From 157315cd6890c3738b95f3d164ffc477e3469425 Mon Sep 17 00:00:00 2001
From: NijikaMyWaifu <139055015+NijikaMyWaifu@users.noreply.github.com>
Date: Fri, 23 May 2025 01:10:53 +0800
Subject: [PATCH] Add Vertex AI express mode support (#3977)

* Add Vertex AI express mode support
Split Google AI Studio and Vertex AI

* Add support for Vertex AI, including updating default models and related settings, modifying frontend HTML to include Vertex AI options, and adjusting request processing logic in the backend API.

* Log API name in the console

* Merge sysprompt toggles back

* Use Gemma tokenizers for Vertex and LearnLM

* AI Studio parity updates

* Add link to express mode doc. Also technically it's not a form

* Split title

* Use array includes

* Add support for Google Vertex AI in image captioning feature

* Specify caption API name, add to compression list

---------

Co-authored-by: Cohee <18619528+Cohee1207@users.noreply.github.com>
---
 default/content/presets/openai/Default.json   |  1 +
 public/index.html                             | 59 +++++++++++++----
 public/script.js                              |  1 +
 public/scripts/RossAscends-mods.js            |  1 +
 public/scripts/extensions/caption/index.js    |  1 +
 .../scripts/extensions/caption/settings.html  |  9 ++-
 public/scripts/extensions/shared.js           | 10 ++-
 public/scripts/openai.js                      | 65 +++++++++++++++++--
 public/scripts/reasoning.js                   |  1 +
 public/scripts/secrets.js                     |  2 +
 public/scripts/slash-commands.js              |  1 +
 public/scripts/tokenizers.js                  |  4 ++
 public/scripts/tool-calling.js                |  1 +
 src/constants.js                              |  1 +
 src/endpoints/backends/chat-completions.js    | 47 ++++++++++----
 src/endpoints/google.js                       | 28 ++++++--
 src/endpoints/secrets.js                      |  1 +
 src/endpoints/tokenizers.js                   |  2 +-
 18 files changed, 193 insertions(+), 42 deletions(-)

