@lobehub/chat 1.15.8 → 1.15.10

package/.i18nrc.js

@@ -24,7 +24,6 @@ module.exports = defineConfig({
   ],
   temperature: 0,
   modelName: 'gpt-4o-mini',
-  splitToken: 2048,
   experimental: {
     jsonMode: true,
   },

package/CHANGELOG.md

@@ -2,6 +2,48 @@
 
 # Changelog
 
+### [Version 1.15.10](https://github.com/lobehub/lobe-chat/compare/v1.15.9...v1.15.10)
+
+<sup>Released on **2024-09-03**</sup>
+
+<br/>
+
+<details>
+<summary><kbd>Improvements and Fixes</kbd></summary>
+
+</details>
+
+<div align="right">
+
+[![](https://img.shields.io/badge/-BACK_TO_TOP-151515?style=flat-square)](#readme-top)
+
+</div>
+
+### [Version 1.15.9](https://github.com/lobehub/lobe-chat/compare/v1.15.8...v1.15.9)
+
+<sup>Released on **2024-09-03**</sup>
+
+#### 🐛 Bug Fixes
+
+- **misc**: Fix speed and rag prompt.
+
+<br/>
+
+<details>
+<summary><kbd>Improvements and Fixes</kbd></summary>
+
+#### What's fixed
+
+- **misc**: Fix speed and rag prompt, closes [#3751](https://github.com/lobehub/lobe-chat/issues/3751) ([dce200c](https://github.com/lobehub/lobe-chat/commit/dce200c))
+
+</details>
+
+<div align="right">
+
+[![](https://img.shields.io/badge/-BACK_TO_TOP-151515?style=flat-square)](#readme-top)
+
+</div>
+
 ### [Version 1.15.8](https://github.com/lobehub/lobe-chat/compare/v1.15.7...v1.15.8)
 
 <sup>Released on **2024-09-03**</sup>

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@lobehub/chat",
-  "version": "1.15.8",
+  "version": "1.15.10",
   "description": "Lobe Chat - an open-source, high-performance chatbot framework that supports speech synthesis, multimodal, and extensible Function Call plugin system. Supports one-click free deployment of your private ChatGPT/LLM web application.",
   "keywords": [
     "framework",

package/src/app/(main)/profile/[[...slugs]]/Client.tsx

@@ -56,6 +56,7 @@ export const useStyles = createStyles(
         border-radius: unset;
       `,
     }) as Partial<{
+      // eslint-disable-next-line unused-imports/no-unused-vars
       [k in keyof ElementsConfig]: any;
     }>,
 );

package/src/app/webapi/tokenizer/index.test.ts

@@ -0,0 +1,32 @@
+// @vitest-environment edge-runtime
+import { describe, expect, it } from 'vitest';
+
+import { POST } from './route';
+
+describe('tokenizer Route', () => {
+  it('count hello world', async () => {
+    const txt = 'Hello, world!';
+    const request = new Request('https://test.com', {
+      method: 'POST',
+      body: txt,
+    });
+
+    const response = await POST(request);
+
+    const data = await response.json();
+    expect(data.count).toEqual(4);
+  });
+
+  it('count Chinese', async () => {
+    const txt = '今天天气真好';
+    const request = new Request('https://test.com', {
+      method: 'POST',
+      body: txt,
+    });
+
+    const response = await POST(request);
+
+    const data = await response.json();
+    expect(data.count).toEqual(5);
+  });
+});

package/src/app/webapi/tokenizer/route.ts

@@ -0,0 +1,8 @@
+import { encode } from 'gpt-tokenizer/encoding/o200k_base';
+import { NextResponse } from 'next/server';
+
+export const POST = async (req: Request) => {
+  const str = await req.text();
+
+  return NextResponse.json({ count: encode(str).length });
+};

package/src/chains/answerWithContext.ts

@@ -11,7 +11,7 @@ export const chainAnswerWithContext = ({
 }): Partial<ChatStreamPayload> => ({
   messages: [
     {
-      content: `You are a helpful assistant good answering questions related to ${knowledge.join('/')}. And you'll be provided with a question and several passages that might be relevant. And your task is to provide answer based on the question and passages.
+      content: `You are aslo a helpful assistant good answering questions related to ${knowledge.join('/')}. And you'll be provided with a question and several passages that might be relevant. And currently your task is to provide answer based on the question and passages.
 
 Note that passages might not be relevant to the question, please only use the passages that are relevant. Or if there is no relevant passage, please answer using your knowledge.
 
