    'audio' => base64_encode($pcm),
+                ]);
+            }, $requestId);
+
+            if (empty($result['aborted'])) {
+                OpenAIService::sseEvent('done', [
+                    'bytes' => (int) ($result['bytes'] ?? 0),
+                ]);
+            }
+        } catch (\Throwable $error) {
+            Log::warning('CosyVoice stream failed: ' . $error->getMessage());
+            OpenAIService::sseEvent('error', [
+                'message' => $error->getMessage(),
+            ]);
+        }
+
+        exit;
+    }
+
+    public function speechCancel()
+    {
+        $this->authUser();
+        $requestId = trim((string) $this->request->post('request_id', ''));
+        if (!preg_match('/^[A-Za-z0-9._:-]{8,128}$/', $requestId)) {
+            return $this->error('语音请求标识无效', 422);
+        }
+
+        return $this->success([
+            'cancelled' => CosyVoiceService::cancelSpeech($requestId),
+            'request_id' => $requestId,
+        ]);
+    }
+
    public function completions()
    {
        $user = $this->authUser();
@@ -29,6 +157,7 @@ class Chat extends BaseApi
        $attachments = $input['attachments'] ?? [];
        $agentId = trim((string) ($input['agent_id'] ?? ''));
        $imageTool = trim((string) ($input['image_tool'] ?? ''));
+        $voiceMode = !empty($input['voice_mode']);
        $stream = ($input['stream'] ?? true) !== false;

        $allowedImageTools = ['enhance', 'erase', 'watermark', 'cutout', 'outpaint', 'replace', 'text', 'restore', 'creative', 'commit'];
@@ -153,7 +282,7 @@ class Chat extends BaseApi
            $history = Message::where('conversation_id', $conversationId)
                ->field('role,content,attachments')
                ->order('id', 'desc')
-                ->limit(50)
+                ->limit($voiceMode ? 16 : 50)
                ->select()
                ->toArray();
            $history = array_reverse($history);
@@ -167,6 +296,16 @@ class Chat extends BaseApi
        }

        $apiMessages = $this->buildApiMessages($history, $model);
+        if ($voiceMode) {
+            $voicePersona = CosyVoiceService::getPersona();
+            $personaName = trim((string) ($voicePersona['name'] ?? 'AI 客服')) ?: 'AI 客服';
+            $personaPrompt = trim((string) ($voicePersona['role_prompt'] ?? ''));
+            array_unshift($apiMessages, [
+                'role' => 'system',
+                'content' => '你是名为“' . $personaName . '”的 AI 客服。人物设定：' . $personaPrompt
+                    . ' 当前正在进行低延迟实时语音对话。像真人客服一样先回应用户的真实诉求，语气口语化、有耐心、有适度共情，不复述问题，不使用 Markdown 列表，不说“作为 AI”。先给结论，通常控制在 1 到 3 句；信息不足时每轮只追问一个最关键的问题，除非用户明确要求详细说明。',
+            ]);
+        }
        $agentImageActionAllowed = false;
        if ($agent) {
            $agentSystemPrompt = $agent['system_prompt'];
@@ -1590,6 +1729,12 @@ class Chat extends BaseApi
            },
            function (string $message) use (&$streamError) {
                $streamError = $message;
+            },
+            function () use ($conversationId, &$resolvedExternalConversationId) {
+                $resolvedExternalConversationId = null;
+                ConversationModel::where('id', $conversationId)->update([
+                    'external_conversation_id' => null,
+                ]);
            }
        );

@@ -1723,6 +1868,10 @@ class Chat extends BaseApi
            ConversationModel::where('id', $conversationId)->update([
                'external_conversation_id' => $resolvedExternalConversationId,
            ]);
+        } elseif (!empty($result['conversation_reset'])) {
+            ConversationModel::where('id', $conversationId)->update([
+                'external_conversation_id' => null,
+            ]);
        }

        $imageAction = $agent