@ui-tars-test/agent-sdk 0.3.9 → 0.3.12

package/dist/GUIAgent.d.ts

@@ -1,4 +1,4 @@
-import { LLMRequestHookPayload } from '@tarko/agent';
+import { LLMRequestHookPayload } from '@ui-tars-test/tarko-agent';
 import { Operator, BaseGUIAgent } from '@ui-tars-test/shared/base';
 import { GUIAgentConfig } from '@ui-tars-test/shared/types';
 export declare class GUIAgent<T extends Operator> extends BaseGUIAgent {

package/dist/GUIAgent.d.ts.map

@@ -1 +1 @@
-{"version":3,"file":"GUIAgent.d.ts","sourceRoot":"","sources":["../src/GUIAgent.ts"],"names":[],"mappings":"AAIA,OAAO,EACL,qBAAqB,EAKtB,MAAM,cAAc,CAAC;AAItB,OAAO,EAAE,QAAQ,EAAE,YAAY,EAAE,MAAM,2BAA2B,CAAC;AACnE,OAAO,EACL,cAAc,EAGf,MAAM,4BAA4B,CAAC;AAapC,qBAAa,QAAQ,CAAC,CAAC,SAAS,QAAQ,CAAE,SAAQ,YAAY;IAC5D,MAAM,CAAC,KAAK,SAAe;IAE3B,OAAO,CAAC,QAAQ,CAAuB;IACvC,OAAO,CAAC,oBAAoB,CAAuB;IACnD,OAAO,CAAC,gBAAgB,CAAwB;IAChD,OAAO,CAAC,gBAAgB,CAAS;gBAErB,MAAM,EAAE,cAAc,CAAC,CAAC,CAAC;IA+C/B,UAAU;IA0CV,YAAY,CAAC,EAAE,EAAE,MAAM,EAAE,OAAO,EAAE,qBAAqB,GAAG,OAAO,CAAC,IAAI,CAAC;IAkDvE,oBAAoB,CAAC,SAAS,EAAE,MAAM;IAItC,cAAc,CAAC,EAAE,EAAE,MAAM,GAAG,OAAO,CAAC,IAAI,CAAC;IAIzC,gBAAgB,CACpB,EAAE,EAAE,MAAM,EACV,QAAQ,EAAE;QAAE,UAAU,EAAE,MAAM,CAAC;QAAC,IAAI,EAAE,MAAM,CAAA;KAAE,EAC9C,IAAI,EAAE,OAAO;IAKT,eAAe,CACnB,EAAE,EAAE,MAAM,EACV,QAAQ,EAAE;QAAE,UAAU,EAAE,MAAM,CAAC;QAAC,IAAI,EAAE,MAAM,CAAA;KAAE,EAC9C,MAAM,EAAE,OAAO,GACd,OAAO,CAAC,OAAO,CAAC;CA4DpB"}
+{"version":3,"file":"GUIAgent.d.ts","sourceRoot":"","sources":["../src/GUIAgent.ts"],"names":[],"mappings":"AAIA,OAAO,EACL,qBAAqB,EAKtB,MAAM,2BAA2B,CAAC;AAInC,OAAO,EAAE,QAAQ,EAAE,YAAY,EAAE,MAAM,2BAA2B,CAAC;AACnE,OAAO,EACL,cAAc,EAGf,MAAM,4BAA4B,CAAC;AAapC,qBAAa,QAAQ,CAAC,CAAC,SAAS,QAAQ,CAAE,SAAQ,YAAY;IAC5D,MAAM,CAAC,KAAK,SAAe;IAE3B,OAAO,CAAC,QAAQ,CAAuB;IACvC,OAAO,CAAC,oBAAoB,CAAuB;IACnD,OAAO,CAAC,gBAAgB,CAAwB;IAChD,OAAO,CAAC,gBAAgB,CAAS;gBAErB,MAAM,EAAE,cAAc,CAAC,CAAC,CAAC;IA+C/B,UAAU;IA0CV,YAAY,CAAC,EAAE,EAAE,MAAM,EAAE,OAAO,EAAE,qBAAqB,GAAG,OAAO,CAAC,IAAI,CAAC;IAkDvE,oBAAoB,CAAC,SAAS,EAAE,MAAM;IAItC,cAAc,CAAC,EAAE,EAAE,MAAM,GAAG,OAAO,CAAC,IAAI,CAAC;IAIzC,gBAAgB,CACpB,EAAE,EAAE,MAAM,EACV,QAAQ,EAAE;QAAE,UAAU,EAAE,MAAM,CAAC;QAAC,IAAI,EAAE,MAAM,CAAA;KAAE,EAC9C,IAAI,EAAE,OAAO;IAKT,eAAe,CACnB,EAAE,EAAE,MAAM,EACV,QAAQ,EAAE;QAAE,UAAU,EAAE,MAAM,CAAC;QAAC,IAAI,EAAE,MAAM,CAAA;KAAE,EAC9C,MAAM,EAAE,OAAO,GACd,OAAO,CAAC,OAAO,CAAC;CA4DpB"}

package/dist/GUIAgent.js

@@ -30,7 +30,7 @@ __webpack_require__.r(__webpack_exports__);
 __webpack_require__.d(__webpack_exports__, {
     GUIAgent: ()=>GUIAgent
 });
-const agent_namespaceObject = require("@tarko/agent");
+const tarko_agent_namespaceObject = require("@ui-tars-test/tarko-agent");
 const external_ToolCallEngine_js_namespaceObject = require("./ToolCallEngine.js");
 const external_prompts_js_namespaceObject = require("./prompts.js");
 const media_utils_namespaceObject = require("@agent-infra/media-utils");
@@ -49,10 +49,10 @@ function _define_property(obj, key, value) {
     else obj[key] = value;
     return obj;
 }
-const defaultLogger = new agent_namespaceObject.ConsoleLogger('[GUIAgent]', agent_namespaceObject.LogLevel.DEBUG);
