Merge 10fec2fed6 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
@@ -40,13 +40,14 @@ English / [简体中文](./README_CN.md)
 
 </div>
 
-## 🥳 Cheer for DeepSeek, China's AI star!
- > Purpose-Built UI for DeepSeek Reasoner Model
+## 🥳 Cheer for NextChat iOS Version Online!
+> [👉 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)
+
  
-<img src="https://github.com/user-attachments/assets/f3952210-3af1-4dc0-9b81-40eaa4847d9a"/>
-
-
-
 ## 🫣 NextChat Support MCP  ! 
 > Before build, please set env ENABLE_MCP=true
 

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/components/chat.tsx

@@ -18,6 +18,7 @@ import ReturnIcon from "../icons/return.svg";
 import CopyIcon from "../icons/copy.svg";
 import SpeakIcon from "../icons/speak.svg";
 import SpeakStopIcon from "../icons/speak-stop.svg";
+import LoadingIcon from "../icons/three-dots.svg";
 import LoadingButtonIcon from "../icons/loading.svg";
 import PromptIcon from "../icons/prompt.svg";
 import MaskIcon from "../icons/mask.svg";
@@ -78,6 +79,8 @@ import {
 
 import { uploadImage as uploadImageRemote } from "@/app/utils/chat";
 
+import dynamic from "next/dynamic";
+
 import { ChatControllerPool } from "../client/controller";
 import { DalleQuality, DalleStyle, ModelSize } from "../typing";
 import { Prompt, usePromptStore } from "../store/prompt";
@@ -122,15 +125,14 @@ import { getModelProvider } from "../utils/model";
 import { RealtimeChat } from "@/app/components/realtime-chat";
 import clsx from "clsx";
 import { getAvailableClientsCount, isMcpEnabled } from "../mcp/actions";
-import { Markdown } from "./markdown";
 
 const localStorage = safeLocalStorage();
 
 const ttsPlayer = createTTSPlayer();
 
-// const Markdown = dynamic(async () => (await import("./markdown")).Markdown, {
-//   loading: () => <LoadingIcon />,
-// });
+const Markdown = dynamic(async () => (await import("./markdown")).Markdown, {
+  loading: () => <LoadingIcon />,
+});
 
 const MCPAction = () => {
   const navigate = useNavigate();
@@ -1982,8 +1984,6 @@ function _Chat() {
                               fontFamily={fontFamily}
                               parentRef={scrollRef}
                               defaultShow={i >= messages.length - 6}
-                              immediatelyRender={i >= messages.length - 3}
-                              streaming={message.streaming}
                             />
                             {getMessageImages(message).length == 1 && (
                               <img