diff --git a/default/content/presets/openai/Default.json b/default/content/presets/openai/Default.json
index 19660ea2d..efd39d5fb 100644
--- a/default/content/presets/openai/Default.json
+++ b/default/content/presets/openai/Default.json
@@ -15,6 +15,7 @@
     "custom_exclude_body": "",
     "custom_include_headers": "",
     "google_model": "gemini-pro",
+    "vertexai_model": "gemini-2.0-flash-001",
     "temperature": 1,
     "frequency_penalty": 0,
     "presence_penalty": 0,
diff --git a/public/index.html b/public/index.html
index 87c789284..ff8d0ff23 100644
--- a/public/index.html
+++ b/public/index.html
@@ -691,7 +691,7 @@
                                         </span>
                                     </div>
                                 </div>
-                                <div class="range-block" data-source="openai,claude,windowai,openrouter,ai21,scale,makersuite,mistralai,custom,cohere,perplexity,groq,01ai,nanogpt,deepseek,xai">
+                                <div class="range-block" data-source="openai,claude,windowai,openrouter,ai21,scale,makersuite,vertexai,mistralai,custom,cohere,perplexity,groq,01ai,nanogpt,deepseek,xai">
                                     <div class="range-block-title" data-i18n="Temperature">
                                         Temperature
                                     </div>
@@ -730,7 +730,7 @@
                                         </div>
                                     </div>
                                 </div>
-                                <div class="range-block" data-source="claude,openrouter,makersuite,cohere,perplexity">
+                                <div class="range-block" data-source="claude,openrouter,makersuite,vertexai,cohere,perplexity">
                                     <div class="range-block-title" data-i18n="Top K">
                                         Top K
                                     </div>
@@ -743,7 +743,7 @@
                                         </div>
                                     </div>
                                 </div>
-                                <div class="range-block" data-source="openai,claude,openrouter,ai21,scale,makersuite,mistralai,custom,cohere,perplexity,groq,01ai,nanogpt,deepseek,xai">
+                                <div class="range-block" data-source="openai,claude,openrouter,ai21,scale,makersuite,vertexai,mistralai,custom,cohere,perplexity,groq,01ai,nanogpt,deepseek,xai">
                                     <div class="range-block-title" data-i18n="Top P">
                                         Top P
                                     </div>
@@ -1969,7 +1969,7 @@
                                             </span>
                                         </div>
                                     </div>
-                                    <div class="range-block" data-source="makersuite,openrouter,claude">
+                                    <div class="range-block" data-source="makersuite,vertexai,openrouter,claude">
                                         <label for="openai_enable_web_search" class="checkbox_label flexWrap widthFreeExpand">
                                             <input id="openai_enable_web_search" type="checkbox" />
                                             <span data-i18n="Enable web search">Enable web search</span>
@@ -1983,7 +1983,7 @@
                                             </b>
                                         </div>
                                     </div>
-                                    <div class="range-block" data-source="openai,cohere,mistralai,custom,claude,openrouter,groq,deepseek,makersuite,ai21,xai,pollinations">
+                                    <div class="range-block" data-source="openai,cohere,mistralai,custom,claude,openrouter,groq,deepseek,makersuite,vertexai,ai21,xai,pollinations">
                                         <label for="openai_function_calling" class="checkbox_label flexWrap widthFreeExpand">
                                             <input id="openai_function_calling" type="checkbox" />
                                             <span data-i18n="Enable function calling">Enable function calling</span>
@@ -1993,7 +1993,7 @@
                                             <span data-i18n="enable_functions_desc_3">Can be utilized by various extensions to provide additional functionality.</span>
                                         </div>
                                     </div>
-                                    <div class="range-block" data-source="openai,openrouter,mistralai,makersuite,claude,custom,01ai,xai,pollinations">
+                                    <div class="range-block" data-source="openai,openrouter,mistralai,makersuite,vertexai,claude,custom,01ai,xai,pollinations">
                                         <label for="openai_image_inlining" class="checkbox_label flexWrap widthFreeExpand">
                                             <input id="openai_image_inlining" type="checkbox" />
                                             <span data-i18n="Send inline images">Send inline images</span>
@@ -2018,7 +2018,7 @@
                                             </div>
                                         </div>
                                     </div>
-                                    <div class="range-block" data-source="makersuite">
+                                    <div class="range-block" data-source="makersuite,vertexai">
                                         <label for="openai_request_images" class="checkbox_label widthFreeExpand">
                                             <input id="openai_request_images" type="checkbox" />
                                             <span>
@@ -2030,12 +2030,12 @@
                                             <span data-i18n="Allows the model to return image attachments.">
                                                 Allows the model to return image attachments.
                                             </span>
-                                            <em data-source="makersuite" data-i18n="Request inline images_desc_2">
+                                            <em data-source="makersuite,vertexai" data-i18n="Request inline images_desc_2">
                                                 Incompatible with the following features: function calling, web search, system prompt.
                                             </em>
                                         </div>
                                     </div>
-                                    <div class="range-block" data-source="makersuite">
+                                    <div class="range-block" data-source="makersuite,vertexai">
                                         <label for="use_makersuite_sysprompt" class="checkbox_label widthFreeExpand">
                                             <input id="use_makersuite_sysprompt" type="checkbox" />
                                             <span>
@@ -2060,12 +2060,12 @@
                                             </span>
                                         </div>
                                     </div>
-                                    <div class="flex-container flexFlowColumn wide100p textAlignCenter marginTop10" data-source="openai,custom,claude,xai,makersuite,openrouter,pollinations">
+                                    <div class="flex-container flexFlowColumn wide100p textAlignCenter marginTop10" data-source="openai,custom,claude,xai,makersuite,vertexai,openrouter,pollinations">
                                         <div class="flex-container oneline-dropdown" title="Constrains effort on reasoning for reasoning models.&#10;Reducing reasoning effort can result in faster responses and fewer tokens used on reasoning in a response." data-i18n="[title]Constrains effort on reasoning for reasoning models.">
                                             <label for="openai_reasoning_effort">
                                                 <span data-i18n="Reasoning Effort">Reasoning Effort</span>
                                                 <i data-source="openai,custom,xai,openrouter" class="opacity50p fa-solid fa-circle-info" title="OpenAI-style options: low, medium, high. Minimum and maximum are aliased to low and high. Auto does not send an effort level." data-i18n="[title]OpenAI-style options: low, medium, high. Minimum and maximum are aliased to low and high. Auto does not send an effort level."></i>
-                                                <i data-source="claude,makersuite" class="opacity50p fa-solid fa-circle-info" title="Allocates a portion of the response length for thinking (low: 10%, medium: 25%, high: 50%). Other options are model-dependent." data-i18n="[title]Allocates a portion of the response length for thinking (low: 10%, medium: 25%, high: 50%). Other options are model-dependent."></i>
+                                                <i data-source="claude,makersuite,vertexai" class="opacity50p fa-solid fa-circle-info" title="Allocates a portion of the response length for thinking (low: 10%, medium: 25%, high: 50%). Other options are model-dependent." data-i18n="[title]Allocates a portion of the response length for thinking (low: 10%, medium: 25%, high: 50%). Other options are model-dependent."></i>
                                             </label>
                                             <select id="openai_reasoning_effort">
                                                 <option data-i18n="openai_reasoning_effort_auto" value="auto">Auto</option>
@@ -2769,6 +2769,7 @@
                                 <option value="deepseek">DeepSeek</option>
                                 <option value="groq">Groq</option>
                                 <option value="makersuite">Google AI Studio</option>
+                                <option value="vertexai">Google Vertex AI (Express mode)</option>
                                 <option value="mistralai">MistralAI</option>
                                 <option value="nanogpt">NanoGPT</option>
                                 <option value="openrouter">OpenRouter</option>
@@ -2779,7 +2780,7 @@
                                 <option value="xai">xAI (Grok)</option>
                             </optgroup>
                         </select>
-                        <div class="inline-drawer wide100p" data-source="openai,claude,mistralai,makersuite,deepseek,xai">
+                        <div class="inline-drawer wide100p" data-source="openai,claude,mistralai,makersuite,vertexai,deepseek,xai">
                             <div class="inline-drawer-toggle inline-drawer-header">
                                 <b data-i18n="Reverse Proxy">Reverse Proxy</b>
                                 <div class="fa-solid fa-circle-chevron-down inline-drawer-icon down"></div>
@@ -2843,7 +2844,7 @@
                                 </div>
                             </div>
                         </div>
-                        <div id="ReverseProxyWarningMessage" data-source="openai,claude,mistralai,makersuite,deepseek,xai">
+                        <div id="ReverseProxyWarningMessage" data-source="openai,claude,mistralai,makersuite,vertexai,deepseek,xai">
                             <div class="reverse_proxy_warning">
                                 <b>
                                     <div data-i18n="Using a proxy that you're not running yourself is a risk to your data privacy.">
@@ -3209,6 +3210,38 @@
                                 </select>
                             </div>
                         </form>
+                        <div id="vertexai_form" data-source="vertexai">
+                            <h4>
+                                <span data-i18n="Google Vertex AI API Key">
+                                    Google Vertex AI API Key
+                                </span>
+                                <a href="https://cloud.google.com/vertex-ai/generative-ai/docs/start/express-mode/overview" data-i18n="(Express mode keys only)" target="_blank" rel="noopener noreferrer">
+                                    (Express mode keys only)
+                                </a>
+                            </h4>
+                            <div class="flex-container">
+                                <input id="api_key_vertexai" name="api_key_vertexai" class="text_pole flex1" value="" type="text" autocomplete="off">
+                                <div title="Clear your API key" data-i18n="[title]Clear your API key" class="menu_button fa-solid fa-circle-xmark clear-api-key" data-key="api_key_vertexai"></div>
+                            </div>
+                            <div data-for="api_key_vertexai" class="neutral_warning" data-i18n="For privacy reasons, your API key will be hidden after you reload the page.">
+                                For privacy reasons, your API key will be hidden after you reload the page.
+                            </div>
+                            <div>
+                                <h4 data-i18n="Google Model">Google Model</h4>
+                                <select id="model_vertexai_select">
+                                    <optgroup label="Gemini 2.5">
+                                        <option value="gemini-2.5-pro-preview-05-06">gemini-2.5-pro-preview-05-06</option>
+                                        <option value="gemini-2.5-pro-preview-03-25">gemini-2.5-pro-preview-03-25</option>
+                                        <option value="gemini-2.5-flash-preview-05-20">gemini-2.5-flash-preview-05-20</option>
+                                        <option value="gemini-2.5-flash-preview-04-17">gemini-2.5-flash-preview-04-17</option>
+                                    </optgroup>
+                                    <optgroup label="Gemini 2.0">
+                                        <option value="gemini-2.0-flash-001">gemini-2.0-flash-001</option>
+                                        <option value="gemini-2.0-flash-lite-001">gemini-2.0-flash-lite-001</option>
+                                    </optgroup>
+                                </select>
+                            </div>
+                        </div>
                         <form id="mistralai_form" data-source="mistralai" action="javascript:void(null);" method="post" enctype="multipart/form-data">
                             <h4 data-i18n="MistralAI API Key">MistralAI API Key</h4>
                             <div class="flex-container">
diff --git a/public/script.js b/public/script.js
index e25c2757f..c91bff5d7 100644
--- a/public/script.js
+++ b/public/script.js
@@ -5941,6 +5941,7 @@ function extractImageFromData(data, { mainApi = null, chatCompletionSource = nul
     switch (mainApi ?? main_api) {
         case 'openai': {
             switch (chatCompletionSource ?? oai_settings.chat_completion_source) {
+                case chat_completion_sources.VERTEXAI:
                 case chat_completion_sources.MAKERSUITE: {
                     const inlineData = data?.responseContent?.parts?.find(x => x.inlineData)?.inlineData;
                     if (inlineData) {
diff --git a/public/scripts/RossAscends-mods.js b/public/scripts/RossAscends-mods.js
index 403be2d85..b3e9e92e7 100644
--- a/public/scripts/RossAscends-mods.js
+++ b/public/scripts/RossAscends-mods.js
@@ -402,6 +402,7 @@ function RA_autoconnect(PrevApi) {
                     || (secret_state[SECRET_KEYS.OPENROUTER] && oai_settings.chat_completion_source == chat_completion_sources.OPENROUTER)
                     || (secret_state[SECRET_KEYS.AI21] && oai_settings.chat_completion_source == chat_completion_sources.AI21)
                     || (secret_state[SECRET_KEYS.MAKERSUITE] && oai_settings.chat_completion_source == chat_completion_sources.MAKERSUITE)
+                    || (secret_state[SECRET_KEYS.VERTEXAI] && oai_settings.chat_completion_source == chat_completion_sources.VERTEXAI)
                     || (secret_state[SECRET_KEYS.MISTRALAI] && oai_settings.chat_completion_source == chat_completion_sources.MISTRALAI)
                     || (secret_state[SECRET_KEYS.COHERE] && oai_settings.chat_completion_source == chat_completion_sources.COHERE)
                     || (secret_state[SECRET_KEYS.PERPLEXITY] && oai_settings.chat_completion_source == chat_completion_sources.PERPLEXITY)
diff --git a/public/scripts/extensions/caption/index.js b/public/scripts/extensions/caption/index.js
index a9af62b2b..f29455af6 100644
--- a/public/scripts/extensions/caption/index.js
+++ b/public/scripts/extensions/caption/index.js
@@ -414,6 +414,7 @@ jQuery(async function () {
                         'openai': SECRET_KEYS.OPENAI,
                         'mistral': SECRET_KEYS.MISTRALAI,
                         'google': SECRET_KEYS.MAKERSUITE,
+                        'vertexai': SECRET_KEYS.VERTEXAI,
                         'anthropic': SECRET_KEYS.CLAUDE,
                     };
 