+const defaultLogger = new tarko_agent_namespaceObject.ConsoleLogger('[GUIAgent]', tarko_agent_namespaceObject.LogLevel.DEBUG);
 class GUIAgent extends base_namespaceObject.BaseGUIAgent {
     async initialize() {
-        this.registerTool(new agent_namespaceObject.Tool({
+        this.registerTool(new tarko_agent_namespaceObject.Tool({
             id: external_constants_js_namespaceObject.GUI_ADAPTED_TOOL_NAME,
             description: 'operator tool',
             parameters: {},
@@ -186,7 +186,7 @@ class GUIAgent extends base_namespaceObject.BaseGUIAgent {
             ...maxLoopCount && {
                 maxIterations: maxLoopCount
             },
-            logLevel: agent_namespaceObject.LogLevel.DEBUG
+            logLevel: tarko_agent_namespaceObject.LogLevel.DEBUG
         }), _define_property(this, "operator", void 0), _define_property(this, "normalizeCoordinates", void 0), _define_property(this, "detailCalculator", void 0), _define_property(this, "loopIntervalInMs", void 0);
         this.operator = operator;
         this.normalizeCoordinates = normalizeCoordinates ?? external_defaultImpls_js_namespaceObject.defaultNormalizeCoords;

package/dist/GUIAgent.js.map

@@ -1 +1 @@
-{"version":3,"file":"GUIAgent.js","sources":["webpack://@ui-tars-test/agent-sdk/webpack/runtime/define_property_getters","webpack://@ui-tars-test/agent-sdk/webpack/runtime/has_own_property","webpack://@ui-tars-test/agent-sdk/webpack/runtime/make_namespace_object","webpack://@ui-tars-test/agent-sdk/./src/GUIAgent.ts"],"sourcesContent":["__webpack_require__.d = (exports, definition) => {\n\tfor(var key in definition) {\n        if(__webpack_require__.o(definition, key) && !__webpack_require__.o(exports, key)) {\n            Object.defineProperty(exports, key, { enumerable: true, get: definition[key] });\n        }\n    }\n};","__webpack_require__.o = (obj, prop) => (Object.prototype.hasOwnProperty.call(obj, prop))","// define __esModule on exports\n__webpack_require__.r = (exports) => {\n\tif(typeof Symbol !== 'undefined' && Symbol.toStringTag) {\n\t\tObject.defineProperty(exports, Symbol.toStringTag, { value: 'Module' });\n\t}\n\tObject.defineProperty(exports, '__esModule', { value: true });\n};","/*\n * Copyright (c) 2025 Bytedance, Inc. and its affiliates.\n * SPDX-License-Identifier: Apache-2.0\n */\nimport {\n  LLMRequestHookPayload,\n  ChatCompletionContentPart,\n  LogLevel,\n  Tool,\n  ConsoleLogger,\n} from '@tarko/agent';\nimport { GUIAgentToolCallEngine } from './ToolCallEngine';\nimport { SYSTEM_PROMPT } from './prompts';\nimport { Base64ImageParser } from '@agent-infra/media-utils';\nimport { Operator, BaseGUIAgent } from '@ui-tars-test/shared/base';\nimport {\n  GUIAgentConfig,\n  NormalizeCoordinates,\n  ImageDetailCalculator,\n} from '@ui-tars-test/shared/types';\nimport {\n  assembleSystemPrompt,\n  isSystemPromptTemplate,\n  normalizeActionCoords,\n  sleep,\n} from '@ui-tars-test/shared/utils';\nimport { GUI_ADAPTED_TOOL_NAME } from './constants';\nimport { convertToAgentUIAction, createGUIErrorResponse } from './utils';\nimport { defaultNormalizeCoords, defaultDetailCalculator } from './defaultImpls';\n\nconst defaultLogger = new ConsoleLogger('[GUIAgent]', LogLevel.DEBUG);\n\nexport class GUIAgent<T extends Operator> extends BaseGUIAgent {\n  static label = 'GUI Agent';\n\n  private operator: Operator | undefined;\n  private normalizeCoordinates: NormalizeCoordinates;\n  private detailCalculator: ImageDetailCalculator;\n  private loopIntervalInMs: number;\n\n  constructor(config: GUIAgentConfig<T>) {\n    const {\n      operator,\n      model,\n      systemPrompt,\n      customeActionParser,\n      normalizeCoordinates,\n      detailCalculator,\n      maxLoopCount,\n      loopIntervalInMs = 500,\n    } = config;\n    let finalSystemPrompt = SYSTEM_PROMPT;\n    if (typeof systemPrompt === 'string') {\n      finalSystemPrompt = systemPrompt;\n    } else if (Array.isArray(systemPrompt)) {\n      finalSystemPrompt = systemPrompt\n        .map((p) => (typeof p === 'string' ? p : p.content))\n        .join('\\n\\n');\n    } else if (systemPrompt && isSystemPromptTemplate(systemPrompt)) {\n      finalSystemPrompt = assembleSystemPrompt(systemPrompt, operator.getSupportedActions());\n    }\n    defaultLogger.debug('final instructions for sp:', finalSystemPrompt);\n\n    // Create a adapted ToolCallEngine constructor that captures the customeActionParser\n    const AdaptedToolCallEngine = class extends GUIAgentToolCallEngine {\n      constructor() {\n        super(customeActionParser);\n      }\n    };\n\n    super({\n      name: GUIAgent.label,\n      instructions: finalSystemPrompt,\n      tools: [],\n      toolCallEngine: AdaptedToolCallEngine,\n      model: model,\n      ...(maxLoopCount && { maxIterations: maxLoopCount }),\n      logLevel: LogLevel.DEBUG,\n    });\n    this.operator = operator;\n    this.normalizeCoordinates = normalizeCoordinates ?? defaultNormalizeCoords;\n    // Default detail calculator implementation\n    this.detailCalculator = detailCalculator ?? defaultDetailCalculator;\n    this.loopIntervalInMs = loopIntervalInMs;\n    this.logger = this.logger.spawn('[GUIAgent]');\n  }\n\n  async initialize() {\n    // Register the GUI tool\n    this.registerTool(\n      new Tool({\n        id: GUI_ADAPTED_TOOL_NAME,\n        description: 'operator tool',\n        parameters: {}, // no need to pass parameters\n        function: async (input) => {\n          this.logger.log(`${GUI_ADAPTED_TOOL_NAME} input:`, input);\n          if (!this.operator) {\n            return createGUIErrorResponse(input.action, 'Operator not initialized');\n          }\n          if (input.errorMessage) {\n            return createGUIErrorResponse(input.action, input.errorMessage);\n          }\n          // normalize coordinates\n          if (input.operator_action) {\n            input.operator_action = normalizeActionCoords(\n              input.operator_action,\n              this.normalizeCoordinates,\n            );\n          }\n          this.logger.info('action to execute:', JSON.stringify(input.operator_action));\n          const result = await this.operator!.doExecute({\n            actions: [input.operator_action],\n          });\n          if (result.errorMessage) {\n            return