Merge pull request #4564 from MrrDrr/gpt4v_remove_max_tokens

remove max_tokens from the official version of gpt4-turbo

app/client/platforms/google.ts

@@ -21,11 +21,10 @@ export class GeminiProApi implements LLMApi {
   }
   async chat(options: ChatOptions): Promise<void> {
     // const apiClient = this;
-    const visionModel = isVisionModel(options.config.model);
     let multimodal = false;
     const messages = options.messages.map((v) => {
       let parts: any[] = [{ text: getMessageTextContent(v) }];
-      if (visionModel) {
+      if (isVisionModel(options.config.model)) {
         const images = getMessageImages(v);
         if (images.length > 0) {
           multimodal = true;
@@ -117,17 +116,12 @@ export class GeminiProApi implements LLMApi {
     const controller = new AbortController();
     options.onController?.(controller);
     try {
-      let googleChatPath = visionModel
-        ? Google.VisionChatPath(modelConfig.model)
-        : Google.ChatPath(modelConfig.model);
-      let chatPath = this.path(googleChatPath);
-
       // let baseUrl = accessStore.googleUrl;
 
       if (!baseUrl) {
         baseUrl = isApp
-          ? DEFAULT_API_HOST + "/api/proxy/google/" + googleChatPath
-          : chatPath;
+          ? DEFAULT_API_HOST + "/api/proxy/google/" + Google.ChatPath(modelConfig.model)
+          : this.path(Google.ChatPath(modelConfig.model));
       }
 
       if (isApp) {
@@ -145,6 +139,7 @@ export class GeminiProApi implements LLMApi {
         () => controller.abort(),
         REQUEST_TIMEOUT_MS,
       );
+      
       if (shouldStream) {
         let responseText = "";
         let remainText = "";

app/client/platforms/openai.ts

@@ -129,7 +129,7 @@ export class ChatGPTApi implements LLMApi {
     };
 
     // add max_tokens to vision model
-    if (visionModel) {
+    if (visionModel && modelConfig.model.includes("preview")) {
       requestPayload["max_tokens"] = Math.max(modelConfig.max_tokens, 4000);
     }
 

app/constant.ts

@@ -99,7 +99,6 @@ export const Azure = {
 export const Google = {
   ExampleEndpoint: "https://generativelanguage.googleapis.com/",
   ChatPath: (modelName: string) => `v1beta/models/${modelName}:generateContent`,
-  VisionChatPath: (modelName: string) => `v1beta/models/${modelName}:generateContent`,
 };
 
 export const DEFAULT_INPUT_TEMPLATE = `{{input}}`; // input / time / model / lang
@@ -128,8 +127,6 @@ export const KnowledgeCutOffDate: Record<string, string> = {
   "gpt-4-turbo": "2023-12",
   "gpt-4-turbo-2024-04-09": "2023-12",
   "gpt-4-turbo-preview": "2023-12",
-  "gpt-4-1106-preview": "2023-04",
-  "gpt-4-0125-preview": "2023-12",
   "gpt-4-vision-preview": "2023-04",
   // After improvements,
   // it's now easier to add "KnowledgeCutOffDate" instead of stupid hardcoding it, as was done previously.
@@ -139,19 +136,11 @@ export const KnowledgeCutOffDate: Record<string, string> = {
 
 const openaiModels = [
   "gpt-3.5-turbo",
-  "gpt-3.5-turbo-0301",
-  "gpt-3.5-turbo-0613",
   "gpt-3.5-turbo-1106",
   "gpt-3.5-turbo-0125",
-  "gpt-3.5-turbo-16k",
-  "gpt-3.5-turbo-16k-0613",
   "gpt-4",
-  "gpt-4-0314",
   "gpt-4-0613",
-  "gpt-4-1106-preview",
-  "gpt-4-0125-preview",
   "gpt-4-32k",
-  "gpt-4-32k-0314",
   "gpt-4-32k-0613",
   "gpt-4-turbo",
   "gpt-4-turbo-preview",

app/styles/globals.scss

@@ -86,6 +86,7 @@
     @include dark;
   }
 }
+
 html {
   height: var(--full-height);
 
@@ -110,6 +111,10 @@ body {
   @media only screen and (max-width: 600px) {
     background-color: var(--second);
   }
+
+  *:focus-visible {
+    outline: none;
+  }
 }
 
 ::-webkit-scrollbar {