@lobehub/chat 1.42.1 → 1.42.2

package/CHANGELOG.md

@@ -2,6 +2,39 @@
 
 # Changelog
 
+### [Version 1.42.2](https://github.com/lobehub/lobe-chat/compare/v1.42.1...v1.42.2)
+
+<sup>Released on **2024-12-31**</sup>
+
+#### ♻ Code Refactoring
+
+- **misc**: Refactor the agent runtime payload.
+
+#### 💄 Styles
+
+- **misc**: Add o1 model in openai and openrouter models.
+
+<br/>
+
+<details>
+<summary><kbd>Improvements and Fixes</kbd></summary>
+
+#### Code refactoring
+
+- **misc**: Refactor the agent runtime payload, closes [#5250](https://github.com/lobehub/lobe-chat/issues/5250) ([e420ab3](https://github.com/lobehub/lobe-chat/commit/e420ab3))
+
+#### Styles
+
+- **misc**: Add o1 model in openai and openrouter models, closes [#5236](https://github.com/lobehub/lobe-chat/issues/5236) ([f733852](https://github.com/lobehub/lobe-chat/commit/f733852))
+
+</details>
+
+<div align="right">
+
+[![](https://img.shields.io/badge/-BACK_TO_TOP-151515?style=flat-square)](#readme-top)
+
+</div>
+
 ### [Version 1.42.1](https://github.com/lobehub/lobe-chat/compare/v1.42.0...v1.42.1)
 
 <sup>Released on **2024-12-29**</sup>

package/changelog/v1.json

@@ -1,4 +1,13 @@
 [
+  {
+    "children": {
+      "improvements": [
+        "Add o1 model in openai and openrouter models."
+      ]
+    },
+    "date": "2024-12-31",
+    "version": "1.42.2"
+  },
   {
     "children": {
       "fixes": [

package/next.config.ts

@@ -25,7 +25,6 @@ const nextConfig: NextConfig = {
       '@icons-pack/react-simple-icons',
       '@lobehub/ui',
       'gpt-tokenizer',
-      'chroma-js',
     ],
     webVitalsAttribution: ['CLS', 'LCP'],
   },

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@lobehub/chat",
-  "version": "1.42.1",
+  "version": "1.42.2",
   "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",
@@ -265,7 +265,7 @@
     "@types/numeral": "^2.0.5",
     "@types/pg": "^8.11.10",
     "@types/react": "18.3.13",
-    "@types/react-dom": "^18.3.1",
+    "@types/react-dom": "^19.0.0",
     "@types/rtl-detect": "^1.0.3",
     "@types/semver": "^7.5.8",
     "@types/systemjs": "^6.15.1",

package/src/app/(main)/settings/llm/components/ProviderModelList/ModelConfigModal/Form.tsx

@@ -2,11 +2,10 @@ import { Checkbox, Form, FormInstance, Input } from 'antd';
 import { memo, useEffect } from 'react';
 import { useTranslation } from 'react-i18next';
 
+import MaxTokenSlider from '@/components/MaxTokenSlider';
 import { useIsMobile } from '@/hooks/useIsMobile';
 import { ChatModelCard } from '@/types/llm';
 
-import MaxTokenSlider from './MaxTokenSlider';
-
 interface ModelConfigFormProps {
   initialValues?: ChatModelCard;
   onFormInstanceReady: (instance: FormInstance) => void;
@@ -66,7 +65,10 @@ const ModelConfigForm = memo<ModelConfigFormProps>(
           >
             <Input placeholder={t('llm.customModelCards.modelConfig.displayName.placeholder')} />
           </Form.Item>
-          <Form.Item label={t('llm.customModelCards.modelConfig.tokens.title')} name={'contextWindowTokens'}>
+          <Form.Item
+            label={t('llm.customModelCards.modelConfig.tokens.title')}
+            name={'contextWindowTokens'}
+          >
             <MaxTokenSlider />
           </Form.Item>
           <Form.Item

package/src/{app/(main)/settings/llm/components/ProviderModelList/ModelConfigModal → components}/MaxTokenSlider.tsx