diff --git a/public/scripts/extensions/caption/settings.html b/public/scripts/extensions/caption/settings.html
index 31f7762a3..5b3102684 100644
--- a/public/scripts/extensions/caption/settings.html
+++ b/public/scripts/extensions/caption/settings.html
@@ -22,6 +22,7 @@
                         <option value="cohere">Cohere</option>
                         <option value="custom" data-i18n="Custom (OpenAI-compatible)">Custom (OpenAI-compatible)</option>
                         <option value="google">Google AI Studio</option>
+                        <option value="vertexai">Google Vertex AI</option>
                         <option value="groq">Groq</option>
                         <option value="koboldcpp">KoboldCpp</option>
                         <option value="llamacpp">llama.cpp</option>
@@ -111,6 +112,12 @@
                         <option data-type="google" value="gemini-1.5-flash-8b-exp-0827">gemini-1.5-flash-8b-exp-0827</option>
                         <option data-type="google" value="learnlm-2.0-flash-experimental">learnlm-2.0-flash-experimental</option>
                         <option data-type="google" value="learnlm-1.5-pro-experimental">learnlm-1.5-pro-experimental</option>
+                        <option data-type="vertexai" value="gemini-2.5-pro-preview-05-06">gemini-2.5-pro-preview-05-06</option>
+                        <option data-type="vertexai" value="gemini-2.5-pro-preview-03-25">gemini-2.5-pro-preview-03-25</option>
+                        <option data-type="vertexai" value="gemini-2.5-flash-preview-05-20">gemini-2.5-flash-preview-05-20</option>
+                        <option data-type="vertexai" value="gemini-2.5-flash-preview-04-17">gemini-2.5-flash-preview-04-17</option>
+                        <option data-type="vertexai" value="gemini-2.0-flash-001">gemini-2.0-flash-001</option>
+                        <option data-type="vertexai" value="gemini-2.0-flash-lite-001">gemini-2.0-flash-lite-001</option>
                         <option data-type="groq" value="llama-3.2-11b-vision-preview">llama-3.2-11b-vision-preview</option>
                         <option data-type="groq" value="llama-3.2-90b-vision-preview">llama-3.2-90b-vision-preview</option>
                         <option data-type="groq" value="llava-v1.5-7b-4096-preview">llava-v1.5-7b-4096-preview</option>
@@ -170,7 +177,7 @@
                 <div data-type="ollama">
                     The model must be downloaded first! Do it with the <code>ollama pull</code> command or <a href="#" id="caption_ollama_pull">click here</a>.
                 </div>
-                <label data-type="openai,anthropic,google,mistral" class="checkbox_label flexBasis100p" for="caption_allow_reverse_proxy" title="Allow using reverse proxy if defined and valid.">
+                <label data-type="openai,anthropic,google,vertexai,mistral" class="checkbox_label flexBasis100p" for="caption_allow_reverse_proxy" title="Allow using reverse proxy if defined and valid.">
                     <input id="caption_allow_reverse_proxy" type="checkbox" class="checkbox">
                     <span data-i18n="Allow reverse proxy">Allow reverse proxy</span>
                 </label>
diff --git a/public/scripts/extensions/shared.js b/public/scripts/extensions/shared.js
index f1ea88085..e3ec2c29e 100644
--- a/public/scripts/extensions/shared.js
+++ b/public/scripts/extensions/shared.js
@@ -15,7 +15,7 @@ import { createThumbnail, isValidUrl } from '../utils.js';
  */
 export async function getMultimodalCaption(base64Img, prompt) {
     const useReverseProxy =
-        (['openai', 'anthropic', 'google', 'mistral'].includes(extension_settings.caption.multimodal_api))
+        (['openai', 'anthropic', 'google', 'mistral', 'vertexai'].includes(extension_settings.caption.multimodal_api))
         && extension_settings.caption.allow_reverse_proxy
         && oai_settings.reverse_proxy
         && isValidUrl(oai_settings.reverse_proxy);
@@ -38,7 +38,8 @@ export async function getMultimodalCaption(base64Img, prompt) {
     const isVllm = extension_settings.caption.multimodal_api === 'vllm';
     const base64Bytes = base64Img.length * 0.75;
     const compressionLimit = 2 * 1024 * 1024;
-    if ((['google', 'openrouter', 'mistral', 'groq'].includes(extension_settings.caption.multimodal_api) && base64Bytes > compressionLimit) || isOoba || isKoboldCpp) {
+    const thumbnailNeeded = ['google', 'openrouter', 'mistral', 'groq', 'vertexai'].includes(extension_settings.caption.multimodal_api);
+    if ((thumbnailNeeded && base64Bytes > compressionLimit) || isOoba || isKoboldCpp) {
         const maxSide = 1024;
         base64Img = await createThumbnail(base64Img, maxSide, maxSide, 'image/jpeg');
     }
@@ -94,6 +95,7 @@ export async function getMultimodalCaption(base64Img, prompt) {
     function getEndpointUrl() {
         switch (extension_settings.caption.multimodal_api) {
             case 'google':
+            case 'vertexai':
                 return '/api/google/caption-image';
             case 'anthropic':
                 return '/api/anthropic/caption-image';
@@ -143,6 +145,10 @@ function throwIfInvalidModel(useReverseProxy) {
         throw new Error('Google AI Studio API key is not set.');
     }
 