createGUIErrorResponse(input.action, result.errorMessage);\n          }\n          // return { action: input.action, status: 'success', result };\n          return {\n            success: true,\n            action: input.action,\n            normalizedAction: convertToAgentUIAction(input.operator_action),\n            observation: undefined, // Reserved for future implementation\n          };\n        },\n      }),\n    );\n    super.initialize();\n  }\n\n  async onLLMRequest(id: string, payload: LLMRequestHookPayload): Promise<void> {\n    try {\n      const safeStringify = (obj: unknown, max = 800): string => {\n        try {\n          const s = JSON.stringify(obj);\n          return s.length > max ? s.slice(0, max) + '…(truncated)' : s;\n        } catch {\n          return '[unserializable]';\n        }\n      };\n\n      const req = payload?.request;\n      const messages = req && Array.isArray(req.messages) ? req.messages : [];\n      const model = req && typeof req.model === 'string' ? req.model : 'unknown';\n\n      const hasImages = (() => {\n        try {\n          let cnt = 0;\n          for (const m of messages as Array<{ content?: unknown }>) {\n            const content = m?.content;\n            if (Array.isArray(content)) {\n              for (const part of content as Array<{ type?: string }>) {\n                if (part?.type === 'image_url') cnt++;\n              }\n            }\n          }\n          return cnt;\n        } catch {\n          return undefined as unknown as number | undefined;\n        }\n      })();\n\n      const summary = {\n        id,\n        model,\n        messagesCount: messages.length,\n        hasImages,\n      };\n\n      this.logger.info('[GUIAgent] onLLMRequest summary:', safeStringify(summary));\n\n      const firstMsg = messages[0];\n      if (firstMsg) {\n        this.logger.debug('[GUIAgent] onLLMRequest first message:', safeStringify(firstMsg, 1200));\n      }\n    } catch (e) {\n      this.logger.error('[GUIAgent] onLLMRequest logging failed:', e);\n    }\n  }\n\n  async onEachAgentLoopStart(sessionId: string) {\n    this.logger.info('onEachAgentLoopStart', sessionId);\n  }\n\n  async onAgentLoopEnd(id: string): Promise<void> {\n    // await this.browserOperator.cleanup();\n  }\n\n  async onBeforeToolCall(\n    id: string,\n    toolCall: { toolCallId: string; name: string },\n    args: unknown,\n  ) {\n    return args;\n  }\n\n  async onAfterToolCall(\n    id: string,\n    toolCall: { toolCallId: string; name: string },\n    result: unknown,\n  ): Promise<unknown> {\n    this.logger.info('onAfterToolCall toolCall', JSON.stringify(toolCall));\n\n    if (toolCall.name !== GUI_ADAPTED_TOOL_NAME) {\n      this.logger.info('onAfterToolCall: skipping screenshot');\n      return;\n    }\n\n    await sleep(this.loopIntervalInMs);\n\n    const output = await this.operator!.doScreenshot();\n    if (!output) {\n      this.logger.error('Failed to get screenshot');\n      return;\n    }\n\n    const base64Tool = new Base64ImageParser(output.base64);\n    const base64Uri = base64Tool.getDataUri();\n    if (!base64Uri) {\n      this.logger.error('Failed to