@@ -21,13 +21,12 @@ The provided passages as context:
 
 <Context>
 ${context.join('\n')}
-</Context>`,
-      role: 'system',
-    },
-    {
-      content: `The question to answer:
+</Context>
+
+The question to answer is:
 
-${question}`,
+${question}
+`,
       role: 'user',
     },
   ],

package/src/database/server/models/session.ts

@@ -170,7 +170,7 @@ export class SessionModel {
 
     if (!result) return;
 
-    // eslint-disable-next-line @typescript-eslint/no-unused-vars
+    // eslint-disable-next-line @typescript-eslint/no-unused-vars,unused-imports/no-unused-vars
     const { agent, clientId, ...session } = result;
     const sessionId = this.genId();
 

package/src/hooks/useTokenCount.ts

@@ -1,20 +1,32 @@
-import { startTransition, useEffect, useState } from 'react';
+import { debounce } from 'lodash-es';
+import { startTransition, useCallback, useEffect, useState } from 'react';
 
 import { encodeAsync } from '@/utils/tokenizer';
 
 export const useTokenCount = (input: string = '') => {
   const [value, setNum] = useState(0);
 
-  useEffect(() => {
-    startTransition(() => {
-      encodeAsync(input || '')
+  const debouncedEncode = useCallback(
+    debounce((text: string) => {
+      encodeAsync(text)
         .then(setNum)
         .catch(() => {
-          // 兜底采用字符数
-          setNum(input.length);
+          setNum(text.length);
         });
+    }, 300),
+    [],
+  );
+
+  useEffect(() => {
+    startTransition(() => {
+      debouncedEncode(input || '');
     });
-  }, [input]);
+
+    // 清理函数
+    return () => {
+      debouncedEncode.cancel();
+    };
+  }, [input, debouncedEncode]);
 
   return value;
 };

package/src/layout/AuthProvider/Clerk/useAppearance.ts

@@ -89,6 +89,7 @@ export const useStyles = createStyles(
         order: -1;
       `,
     }) as Partial<{
+      // eslint-disable-next-line unused-imports/no-unused-vars
       [k in keyof ElementsConfig]: any;
     }>,
 );

package/src/server/context.ts

@@ -62,7 +62,9 @@ export const createContext = async (request: NextRequest): Promise<Context> => {
         userId = session.user.id;
       }
       return createContextInner({ authorizationHeader: authorization, nextAuth: auth, userId });
-    } catch {}
+    } catch (e) {
+      console.error('next auth err', e);
+    }
   }
 
   return createContextInner({ authorizationHeader: authorization, userId });

package/src/server/routers/edge/index.ts

@@ -1,5 +1,5 @@
 /**
- * This file contains the root router of Lobe Chat tRPC-backend
+ * This file contains the edge router of Lobe Chat tRPC-backend
  */
 import { publicProcedure, router } from '@/libs/trpc';
 

package/src/services/upload.ts

@@ -35,7 +35,7 @@ class UploadService {
           // so make it as 99.9 and let users think it's still uploading
           progress: progress === 100 ? 99.9 : progress,
           restTime: (event.total - event.loaded) / speedInByte,
-          speed: speedInByte / 1024,
+          speed: speedInByte,
         });
       }
     });

package/src/services/user/client.ts

@@ -50,7 +50,7 @@ export class ClientService implements IUserService {
     await this.preferenceStorage.saveToLocalStorage(preference);
   }
 
-  // eslint-disable-next-line @typescript-eslint/no-unused-vars
+  // eslint-disable-next-line @typescript-eslint/no-unused-vars,unused-imports/no-unused-vars
   async updateGuide(guide: Partial<UserGuide>) {
     throw new Error('Method not implemented.');
   }

package/src/types/files/upload.ts

@@ -9,7 +9,7 @@ export interface FileUploadState {
    */
   restTime: number;
   /**
-   * upload speed in KB/s
+   * upload speed in Byte/s
    */
   speed: number;
 }

package/src/types/worker.d.ts

@@ -0,0 +1,7 @@
+declare module '*.worker.ts' {
+  class WebpackWorker extends Worker {
+    constructor();
+  }
+
+  export default WebpackWorker;
+}

package/src/utils/fetch/__tests__/fetchSSE.test.ts

@@ -139,8 +139,8 @@ describe('fetchSSE', () => {
       onFinish: mockOnFinish,
     });
 
-    expect(mockOnMessageHandle).toHaveBeenNthCalledWith(1, { text: 'He', type: 'text' });
-    expect(mockOnMessageHandle).toHaveBeenNthCalledWith(2, { text: 'llo World', type: 'text' });
+    expect(mockOnMessageHandle).toHaveBeenNthCalledWith(1, { text: 'Hell', type: 'text' });
+    expect(mockOnMessageHandle).toHaveBeenNthCalledWith(2, { text: 'o World', type: 'text' });
     // more assertions for each character...
     expect(mockOnFinish).toHaveBeenCalledWith('Hello World', {
       observationId: null,
@@ -232,8 +232,8 @@ describe('fetchSSE', () => {
       signal: abortController.signal,
     });
 
-    expect(mockOnMessageHandle).toHaveBeenNthCalledWith(1, { text: 'He', type: 'text' });
-    expect(mockOnMessageHandle).toHaveBeenNthCalledWith(2, { text: 'llo World', type: 'text' });
+    expect(mockOnMessageHandle).toHaveBeenNthCalledWith(1, { text: 'Hell', type: 'text' });
+    expect(mockOnMessageHandle).toHaveBeenNthCalledWith(2, { text: 'o World', type: 'text' });
 
     expect(mockOnFinish).toHaveBeenCalledWith('Hello World', {
       type: 'done',

package/src/utils/fetch/fetchSSE.ts

@@ -44,6 +44,10 @@ export interface FetchSSEOptions {
   smoothing?: boolean;
 }
 
+const START_ANIMATION_SPEED = 4;
+
+const END_ANIMATION_SPEED = 15;
+
 const createSmoothMessage = (params: { onTextUpdate: (delta: string, text: string) => void }) => {
   let buffer = '';
   // why use queue: https://shareg.pt/GLBrjpK
@@ -64,7 +68,7 @@ const createSmoothMessage = (params: { onTextUpdate: (delta: string, text: strin
 
