refactor: clean up chat and markdown components, improve markdown rendering with lazy loading and streaming support

Update README.md

README.md

@@ -7,7 +7,7 @@
 
 
 
-<h1 align="center">NextChat (ChatGPT Next Web)</h1>
+<h1 align="center">NextChat</h1>
 
 English / [简体中文](./README_CN.md)
 
@@ -22,7 +22,6 @@ English / [简体中文](./README_CN.md)
 [![MacOS][MacOS-image]][download-url]
 [![Linux][Linux-image]][download-url]
 
-[NextChatAI](https://nextchat.dev/chat?utm_source=readme) / [Web App Demo](https://app.nextchat.dev) / [Desktop App](https://github.com/Yidadaa/ChatGPT-Next-Web/releases) 
 [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)
 
 
@@ -41,24 +40,6 @@ English / [简体中文](./README_CN.md)
 
 </div>
 
-## 👋 Hey, NextChat is going to develop a native app!
-
-> This week we are going to start working on iOS and Android APP, and we want to find some reliable friends to do it together!
-
-
-✨ Several key points:
-
-- Starting from 0, you are a veteran
-- Completely open source, not hidden
-- Native development, pursuing the ultimate experience
-
-Will you come and do something together? 😎
-
-https://github.com/ChatGPTNextWeb/NextChat/issues/6269
-
-#Seeking for talents is thirsty #lack of people
-
-
 ## 🥳 Cheer for DeepSeek, China's AI star!
  > Purpose-Built UI for DeepSeek Reasoner Model
  

app/client/api.ts

@@ -40,6 +40,11 @@ export interface MultimodalContent {
   };
 }
 
+export interface MultimodalContentForAlibaba {
+  text?: string;
+  image?: string;
+}
+
 export interface RequestMessage {
   role: MessageRole;
   content: string | MultimodalContent[];

app/client/platforms/alibaba.ts

@@ -7,7 +7,10 @@ import {
   ChatMessageTool,
   usePluginStore,
 } from "@/app/store";
-import { streamWithThink } from "@/app/utils/chat";
+import {
+  preProcessImageContentForAlibabaDashScope,
+  streamWithThink,
+} from "@/app/utils/chat";
 import {
   ChatOptions,
   getHeaders,
@@ -15,12 +18,14 @@ import {
   LLMModel,
   SpeechOptions,
   MultimodalContent,
+  MultimodalContentForAlibaba,
 } from "../api";
 import { getClientConfig } from "@/app/config/client";
 import {
   getMessageTextContent,
   getMessageTextContentWithoutThinking,
   getTimeoutMSByModel,
+  isVisionModel,
 } from "@/app/utils";
 import { fetch } from "@/app/utils/stream";
 
@@ -89,14 +94,6 @@ export class QwenApi implements LLMApi {
   }
 
   async chat(options: ChatOptions) {
-    const messages = options.messages.map((v) => ({
-      role: v.role,
-      content:
-        v.role === "assistant"
-          ? getMessageTextContentWithoutThinking(v)
-          : getMessageTextContent(v),
-    }));
-
     const modelConfig = {
       ...useAppConfig.getState().modelConfig,
       ...useChatStore.getState().currentSession().mask.modelConfig,
@@ -105,6 +102,21 @@ export class QwenApi implements LLMApi {
       },
     };
 
+    const visionModel = isVisionModel(options.config.model);
+
+    const messages: ChatOptions["messages"] = [];
+    for (const v of options.messages) {
+      const content = (
+        visionModel
+          ? await preProcessImageContentForAlibabaDashScope(v.content)
+          : v.role === "assistant"
+          ? getMessageTextContentWithoutThinking(v)
+          : getMessageTextContent(v)
+      ) as any;
+
+      messages.push({ role: v.role, content });
+    }
+
     const shouldStream = !!options.config.stream;
     const requestPayload: RequestPayload = {
       model: modelConfig.model,
@@ -129,7 +141,7 @@ export class QwenApi implements LLMApi {
         "X-DashScope-SSE": shouldStream ? "enable" : "disable",
       };
 
-      const chatPath = this.path(Alibaba.ChatPath);
+      const chatPath = this.path(Alibaba.ChatPath(modelConfig.model));
       const chatPayload = {
         method: "POST",
         body: JSON.stringify(requestPayload),
@@ -162,7 +174,7 @@ export class QwenApi implements LLMApi {
             const json = JSON.parse(text);
             const choices = json.output.choices as Array<{
               message: {
-                content: string | null;
+                content: string | null | MultimodalContentForAlibaba[];
                 tool_calls: ChatMessageTool[];
                 reasoning_content: string | null;
               };
@@ -212,7 +224,9 @@ export class QwenApi implements LLMApi {
             } else if (content && content.length > 0) {
               return {
                 isThinking: false,
-                content: content,
+                content: Array.isArray(content)
+                  ? content.map((item) => item.text).join(",")
+                  : content,
               };
             }
 

app/components/chat.tsx

@@ -18,7 +18,6 @@ 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";
@@ -79,8 +78,6 @@ 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";
@@ -125,14 +122,15 @@ 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();
@@ -1984,6 +1982,8 @@ 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

@@ -267,6 +267,136 @@ 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));
@@ -326,9 +456,27 @@ 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
@@ -344,6 +492,24 @@ 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