+    if (extension_settings.caption.multimodal_api === 'vertexai' && !secret_state[SECRET_KEYS.VERTEXAI] && !useReverseProxy) {
+        throw new Error('Google Vertex AI API key is not set.');
+    }
+
     if (extension_settings.caption.multimodal_api === 'mistral' && !secret_state[SECRET_KEYS.MISTRALAI] && !useReverseProxy) {
         throw new Error('Mistral AI API key is not set.');
     }
diff --git a/public/scripts/openai.js b/public/scripts/openai.js
index 1fdd0d9d5..db88ea37e 100644
--- a/public/scripts/openai.js
+++ b/public/scripts/openai.js
@@ -176,6 +176,7 @@ export const chat_completion_sources = {
     OPENROUTER: 'openrouter',
     AI21: 'ai21',
     MAKERSUITE: 'makersuite',
+    VERTEXAI: 'vertexai',
     MISTRALAI: 'mistralai',
     CUSTOM: 'custom',
     COHERE: 'cohere',
@@ -277,6 +278,7 @@ export const settingsToUpdate = {
     custom_include_headers: ['#custom_include_headers', 'custom_include_headers', false, true],
     custom_prompt_post_processing: ['#custom_prompt_post_processing', 'custom_prompt_post_processing', false, true],
     google_model: ['#model_google_select', 'google_model', false, true],
+    vertexai_model: ['#model_vertexai_select', 'vertexai_model', false, true],
     openai_max_context: ['#openai_max_context', 'openai_max_context', false, false],
     openai_max_tokens: ['#openai_max_tokens', 'openai_max_tokens', false, false],
     wrap_in_quotes: ['#wrap_in_quotes', 'wrap_in_quotes', true, false],
@@ -350,6 +352,7 @@ const default_settings = {
     openai_model: 'gpt-4-turbo',
     claude_model: 'claude-3-5-sonnet-20240620',
     google_model: 'gemini-1.5-pro',
+    vertexai_model: 'gemini-2.0-flash-001',
     ai21_model: 'jamba-1.6-large',
     mistralai_model: 'mistral-large-latest',
     cohere_model: 'command-r-plus',
@@ -433,6 +436,7 @@ const oai_settings = {
     openai_model: 'gpt-4-turbo',
     claude_model: 'claude-3-5-sonnet-20240620',
     google_model: 'gemini-1.5-pro',
+    vertexai_model: 'gemini-2.0-flash-001',
     ai21_model: 'jamba-1.6-large',
     mistralai_model: 'mistral-large-latest',
     cohere_model: 'command-r-plus',
@@ -1666,6 +1670,8 @@ export function getChatCompletionModel(source = null) {
             return '';
         case chat_completion_sources.MAKERSUITE:
             return oai_settings.google_model;
+        case chat_completion_sources.VERTEXAI:
+            return oai_settings.vertexai_model;
         case chat_completion_sources.OPENROUTER:
             return oai_settings.openrouter_model !== openrouter_website_model ? oai_settings.openrouter_model : null;
         case chat_completion_sources.AI21:
@@ -2048,6 +2054,7 @@ async function sendOpenAIRequest(type, messages, signal) {
     const isOpenRouter = oai_settings.chat_completion_source == chat_completion_sources.OPENROUTER;
     const isScale = oai_settings.chat_completion_source == chat_completion_sources.SCALE;
     const isGoogle = oai_settings.chat_completion_source == chat_completion_sources.MAKERSUITE;
+    const isVertexAI = oai_settings.chat_completion_source == chat_completion_sources.VERTEXAI;
     const isOAI = oai_settings.chat_completion_source == chat_completion_sources.OPENAI;
     const isMistral = oai_settings.chat_completion_source == chat_completion_sources.MISTRALAI;
     const isCustom = oai_settings.chat_completion_source == chat_completion_sources.CUSTOM;
@@ -2123,8 +2130,8 @@ async function sendOpenAIRequest(type, messages, signal) {
         delete generate_data.stop;
     }
 
