feat(omni-agent): update agent mode config structure

multimodal/omni-tars/core/src/AgentComposer.ts

@@ -32,7 +32,9 @@ export class AgentComposer {
    */
   async initialize(): Promise<void> {
     for (const plugin of this.plugins) {
+      const start = Date.now();
       await plugin.initialize?.();
+      this.logger.info(`initialize agent plugin ${plugin.name} cost: `, Date.now() - start);
     }
   }
 

multimodal/omni-tars/mcp-agent/src/McpAgentPlugin.ts

@@ -30,7 +30,8 @@ export class McpAgentPlugin extends AgentPlugin {
   }
 
   async initialize(): Promise<void> {
-    await this.mcpManager.init();
+    //FIXME:Temporarily remove await to speed up the agent initialization process; the logic of mcpManager.getClient() needs to be added later
+    this.mcpManager.init();
 
     // Initialize tools
     this.tools = [

multimodal/omni-tars/omni-agent/src/index.ts

@@ -4,7 +4,7 @@
  */
 
 import { ComposableAgent } from '@omni-tars/core';
-import {  AgentWebUIImplementation } from '@tarko/interface';
+import { AgentWebUIImplementation } from '@tarko/interface';
 import { getComposableOption, OmniTarsOption } from './options';
 
 const sandboxBaseUrl = process.env.AIO_SANDBOX_URL ?? '.';

multimodal/omni-tars/omni-agent/src/options.ts

@@ -3,66 +3,87 @@
  * SPDX-License-Identifier: Apache-2.0
  */
 
-import { CodeAgentExtraOption, codePluginBuilder, CodeToolCallEngineProvider } from "@omni-tars/code-agent";
-import { AgentMode, ComposableAgentOptions, createComposableToolCallEngineFactory } from "@omni-tars/core";
-import { GuiAgentPlugin, GuiToolCallEngineProvider, OperatorManager } from "@omni-tars/gui-agent";
-import { mcpPluginBuilder, MCPTarsExtraOption, McpToolCallEngineProvider } from "@omni-tars/mcp-agent";
-import { AgentAppConfig } from "@tarko/interface";
-
+import {
+  CodeAgentExtraOption,
+  codePluginBuilder,
+  CodeToolCallEngineProvider,
+} from '@omni-tars/code-agent';
+import {
+  AgentMode,
+  ComposableAgentOptions,
+  createComposableToolCallEngineFactory,
+} from '@omni-tars/core';
+import { GuiAgentPlugin, GuiToolCallEngineProvider, OperatorManager } from '@omni-tars/gui-agent';
+import {
+  mcpPluginBuilder,
+  MCPTarsExtraOption,
+  McpToolCallEngineProvider,
+} from '@omni-tars/mcp-agent';
+import { AgentAppConfig } from '@tarko/interface';
 
 const mcpToolCallEngine = new McpToolCallEngineProvider();
 
 const omniToolCallEngine = createComposableToolCallEngineFactory({
-  engines: [new GuiToolCallEngineProvider('omni'), mcpToolCallEngine, new CodeToolCallEngineProvider()],
+  engines: [
+    new GuiToolCallEngineProvider('omni'),
+    mcpToolCallEngine,
+    new CodeToolCallEngineProvider(),
+  ],
   defaultEngine: mcpToolCallEngine,
 });
 
-const guiToolCallEngine = createComposableToolCallEngineFactory({ engines: [new GuiToolCallEngineProvider('gui')] });
+const guiToolCallEngine = createComposableToolCallEngineFactory({
+  engines: [new GuiToolCallEngineProvider('gui')],
+});
+
+export type OmniTarsOption = AgentAppConfig &
+  MCPTarsExtraOption &
+  CodeAgentExtraOption & {
+    agentMode: AgentMode;
+    browserMode: 'dom' | 'visual-grounding' | 'hybrid';
+  };
 