@@ -221,7 +221,12 @@ export const ByteDance = {
 
 export const Alibaba = {
   ExampleEndpoint: ALIBABA_BASE_URL,
-  ChatPath: "v1/services/aigc/text-generation/generation",
+  ChatPath: (modelName: string) => {
+    if (modelName.includes("vl") || modelName.includes("omni")) {
+      return "v1/services/aigc/multimodal-generation/generation";
+    }
+    return `v1/services/aigc/text-generation/generation`;
+  },
 };
 
 export const Tencent = {
@@ -570,6 +575,9 @@ const alibabaModes = [
   "qwen-max-0403",
   "qwen-max-0107",
   "qwen-max-longcontext",
+  "qwen-omni-turbo",
+  "qwen-vl-plus",
+  "qwen-vl-max",
 ];
 
 const tencentModels = [

app/styles/markdown.scss

@@ -99,6 +99,7 @@
   font-size: 14px;
   line-height: 1.5;
   word-wrap: break-word;
+  margin-bottom: 0;
 }
 
 .light {
@@ -358,8 +359,14 @@
 .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;
@@ -448,16 +455,28 @@
 .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;
 }
@@ -1130,3 +1149,87 @@
 #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;
+  }
+}

app/utils/chat.ts

@@ -3,7 +3,7 @@ import {
   UPLOAD_URL,
   REQUEST_TIMEOUT_MS,
 } from "@/app/constant";
-import { RequestMessage } from "@/app/client/api";
+import { MultimodalContent, RequestMessage } from "@/app/client/api";
 import Locale from "@/app/locales";
 import {
   EventStreamContentType,
@@ -70,8 +70,9 @@ export function compressImage(file: Blob, maxSize: number): Promise<string> {
   });
 }
 
-export async function preProcessImageContent(
+export async function preProcessImageContentBase(
   content: RequestMessage["content"],
+  transformImageUrl: (url: string) => Promise<{ [key: string]: any }>,
 ) {
   if (typeof content === "string") {
     return content;
@@ -81,7 +82,7 @@ export async function preProcessImageContent(
     if (part?.type == "image_url" && part?.image_url?.url) {
       try {
         const url = await cacheImageToBase64Image(part?.image_url?.url);
-        result.push({ type: part.type, image_url: { url } });
+        result.push(await transformImageUrl(url));
       } catch (error) {
         console.error("Error processing image URL:", error);
       }
@@ -92,6 +93,23 @@ export async function preProcessImageContent(
   return result;
 }
 
+export async function preProcessImageContent(
+  content: RequestMessage["content"],
+) {
+  return preProcessImageContentBase(content, async (url) => ({
+    type: "image_url",
+    image_url: { url },
+  })) as Promise<MultimodalContent[] | string>;
+}
+
+export async function preProcessImageContentForAlibabaDashScope(
+  content: RequestMessage["content"],
+) {
+  return preProcessImageContentBase(content, async (url) => ({
+    image: url,
+  }));
+}
+
 const imageCaches: Record<string, string> = {};
 export function cacheImageToBase64Image(imageUrl: string) {
   if (imageUrl.includes(CACHE_URL_PREFIX)) {

jest.config.ts

@@ -15,6 +15,8 @@ const config: Config = {
   moduleNameMapper: {
     "^@/(.*)$": "<rootDir>/$1",
   },
+  extensionsToTreatAsEsm: [".ts", ".tsx"],
+  injectGlobals: true,
 };
 
 // createJestConfig is exported this way to ensure that next/jest can load the Next.js config which is async

jest.setup.ts

@@ -1,24 +1,22 @@
 // Learn more: https://github.com/testing-library/jest-dom
 import "@testing-library/jest-dom";
+import { jest } from "@jest/globals";
 
 global.fetch = jest.fn(() =>
   Promise.resolve({
     ok: true,
     status: 200,
-    json: () => Promise.resolve({}),
+    json: () => Promise.resolve([]),
     headers: new Headers(),
     redirected: false,
     statusText: "OK",
     type: "basic",
     url: "",
-    clone: function () {
-      return this;
-    },
     body: null,
     bodyUsed: false,
     arrayBuffer: () => Promise.resolve(new ArrayBuffer(0)),
     blob: () => Promise.resolve(new Blob()),
     formData: () => Promise.resolve(new FormData()),
     text: () => Promise.resolve(""),
-  }),
+  } as Response),
 );

package.json

@@ -17,8 +17,8 @@
     "prompts": "node ./scripts/fetch-prompts.mjs",
     "prepare": "husky install",
     "proxy-dev": "sh ./scripts/init-proxy.sh && proxychains -f ./scripts/proxychains.conf yarn dev",
-    "test": "jest --watch",
-    "test:ci": "jest --ci"
+    "test": "node --no-warnings --experimental-vm-modules $(yarn bin jest) --watch",
+    "test:ci": "node --no-warnings --experimental-vm-modules $(yarn bin jest) --ci"
   },
   "dependencies": {
     "@fortaine/fetch-event-source": "^3.0.6",

test/vision-model-checker.test.ts

@@ -1,3 +1,4 @@
+import { jest } from "@jest/globals";
 import { isVisionModel } from "../app/utils";
 
 describe("isVisionModel", () => {
@@ -50,7 +51,7 @@ describe("isVisionModel", () => {
 
   test("should identify models from VISION_MODELS env var", () => {
     process.env.VISION_MODELS = "custom-vision-model,another-vision-model";
-    
+
     expect(isVisionModel("custom-vision-model")).toBe(true);
     expect(isVisionModel("another-vision-model")).toBe(true);
     expect(isVisionModel("unrelated-model")).toBe(false);
@@ -64,4 +65,4 @@ describe("isVisionModel", () => {
     expect(isVisionModel("unrelated-model")).toBe(false);
     expect(isVisionModel("gpt-4-vision")).toBe(true);
   });
-});
+});