-    // Proxy is only supported for Claude, OpenAI, Mistral, and Google MakerSuite
-    if (oai_settings.reverse_proxy && [chat_completion_sources.CLAUDE, chat_completion_sources.OPENAI, chat_completion_sources.MISTRALAI, chat_completion_sources.MAKERSUITE, chat_completion_sources.DEEPSEEK, chat_completion_sources.XAI].includes(oai_settings.chat_completion_source)) {
+    // Proxy is only supported for Claude, OpenAI, Mistral, Google MakerSuite, and Vertex AI
+    if (oai_settings.reverse_proxy && [chat_completion_sources.CLAUDE, chat_completion_sources.OPENAI, chat_completion_sources.MISTRALAI, chat_completion_sources.MAKERSUITE, chat_completion_sources.VERTEXAI, chat_completion_sources.DEEPSEEK, chat_completion_sources.XAI].includes(oai_settings.chat_completion_source)) {
         await validateReverseProxy();
         generate_data['reverse_proxy'] = oai_settings.reverse_proxy;
         generate_data['proxy_password'] = oai_settings.proxy_password;
@@ -2175,7 +2182,7 @@ async function sendOpenAIRequest(type, messages, signal) {
         generate_data['api_url_scale'] = oai_settings.api_url_scale;
     }
 
-    if (isGoogle) {
+    if (isGoogle || isVertexAI) {
         const stopStringsLimit = 5;
         generate_data['top_k'] = Number(oai_settings.top_k_openai);
         generate_data['stop'] = getCustomStoppingStrings(stopStringsLimit).slice(0, stopStringsLimit).filter(x => x.length >= 1 && x.length <= 16);
@@ -2382,7 +2389,7 @@ export function getStreamingReply(data, state, { chatCompletionSource = null, ov
             state.reasoning += data?.delta?.thinking || '';
         }
         return data?.delta?.text || '';
-    } else if (chat_completion_source === chat_completion_sources.MAKERSUITE) {
+    } else if ([chat_completion_sources.MAKERSUITE, chat_completion_sources.VERTEXAI].includes(chat_completion_source)) {
         const inlineData = data?.candidates?.[0]?.content?.parts?.find(x => x.inlineData)?.inlineData;
         if (inlineData) {
             state.image = `data:${inlineData.mimeType};base64,${inlineData.data}`;
@@ -2772,7 +2779,13 @@ class Message {
      * @returns {Promise<string>} Compressed image as a Data URL.
      */
     async compressImage(image) {
-        if ([chat_completion_sources.OPENROUTER, chat_completion_sources.MAKERSUITE, chat_completion_sources.MISTRALAI].includes(oai_settings.chat_completion_source)) {
+        const compressImageSources = [
+            chat_completion_sources.OPENROUTER,
+            chat_completion_sources.MAKERSUITE,
+            chat_completion_sources.MISTRALAI,
+            chat_completion_sources.VERTEXAI,
+        ];
+        if (compressImageSources.includes(oai_settings.chat_completion_source)) {
             const sizeThreshold = 2 * 1024 * 1024;
             const dataSize = image.length * 0.75;
             const maxSide = 1024;
@@ -3368,6 +3381,7 @@ function loadOpenAISettings(data, settings) {
     oai_settings.custom_include_headers = settings.custom_include_headers ?? default_settings.custom_include_headers;
     oai_settings.custom_prompt_post_processing = settings.custom_prompt_post_processing ?? default_settings.custom_prompt_post_processing;
     oai_settings.google_model = settings.google_model ?? default_settings.google_model;
+    oai_settings.vertexai_model = settings.vertexai_model ?? default_settings.vertexai_model;
     oai_settings.chat_completion_source = settings.chat_completion_source ?? default_settings.chat_completion_source;
     oai_settings.api_url_scale = settings.api_url_scale ?? default_settings.api_url_scale;
     oai_settings.show_external_models = settings.show_external_models ?? default_settings.show_external_models;
@@ -3432,6 +3446,8 @@ function loadOpenAISettings(data, settings) {
     $(`#model_windowai_select option[value="${oai_settings.windowai_model}"`).prop('selected', true);
     $('#model_google_select').val(oai_settings.google_model);
     $(`#model_google_select option[value="${oai_settings.google_model}"`).prop('selected', true);
+    $('#model_vertexai_select').val(oai_settings.vertexai_model);
+    $(`#model_vertexai_select option[value="${oai_settings.vertexai_model}"`).prop('selected', true);
     $('#model_ai21_select').val(oai_settings.ai21_model);
     $(`#model_ai21_select option[value="${oai_settings.ai21_model}"`).prop('selected', true);
     $('#model_mistralai_select').val(oai_settings.mistralai_model);
@@ -3627,6 +3643,7 @@ async function getStatusOpen() {
         chat_completion_sources.CLAUDE,
         chat_completion_sources.AI21,
         chat_completion_sources.MAKERSUITE,
+        chat_completion_sources.VERTEXAI,
         chat_completion_sources.PERPLEXITY,
         chat_completion_sources.GROQ,
     ];
@@ -3648,7 +3665,16 @@ async function getStatusOpen() {
         chat_completion_source: oai_settings.chat_completion_source,
     };
 
-    if (oai_settings.reverse_proxy && [chat_completion_sources.CLAUDE, chat_completion_sources.OPENAI, chat_completion_sources.MISTRALAI, chat_completion_sources.MAKERSUITE, chat_completion_sources.DEEPSEEK, chat_completion_sources.XAI].includes(oai_settings.chat_completion_source)) {
+    const validateProxySources = [
+        chat_completion_sources.CLAUDE,
+        chat_completion_sources.OPENAI,
+        chat_completion_sources.MISTRALAI,
+        chat_completion_sources.MAKERSUITE,
+        chat_completion_sources.VERTEXAI,
+        chat_completion_sources.DEEPSEEK,
+        chat_completion_sources.XAI,
+    ];
+    if (oai_settings.reverse_proxy && validateProxySources.includes(oai_settings.chat_completion_source)) {
         await validateReverseProxy();
     }
 
@@ -3740,6 +3766,7 @@ async function saveOpenAIPreset(name, settings, triggerUi = true) {
         custom_include_headers: settings.custom_include_headers,
         custom_prompt_post_processing: settings.custom_prompt_post_processing,
         google_model: settings.google_model,
+        vertexai_model: settings.vertexai_model,
         temperature: settings.temp_openai,
         frequency_penalty: settings.freq_pen_openai,
         presence_penalty: settings.pres_pen_openai,
@@ -4495,6 +4522,11 @@ async function onModelChange() {
         oai_settings.google_model = value;
     }
 
+    if ($(this).is('#model_vertexai_select')) {
+        console.log('Vertex AI model changed to', value);
+        oai_settings.vertexai_model = value;
+    }
+
     if ($(this).is('#model_mistralai_select')) {
         // Upgrade old mistral models to new naming scheme
         // would have done this in loadOpenAISettings, but it wasn't updating on preset change?
@@ -4575,7 +4607,7 @@ async function onModelChange() {
         $('#temp_openai').attr('max', oai_max_temp).val(oai_settings.temp_openai).trigger('input');
     }
 
