refactor(gui-agent): enhance action parser and add comprehensive utilities

multimodal/gui-agent/action-parser/src/DefaultActionParser.ts

@@ -31,29 +31,15 @@ export class DefaultActionParser extends BaseActionParser {
     input = input.trim();
 
     let reasoningContent = null;
-    let rawActionStrings = null;
-    let actions = null;
+    let rawActionStrings: string[] | undefined = undefined;
+    let actions: BaseAction[] | undefined = undefined;
     try {
       ({ reasoningContent, rawActionStrings, actions } = this.extractActionStrings(input));
     } catch (error) {
-      return {
-        errorMessage: (error as Error).message,
-        rawContent: originInput,
-        rawActionStrings: [],
-        actions: [],
-      };
-    }
-
-    if (!rawActionStrings || rawActionStrings.length <= 0) {
-      return {
-        errorMessage: 'There is no GUI action detected',
-        rawContent: originInput,
-        reasoningContent,
-        rawActionStrings: [],
-        actions: [],
-      };
+      return this.createErrorResponse((error as Error).message, originInput);
     }
 
+    // if actions has prased, just return it
     if (actions && actions.length > 0) {
       return {
         rawContent: originInput,
@@ -63,28 +49,22 @@ export class DefaultActionParser extends BaseActionParser {
       };
     }
 
+    if (!rawActionStrings || rawActionStrings.length <= 0) {
+      return this.createErrorResponse('There is no GUI action detected', originInput);
+    }
+
     actions = [];
     try {
       for (const actionString of rawActionStrings) {
-        const action = this.helper.parseActionFromString(actionString);
+        const action = this.helper.parseActionCallString(actionString);
         if (action) actions.push(action);
       }
     } catch (error) {
-      return {
-        errorMessage: (error as Error).message,
-        rawContent: originInput,
-        reasoningContent,
-        rawActionStrings,
-        actions: [],
-      };
+      return this.createErrorResponse((error as Error).message, originInput);
     }
 
-    this.logger.debug(
-      '[parsePrediction] final result: reasoningContent:',
-      reasoningContent,
-      ', actions lenth:',
-      actions.length,
-    );
+    this.logger.debug('[parsePrediction] reasoningContent:', reasoningContent);
+    this.logger.debug('[parsePrediction] actions lenth:', actions.length);
 
     return {
       rawContent: originInput,
@@ -118,4 +98,12 @@ export class DefaultActionParser extends BaseActionParser {
       actions: actions || undefined,
     };
   }
+
+  private createErrorResponse(errorMessage: string, rawContent: string): ParsedGUIResponse {
+    return {
+      errorMessage,
+      rawContent,
+      actions: [],
+    };
+  }
 }

multimodal/gui-agent/action-parser/src/FomatParsers.ts

@@ -2,17 +2,18 @@
  * Copyright (c) 2025 Bytedance, Inc. and its affiliates.
  * SPDX-License-Identifier: Apache-2.0
  */
-import { ConsoleLogger } from '@agent-infra/logger';
+import { ConsoleLogger, LogLevel } from '@agent-infra/logger';
 import { BaseAction } from '@gui-agent/shared/types';
 import { XMLBuilder, XMLParser } from 'fast-xml-parser';
 // Remove circular dependency
 import { ActionParserHelper } from './ActionParserHelper';
+import { serializeAction } from '@gui-agent/shared/utils';
 
 export interface FormatParser {
   parse(text: string): {
     reasoningContent: string | null;
-    rawActionStrings: string[] | null;
-    actions: BaseAction[] | null;
+    rawActionStrings: string[] | undefined;
+    actions: BaseAction[] | undefined;
   } | null;
 }
 
@@ -24,8 +25,8 @@ export class XMLFormatParser implements FormatParser {
 
   parse(text: string): {
     reasoningContent: string | null;
-    rawActionStrings: string[] | null;
-    actions: BaseAction[] | null;
+    rawActionStrings: string[] | undefined;
+    actions: BaseAction[] | undefined;
   } | null {
     if (text.includes('computer_env')) {
       // The text is omni format, not a solid XML format, refuse parse
@@ -71,7 +72,7 @@ export class XMLFormatParser implements FormatParser {
       }
       if (key === 'seed:tool_call') {
         canParse = true;
-        actions.push(...this.helper.standardizeGUIActions(value));
+        actions.push(...this.helper.standardizeGUIActionsFromXMLObject(value));
         continue;
       }
     }
@@ -85,9 +86,7 @@ export class XMLFormatParser implements FormatParser {
 