@@ -4,8 +4,7 @@ import { useTranslation } from 'react-i18next';
 import { Flexbox } from 'react-layout-kit';
 import useMergeState from 'use-merge-value';
 
-import { useServerConfigStore } from '@/store/serverConfig';
-import { serverConfigSelectors } from '@/store/serverConfig/selectors';
+import { useIsMobile } from '@/hooks/useIsMobile';
 
 const Kibi = 1024;
 
@@ -20,7 +19,7 @@ interface MaxTokenSliderProps {
 }
 
 const MaxTokenSlider = memo<MaxTokenSliderProps>(({ value, onChange, defaultValue }) => {
-  const { t } = useTranslation('setting');
+  const { t } = useTranslation('components');
 
   const [token, setTokens] = useMergeState(0, {
     defaultValue,
@@ -45,7 +44,7 @@ const MaxTokenSlider = memo<MaxTokenSliderProps>(({ value, onChange, defaultValu
     setPowValue(exponent(value / Kibi));
   };
 
-  const isMobile = useServerConfigStore(serverConfigSelectors.isMobile);
+  const isMobile = useIsMobile();
 
   const marks = useMemo(() => {
     return {
@@ -74,7 +73,7 @@ const MaxTokenSlider = memo<MaxTokenSliderProps>(({ value, onChange, defaultValu
           tooltip={{
             formatter: (x) => {
               if (typeof x === 'undefined') return;
-              if (x === 0) return t('llm.customModelCards.modelConfig.tokens.unlimited');
+              if (x === 0) return t('MaxTokenSlider.unlimited');
 
               let value = getRealValue(x);
               if (value < 125) return value.toFixed(0) + 'K';

package/src/components/ModelSelect/index.tsx

@@ -9,6 +9,7 @@ import { FC, memo } from 'react';
 import { useTranslation } from 'react-i18next';
 import { Center, Flexbox } from 'react-layout-kit';
 
+import { ModelAbilities } from '@/types/aiModel';
 import { ChatModelCard } from '@/types/llm';
 import { formatTokenNumber } from '@/utils/format';
 
@@ -57,8 +58,10 @@ const useStyles = createStyles(({ css, token }) => ({
   `,
 }));
 
-interface ModelInfoTagsProps extends ChatModelCard {
+interface ModelInfoTagsProps extends ModelAbilities {
+  contextWindowTokens?: number | null;
   directionReverse?: boolean;
+  isCustom?: boolean;
   placement?: 'top' | 'right';
 }
 
@@ -102,7 +105,7 @@ export const ModelInfoTags = memo<ModelInfoTagsProps>(
             </div>
           </Tooltip>
         )}
-        {model.contextWindowTokens !== undefined && (
+        {typeof model.contextWindowTokens === 'number' && (
           <Tooltip
             overlayStyle={{ maxWidth: 'unset', pointerEvents: 'none' }}
             placement={placement}
@@ -117,7 +120,7 @@ export const ModelInfoTags = memo<ModelInfoTagsProps>(
               {model.contextWindowTokens === 0 ? (
                 <Infinity size={17} strokeWidth={1.6} />
               ) : (
-                formatTokenNumber(model.contextWindowTokens)
+                formatTokenNumber(model.contextWindowTokens as number)
               )}
             </Center>
           </Tooltip>

package/src/components/NProgress/index.tsx

@@ -6,7 +6,15 @@ import { memo } from 'react';
 
 const NProgress = memo(() => {
   const theme = useTheme();
-  return <NextTopLoader color={theme.colorText} height={2} shadow={false} showSpinner={false} />;
+  return (
+    <NextTopLoader
+      color={theme.colorText}
+      height={2}
+      shadow={false}
+      showSpinner={false}
+      zIndex={1000}
+    />
+  );
 });
 