app/components/markdown.tsx

@@ -267,136 +267,6 @@ function tryWrapHtmlCode(text: string) {
     );
 }
 
-// Split content into paragraphs while preserving code blocks
-function splitContentIntoParagraphs(content: string) {
-  // Check for unclosed code blocks
-  const codeBlockStartCount = (content.match(/```/g) || []).length;
-  let processedContent = content;
-
-  // Add closing tag if there's an odd number of code block markers
-  if (codeBlockStartCount % 2 !== 0) {
-    processedContent = content + "\n```";
-  }
-
-  // Extract code blocks
-  const codeBlockRegex = /```[\s\S]*?```/g;
-  const codeBlocks: string[] = [];
-  let codeBlockCounter = 0;
-
-  // Replace code blocks with placeholders
-  const contentWithPlaceholders = processedContent.replace(
-    codeBlockRegex,
-    (match) => {
-      codeBlocks.push(match);
-      const placeholder = `__CODE_BLOCK_${codeBlockCounter++}__`;
-      return placeholder;
-    },
-  );
-
-  // Split by double newlines
-  const paragraphs = contentWithPlaceholders
-    .split(/\n\n+/)
-    .filter((p) => p.trim());
-
-  // Restore code blocks
-  return paragraphs.map((p) => {
-    if (p.match(/__CODE_BLOCK_\d+__/)) {
-      return p.replace(/__CODE_BLOCK_\d+__/g, (match) => {
-        const index = parseInt(match.match(/\d+/)?.[0] || "0");
-        return codeBlocks[index] || match;
-      });
-    }
-    return p;
-  });
-}
-
-// Lazy-loaded paragraph component
-function MarkdownParagraph({
-  content,
-  onLoad,
-}: {
-  content: string;
-  onLoad?: () => void;
-}) {
-  const [isLoaded, setIsLoaded] = useState(false);
-  const placeholderRef = useRef<HTMLDivElement>(null);
-  const [isVisible, setIsVisible] = useState(false);
-
-  useEffect(() => {
-    let observer: IntersectionObserver;
-    if (placeholderRef.current) {
-      observer = new IntersectionObserver(
-        (entries) => {
-          if (entries[0].isIntersecting) {
-            setIsVisible(true);
-          }
-        },
-        { threshold: 0.1, rootMargin: "200px 0px" },
-      );
-      observer.observe(placeholderRef.current);
-    }
-    return () => observer?.disconnect();
-  }, []);
-
-  useEffect(() => {
-    if (isVisible && !isLoaded) {
-      setIsLoaded(true);
-      onLoad?.();
-    }
-  }, [isVisible, isLoaded, onLoad]);
-
-  // Generate preview content
-  const previewContent = useMemo(() => {
-    if (content.startsWith("```")) {
-      return "```" + (content.split("\n")[0] || "").slice(3) + "...```";
-    }
-    return content.length > 60 ? content.slice(0, 60) + "..." : content;
-  }, [content]);
-
-  return (
-    <div className="markdown-paragraph" ref={placeholderRef}>
-      {!isLoaded ? (
-        <div className="markdown-paragraph-placeholder">{previewContent}</div>
-      ) : (
-        <_MarkDownContent content={content} />
-      )}
-    </div>
-  );
-}
-
-// Memoized paragraph component to prevent unnecessary re-renders
-const MemoizedMarkdownParagraph = React.memo(
-  ({ content }: { content: string }) => {
-    return <_MarkDownContent content={content} />;
-  },
-  (prevProps, nextProps) => prevProps.content === nextProps.content,
-);
-
-MemoizedMarkdownParagraph.displayName = "MemoizedMarkdownParagraph";
-
-// Specialized component for streaming content
-function StreamingMarkdownContent({ content }: { content: string }) {
-  const paragraphs = useMemo(
-    () => splitContentIntoParagraphs(content),
-    [content],
-  );
-  const lastParagraphRef = useRef<HTMLDivElement>(null);
-
-  return (
-    <div className="markdown-streaming-content">
-      {paragraphs.map((paragraph, index) => (
-        <div
-          key={`p-${index}-${paragraph.substring(0, 20)}`}
-          className="markdown-paragraph markdown-streaming-paragraph"
-          ref={index === paragraphs.length - 1 ? lastParagraphRef : null}
-        >
-          <MemoizedMarkdownParagraph content={paragraph} />
-        </div>
-      ))}
-    </div>
-  );
-}
-
 function _MarkDownContent(props: { content: string }) {
   const escapedContent = useMemo(() => {
     return tryWrapHtmlCode(escapeBrackets(props.content));
@@ -456,27 +326,9 @@ export function Markdown(
     fontFamily?: string;
     parentRef?: RefObject<HTMLDivElement>;
     defaultShow?: boolean;
-    immediatelyRender?: boolean;
-    streaming?: boolean; // Whether this is a streaming response
   } & React.DOMAttributes<HTMLDivElement>,
 ) {
   const mdRef = useRef<HTMLDivElement>(null);
-  const paragraphs = useMemo(
-    () => splitContentIntoParagraphs(props.content),
-    [props.content],
-  );
-  const [loadedCount, setLoadedCount] = useState(0);
-
-  // Determine rendering strategy based on props
-  const shouldAsyncRender =
-    !props.immediatelyRender && !props.streaming && paragraphs.length > 1;
-
-  useEffect(() => {
-    // Immediately render all paragraphs if specified
-    if (props.immediatelyRender) {
-      setLoadedCount(paragraphs.length);
-    }
-  }, [props.immediatelyRender, paragraphs.length]);
 
   return (
     <div
@@ -492,24 +344,6 @@ export function Markdown(
     >
       {props.loading ? (
         <LoadingIcon />
-      ) : props.streaming ? (
-        // Use specialized component for streaming content
-        <StreamingMarkdownContent content={props.content} />
-      ) : shouldAsyncRender ? (
-        <div className="markdown-content">
-          {paragraphs.map((paragraph, index) => (
-            <MarkdownParagraph
-              key={index}
-              content={paragraph}
-              onLoad={() => setLoadedCount((prev) => prev + 1)}
-            />
-          ))}
-          {loadedCount < paragraphs.length && loadedCount > 0 && (
-            <div className="markdown-paragraph-loading">
-              <LoadingIcon />
-            </div>
-          )}
-        </div>
       ) : (
         <MarkdownContent content={props.content} />
       )}