     const rawActionStrings: string[] = [];
     for (const action of actions) {
-      rawActionStrings.push(
-        this.helper.convertRoughActionInputsToLegacyActionString(action.type, action.inputs),
-      );
+      rawActionStrings.push(serializeAction(action));
     }
 
     return {
@@ -123,8 +122,8 @@ export class OmniFormatParser implements FormatParser {
 
   parse(text: string): {
     reasoningContent: string | null;
-    rawActionStrings: string[] | null;
-    actions: BaseAction[] | null;
+    rawActionStrings: string[] | undefined;
+    actions: BaseAction[] | undefined;
   } | null {
     if (!this.canParse(text)) {
       return null;
@@ -148,7 +147,7 @@ export class OmniFormatParser implements FormatParser {
     const result = {
       reasoningContent,
       rawActionStrings: actionStr.split('\n\n').filter((action) => action.trim() !== ''),
-      actions: null,
+      actions: undefined,
     };
     return result;
   }
@@ -187,8 +186,8 @@ export class UnifiedBCFormatParser implements FormatParser {
 
   parse(text: string): {
     reasoningContent: string | null;
-    rawActionStrings: string[] | null;
-    actions: BaseAction[] | null;
+    rawActionStrings: string[] | undefined;
+    actions: BaseAction[] | undefined;
   } | null {
     if (!this.canParse(text)) {
       return null;
@@ -220,7 +219,7 @@ export class UnifiedBCFormatParser implements FormatParser {
     return {
       reasoningContent,
       rawActionStrings: actionStr.split('\n\n').filter((action) => action.trim() !== ''),
-      actions: null,
+      actions: undefined,
     };
   }
 }
@@ -245,8 +244,8 @@ class BCComplexFormatParser implements FormatParser {
 
   parse(text: string): {
     reasoningContent: string | null;
-    rawActionStrings: string[] | null;
-    actions: BaseAction[] | null;
+    rawActionStrings: string[] | undefined;
+    actions: BaseAction[] | undefined;
   } | null {
     if (!this.canParse(text)) {
       return null;
@@ -282,7 +281,7 @@ class BCComplexFormatParser implements FormatParser {
       reasoningContent:
         reflection && thought ? `${reflection}, ${thought}` : (thought ?? reflection),
       rawActionStrings: actionStr.split('\n\n').filter((action) => action.trim() !== ''),
-      actions: null,
+      actions: undefined,
     };
   }
 }
@@ -306,8 +305,8 @@ class O1FormatParser implements FormatParser {
 
   parse(text: string): {
     reasoningContent: string | null;
-    rawActionStrings: string[] | null;
-    actions: BaseAction[] | null;
+    rawActionStrings: string[] | undefined;
+    actions: BaseAction[] | undefined;
   } | null {
     // this.logger.debug('[O1FormatParser] start...');
     if (!this.canParse(text)) {
@@ -333,7 +332,7 @@ class O1FormatParser implements FormatParser {
     return {
       reasoningContent: thought,
       rawActionStrings: actionContent.split('\n\n').filter((action) => action.trim() !== ''),
-      actions: null,
+      actions: undefined,
     };
   }
 }
@@ -350,8 +349,8 @@ class FallbackFormatParser implements FormatParser {
 
   parse(text: string): {
     reasoningContent: string | null;
-    rawActionStrings: string[] | null;
-    actions: BaseAction[] | null;
+    rawActionStrings: string[] | undefined;
+    actions: BaseAction[] | undefined;
   } | null {
     this.logger.debug('[FallbackFormatParser] canParse: always true');
 
@@ -386,15 +385,19 @@ class FallbackFormatParser implements FormatParser {
     return {
       reasoningContent: thoughtStr,
       rawActionStrings: actionStr.split('\n\n').filter((action) => action.trim() !== ''),
-      actions: null,
+      actions: undefined,
     };
   }
 }
 
+const defaultLogger = new ConsoleLogger(undefined, LogLevel.DEBUG);
+
 export class FormatParserChain {
+  private logger: ConsoleLogger;
   private parsers: FormatParser[];
 
-  constructor(private logger: ConsoleLogger) {
+  constructor(logger: ConsoleLogger = defaultLogger) {
+    this.logger = logger.spawn('[FormatParserChain]');
     this.parsers = [
       new XMLFormatParser(this.logger),
       new OmniFormatParser(this.logger),
@@ -407,8 +410,8 @@ export class FormatParserChain {
 
   parse(text: string): {
     reasoningContent: string | null;
-    rawActionStrings: string[] | null;
-    actions: BaseAction[] | null;
+    rawActionStrings: string[] | undefined;
+    actions: BaseAction[] | undefined;
   } {
     this.logger.debug('[FormatParserChain] start...');