 export default NProgress;

package/src/config/modelProviders/openai.ts

@@ -17,6 +17,21 @@ const OpenAI: ModelProviderCard = {
       },
       releasedAt: '2024-09-12',
     },
+    {
+      contextWindowTokens: 200_000,
+      description:
+        'o1是OpenAI新的推理模型，支持图文输入并输出文本，适用于需要广泛通用知识的复杂任务。该模型具有200K上下文和2023年10月的知识截止日期。',
+      displayName: 'OpenAI o1',
+      enabled: true,
+      id: 'o1-2024-12-17',
+      maxOutput: 100_000,
+      pricing: {
+        input: 15,
+        output: 60,
+      },
+      releasedAt: '2024-12-17',      
+      vision: true,
+    },    
     {
       contextWindowTokens: 128_000,
       description:

package/src/config/modelProviders/openrouter.ts

@@ -27,6 +27,21 @@ const OpenRouter: ModelProviderCard = {
       },
       releasedAt: '2024-09-12',
     },
+    {
+      contextWindowTokens: 200_000,
+      description:
+        'o1是OpenAI新的推理模型，支持图文输入并输出文本，适用于需要广泛通用知识的复杂任务。该模型具有200K上下文和2023年10月的知识截止日期。',
+      displayName: 'OpenAI o1',
+      enabled: true,
+      id: 'openai/o1',
+      maxOutput: 100_000,
+      pricing: {
+        input: 15,
+        output: 60,
+      },
+      releasedAt: '2024-12-17',
+      vision: true,
+    },    
     {
       contextWindowTokens: 128_000,
       description:

package/src/const/auth.ts

@@ -28,7 +28,7 @@ export interface JWTPayload {
   /**
    * Represents the endpoint of provider
    */
-  endpoint?: string;
+  baseURL?: string;
 
   azureApiVersion?: string;
 

package/src/database/server/models/__tests__/user.test.ts

@@ -130,6 +130,17 @@ describe('UserModel', () => {
     });
   });
 
+  describe('getUserSettings', () => {
+    it('should get user settings', async () => {
+      await serverDB.insert(users).values({ id: userId });
+      await serverDB.insert(userSettings).values({ id: userId, general: { language: 'en-US' } });
+
+      const data = await userModel.getUserSettings();
+
+      expect(data).toMatchObject({ id: userId, general: { language: 'en-US' } });
+    });
+  });
+
   describe('deleteSetting', () => {
     it('should delete user settings', async () => {
       await serverDB.insert(users).values({ id: userId });

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

@@ -75,6 +75,10 @@ export class UserModel {
     };
   };
 
+  getUserSettings = async () => {
+    return this.db.query.userSettings.findFirst({ where: eq(userSettings.id, this.userId) });
+  };
+
   updateUser = async (value: Partial<UserItem>) => {
     return this.db
       .update(users)

package/src/libs/agent-runtime/AgentRuntime.test.ts

@@ -75,8 +75,8 @@ describe('AgentRuntime', () => {
     describe('Azure OpenAI provider', () => {
       it('should initialize correctly', async () => {
         const jwtPayload = {
-          apikey: 'user-azure-key',
-          endpoint: 'user-azure-endpoint',
+          apiKey: 'user-azure-key',
+          baseURL: 'user-azure-endpoint',
           apiVersion: '2024-06-01',
         };
 
@@ -90,8 +90,8 @@ describe('AgentRuntime', () => {
       });
       it('should initialize with azureOpenAIParams correctly', async () => {
         const jwtPayload = {
-          apikey: 'user-openai-key',
-          endpoint: 'user-endpoint',
+          apiKey: 'user-openai-key',
+          baseURL: 'user-endpoint',
           apiVersion: 'custom-version',
         };
 