app/constant.ts

@@ -417,6 +417,14 @@ 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-4.1": "2024-06",
+  "gpt-4.1-2025-04-14": "2024-06",
+  "gpt-4.1-mini": "2024-06",
+  "gpt-4.1-mini-2025-04-14": "2024-06",
+  "gpt-4.1-nano": "2024-06",
+  "gpt-4.1-nano-2025-04-14": "2024-06",
+  "gpt-4.5-preview": "2023-10",
+  "gpt-4.5-preview-2025-02-27": "2023-10",
   "gpt-4o": "2023-10",
   "gpt-4o-2024-05-13": "2023-10",
   "gpt-4o-2024-08-06": "2023-10",
@@ -458,6 +466,7 @@ export const DEFAULT_TTS_VOICES = [
 export const VISION_MODEL_REGEXES = [
   /vision/,
   /gpt-4o/,
+  /gpt-4\.1/,
   /claude-3/,
   /gemini-1\.5/,
   /gemini-exp/,
@@ -469,6 +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/];
@@ -485,6 +496,14 @@ const openaiModels = [
   "gpt-4-32k-0613",
   "gpt-4-turbo",
   "gpt-4-turbo-preview",
+  "gpt-4.1",
+  "gpt-4.1-2025-04-14",
+  "gpt-4.1-mini",
+  "gpt-4.1-mini-2025-04-14",
+  "gpt-4.1-nano",
+  "gpt-4.1-nano-2025-04-14",
+  "gpt-4.5-preview",
+  "gpt-4.5-preview-2025-02-27",
   "gpt-4o",
   "gpt-4o-2024-05-13",
   "gpt-4o-2024-08-06",
@@ -499,6 +518,8 @@ const openaiModels = [
   "o1-mini",
   "o1-preview",
   "o3-mini",
+  "o3",
+  "o4-mini",
 ];
 
 const googleModels = [
@@ -525,6 +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 = [
@@ -611,6 +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/styles/markdown.scss

@@ -99,7 +99,6 @@
   font-size: 14px;
   line-height: 1.5;
   word-wrap: break-word;
-  margin-bottom: 0;
 }
 
 .light {
@@ -359,14 +358,8 @@
 .markdown-body kbd {
   display: inline-block;
   padding: 3px 5px;
-  font:
-    11px ui-monospace,
-    SFMono-Regular,
-    SF Mono,
-    Menlo,
-    Consolas,
-    Liberation Mono,
-    monospace;
+  font: 11px ui-monospace, SFMono-Regular, SF Mono, Menlo, Consolas,
+    Liberation Mono, monospace;
   line-height: 10px;
   color: var(--color-fg-default);
   vertical-align: middle;
@@ -455,28 +448,16 @@
 .markdown-body tt,
 .markdown-body code,
 .markdown-body samp {
-  font-family:
-    ui-monospace,
-    SFMono-Regular,
-    SF Mono,
-    Menlo,
-    Consolas,
-    Liberation Mono,
-    monospace;
+  font-family: ui-monospace, SFMono-Regular, SF Mono, Menlo, Consolas,
+    Liberation Mono, monospace;
   font-size: 12px;
 }
 
 .markdown-body pre {
   margin-top: 0;
   margin-bottom: 0;
-  font-family:
-    ui-monospace,
-    SFMono-Regular,
-    SF Mono,
-    Menlo,
-    Consolas,
-    Liberation Mono,
-    monospace;
+  font-family: ui-monospace, SFMono-Regular, SF Mono, Menlo, Consolas,
+    Liberation Mono, monospace;
   font-size: 12px;
   word-wrap: normal;
 }
@@ -1149,87 +1130,3 @@
 #dmermaid {
   display: none;
 }
-
-.markdown-content {
-  width: 100%;
-}
-
-.markdown-paragraph {
-  transition: opacity 0.3s ease;
-  margin-bottom: 0.5em;
-
-  &.markdown-paragraph-visible {
-    opacity: 1;
-  }
-
-  &.markdown-paragraph-hidden {
-    opacity: 0.7;
-  }
-}
-
-.markdown-paragraph-placeholder {
-  padding: 8px;
-  color: var(--color-fg-subtle);
-  background-color: var(--color-canvas-subtle);
-  border-radius: 6px;
-  border-left: 3px solid var(--color-border-muted);
-  white-space: nowrap;
-  overflow: hidden;
-  text-overflow: ellipsis;
-  font-family: var(--font-family-sans);
-  font-size: 14px;
-  min-height: 1.2em;
-}
-
-.markdown-paragraph-loading {
-  height: 20px;
-  background-color: var(--color-canvas-subtle);
-  border-radius: 6px;
-  margin-bottom: 8px;
-  position: relative;
-  overflow: hidden;
-
-  &::after {
-    content: "";
-    position: absolute;
-    top: 0;
-    left: 0;
-    width: 30%;
-    height: 100%;
-    background: linear-gradient(
-      90deg,
-      transparent,
-      rgba(255, 255, 255, 0.1),
-      transparent
-    );
-    animation: shimmer 1.5s infinite;
-  }
-}
-
-@keyframes shimmer {
-  0% {
-    transform: translateX(-100%);
-  }
-  100% {
-    transform: translateX(200%);
-  }
-}
-
-.markdown-streaming-content {
-  width: 100%;
-}
-
-.markdown-streaming-paragraph {
-  opacity: 1;
-  animation: fadeIn 0.3s ease-in-out;
-  margin-bottom: 0.5em;
-}
-
-@keyframes fadeIn {
-  from {
-    opacity: 0.5;
-  }
-  to {
-    opacity: 1;
-  }
-}