🎙️ fix: Align Speech Engine Configuration With Runtime (#14736)

* fix: align speech engine configuration with runtime

* fix: guard speech recording shortcuts

* fix: reconcile speech engine availability

---------

Co-authored-by: Danny Avila <danny@librechat.ai>

api/server/services/Files/Audio/getCustomConfigSpeech.js

@@ -1,6 +1,21 @@
 const { logger } = require('@librechat/data-schemas');
 const { getAppConfig } = require('~/server/services/Config');
 
+const LEGACY_EXTERNAL_STT_ENGINES = new Set(['openai', 'azureOpenAI']);
+const LEGACY_EXTERNAL_TTS_ENGINES = new Set(['openai', 'azureOpenAI', 'elevenlabs', 'localai']);
+
+function normalizeSpeechEngine(key, value) {
+  if (key === 'engineSTT' && LEGACY_EXTERNAL_STT_ENGINES.has(value)) {
+    return 'external';
+  }
+
+  if (key === 'engineTTS' && LEGACY_EXTERNAL_TTS_ENGINES.has(value)) {
+    return 'external';
+  }
+
+  return value;
+}
+
 /**
  * This function retrieves the speechTab settings from the custom configuration
  * It first fetches the custom configuration
@@ -52,7 +67,8 @@ async function getCustomConfigSpeech(req, res) {
       } else {
         for (const key in speechTab.speechToText) {
           if (speechTab.speechToText[key] !== undefined) {
-            settings[key] = speechTab.speechToText[key];
+            const value = speechTab.speechToText[key];
+            settings[key] = normalizeSpeechEngine(key, value);
           }
         }
       }
@@ -64,7 +80,8 @@ async function getCustomConfigSpeech(req, res) {
       } else {
         for (const key in speechTab.textToSpeech) {
           if (speechTab.textToSpeech[key] !== undefined) {
-            settings[key] = speechTab.textToSpeech[key];
+            const value = speechTab.textToSpeech[key];
+            settings[key] = normalizeSpeechEngine(key, value);
           }
         }
       }

api/server/services/Files/Audio/getCustomConfigSpeech.spec.js

@@ -0,0 +1,95 @@
+jest.mock('@librechat/data-schemas', () => ({ logger: { error: jest.fn() } }));
+jest.mock('~/server/services/Config', () => ({ getAppConfig: jest.fn() }));
+
+const getCustomConfigSpeech = require('./getCustomConfigSpeech');
+
+const createResponse = () => {
+  const res = {
+    send: jest.fn(),
+    status: jest.fn(),
+  };
+  res.status.mockReturnValue(res);
+  return res;
+};
+
+describe('getCustomConfigSpeech', () => {
+  it.each(['openai', 'azureOpenAI'])(
+    'normalizes the legacy STT provider "%s" to the external engine',
+    async (engineSTT) => {
+      const req = {
+        config: {
+          speech: {
+            stt: { openai: {} },
+            speechTab: { speechToText: { engineSTT } },
+          },
+        },
+      };
+      const res = createResponse();
+
+      await getCustomConfigSpeech(req, res);
+
+      expect(res.status).toHaveBeenCalledWith(200);
+      expect(res.send).toHaveBeenCalledWith({
+        sttExternal: true,
+        ttsExternal: false,
+        engineSTT: 'external',
+      });
+    },
+  );
+
+  it.each(['browser', 'external'])('preserves the runtime STT engine "%s"', async (engineSTT) => {
+    const req = {
+      config: {
+        speech: {
+          stt: { openai: {} },
+          speechTab: { speechToText: { engineSTT } },
+        },
+      },
+    };
+    const res = createResponse();
+
+    await getCustomConfigSpeech(req, res);
+
+    expect(res.send).toHaveBeenCalledWith(expect.objectContaining({ engineSTT }));
+  });
+
+  it.each(['openai', 'azureOpenAI', 'elevenlabs', 'localai'])(
+    'normalizes the legacy TTS provider "%s" to the external engine',
+    async (engineTTS) => {
+      const req = {
+        config: {
+          speech: {
+            tts: { openai: {} },
+            speechTab: { textToSpeech: { engineTTS } },
+          },
+        },
+      };
+      const res = createResponse();
+
+      await getCustomConfigSpeech(req, res);
+
+      expect(res.status).toHaveBeenCalledWith(200);
+      expect(res.send).toHaveBeenCalledWith({
+        sttExternal: false,
+        ttsExternal: true,
+        engineTTS: 'external',
+      });
+    },
+  );
+
+  it.each(['browser', 'external'])('preserves the runtime TTS engine "%s"', async (engineTTS) => {
+    const req = {
+      config: {
+        speech: {
+          tts: { openai: {} },
+          speechTab: { textToSpeech: { engineTTS } },
+        },
+      },
+    };
+    const res = createResponse();
+
+    await getCustomConfigSpeech(req, res);
+
+    expect(res.send).toHaveBeenCalledWith(expect.objectContaining({ engineTTS }));
+  });
+});