-    if (oai_settings.chat_completion_source == chat_completion_sources.MAKERSUITE) {
+    if ([chat_completion_sources.MAKERSUITE, chat_completion_sources.VERTEXAI].includes(oai_settings.chat_completion_source)) {
         if (oai_settings.max_context_unlocked) {
             $('#openai_max_context').attr('max', max_2mil);
         } else if (value.includes('gemini-1.5-pro')) {
@@ -4943,6 +4975,19 @@ async function onConnectButtonClick(e) {
         }
     }
 
+    if (oai_settings.chat_completion_source == chat_completion_sources.VERTEXAI) {
+        const api_key_vertexai = String($('#api_key_vertexai').val()).trim();
+
+        if (api_key_vertexai.length) {
+            await writeSecret(SECRET_KEYS.VERTEXAI, api_key_vertexai);
+        }
+
+        if (!secret_state[SECRET_KEYS.VERTEXAI] && !oai_settings.reverse_proxy) {
+            console.log('No secret key saved for Vertex AI');
+            return;
+        }
+    }
+
     if (oai_settings.chat_completion_source == chat_completion_sources.CLAUDE) {
         const api_key_claude = String($('#api_key_claude').val()).trim();
 
@@ -5120,6 +5165,9 @@ function toggleChatCompletionForms() {
     else if (oai_settings.chat_completion_source == chat_completion_sources.MAKERSUITE) {
         $('#model_google_select').trigger('change');
     }
+    else if (oai_settings.chat_completion_source == chat_completion_sources.VERTEXAI) {
+        $('#model_vertexai_select').trigger('change');
+    }
     else if (oai_settings.chat_completion_source == chat_completion_sources.OPENROUTER) {
         $('#model_openrouter_select').trigger('change');
     }
@@ -5281,6 +5329,8 @@ export function isImageInliningSupported() {
             );
         case chat_completion_sources.MAKERSUITE:
             return visionSupportedModels.some(model => oai_settings.google_model.includes(model));
+        case chat_completion_sources.VERTEXAI:
+            return visionSupportedModels.some(model => oai_settings.vertexai_model.includes(model));
         case chat_completion_sources.CLAUDE:
             return visionSupportedModels.some(model => oai_settings.claude_model.includes(model));
         case chat_completion_sources.OPENROUTER:
@@ -5887,6 +5937,7 @@ export function initOpenAI() {
     $('#model_windowai_select').on('change', onModelChange);
     $('#model_scale_select').on('change', onModelChange);
     $('#model_google_select').on('change', onModelChange);
+    $('#model_vertexai_select').on('change', onModelChange);
     $('#model_openrouter_select').on('change', onModelChange);
     $('#openrouter_group_models').on('change', onOpenrouterModelSortChange);
     $('#openrouter_sort_models').on('change', onOpenrouterModelSortChange);
diff --git a/public/scripts/reasoning.js b/public/scripts/reasoning.js
index c17be63eb..91c12e86a 100644
--- a/public/scripts/reasoning.js
+++ b/public/scripts/reasoning.js
@@ -114,6 +114,7 @@ export function extractReasoningFromData(data, {
                 case chat_completion_sources.OPENROUTER:
                     return data?.choices?.[0]?.message?.reasoning ?? '';
                 case chat_completion_sources.MAKERSUITE:
+                case chat_completion_sources.VERTEXAI:
                     return data?.responseContent?.parts?.filter(part => part.thought)?.map(part => part.text)?.join('\n\n') ?? '';
                 case chat_completion_sources.CLAUDE:
                     return data?.content?.find(part => part.type === 'thinking')?.thinking ?? '';
diff --git a/public/scripts/secrets.js b/public/scripts/secrets.js
index 53cf283df..b0cc3d018 100644
--- a/public/scripts/secrets.js
+++ b/public/scripts/secrets.js
@@ -16,6 +16,7 @@ export const SECRET_KEYS = {
     AI21: 'api_key_ai21',
     SCALE_COOKIE: 'scale_cookie',
     MAKERSUITE: 'api_key_makersuite',
+    VERTEXAI: 'api_key_vertexai',
     SERPAPI: 'api_key_serpapi',
     MISTRALAI: 'api_key_mistralai',
     TOGETHERAI: 'api_key_togetherai',
@@ -56,6 +57,7 @@ const INPUT_MAP = {
     [SECRET_KEYS.AI21]: '#api_key_ai21',
     [SECRET_KEYS.SCALE_COOKIE]: '#scale_cookie',
     [SECRET_KEYS.MAKERSUITE]: '#api_key_makersuite',
+    [SECRET_KEYS.VERTEXAI]: '#api_key_vertexai',
     [SECRET_KEYS.VLLM]: '#api_key_vllm',
     [SECRET_KEYS.APHRODITE]: '#api_key_aphrodite',
     [SECRET_KEYS.TABBY]: '#api_key_tabby',
diff --git a/public/scripts/slash-commands.js b/public/scripts/slash-commands.js
index dc239caa9..5e4853def 100644
--- a/public/scripts/slash-commands.js
+++ b/public/scripts/slash-commands.js
@@ -4127,6 +4127,7 @@ function getModelOptions(quiet) {
         { id: 'model_openrouter_select', api: 'openai', type: chat_completion_sources.OPENROUTER },
         { id: 'model_ai21_select', api: 'openai', type: chat_completion_sources.AI21 },
         { id: 'model_google_select', api: 'openai', type: chat_completion_sources.MAKERSUITE },
+        { id: 'model_vertexai_select', api: 'openai', type: chat_completion_sources.VERTEXAI },
         { id: 'model_mistralai_select', api: 'openai', type: chat_completion_sources.MISTRALAI },
         { id: 'custom_model_id', api: 'openai', type: chat_completion_sources.CUSTOM },
         { id: 'model_cohere_select', api: 'openai', type: chat_completion_sources.COHERE },
diff --git a/public/scripts/tokenizers.js b/public/scripts/tokenizers.js
index 5bec8961c..1545787fd 100644
--- a/public/scripts/tokenizers.js
+++ b/public/scripts/tokenizers.js
@@ -676,6 +676,10 @@ export function getTokenizerModel() {
         return gemmaTokenizer;
     }
 