get base64 image uri');\n      return;\n    }\n\n    const { width: imageWidth, height: imageHeight } = base64Tool.getDimensions() || {\n      width: -1,\n      height: -1,\n    };\n\n    const content: ChatCompletionContentPart[] = [\n      {\n        type: 'image_url',\n        image_url: {\n          url: base64Uri,\n          detail: this.detailCalculator(imageWidth, imageHeight),\n        },\n      },\n    ];\n\n    if (output?.url) {\n      content.push({\n        type: 'text',\n        text: `The current page's url: ${output.url}`,\n      });\n    }\n\n    const eventStream = this.getEventStream();\n    const events = eventStream.getEvents();\n    this.logger.info('onAfterToolCall events length:', events.length);\n\n    const event = eventStream.createEvent('environment_input', {\n      description: 'Browser Screenshot',\n      content,\n      metadata: {\n        type: 'screenshot',\n        url: output?.url,\n      },\n    });\n    eventStream.sendEvent(event);\n    return result;\n  }\n}\n"],"names":["__webpack_require__","definition","key","Object","obj","prop","Symbol","defaultLogger","ConsoleLogger","LogLevel","GUIAgent","BaseGUIAgent","Tool","GUI_ADAPTED_TOOL_NAME","input","createGUIErrorResponse","normalizeActionCoords","JSON","result","convertToAgentUIAction","undefined","id","payload","safeStringify","max","s","req","messages","Array","model","hasImages","cnt","m","content","part","summary","firstMsg","e","sessionId","toolCall","args","sleep","output","base64Tool","Base64ImageParser","base64Uri","imageWidth","imageHeight","eventStream","events","event","config","operator","systemPrompt","customeActionParser","normalizeCoordinates","detailCalculator","maxLoopCount","loopIntervalInMs","finalSystemPrompt","SYSTEM_PROMPT","p","isSystemPromptTemplate","assembleSystemPrompt","AdaptedToolCallEngine","GUIAgentToolCallEngine","defaultNormalizeCoords","defaultDetailCalculator"],"mappings":";;;;;;;IAAAA,oBAAoB,CAAC,GAAG,CAAC,UAASC;QACjC,IAAI,IAAIC,OAAOD,WACR,IAAGD,oBAAoB,CAAC,CAACC,YAAYC,QAAQ,CAACF,oBAAoB,CAAC,CAAC,UAASE,MACzEC,OAAO,cAAc,CAAC,UAASD,KAAK;YAAE,YAAY;YAAM,KAAKD,UAAU,CAACC,IAAI;QAAC;IAGzF;;;ICNAF,oBAAoB,CAAC,GAAG,CAACI,KAAKC,OAAUF,OAAO,SAAS,CAAC,cAAc,CAAC,IAAI,CAACC,KAAKC;;;ICClFL,oBAAoB,CAAC,GAAG,CAAC;QACxB,IAAG,AAAkB,eAAlB,OAAOM,UAA0BA,OAAO,WAAW,EACrDH,OAAO,cAAc,CAAC,UAASG,OAAO,WAAW,EAAE;YAAE,OAAO;QAAS;QAEtEH,OAAO,cAAc,CAAC,UAAS,cAAc;YAAE,OAAO;QAAK;IAC5D;;;;;;;;;;;;;;;;ACHC;;;;;;;;;;AA2BD,MAAMI,gBAAgB,IAAIC,sBAAAA,aAAaA,CAAC,cAAcC,sBAAAA,QAAAA,CAAAA,KAAc;AAE7D,MAAMC,iBAAqCC,qBAAAA,YAAYA;IAuD5D,MAAM,aAAa;QAEjB,IAAI,CAAC,YAAY,CACf,IAAIC,sBAAAA,IAAIA,CAAC;YACP,IAAIC,sCAAAA,qBAAqBA;YACzB,aAAa;YACb,YAAY,CAAC;YACb,UAAU,OAAOC;gBACf,IAAI,CAAC,MAAM,CAAC,GAAG,CAAC,GAAGD,sCAAAA,qBAAqBA,CAAC,OAAO,CAAC,EAAEC;gBACnD,IAAI,CAAC,IAAI,CAAC,QAAQ,EAChB,OAAOC,AAAAA,IAAAA,kCAAAA,sBAAAA,AAAAA,EAAuBD,MAAM,MAAM,EAAE;gBAE9C,IAAIA,MAAM,YAAY,EACpB,OAAOC,AAAAA,IAAAA,kCAAAA,sBAAAA,AAAAA,EAAuBD,MAAM,MAAM,EAAEA,MAAM,YAAY;gBAGhE,IAAIA,MAAM,eAAe,EACvBA,MAAM,eAAe,GAAGE,AAAAA,IAAAA,sBAAAA,qBAAAA,AAAAA,EACtBF,MAAM,eAAe,EACrB,IAAI,CAAC,oBAAoB;gBAG7B,IAAI,CAAC,MAAM,CAAC,IAAI,CAAC,sBAAsBG,KAAK,SAAS,