Merge bc53c17a8c into c30ddfbb07

Fix: Encourage markdown inclusion in model responses for o1/o3

README.md

@@ -22,7 +22,7 @@ English / [简体中文](./README_CN.md)
 [![MacOS][MacOS-image]][download-url]
 [![Linux][Linux-image]][download-url]
 
-[NextChatAI](https://nextchat.club?utm_source=readme) / [Web App Demo](https://app.nextchat.dev) / [Desktop App](https://github.com/Yidadaa/ChatGPT-Next-Web/releases) / [Discord](https://discord.gg/YCkeafCafC) / [Enterprise Edition](#enterprise-edition) / [Twitter](https://twitter.com/NextChatDev)
+[NextChatAI](https://nextchat.club?utm_source=readme) / [iOS APP](https://apps.apple.com/us/app/nextchat-ai/id6743085599) / [Web App Demo](https://app.nextchat.dev) / [Desktop App](https://github.com/Yidadaa/ChatGPT-Next-Web/releases) / [Enterprise Edition](#enterprise-edition) 
 
 
 [saas-url]: https://nextchat.club?utm_source=readme
@@ -41,7 +41,9 @@ English / [简体中文](./README_CN.md)
 </div>
 
 ## 🥳 Cheer for NextChat iOS Version Online!
- > [ 👉 Click Here Install Now](https://apps.apple.com/us/app/nextchat-ai/id6743085599)
+> [👉 Click Here to Install Now](https://apps.apple.com/us/app/nextchat-ai/id6743085599)
+
+> [❤️ Source Code Coming Soon](https://github.com/ChatGPTNextWeb/NextChat-iOS)
 
 ![Github iOS Image](https://github.com/user-attachments/assets/e0aa334f-4c13-4dc9-8310-e3b09fa4b9f3)
 

app/client/platforms/openai.ts

@@ -56,7 +56,7 @@ export interface OpenAIListModelResponse {
 
 export interface RequestPayload {
   messages: {
-    role: "system" | "user" | "assistant";
+    role: "developer" | "system" | "user" | "assistant";
     content: string | MultimodalContent[];
   }[];
   stream?: boolean;
@@ -198,7 +198,8 @@ export class ChatGPTApi implements LLMApi {
     const isDalle3 = _isDalle3(options.config.model);
     const isO1OrO3 =
       options.config.model.startsWith("o1") ||
-      options.config.model.startsWith("o3");
+      options.config.model.startsWith("o3") ||
+      options.config.model.startsWith("o4-mini");
     if (isDalle3) {
       const prompt = getMessageTextContent(
         options.messages.slice(-1)?.pop() as any,
@@ -237,13 +238,21 @@ export class ChatGPTApi implements LLMApi {
         // Please do not ask me why not send max_tokens, no reason, this param is just shit, I dont want to explain anymore.
       };
 
-      // O1 使用 max_completion_tokens 控制token数 (https://platform.openai.com/docs/guides/reasoning#controlling-costs)
       if (isO1OrO3) {
+        // by default the o1/o3 models will not attempt to produce output that includes markdown formatting
+        // manually add "Formatting re-enabled" developer message to encourage markdown inclusion in model responses
+        // (https://learn.microsoft.com/en-us/azure/ai-services/openai/how-to/reasoning?tabs=python-secure#markdown-output)
+        requestPayload["messages"].unshift({
+          role: "developer",
+          content: "Formatting re-enabled",
+        });
+
+        // o1/o3 uses max_completion_tokens to control the number of tokens (https://platform.openai.com/docs/guides/reasoning#controlling-costs)
         requestPayload["max_completion_tokens"] = modelConfig.max_tokens;
       }
 
       // add max_tokens to vision model
-      if (visionModel) {
+      if (visionModel && !isO1OrO3) {
         requestPayload["max_tokens"] = Math.max(modelConfig.max_tokens, 4000);
       }
     }

app/constant.ts

@@ -478,6 +478,8 @@ export const VISION_MODEL_REGEXES = [
   /^dall-e-3$/, // Matches exactly "dall-e-3"
   /glm-4v/,
   /vl/i,
+  /o3/,
+  /o4-mini/,
 ];
 
 export const EXCLUDE_VISION_MODEL_REGEXES = [/claude-3-5-haiku-20241022/];
@@ -516,6 +518,8 @@ const openaiModels = [
   "o1-mini",
   "o1-preview",
   "o3-mini",
+  "o3",
+  "o4-mini",
 ];
 
 const googleModels = [
@@ -542,6 +546,7 @@ const googleModels = [
   "gemini-2.0-flash-thinking-exp-01-21",
   "gemini-2.0-pro-exp",
   "gemini-2.0-pro-exp-02-05",
+  "gemini-2.5-pro-preview-06-05",
 ];
 
 const anthropicModels = [
@@ -628,6 +633,18 @@ const xAIModes = [
   "grok-2-vision-1212",
   "grok-2-vision",
   "grok-2-vision-latest",
+  "grok-3-mini-fast-beta",
+  "grok-3-mini-fast",
+  "grok-3-mini-fast-latest",
+  "grok-3-mini-beta",
+  "grok-3-mini",
+  "grok-3-mini-latest",
+  "grok-3-fast-beta",
+  "grok-3-fast",
+  "grok-3-fast-latest",
+  "grok-3-beta",
+  "grok-3",
+  "grok-3-latest",
 ];
 
 const chatglmModels = [

app/store/prompt.ts

@@ -159,10 +159,10 @@ export const usePromptStore = createPersistStore(
       fetch(PROMPT_URL)
         .then((res) => res.json())
         .then((res) => {
-          let fetchPrompts = [res.en, res.tw, res.cn];
-          if (getLang() === "cn") {
-            fetchPrompts = fetchPrompts.reverse();
-          }
+          const lang = getLang();
+          const fetchPrompts = [res[lang], res.en, res.tw, res.cn].filter(
+            Boolean,
+          );
           const builtinPrompts = fetchPrompts.map((promptList: PromptList) => {
             return promptList.map(
               ([title, content]) =>
@@ -180,8 +180,9 @@ export const usePromptStore = createPersistStore(
           const allPromptsForSearch = builtinPrompts
             .reduce((pre, cur) => pre.concat(cur), [])
             .filter((v) => !!v.title && !!v.content);
-          SearchService.count.builtin =
-            res.en.length + res.cn.length + res.tw.length;
+          SearchService.count.builtin = Object.values(res)
+            .filter(Array.isArray)
+            .reduce((total, promptList) => total + promptList.length, 0);
           SearchService.init(allPromptsForSearch, userPrompts);
         });
     },