🎯 fix: Google AI Client Stability; feat: gemini-exp models (#4781)

* fix: Google timing out and issuing AbortError, bump package, and use `@google/generative-ai` explicitly for latest models

* feat: gemini-exp-

.env.example

@@ -140,7 +140,7 @@ GOOGLE_KEY=user_provided
 # GOOGLE_REVERSE_PROXY=
 
 # Gemini API (AI Studio)
-# GOOGLE_MODELS=gemini-1.5-flash-latest,gemini-1.0-pro,gemini-1.0-pro-001,gemini-1.0-pro-latest,gemini-1.0-pro-vision-latest,gemini-1.5-pro-latest,gemini-pro,gemini-pro-vision
+# GOOGLE_MODELS=gemini-exp-1121,gemini-exp-1114,gemini-1.5-flash-latest,gemini-1.0-pro,gemini-1.0-pro-001,gemini-1.0-pro-latest,gemini-1.0-pro-vision-latest,gemini-1.5-pro-latest,gemini-pro,gemini-pro-vision
 
 # Vertex AI
 # GOOGLE_MODELS=gemini-1.5-flash-preview-0514,gemini-1.5-pro-preview-0514,gemini-1.0-pro-vision-001,gemini-1.0-pro-002,gemini-1.0-pro-001,gemini-pro-vision,gemini-1.0-pro

api/app/clients/GoogleClient.js

@@ -35,6 +35,7 @@ const endpointPrefix = `https://${loc}-aiplatform.googleapis.com`;
 const tokenizersCache = {};
 
 const settings = endpointSettings[EModelEndpoint.google];
+const EXCLUDED_GENAI_MODELS = /gemini-(?:1\.0|1-0|pro)/;
 
 class GoogleClient extends BaseClient {
   constructor(credentials, options = {}) {
@@ -366,7 +367,7 @@ class GoogleClient extends BaseClient {
       );
     }
 
-    if (!this.project_id && this.modelOptions.model.includes('1.5')) {
+    if (!this.project_id && !EXCLUDED_GENAI_MODELS.test(this.modelOptions.model)) {
       return await this.buildGenerativeMessages(messages);
     }
 
@@ -604,15 +605,12 @@ class GoogleClient extends BaseClient {
     } else if (this.project_id) {
       logger.debug('Creating VertexAI client');
       return new ChatVertexAI(clientOptions);
-    } else if (model.includes('1.5')) {
+    } else if (!EXCLUDED_GENAI_MODELS.test(model)) {
       logger.debug('Creating GenAI client');
-      return new GenAI(this.apiKey).getGenerativeModel(
-        {
-          ...clientOptions,
-          model,
-        },
-        { apiVersion: 'v1beta' },
-      );
+      return new GenAI(this.apiKey).getGenerativeModel({
+        ...clientOptions,
+        model,
+      });
     }
 
     logger.debug('Creating Chat Google Generative AI client');
@@ -674,7 +672,7 @@ class GoogleClient extends BaseClient {
     }
 
     const modelName = clientOptions.modelName ?? clientOptions.model ?? '';
-    if (modelName?.includes('1.5') && !this.project_id) {
+    if (!EXCLUDED_GENAI_MODELS.test(modelName) && !this.project_id) {
       const client = model;
       const requestOptions = {
         contents: _payload,
@@ -697,7 +695,7 @@ class GoogleClient extends BaseClient {
 
       requestOptions.safetySettings = _payload.safetySettings;
 
-      const delay = modelName.includes('flash') ? 8 : 14;
+      const delay = modelName.includes('flash') ? 8 : 15;
       const result = await client.generateContentStream(requestOptions);
       for await (const chunk of result.stream) {
         const chunkText = chunk.text();
@@ -712,7 +710,6 @@ class GoogleClient extends BaseClient {
 
     const stream = await model.stream(messages, {
       signal: abortController.signal,
-      timeout: 7000,
       safetySettings: _payload.safetySettings,
     });
 
@@ -720,7 +717,7 @@ class GoogleClient extends BaseClient {
 
     if (!this.options.streamRate) {
       if (this.isGenerativeModel) {
-        delay = 12;
+        delay = 15;
       }
       if (modelName.includes('flash')) {
         delay = 5;
@@ -774,8 +771,8 @@ class GoogleClient extends BaseClient {
     const messages = this.isTextModel ? _payload.trim() : _messages;
 
     const modelName = clientOptions.modelName ?? clientOptions.model ?? '';
-    if (modelName?.includes('1.5') && !this.project_id) {
-      logger.debug('Identified titling model as 1.5 version');
+    if (!EXCLUDED_GENAI_MODELS.test(modelName) && !this.project_id) {
+      logger.debug('Identified titling model as GenAI version');
       /** @type {GenerativeModel} */
       const client = model;
       const requestOptions = {

api/package.json

@@ -41,7 +41,7 @@
     "@keyv/redis": "^2.8.1",
     "@langchain/community": "^0.3.13",
     "@langchain/core": "^0.3.17",
-    "@langchain/google-genai": "^0.1.3",
+    "@langchain/google-genai": "^0.1.4",
     "@langchain/google-vertexai": "^0.1.2",
     "@langchain/textsplitters": "^0.1.0",
     "@librechat/agents": "^1.7.7",

api/utils/tokens.js

@@ -49,6 +49,7 @@ const googleModels = {
   /* Max I/O is combined so we subtract the amount from max response tokens for actual total */
   gemini: 30720, // -2048 from max
   'gemini-pro-vision': 12288, // -4096 from max
+  'gemini-exp': 8000,
   'gemini-1.5': 1048576, // -8192 from max
   'text-bison-32k': 32758, // -10 from max
   'chat-bison-32k': 32758, // -10 from max

package-lock.json

@@ -50,7 +50,7 @@
         "@keyv/redis": "^2.8.1",
         "@langchain/community": "^0.3.13",
         "@langchain/core": "^0.3.17",
-        "@langchain/google-genai": "^0.1.3",
+        "@langchain/google-genai": "^0.1.4",
         "@langchain/google-vertexai": "^0.1.2",
         "@langchain/textsplitters": "^0.1.0",
         "@librechat/agents": "^1.7.7",
@@ -218,10 +218,9 @@
       }
     },
     "api/node_modules/@langchain/google-genai": {
-      "version": "0.1.3",
-      "resolved": "https://registry.npmjs.org/@langchain/google-genai/-/google-genai-0.1.3.tgz",
-      "integrity": "sha512-GHZV4qEMoi+rnqSM5I+ADXwUSBRSD0hsmlS1lTQEGW9HmvzPu3zryvYjuRAoelZSENTmZmBatdM+kgiV8H2+JA==",
-      "license": "MIT",
+      "version": "0.1.4",
+      "resolved": "https://registry.npmjs.org/@langchain/google-genai/-/google-genai-0.1.4.tgz",
+      "integrity": "sha512-b8qrqnHYbNseaAikrWyxuDTww6CUIse82F5/BmF2GtWVR25yJrNUWETfTp7o7iIMxhFR0PuQag4gEZOL74F5Tw==",
       "dependencies": {
         "@google/generative-ai": "^0.21.0",
         "zod-to-json-schema": "^3.22.4"