CAACH,MAAM,eAAe;gBAC3E,MAAMI,SAAS,MAAM,IAAI,CAAC,QAAQ,CAAE,SAAS,CAAC;oBAC5C,SAAS;wBAACJ,MAAM,eAAe;qBAAC;gBAClC;gBACA,IAAII,OAAO,YAAY,EACrB,OAAOH,AAAAA,IAAAA,kCAAAA,sBAAAA,AAAAA,EAAuBD,MAAM,MAAM,EAAEI,OAAO,YAAY;gBAGjE,OAAO;oBACL,SAAS;oBACT,QAAQJ,MAAM,MAAM;oBACpB,kBAAkBK,AAAAA,IAAAA,kCAAAA,sBAAAA,AAAAA,EAAuBL,MAAM,eAAe;oBAC9D,aAAaM;gBACf;YACF;QACF;QAEF,KAAK,CAAC;IACR;IAEA,MAAM,aAAaC,EAAU,EAAEC,OAA8B,EAAiB;QAC5E,IAAI;YACF,MAAMC,gBAAgB,CAACnB,KAAcoB,MAAM,GAAG;gBAC5C,IAAI;oBACF,MAAMC,IAAIR,KAAK,SAAS,CAACb;oBACzB,OAAOqB,EAAE,MAAM,GAAGD,MAAMC,EAAE,KAAK,CAAC,GAAGD,OAAO,sBAAiBC;gBAC7D,EAAE,OAAM;oBACN,OAAO;gBACT;YACF;YAEA,MAAMC,MAAMJ,QAAAA,UAAAA,KAAAA,IAAAA,QAAS,OAAO;YAC5B,MAAMK,WAAWD,OAAOE,MAAM,OAAO,CAACF,IAAI,QAAQ,IAAIA,IAAI,QAAQ,GAAG,EAAE;YACvE,MAAMG,QAAQH,OAAO,AAAqB,YAArB,OAAOA,IAAI,KAAK,GAAgBA,IAAI,KAAK,GAAG;YAEjE,MAAMI,YAAa,AAAC;gBAClB,IAAI;oBACF,IAAIC,MAAM;oBACV,KAAK,MAAMC,KAAKL,SAA0C;wBACxD,MAAMM,UAAUD,QAAAA,IAAAA,KAAAA,IAAAA,EAAG,OAAO;wBAC1B,IAAIJ,MAAM,OAAO,CAACK,UAChB;4BAAA,KAAK,MAAMC,QAAQD,QACjB,IAAIC,AAAAA,CAAAA,QAAAA,OAAAA,KAAAA,IAAAA,KAAM,IAAI,AAAD,MAAM,aAAaH;wBAClC;oBAEJ;oBACA,OAAOA;gBACT,EAAE,OAAM;oBACN;gBACF;YACF;YAEA,MAAMI,UAAU;gBACdd;gBACAQ;gBACA,eAAeF,SAAS,MAAM;gBAC9BG;YACF;YAEA,IAAI,CAAC,MAAM,CAAC,IAAI,CAAC,oCAAoCP,cAAcY;YAEnE,MAAMC,WAAWT,QAAQ,CAAC,EAAE;YAC5B,IAAIS,UACF,IAAI,CAAC,MAAM,CAAC,KAAK,CAAC,0CAA0Cb,cAAca,UAAU;QAExF,EAAE,OAAOC,GAAG;YACV,IAAI,CAAC,MAAM,CAAC,KAAK,CAAC,2CAA2CA;QAC/D;IACF;IAEA,MAAM,qBAAqBC,SAAiB,EAAE;QAC5C,IAAI,CAAC,MAAM,CAAC,IAAI,CAAC,wBAAwBA;IAC3C;IAEA,MAAM,eAAejB,EAAU,EAAiB,CAEhD;IAEA,MAAM,iBACJA,EAAU,EACVkB,QAA8C,EAC9CC,IAAa,EACb;QACA,OAAOA;IACT;IAEA,MAAM,gBACJnB,EAAU,EACVkB,QAA8C,EAC9CrB,MAAe,EACG;QAClB,IAAI,CAAC,MAAM,CAAC,IAAI,CAAC,4BAA4BD,KAAK,SAAS,CAACsB;QAE5D,IAAIA,SAAS,IAAI,KAAK1B,sCAAAA,qBAAqBA,EAAE,YAC3C,IAAI,CAAC,MAAM,CAAC,IAAI,CAAC;QAInB,MAAM4B,AAAAA,IAAAA,sBAAAA,KAAAA,AAAAA,EAAM,IAAI,CAAC,gBAAgB;QAEjC,MAAMC,SAAS,MAAM,IAAI,CAAC,QAAQ,CAAE,YAAY;QAChD,IAAI,CAACA,QAAQ,YACX,IAAI,CAAC,MAAM,CAAC,KAAK,CAAC;QAIpB,MAAMC,aAAa,IAAIC,4BAAAA,iBAAiBA,CAACF,OAAO,MAAM;QACtD,MAAMG,YAAYF,WAAW,UAAU;QACvC,IAAI,CAACE,WAAW,YACd,IAAI,CAAC,MAAM,CAAC,KAAK,CAAC;QAIpB,MAAM,EAAE,OAAOC,UAAU,EAAE,QAAQC,WAAW,EAAE,GAAGJ,WAAW,aAAa,MAAM;YAC/E,OAAO;YACP,QAAQ;QACV;QAEA,MAAMV,UAAuC;YAC3C;gBACE,MAAM;gBACN,WAAW;oBACT,KAAKY;oBACL,QAAQ,IAAI,CAAC,gBAAgB,CAACC,YAAYC;gBAC5C;YACF;SACD;QAED,IAAIL,QAAAA,SAAAA,KAAAA,IAAAA,OAAQ,GAAG,EACbT,QAAQ,IAAI,CAAC;YACX,MAAM;YACN,MAAM,CAAC,wBAAwB,EAAES,OAAO,GAAG,EAAE;QAC/C;QAGF,MAAMM,cAAc,IAAI,CAAC,cAAc;QACvC,MAAMC,SAASD,YAAY,SAAS;QACpC,IAAI,CAAC,MAAM,CAAC,IAAI,CAAC,kCAAkCC,OAAO,MAAM;QAEhE,MAAMC,QAAQF,YAAY,WAAW,CAAC,qBAAqB;YACzD,aAAa;YACbf;YACA,UAAU;gBACR,MAAM;gBACN,KAAKS,QAAAA,SAAAA,KAAAA,IAAAA,OAAQ,GAAG;YAClB;QACF;QACAM,YAAY,SAAS,CAACE;QACtB,OAAOhC;IACT;IA1NA,YAAYiC,MAAyB,CAAE;QACrC,MAAM,EACJC,QAAQ,EACRvB,KAAK,EACLwB,YAAY,EACZC,mBAAmB,EACnBC,oBAAoB,EACpBC,gBAAgB,EAChBC,YAAY,EACZC,mBAAmB,GAAG,EACvB,GAAGP;QACJ,IAAIQ,oBAAoBC,oCAAAA,aAAaA;QACrC,IAAI,AAAwB,YAAxB,OAAOP,cACTM,oBAAoBN;aACf,IAAIzB,MAAM,OAAO,CAACyB,eACvBM,oBAAoBN,aACjB,GAAG,CAAC,CAACQ,IAAO,AAAa,YAAb,OAAOA,IAAiBA,IAAIA,EAAE,OAAO,EACjD,IAAI,CAAC;aACH,IAAIR,gBAAgBS,AAAAA,IAAAA,sBAAAA,sBAAAA