+    if (oai_settings.chat_completion_source == chat_completion_sources.VERTEXAI) {
+        return gemmaTokenizer;
+    }
+
     if (oai_settings.chat_completion_source == chat_completion_sources.AI21) {
         return jambaTokenizer;
     }
diff --git a/public/scripts/tool-calling.js b/public/scripts/tool-calling.js
index 451a3d41a..2dc2f11d8 100644
--- a/public/scripts/tool-calling.js
+++ b/public/scripts/tool-calling.js
@@ -592,6 +592,7 @@ export class ToolManager {
             chat_completion_sources.COHERE,
             chat_completion_sources.DEEPSEEK,
             chat_completion_sources.MAKERSUITE,
+            chat_completion_sources.VERTEXAI,
             chat_completion_sources.AI21,
             chat_completion_sources.XAI,
             chat_completion_sources.POLLINATIONS,
diff --git a/src/constants.js b/src/constants.js
index d68aa5c4e..0c9720829 100644
--- a/src/constants.js
+++ b/src/constants.js
@@ -168,6 +168,7 @@ export const CHAT_COMPLETION_SOURCES = {
     OPENROUTER: 'openrouter',
     AI21: 'ai21',
     MAKERSUITE: 'makersuite',
+    VERTEXAI: 'vertexai',
     MISTRALAI: 'mistralai',
     CUSTOM: 'custom',
     COHERE: 'cohere',
diff --git a/src/endpoints/backends/chat-completions.js b/src/endpoints/backends/chat-completions.js
index bd30701dd..35c4cd43c 100644
--- a/src/endpoints/backends/chat-completions.js
+++ b/src/endpoints/backends/chat-completions.js
@@ -52,6 +52,7 @@ const API_COHERE_V2 = 'https://api.cohere.ai/v2';
 const API_PERPLEXITY = 'https://api.perplexity.ai';
 const API_GROQ = 'https://api.groq.com/openai/v1';
 const API_MAKERSUITE = 'https://generativelanguage.googleapis.com';
+const API_VERTEX_AI = 'https://us-central1-aiplatform.googleapis.com';
 const API_01AI = 'https://api.lingyiwanwu.com/v1';
 const API_AI21 = 'https://api.ai21.com/studio/v1';
 const API_NANOGPT = 'https://nano-gpt.com/api/v1';
@@ -337,12 +338,27 @@ async function sendScaleRequest(request, response) {
  * @param {express.Response} response Express response
  */
 async function sendMakerSuiteRequest(request, response) {
-    const apiUrl = new URL(request.body.reverse_proxy || API_MAKERSUITE);
-    const apiKey = request.body.reverse_proxy ? request.body.proxy_password : readSecret(request.user.directories, SECRET_KEYS.MAKERSUITE);
+    const useVertexAi = request.body.chat_completion_source === CHAT_COMPLETION_SOURCES.VERTEXAI;
+    const apiName = useVertexAi ? 'Google Vertex AI' : 'Google AI Studio';
+    let apiUrl;
+    let apiKey;
 
-    if (!request.body.reverse_proxy && !apiKey) {
-        console.warn('Google AI Studio API key is missing.');
-        return response.status(400).send({ error: true });
+    if (useVertexAi) {
+        apiUrl = new URL(request.body.reverse_proxy || API_VERTEX_AI);
+        apiKey = request.body.reverse_proxy ? request.body.proxy_password : readSecret(request.user.directories, SECRET_KEYS.VERTEXAI);
+
+        if (!request.body.reverse_proxy && !apiKey) {
+            console.warn(`${apiName} API key is missing.`);
+            return response.status(400).send({ error: true });
+        }
+    } else {
+        apiUrl = new URL(request.body.reverse_proxy || API_MAKERSUITE);
+        apiKey = request.body.reverse_proxy ? request.body.proxy_password : readSecret(request.user.directories, SECRET_KEYS.MAKERSUITE);
+
+        if (!request.body.reverse_proxy && !apiKey) {
+            console.warn(`${apiName} API key is missing.`);
+            return response.status(400).send({ error: true });
+        }
     }
 
     const model = String(request.body.model);
@@ -465,7 +481,7 @@ async function sendMakerSuiteRequest(request, response) {
     }
 
     const body = getGeminiBody();
-    console.debug('Google AI Studio request:', body);
+    console.debug(`${apiName} request:`, body);
 
     try {
         const controller = new AbortController();
@@ -477,7 +493,13 @@ async function sendMakerSuiteRequest(request, response) {
         const apiVersion = getConfigValue('gemini.apiVersion', 'v1beta');
         const responseType = (stream ? 'streamGenerateContent' : 'generateContent');
 
-        const generateResponse = await fetch(`${apiUrl.toString().replace(/\/$/, '')}/${apiVersion}/models/${model}:${responseType}?key=${apiKey}${stream ? '&alt=sse' : ''}`, {
+        let url;
+        if (useVertexAi) {
+            url = `${apiUrl.toString().replace(/\/$/, '')}/v1/publishers/google/models/${model}:${responseType}?key=${apiKey}${stream ? '&alt=sse' : ''}`;
+        } else {
+            url = `${apiUrl.toString().replace(/\/$/, '')}/${apiVersion}/models/${model}:${responseType}?key=${apiKey}${stream ? '&alt=sse' : ''}`;
+        }
+        const generateResponse = await fetch(url, {
             body: JSON.stringify(body),
             method: 'POST',
             headers: {
@@ -498,7 +520,7 @@ async function sendMakerSuiteRequest(request, response) {
             }
         } else {
             if (!generateResponse.ok) {
-                console.warn(`Google AI Studio API returned error: ${generateResponse.status} ${generateResponse.statusText} ${await generateResponse.text()}`);
+                console.warn(`${apiName} API returned error: ${generateResponse.status} ${generateResponse.statusText} ${await generateResponse.text()}`);
                 return response.status(500).send({ error: true });
             }
 
