feat(tarko): add gui agent screenshot render strategy config (#1296)

Author: ulivz

multimodal/agent-tars/core/src/agent-tars.ts

@@ -63,6 +63,11 @@ export class AgentTARS<T extends AgentTARSOptions = AgentTARSOptions> extends MC
       'Please book me the earliest flight from Hangzhou to Shenzhen on 10.1',
     ],
     enableContextualSelector: true,
+    guiAgent: {
+      defaultScreenshotRenderStrategy: 'beforeAction',
+      enableScreenshotRenderStrategySwitch: true,
+      renderGUIAction: true,
+    },
   };
   private workspace: string;
   // FIXME: remove it since options is strict type already

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

@@ -56,6 +56,11 @@ export default class OmniTARSAgent extends ComposableAgent {
         },
       ],
     },
+    guiAgent: {
+      defaultScreenshotRenderStrategy: 'afterAction',
+      enableScreenshotRenderStrategySwitch: true,
+      renderGUIAction: true,
+    },
   };
 
   constructor(options: OmniTarsOption) {

multimodal/tarko/agent-web-ui/src/common/constants/index.ts

@@ -90,4 +90,5 @@ export {
   isContextualSelectorEnabled,
   getLogoUrl,
   getWorkspaceNavItems,
+  getGUIAgentConfig,
 } from './shared';

multimodal/tarko/agent-web-ui/src/common/constants/shared.ts

@@ -40,3 +40,14 @@ export function getLogoUrl(): string {
 export function getWorkspaceNavItems() {
   return getWebUIConfig().workspace?.navItems || [];
 }
+
+/**
+ * Get GUI Agent configuration from web UI config
+ */
+export function getGUIAgentConfig() {
+  return getWebUIConfig().guiAgent || {
+    defaultScreenshotRenderStrategy: 'afterAction',
+    enableScreenshotRenderStrategySwitch: false,
+    renderGUIAction: true,
+  };
+}