@@ -106,8 +106,8 @@ describe('AgentRuntime', () => {
 
       it('should initialize with AzureAI correctly', async () => {
         const jwtPayload = {
-          apikey: 'user-azure-key',
-          endpoint: 'user-azure-endpoint',
+          apiKey: 'user-azure-key',
+          baseURL: 'user-azure-endpoint',
         };
         const runtime = await AgentRuntime.initializeWithProviderOptions(ModelProvider.Azure, {
           azure: jwtPayload,
@@ -171,7 +171,7 @@ describe('AgentRuntime', () => {
 
     describe('Ollama provider', () => {
       it('should initialize correctly', async () => {
-        const jwtPayload: JWTPayload = { endpoint: 'user-ollama-url' };
+        const jwtPayload: JWTPayload = { baseURL: 'https://user-ollama-url' };
         const runtime = await AgentRuntime.initializeWithProviderOptions(ModelProvider.Ollama, {
           ollama: jwtPayload,
         });
@@ -255,7 +255,7 @@ describe('AgentRuntime', () => {
 
   describe('AgentRuntime chat method', () => {
     it('should run correctly', async () => {
-      const jwtPayload: JWTPayload = { apiKey: 'user-openai-key', endpoint: 'user-endpoint' };
+      const jwtPayload: JWTPayload = { apiKey: 'user-openai-key', baseURL: 'user-endpoint' };
       const runtime = await AgentRuntime.initializeWithProviderOptions(ModelProvider.OpenAI, {
         openai: jwtPayload,
       });
@@ -271,7 +271,7 @@ describe('AgentRuntime', () => {
       await runtime.chat(payload);
     });
     it('should handle options correctly', async () => {
-      const jwtPayload: JWTPayload = { apiKey: 'user-openai-key', endpoint: 'user-endpoint' };
+      const jwtPayload: JWTPayload = { apiKey: 'user-openai-key', baseURL: 'user-endpoint' };
       const runtime = await AgentRuntime.initializeWithProviderOptions(ModelProvider.OpenAI, {
         openai: jwtPayload,
       });
@@ -300,7 +300,7 @@ describe('AgentRuntime', () => {
     });
 
     describe('callback', async () => {
-      const jwtPayload: JWTPayload = { apiKey: 'user-openai-key', endpoint: 'user-endpoint' };
+      const jwtPayload: JWTPayload = { apiKey: 'user-openai-key', baseURL: 'user-endpoint' };
       const runtime = await AgentRuntime.initializeWithProviderOptions(ModelProvider.OpenAI, {
         openai: jwtPayload,
       });

package/src/libs/agent-runtime/AgentRuntime.ts

@@ -133,7 +133,7 @@ class AgentRuntime {
       ai21: Partial<ClientOptions>;
       ai360: Partial<ClientOptions>;
       anthropic: Partial<ClientOptions>;
-      azure: { apiVersion?: string; apikey?: string; endpoint?: string };
+      azure: { apiKey?: string; apiVersion?: string; baseURL?: string };
       baichuan: Partial<ClientOptions>;
       bedrock: Partial<LobeBedrockAIParams>;
       cloudflare: Partial<LobeCloudflareParams>;
@@ -180,8 +180,8 @@ class AgentRuntime {
 
       case ModelProvider.Azure: {
         runtimeModel = new LobeAzureOpenAI(
-          params.azure?.endpoint,
-          params.azure?.apikey,
+          params.azure?.baseURL,
+          params.azure?.apiKey,
           params.azure?.apiVersion,
         );
         break;

package/src/libs/agent-runtime/ollama/index.test.ts

@@ -29,7 +29,10 @@ describe('LobeOllamaAI', () => {
       try {
         new LobeOllamaAI({ baseURL: 'invalid-url' });
       } catch (e) {
-        expect(e).toEqual(AgentRuntimeError.createError(AgentRuntimeErrorType.InvalidOllamaArgs));
+        expect(e).toEqual({
+          error: new TypeError('Invalid URL'),
+          errorType: 'InvalidOllamaArgs',
+        });
       }
     });
   });