+export function getComposableOption(options: OmniTarsOption) {
+  const {
+    tavilyApiKey,
+    googleApiKey,
+    googleMcpUrl,
+    sandboxUrl,
+    ignoreSandboxCheck,
+    linkReaderAK,
+    linkReaderMcpUrl,
+    agentMode = 'omni',
+    browserMode = 'hybrid',
+    ...restOptions
+  } = options;
 
-export type OmniTarsOption = AgentAppConfig & MCPTarsExtraOption & CodeAgentExtraOption & {
-  agentMode: AgentMode
-};
+  const baseOptions: Partial<ComposableAgentOptions> = {
+    ...restOptions,
+    maxTokens: 32768,
+    enableStreamingToolCallEvents: true,
+  };
 
+  const guiPlugin = new GuiAgentPlugin({
+    operatorManager: OperatorManager.createHybird(options.sandboxUrl),
+  });
 
-export function getComposableOption(options: OmniTarsOption) {
-    const {
+  if (agentMode === 'gui') {
+    baseOptions.toolCallEngine = guiToolCallEngine;
+    baseOptions.plugins = [guiPlugin];
+  } else if (agentMode === 'omni') {
+    baseOptions.toolCallEngine = omniToolCallEngine;
+    baseOptions.plugins = [
+      mcpPluginBuilder({
         tavilyApiKey,
         googleApiKey,
         googleMcpUrl,
-        sandboxUrl,
-        ignoreSandboxCheck,
         linkReaderAK,
         linkReaderMcpUrl,
-        agentMode = 'omni',
-        ...restOptions
-      } = options;
-
-    const baseOptions: Partial<ComposableAgentOptions> = {
-      ...restOptions,
-      maxTokens: 32768,
-      enableStreamingToolCallEvents: true,
-    };
+      }),
+      codePluginBuilder({ sandboxUrl, ignoreSandboxCheck }),
+      guiPlugin,
+    ];
+  }
 
-    if(agentMode === 'gui') {
-      baseOptions.toolCallEngine = guiToolCallEngine;
-      baseOptions.plugins =  [
-        new GuiAgentPlugin({ operatorManager: OperatorManager.createHybird(options.sandboxUrl) }),
-      ];
-    } else if(agentMode === 'omni') {
-      baseOptions.toolCallEngine = omniToolCallEngine;
-      baseOptions.plugins =  [
-        mcpPluginBuilder({
-          tavilyApiKey,
-          googleApiKey,
-          googleMcpUrl,
-          linkReaderAK,
-          linkReaderMcpUrl,
-        }),
-        codePluginBuilder({ sandboxUrl, ignoreSandboxCheck }),
-        new GuiAgentPlugin({ operatorManager: OperatorManager.createHybird(options.sandboxUrl) }),
-      ];
-    }
-
-    return baseOptions as ComposableAgentOptions;
-}
+  return baseOptions as ComposableAgentOptions;
+}

multimodal/tarko/agent-server-next/examples/bootstrap.ts

@@ -3,7 +3,7 @@
  * SPDX-License-Identifier: Apache-2.0
  */
 import { getContext } from 'hono/context-storage';
-import { AuthHook, CorsHook, AgentServer,  ContextStorageHook } from '../src/index';
+import { AuthHook, CorsHook, AgentServer, ContextStorageHook } from '../src/index';
 import { resolve } from 'path';
 import { ContextVariables } from '../src/types';
 