,AAAAA,EAAuBT,eAChDM,oBAAoBI,AAAAA,IAAAA,sBAAAA,oBAAAA,AAAAA,EAAqBV,cAAcD,SAAS,mBAAmB;QAErF7C,cAAc,KAAK,CAAC,8BAA8BoD;QAGlD,MAAMK,wBAAwB,cAAcC,2CAAAA,sBAAsBA;YAChE,aAAc;gBACZ,KAAK,CAACX;YACR;QACF;QAEA,KAAK,CAAC;YACJ,MAAM5C,SAAS,KAAK;YACpB,cAAciD;YACd,OAAO,EAAE;YACT,gBAAgBK;YAChB,OAAOnC;YACP,GAAI4B,gBAAgB;gBAAE,eAAeA;YAAa,CAAC;YACnD,UAAUhD,sBAAAA,QAAAA,CAAAA,KAAc;QAC1B,IA3CF,uBAAQ,YAAR,SACA,uBAAQ,wBAAR,SACA,uBAAQ,oBAAR,SACA,uBAAQ,oBAAR;QAyCE,IAAI,CAAC,QAAQ,GAAG2C;QAChB,IAAI,CAAC,oBAAoB,GAAGG,wBAAwBW,yCAAAA,sBAAsBA;QAE1E,IAAI,CAAC,gBAAgB,GAAGV,oBAAoBW,yCAAAA,uBAAuBA;QACnE,IAAI,CAAC,gBAAgB,GAAGT;QACxB,IAAI,CAAC,MAAM,GAAG,IAAI,CAAC,MAAM,CAAC,KAAK,CAAC;IAClC;AA8KF;AAlOE,iBADWhD,UACJ,SAAQ"}
+{"version":3,"file":"GUIAgent.js","sources":["webpack://@ui-tars-test/agent-sdk/webpack/runtime/define_property_getters","webpack://@ui-tars-test/agent-sdk/webpack/runtime/has_own_property","webpack://@ui-tars-test/agent-sdk/webpack/runtime/make_namespace_object","webpack://@ui-tars-test/agent-sdk/./src/GUIAgent.ts"],"sourcesContent":["__webpack_require__.d = (exports, definition) => {\n\tfor(var key in definition) {\n        if(__webpack_require__.o(definition, key) && !__webpack_require__.o(exports, key)) {\n            Object.defineProperty(exports, key, { enumerable: true, get: definition[key] });\n        }\n    }\n};","__webpack_require__.o = (obj, prop) => (Object.prototype.hasOwnProperty.call(obj, prop))","// define __esModule on exports\n__webpack_require__.r = (exports) => {\n\tif(typeof Symbol !== 'undefined' && Symbol.toStringTag) {\n\t\tObject.defineProperty(exports, Symbol.toStringTag, { value: 'Module' });\n\t}\n\tObject.defineProperty(exports, '__esModule', { value: true });\n};","/*\n * Copyright (c) 2025 Bytedance, Inc. and its affiliates.\n * SPDX-License-Identifier: Apache-2.0\n */\nimport {\n  LLMRequestHookPayload,\n  ChatCompletionContentPart,\n  LogLevel,\n  Tool,\n  ConsoleLogger,\n} from '@ui-tars-test/tarko-agent';\nimport { GUIAgentToolCallEngine } from './ToolCallEngine';\nimport { SYSTEM_PROMPT } from './prompts';\nimport { Base64ImageParser } from '@agent-infra/media-utils';\nimport { Operator, BaseGUIAgent } from '@ui-tars-test/shared/base';\nimport {\n  GUIAgentConfig,\n  NormalizeCoordinates,\n  ImageDetailCalculator,\n} from '@ui-tars-test/shared/types';\nimport {\n  assembleSystemPrompt,\n  isSystemPromptTemplate,\n  normalizeActionCoords,\n  sleep,\n} from '@ui-tars-test/shared/utils';\nimport { GUI_ADAPTED_TOOL_NAME } from './constants';\nimport { convertToAgentUIAction, createGUIErrorResponse } from './utils';\nimport { defaultNormalizeCoords, defaultDetailCalculator } from './defaultImpls';\n\nconst defaultLogger = new ConsoleLogger('[GUIAgent]', LogLevel.DEBUG);\n\nexport class GUIAgent<T extends Operator> extends BaseGUIAgent {\n  static label = 'GUI Agent';\n\n  private operator: Operator | undefined;\n  private normalizeCoordinates: NormalizeCoordinates;\n  private detailCalculator: ImageDetailCalculator;\n  private loopIntervalInMs: number;\n\n  constructor(config: GUIAgentConfig<T>) {\n    const {\n      operator,\n      model,\n      systemPrompt,\n      