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

@@ -22,8 +22,8 @@ export class LobeOllamaAI implements LobeRuntimeAI {
   constructor({ baseURL }: ClientOptions = {}) {
     try {
       if (baseURL) new URL(baseURL);
-    } catch {
-      throw AgentRuntimeError.createError(AgentRuntimeErrorType.InvalidOllamaArgs);
+    } catch (e) {
+      throw AgentRuntimeError.createError(AgentRuntimeErrorType.InvalidOllamaArgs, e);
     }
 
     this.client = new Ollama(!baseURL ? undefined : { host: baseURL });

package/src/libs/agent-runtime/openai/__snapshots__/index.test.ts.snap

@@ -12,6 +12,7 @@ exports[`LobeOpenAI > models > should get models 1`] = `
       "input": 0.5,
       "output": 1.5,
     },
+    "releasedAt": "2023-02-28",
   },
   {
     "id": "gpt-3.5-turbo-16k",
@@ -35,6 +36,7 @@ exports[`LobeOpenAI > models > should get models 1`] = `
       "input": 10,
       "output": 30,
     },
+    "releasedAt": "2024-01-23",
   },
   {
     "contextWindowTokens": 128000,
@@ -46,6 +48,7 @@ exports[`LobeOpenAI > models > should get models 1`] = `
       "input": 10,
       "output": 30,
     },
+    "releasedAt": "2024-01-23",
   },
   {
     "contextWindowTokens": 4096,
@@ -56,6 +59,7 @@ exports[`LobeOpenAI > models > should get models 1`] = `
       "input": 1.5,
       "output": 2,
     },
+    "releasedAt": "2023-08-24",
   },
   {
     "id": "gpt-3.5-turbo-0301",
@@ -73,6 +77,7 @@ exports[`LobeOpenAI > models > should get models 1`] = `
       "input": 1,
       "output": 2,
     },
+    "releasedAt": "2023-11-02",
   },
   {
     "contextWindowTokens": 128000,
@@ -84,6 +89,7 @@ exports[`LobeOpenAI > models > should get models 1`] = `
       "input": 10,
       "output": 30,
     },
+    "releasedAt": "2023-11-02",
   },
   {
     "contextWindowTokens": 128000,
@@ -91,6 +97,7 @@ exports[`LobeOpenAI > models > should get models 1`] = `
     "description": "GPT-4 视觉预览版，专为图像分析和处理任务设计。",
     "displayName": "GPT 4 Turbo with Vision Preview",
     "id": "gpt-4-vision-preview",
+    "releasedAt": "2023-11-02",
     "vision": true,
   },
   {
@@ -103,6 +110,7 @@ exports[`LobeOpenAI > models > should get models 1`] = `
       "input": 30,
       "output": 60,
     },
+    "releasedAt": "2023-06-27",
   },
   {
     "contextWindowTokens": 16385,
@@ -114,6 +122,7 @@ exports[`LobeOpenAI > models > should get models 1`] = `
       "input": 0.5,
       "output": 1.5,
     },
+    "releasedAt": "2024-01-23",
   },
   {
     "contextWindowTokens": 8192,
@@ -125,6 +134,7 @@ exports[`LobeOpenAI > models > should get models 1`] = `
       "input": 30,
       "output": 60,
     },
+    "releasedAt": "2023-06-12",
   },
 ]
 `;

package/src/libs/agent-runtime/utils/openaiCompatibleFactory/index.ts

@@ -1,3 +1,5 @@
+import dayjs from 'dayjs';
+import utc from 'dayjs/plugin/utc';
 import OpenAI, { ClientOptions } from 'openai';
 import { Stream } from 'openai/streaming';
 