@@ -42,7 +42,22 @@ const server = new AgentServer({
               type: 'string',
               title: 'Agent Mode',
               enum: ['omni', 'gui'],
+              enumLabels: ['Omni', 'GUI'],
               default: 'omni',
+              placement: 'chat-bottom',
+            },
+
+            browserMode: {
+              type: 'string',
+              title: 'Browser Control',
+              enum: ['hybrid'],
+              enumLabels: ['混合模式'],
+              default: 'hybrid',
+              placement: 'chat-bottom',
+              visible: {
+                dependsOn: 'agentMode',
+                when: 'gui',
+              },
             },
           },
         },
@@ -60,16 +75,16 @@ const server = new AgentServer({
       },
       models: [
         {
-          id: "ep-20250909173748-wcfb2",
-          provider: "volcengine",
-          displayName: "T6-SFT",
+          id: 'ep-20250909173748-wcfb2',
+          provider: 'volcengine',
+          displayName: 'T6-SFT',
           baseURL: process.env.OMNI_TARS_BASE_URL,
           apiKey: process.env.OMNI_TARS_API_KEY,
         },
         {
-          id: "ep-20250905175225-hlrvd",
-          provider: "volcengine",
-          displayName: "T5-RL",
+          id: 'ep-20250905175225-hlrvd',
+          provider: 'volcengine',
+          displayName: 'T5-RL',
           baseURL: process.env.OMNI_TARS_BASE_URL,
           apiKey: process.env.OMNI_TARS_API_KEY,
         },
@@ -101,15 +116,15 @@ const logger = {
   name: 'custom',
   info: (...splat) => {
     const requestId = getContext<{ Variables: ContextVariables }>().var.requestId;
-    console.log(`[CUSTOM LOGGER] ${requestId}`, ...splat)
+    console.log(`[CUSTOM LOGGER] ${requestId}`, ...splat);
   },
   warn: console.warn,
   error: console.error,
   debug: console.debug,
-  setLevel: () => {  }
-}
+  setLevel: () => {},
+};
 
-server.setLogger(logger)
+server.setLogger(logger);
 server.registerHook(AuthHook);
 server.registerHook(CorsHook);
 server.registerHook(ContextStorageHook);

multimodal/tarko/agent-server-next/src/middlewares/session-restore.ts

@@ -32,7 +32,7 @@ export async function sessionRestoreMiddleware(
 
     // If not exist, restored the AgentSession instance based on the database data.
     if (!session) {
-      logger.info('session instance not exist, prepare to restore')
+      logger.info('session instance not exist, prepare to restore');
 
       const start = Date.now();
 
@@ -47,6 +47,8 @@ export async function sessionRestoreMiddleware(
         restored.storageUnsubscribe &&
           (server.storageUnsubscribes[sessionId] = restored.storageUnsubscribe);
       }
+    } else {
+      logger.info('get session from sessionPool, sessionId:', session.id);
     }
 
     if (!session) {

multimodal/tarko/agent-server-next/src/services/session/AgentSession.ts

@@ -127,7 +127,7 @@ export class AgentSession {
       throw new Error('Cannot found available resolved agent');
     }
 
-       // Get stored events for this session before creating the agent
+    // Get stored events for this session before creating the agent
     const storedEvents = this.server.daoFactory
       ? await this.server.daoFactory.getSessionEvents(this.id)
       : [];
@@ -187,15 +187,26 @@ export class AgentSession {
       // Log AGIO initialization
       console.debug('AGIO collector initialized', { provider: agentOptions.agio.provider });
     }
-
-    this.logger.info('create new agent with config: ',  JSON.stringify({
-      agent: agentOptions.agent,
-      share: agentOptions.share,
-      workspace: agentOptions.workspace,
-      thinking: agentOptions.thinking,
-      name: agentOptions.name,
-      runtimeSettings: transformedOptions
-    }, null, 2));
+
+    this.logger.info(
+      'create new agent with config: ',
+      JSON.stringify(
+        {
+          agent: agentOptions.agent,
+          share: agentOptions.share,
+          workspace: agentOptions.workspace,
+          thinking: agentOptions.thinking,
+          name: agentOptions.name,
+          model: {
+            id: agentOptions.model?.id,
+            provider: agentOptions.model?.provider,
+          },
+          runtimeSettings: transformedOptions,
+        },
+        null,
+        2,
+      ),
+    );
 
     return wrappedAgent;
   }
@@ -552,6 +563,4 @@ export class AgentSession {
   }
 }
 
-
-
 export default AgentSession;