   // define startAnimation function to display the text in buffer smooth
   // when you need to start the animation, call this function
-  const startAnimation = (speed = 2) =>
+  const startAnimation = (speed = START_ANIMATION_SPEED) =>
     new Promise<void>((resolve) => {
       if (isAnimationActive) {
         resolve();
@@ -137,7 +141,7 @@ const createSmoothToolCalls = (params: {
     }
   };
 
-  const startAnimation = (index: number, speed = 2) =>
+  const startAnimation = (index: number, speed = START_ANIMATION_SPEED) =>
     new Promise<void>((resolve) => {
       if (isAnimationActives[index]) {
         resolve();
@@ -191,7 +195,7 @@ const createSmoothToolCalls = (params: {
     });
   };
 
-  const startAnimations = async (speed = 2) => {
+  const startAnimations = async (speed = START_ANIMATION_SPEED) => {
     const pools = toolCallsBuffer.map(async (_, index) => {
       if (outputQueues[index].length > 0 && !isAnimationActives[index]) {
         await startAnimation(index, speed);
@@ -365,11 +369,11 @@ export const fetchSSE = async (url: string, options: RequestInit & FetchSSEOptio
       const observationId = response.headers.get(LOBE_CHAT_OBSERVATION_ID);
 
       if (textController.isTokenRemain()) {
-        await textController.startAnimation(15);
+        await textController.startAnimation(END_ANIMATION_SPEED);
       }
 
       if (toolCallsController.isTokenRemain()) {
-        await toolCallsController.startAnimations(15);
+        await toolCallsController.startAnimations(END_ANIMATION_SPEED);
       }
 
       await options?.onFinish?.(output, { observationId, toolCalls, traceId, type: finishedType });

package/src/utils/tokenizer/client.ts

@@ -0,0 +1,35 @@
+let worker: Worker | null = null;
+
+const getWorker = () => {
+  if (!worker && typeof Worker !== 'undefined') {
+    worker = new Worker(new URL('tokenizer.worker.ts', import.meta.url));
+  }
+  return worker;
+};
+
+export const clientEncodeAsync = (str: string): Promise<number> =>
+  new Promise((resolve, reject) => {
+    const worker = getWorker();
+
+    if (!worker) {
+      // 如果 WebWorker 不可用，回退到字符串计算
+      resolve(str.length);
+      return;
+    }
+
+    const id = Date.now().toString();
+
+    const handleMessage = (event: MessageEvent) => {
+      if (event.data.id === id) {
+        worker.removeEventListener('message', handleMessage);
+        if (event.data.error) {
+          reject(new Error(event.data.error));
+        } else {
+          resolve(event.data.result);
+        }
+      }
+    };
+
+    worker.addEventListener('message', handleMessage);
+    worker.postMessage({ id, str });
+  });

package/src/utils/tokenizer/index.ts

@@ -0,0 +1,15 @@
+export const encodeAsync = async (str: string): Promise<number> => {
+  if (str.length === 0) return 0;
+
+  // 50_000 is the limit of the client
+  // if the string is longer than 100_000, we will use the server
+  if (str.length <= 50_000) {
+    const { clientEncodeAsync } = await import('./client');
+
+    return await clientEncodeAsync(str);
+  } else {
+    const { serverEncodeAsync } = await import('./server');
+
+    return await serverEncodeAsync(str);
+  }
+};

package/src/utils/tokenizer/server.ts

@@ -0,0 +1,11 @@
+export const serverEncodeAsync = async (str: string): Promise<number> => {
+  try {
+    const res = await fetch('/webapi/tokenizer', { body: str, method: 'POST' });
+    const data = await res.json();
+
+    return data.count;
+  } catch (e) {
+    console.error('serverEncodeAsync:', e);
+    return str.length;
+  }
+};

package/src/utils/tokenizer/tokenizer.worker.ts

@@ -0,0 +1,14 @@
+addEventListener('message', async (event) => {
+  const { id, str } = event.data;
+  try {
+    const { encode } = await import('gpt-tokenizer');
+
+    console.time('client tokenizer');
+    const tokenCount = encode(str).length;
+    console.timeEnd('client tokenizer');
+
+    postMessage({ id, result: tokenCount });
+  } catch (error) {
+    postMessage({ error: (error as Error).message, id });
+  }
+});

package/src/utils/tokenizer.ts

@@ -1,5 +0,0 @@
-export const encodeAsync = async (str: string) => {
-  const { encode } = await import('gpt-tokenizer');
-
-  return encode(str).length;
-};