@@ -18,6 +20,7 @@ import type {
   TextToSpeechOptions,
   TextToSpeechPayload,
 } from '../../types';
+import { ChatStreamCallbacks } from '../../types';
 import { AgentRuntimeError } from '../createError';
 import { debugResponse, debugStream } from '../debugStream';
 import { desensitizeUrl } from '../desensitizeUrl';
@@ -25,7 +28,6 @@ import { handleOpenAIError } from '../handleOpenAIError';
 import { convertOpenAIMessages } from '../openaiHelpers';
 import { StreamingResponse } from '../response';
 import { OpenAIStream, OpenAIStreamOptions } from '../streams';
-import { ChatStreamCallbacks } from '../../types';
 
 // the model contains the following keywords is not a chat model, so we should filter them out
 export const CHAT_MODELS_BLOCK_LIST = [
@@ -248,7 +250,8 @@ export const LobeOpenAICompatibleFactory = <T extends Record<string, any> = any>
 
         if (responseMode === 'json') return Response.json(response);
 
-        const transformHandler = chatCompletion?.handleTransformResponseToStream || transformResponseToStream;
+        const transformHandler =
+          chatCompletion?.handleTransformResponseToStream || transformResponseToStream;
         const stream = transformHandler(response as unknown as OpenAI.ChatCompletion);
 
         const streamHandler = chatCompletion?.handleStream || OpenAIStream;
@@ -278,7 +281,15 @@ export const LobeOpenAICompatibleFactory = <T extends Record<string, any> = any>
 
           const knownModel = LOBE_DEFAULT_MODEL_LIST.find((model) => model.id === item.id);
 
-          if (knownModel) return knownModel;
+          if (knownModel) {
+            dayjs.extend(utc);
+
+            return {
+              ...knownModel,
+              releasedAt:
+                knownModel.releasedAt ?? dayjs.utc(item.created * 1000).format('YYYY-MM-DD'),
+            };
+          }
 
           return { id: item.id };
         })

package/src/locales/default/components.ts

@@ -70,6 +70,9 @@ export default {
   GoBack: {
     back: '返回',
   },
+  MaxTokenSlider: {
+    unlimited: '无限制',
+  },
   ModelSelect: {
     featureTag: {
       custom: '自定义模型，默认设定同时支持函数调用与视觉识别，请根据实际情况验证上述能力的可用性',

package/src/locales/default/setting.ts

@@ -86,7 +86,6 @@ export default {
         modalTitle: '自定义模型配置',
         tokens: {
           title: '最大 token 数',
-          unlimited: '无限制',
         },
         vision: {
           extra:

package/src/server/modules/AgentRuntime/index.test.ts

@@ -70,23 +70,23 @@ vi.mock('@/config/llm', () => ({
 describe('initAgentRuntimeWithUserPayload method', () => {
   describe('should initialize with options correctly', () => {
     it('OpenAI provider: with apikey and endpoint', async () => {
-      const jwtPayload: JWTPayload = { apiKey: 'user-openai-key', endpoint: 'user-endpoint' };
+      const jwtPayload: JWTPayload = { apiKey: 'user-openai-key', baseURL: 'user-endpoint' };
       const runtime = await initAgentRuntimeWithUserPayload(ModelProvider.OpenAI, jwtPayload);
       expect(runtime).toBeInstanceOf(AgentRuntime);
       expect(runtime['_runtime']).toBeInstanceOf(LobeOpenAI);
-      expect(runtime['_runtime'].baseURL).toBe(jwtPayload.endpoint);
+      expect(runtime['_runtime'].baseURL).toBe(jwtPayload.baseURL);
     });
 
