@lobehub/chat 0.151.8 → 0.151.9

package/CHANGELOG.md

@@ -2,6 +2,31 @@
 
 # Changelog
 
+### [Version 0.151.9](https://github.com/lobehub/lobe-chat/compare/v0.151.8...v0.151.9)
+
+<sup>Released on **2024-04-30**</sup>
+
+#### 🐛 Bug Fixes
+
+- **misc**: Minimax truncationed output.
+
+<br/>
+
+<details>
+<summary><kbd>Improvements and Fixes</kbd></summary>
+
+#### What's fixed
+
+- **misc**: Minimax truncationed output, closes [#2308](https://github.com/lobehub/lobe-chat/issues/2308) ([488f319](https://github.com/lobehub/lobe-chat/commit/488f319))
+
+</details>
+
+<div align="right">
+
+[![](https://img.shields.io/badge/-BACK_TO_TOP-151515?style=flat-square)](#readme-top)
+
+</div>
+
 ### [Version 0.151.8](https://github.com/lobehub/lobe-chat/compare/v0.151.7...v0.151.8)
 
 <sup>Released on **2024-04-30**</sup>

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@lobehub/chat",
-  "version": "0.151.8",
+  "version": "0.151.9",
   "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/libs/agent-runtime/minimax/index.test.ts

@@ -223,7 +223,6 @@ describe('LobeMinimaxAI', () => {
           model: 'text-davinci-003',
           temperature: 0.5,
           top_p: 0.8,
-          max_tokens: 100,
         };
 
         const result = instance['buildCompletionsParams'](payload);
@@ -234,7 +233,6 @@ describe('LobeMinimaxAI', () => {
           stream: true,
           temperature: 0.5,
           top_p: 0.8,
-          max_tokens: 100,
         });
       });
 
@@ -244,7 +242,6 @@ describe('LobeMinimaxAI', () => {
           model: 'text-davinci-003',
           temperature: 0,
           top_p: 0,
-          max_tokens: 100,
         };
 
         const result = instance['buildCompletionsParams'](payload);
@@ -253,7 +250,24 @@ describe('LobeMinimaxAI', () => {
           messages: [{ content: 'Hello', role: 'user' }],
           model: 'text-davinci-003',
           stream: true,
-          max_tokens: 100,
+        });
+      });
+
+      it('should include max tokens when model is abab6.5-chat', () => {
+        const payload: ChatStreamPayload = {
+          messages: [{ content: 'Hello', role: 'user' }],
+          model: 'abab6.5-chat',
+          temperature: 0,
+          top_p: 0,
+        };
+
+        const result = instance['buildCompletionsParams'](payload);
+
+        expect(result).toEqual({
+          messages: [{ content: 'Hello', role: 'user' }],
+          model: 'abab6.5-chat',
+          stream: true,
+          max_tokens: 2048,
         });
       });
     });

package/src/libs/agent-runtime/minimax/index.ts

@@ -49,7 +49,7 @@ function throwIfErrorResponse(data: MinimaxResponse) {
   });
 }
 
-function parseMinimaxResponse(chunk: string): string | undefined {
+function parseMinimaxResponse(chunk: string): MinimaxResponse | undefined {
   let body = chunk;
   if (body.startsWith('data:')) {
     body = body.slice(5).trim();
@@ -57,10 +57,7 @@ function parseMinimaxResponse(chunk: string): string | undefined {
   if (isEmpty(body)) {
     return;
   }
-  const data = JSON.parse(body) as MinimaxResponse;
-  if (data.choices?.at(0)?.delta?.content) {
-    return data.choices.at(0)?.delta.content || undefined;
-  }
+  return JSON.parse(body) as MinimaxResponse;
 }
 
 export class LobeMinimaxAI implements LobeRuntimeAI {
@@ -136,11 +133,25 @@ export class LobeMinimaxAI implements LobeRuntimeAI {
     }
   }
 
+  // the document gives the default value of max tokens, but abab6.5 and abab6.5s
+  // will meet length finished error, and output is truncationed
+  // so here fill the max tokens number to fix it
+  // https://www.minimaxi.com/document/guides/chat-model/V2
+  private getMaxTokens(model: string): number | undefined {
+    switch (model) {
+      case 'abab6.5-chat':
+      case 'abab6.5s-chat': {
+        return 2048;
+      }
+    }
+  }
+
   private buildCompletionsParams(payload: ChatStreamPayload) {
     const { temperature, top_p, ...params } = payload;
 
     return {
       ...params,
+      max_tokens: this.getMaxTokens(payload.model),
       stream: true,
       temperature: temperature === 0 ? undefined : temperature,
       top_p: top_p === 0 ? undefined : top_p,
@@ -159,7 +170,8 @@ export class LobeMinimaxAI implements LobeRuntimeAI {
       const { value, done: doneReading } = await reader.read();
       done = doneReading;
       const chunkValue = decoder.decode(value, { stream: true });
-      const text = parseMinimaxResponse(chunkValue);
+      const data = parseMinimaxResponse(chunkValue);
+      const text = data?.choices?.at(0)?.delta?.content || undefined;
       streamController?.enqueue(encoder.encode(text));
     }
 
@@ -173,12 +185,14 @@ export class LobeMinimaxAI implements LobeRuntimeAI {
     const chunkValue = decoder.decode(value, { stream: true });
     let data;
     try {
-      data = JSON.parse(chunkValue) as MinimaxResponse;
+      data = parseMinimaxResponse(chunkValue);
     } catch {
       // parse error, skip it
       return;
     }
-    throwIfErrorResponse(data);
+    if (data) {
+      throwIfErrorResponse(data);
+    }
   }
 }
 

package/src/server/globalConfig/index.ts

@@ -25,6 +25,7 @@ export const getServerGlobalConfig = () => {
     ENABLED_GROQ,
     ENABLED_PERPLEXITY,
     ENABLED_ANTHROPIC,
+    ENABLED_MINIMAX,
     ENABLED_MISTRAL,
 
     ENABLED_AZURE_OPENAI,
@@ -64,6 +65,7 @@ export const getServerGlobalConfig = () => {
       bedrock: { enabled: ENABLED_AWS_BEDROCK },
       google: { enabled: ENABLED_GOOGLE },
       groq: { enabled: ENABLED_GROQ },
+      minimax: { enabled: ENABLED_MINIMAX },
       mistral: { enabled: ENABLED_MISTRAL },
       moonshot: { enabled: ENABLED_MOONSHOT },
       ollama: {