@@ -507,7 +529,7 @@ async function sendMakerSuiteRequest(request, response) {
 
             const candidates = generateResponseJson?.candidates;
             if (!candidates || candidates.length === 0) {
-                let message = 'Google AI Studio API returned no candidate';
+                let message = `${apiName} API returned no candidate`;
                 console.warn(message, generateResponseJson);
                 if (generateResponseJson?.promptFeedback?.blockReason) {
                     message += `\nPrompt was blocked due to : ${generateResponseJson.promptFeedback.blockReason}`;
@@ -518,11 +540,11 @@ async function sendMakerSuiteRequest(request, response) {
             const responseContent = candidates[0].content ?? candidates[0].output;
             const functionCall = (candidates?.[0]?.content?.parts ?? []).some(part => part.functionCall);
             const inlineData = (candidates?.[0]?.content?.parts ?? []).some(part => part.inlineData);
-            console.debug('Google AI Studio response:', util.inspect(generateResponseJson, { depth: 5, colors: true }));
+            console.debug(`${apiName} response:`, util.inspect(generateResponseJson, { depth: 5, colors: true }));
 
             const responseText = typeof responseContent === 'string' ? responseContent : responseContent?.parts?.filter(part => !part.thought)?.map(part => part.text)?.join('\n\n');
             if (!responseText && !functionCall && !inlineData) {
-                let message = 'Google AI Studio Candidate text empty';
+                let message = `${apiName} Candidate text empty`;
                 console.warn(message, generateResponseJson);
                 return response.send({ error: { message } });
             }
@@ -532,7 +554,7 @@ async function sendMakerSuiteRequest(request, response) {
             return response.send(reply);
         }
     } catch (error) {
-        console.error('Error communicating with Google AI Studio API: ', error);
+        console.error(`Error communicating with ${apiName} API:`, error);
         if (!response.headersSent) {
             return response.status(500).send({ error: true });
         }
@@ -1189,6 +1211,7 @@ router.post('/generate', function (request, response) {
         case CHAT_COMPLETION_SOURCES.SCALE: return sendScaleRequest(request, response);
         case CHAT_COMPLETION_SOURCES.AI21: return sendAI21Request(request, response);
         case CHAT_COMPLETION_SOURCES.MAKERSUITE: return sendMakerSuiteRequest(request, response);
+        case CHAT_COMPLETION_SOURCES.VERTEXAI: return sendMakerSuiteRequest(request, response);
         case CHAT_COMPLETION_SOURCES.MISTRALAI: return sendMistralAIRequest(request, response);
         case CHAT_COMPLETION_SOURCES.COHERE: return sendCohereRequest(request, response);
         case CHAT_COMPLETION_SOURCES.DEEPSEEK: return sendDeepSeekRequest(request, response);
diff --git a/src/endpoints/google.js b/src/endpoints/google.js
index fa8a8b5f8..435a3d388 100644
--- a/src/endpoints/google.js
+++ b/src/endpoints/google.js
@@ -7,6 +7,7 @@ import { readSecret, SECRET_KEYS } from './secrets.js';
 import { GEMINI_SAFETY } from '../constants.js';
 
 const API_MAKERSUITE = 'https://generativelanguage.googleapis.com';
+const API_VERTEX_AI = 'https://us-central1-aiplatform.googleapis.com';
 
 export const router = express.Router();
 
@@ -14,12 +15,27 @@ router.post('/caption-image', async (request, response) => {
     try {
         const mimeType = request.body.image.split(';')[0].split(':')[1];
         const base64Data = request.body.image.split(',')[1];
-        const apiKey = request.body.reverse_proxy ? request.body.proxy_password : readSecret(request.user.directories, SECRET_KEYS.MAKERSUITE);
-        const apiUrl = new URL(request.body.reverse_proxy || API_MAKERSUITE);
+        const useVertexAi = request.body.api === 'vertexai';
+        const apiName = useVertexAi ? 'Google Vertex AI' : 'Google AI Studio';
+        let apiKey;
+        let apiUrl;
+        if (useVertexAi) {
+            apiKey = request.body.reverse_proxy ? request.body.proxy_password : readSecret(request.user.directories, SECRET_KEYS.VERTEXAI);
+            apiUrl = new URL(request.body.reverse_proxy || API_VERTEX_AI);
+        } else {
+            apiKey = request.body.reverse_proxy ? request.body.proxy_password : readSecret(request.user.directories, SECRET_KEYS.MAKERSUITE);
+            apiUrl = new URL(request.body.reverse_proxy || API_MAKERSUITE);
+        }
         const model = request.body.model || 'gemini-2.0-flash';
-        const url = `${apiUrl.origin}/v1beta/models/${model}:generateContent?key=${apiKey}`;
+        let url;
+        if (useVertexAi) {
+            url = `${apiUrl.origin}/v1/publishers/google/models/${model}:generateContent?key=${apiKey}`;
+        } else {
+            url = `${apiUrl.origin}/v1beta/models/${model}:generateContent?key=${apiKey}`;
+        }
         const body = {
             contents: [{
+                role: 'user',
                 parts: [
                     { text: request.body.prompt },
                     {
@@ -32,7 +48,7 @@ router.post('/caption-image', async (request, response) => {
             safetySettings: GEMINI_SAFETY,
         };
 
-        console.debug('Multimodal captioning request', model, body);
+        console.debug(`${apiName} captioning request`, model, body);
 
         const result = await fetch(url, {
             body: JSON.stringify(body),
@@ -44,13 +60,13 @@ router.post('/caption-image', async (request, response) => {
 
         if (!result.ok) {
             const error = await result.json();
-            console.error(`Google AI Studio API returned error: ${result.status} ${result.statusText}`, error);
+            console.error(`${apiName} API returned error: ${result.status} ${result.statusText}`, error);
             return response.status(500).send({ error: true });
         }
 
         /** @type {any} */
         const data = await result.json();
-        console.info('Multimodal captioning response', data);
+        console.info(`${apiName} captioning response`, data);
 
         const candidates = data?.candidates;
         if (!candidates) {
diff --git a/src/endpoints/secrets.js b/src/endpoints/secrets.js
index 20d0b008a..0eed2eb7f 100644
--- a/src/endpoints/secrets.js
+++ b/src/endpoints/secrets.js
@@ -26,6 +26,7 @@ export const SECRET_KEYS = {
     ONERING_URL: 'oneringtranslator_url',
     DEEPLX_URL: 'deeplx_url',
     MAKERSUITE: 'api_key_makersuite',
+    VERTEXAI: 'api_key_vertexai',
     SERPAPI: 'api_key_serpapi',
     TOGETHERAI: 'api_key_togetherai',
     MISTRALAI: 'api_key_mistralai',
diff --git a/src/endpoints/tokenizers.js b/src/endpoints/tokenizers.js
index deb6a6482..bcdae1f17 100644
--- a/src/endpoints/tokenizers.js
+++ b/src/endpoints/tokenizers.js
@@ -463,7 +463,7 @@ export function getTokenizerModel(requestModel) {
         return 'deepseek';
     }
 
-    if (requestModel.includes('gemma') || requestModel.includes('gemini')) {
+    if (requestModel.includes('gemma') || requestModel.includes('gemini') || requestModel.includes('learnlm')) {
         return 'gemma';
     }