     it('Azure AI provider: with apikey, endpoint and apiversion', async () => {
       const jwtPayload: JWTPayload = {
         apiKey: 'user-azure-key',
-        endpoint: 'user-azure-endpoint',
+        baseURL: 'user-azure-endpoint',
         azureApiVersion: '2024-06-01',
       };
       const runtime = await initAgentRuntimeWithUserPayload(ModelProvider.Azure, jwtPayload);
       expect(runtime).toBeInstanceOf(AgentRuntime);
       expect(runtime['_runtime']).toBeInstanceOf(LobeAzureOpenAI);
-      expect(runtime['_runtime'].baseURL).toBe(jwtPayload.endpoint);
+      expect(runtime['_runtime'].baseURL).toBe(jwtPayload.baseURL);
     });
 
     it('ZhiPu AI provider: with apikey', async () => {
@@ -130,11 +130,11 @@ describe('initAgentRuntimeWithUserPayload method', () => {
     });
 
     it('Ollama provider: with endpoint', async () => {
-      const jwtPayload: JWTPayload = { endpoint: 'http://user-ollama-url' };
+      const jwtPayload: JWTPayload = { baseURL: 'http://user-ollama-url' };
       const runtime = await initAgentRuntimeWithUserPayload(ModelProvider.Ollama, jwtPayload);
       expect(runtime).toBeInstanceOf(AgentRuntime);
       expect(runtime['_runtime']).toBeInstanceOf(LobeOllamaAI);
-      expect(runtime['_runtime']['baseURL']).toEqual(jwtPayload.endpoint);
+      expect(runtime['_runtime']['baseURL']).toEqual(jwtPayload.baseURL);
     });
 
     it('Perplexity AI provider: with apikey', async () => {
@@ -220,12 +220,12 @@ describe('initAgentRuntimeWithUserPayload method', () => {
     it('Unknown Provider: with apikey and endpoint, should initialize to OpenAi', async () => {
       const jwtPayload: JWTPayload = {
         apiKey: 'user-unknown-key',
-        endpoint: 'user-unknown-endpoint',
+        baseURL: 'user-unknown-endpoint',
       };
       const runtime = await initAgentRuntimeWithUserPayload('unknown', jwtPayload);
       expect(runtime).toBeInstanceOf(AgentRuntime);
       expect(runtime['_runtime']).toBeInstanceOf(LobeOpenAI);
-      expect(runtime['_runtime'].baseURL).toBe(jwtPayload.endpoint);
+      expect(runtime['_runtime'].baseURL).toBe(jwtPayload.baseURL);
     });
   });
 

package/src/server/modules/AgentRuntime/index.ts

@@ -38,23 +38,23 @@ const getLlmOptionsFromPayload = (provider: string, payload: JWTPayload) => {
       }
 
       const apiKey = apiKeyManager.pick(payload?.apiKey || llmConfig[`${upperProvider}_API_KEY`]);
-      const baseURL = payload?.endpoint || process.env[`${upperProvider}_PROXY_URL`];
+      const baseURL = payload?.baseURL || process.env[`${upperProvider}_PROXY_URL`];
 
       return baseURL ? { apiKey, baseURL } : { apiKey };
     }
 
     case ModelProvider.Ollama: {
-      const baseURL = payload?.endpoint || process.env.OLLAMA_PROXY_URL;
+      const baseURL = payload?.baseURL || process.env.OLLAMA_PROXY_URL;
 
       return { baseURL };
     }
 
     case ModelProvider.Azure: {
       const { AZURE_API_KEY, AZURE_API_VERSION, AZURE_ENDPOINT } = llmConfig;
-      const apikey = apiKeyManager.pick(payload?.apiKey || AZURE_API_KEY);
-      const endpoint = payload?.endpoint || AZURE_ENDPOINT;
+      const apiKey = apiKeyManager.pick(payload?.apiKey || AZURE_API_KEY);
+      const baseURL = payload?.baseURL || AZURE_ENDPOINT;
       const apiVersion = payload?.azureApiVersion || AZURE_API_VERSION;
-      return { apiVersion, apikey, endpoint };
+      return { apiKey, apiVersion, baseURL };
     }
 
     case ModelProvider.Bedrock: {