customeActionParser,\n      normalizeCoordinates,\n      detailCalculator,\n      maxLoopCount,\n      loopIntervalInMs = 500,\n    } = config;\n    let finalSystemPrompt = SYSTEM_PROMPT;\n    if (typeof systemPrompt === 'string') {\n      finalSystemPrompt = systemPrompt;\n    } else if (Array.isArray(systemPrompt)) {\n      finalSystemPrompt = systemPrompt\n        .map((p) => (typeof p === 'string' ? p : p.content))\n        .join('\\n\\n');\n    } else if (systemPrompt && isSystemPromptTemplate(systemPrompt)) {\n      finalSystemPrompt = assembleSystemPrompt(systemPrompt, operator.getSupportedActions());\n    }\n    defaultLogger.debug('final instructions for sp:', finalSystemPrompt);\n\n    // Create a adapted ToolCallEngine constructor that captures the customeActionParser\n    const AdaptedToolCallEngine = class extends GUIAgentToolCallEngine {\n      constructor() {\n        super(customeActionParser);\n      }\n    };\n\n    super({\n      name: GUIAgent.label,\n      instructions: finalSystemPrompt,\n      tools: [],\n      toolCallEngine: AdaptedToolCallEngine,\n      model: model,\n      ...(maxLoopCount && { maxIterations: maxLoopCount }),\n      logLevel: LogLevel.DEBUG,\n    });\n    this.operator = operator;\n    this.normalizeCoordinates = normalizeCoordinates ?? defaultNormalizeCoords;\n    // Default detail calculator implementation\n    this.detailCalculator = detailCalculator ?? defaultDetailCalculator;\n    this.loopIntervalInMs = loopIntervalInMs;\n    this.logger = this.logger.spawn('[GUIAgent]');\n  }\n\n  async initialize() {\n    // Register the GUI tool\n    this.registerTool(\n      new Tool({\n        id: GUI_ADAPTED_TOOL_NAME,\n        description: 'operator tool',\n        parameters: {}, // no need to pass parameters\n        function: async (input) => {\n          this.logger.log(`${GUI_ADAPTED_TOOL_NAME} input:`, input);\n          if (!this.operator) {\n            return createGUIErrorResponse(input.action, 'Operator not initialized');\n          }\n          if (input.errorMessage) {\n            return createGUIErrorResponse(input.action, input.errorMessage);\n          }\n          // normalize coordinates\n          if (input.operator_action) {\n            input.operator_action = normalizeActionCoords(\n              input.operator_action,\n              this.normalizeCoordinates,\n            );\n          }\n          this.logger.info('action to execute:', JSON.stringify(input.operator_action));\n          const result = await this.operator!.doExecute({\n            actions: [input.operator_action],\n          });\n          if (result.errorMessage) {\n            return createGUIErrorResponse(input.action, result.errorMessage);\n          }\n          // return { action: input.action, status: 'success', result };\n          return {\n            success: true,\n            action: input.action,\n            normalizedAction: convertToAgentUIAction(input.operator_action),\n            observation: undefined, // Reserved for future implementation\n          };\n        },\n      }),\n    );\n    super.initialize();\n  }\n\n  async onLLMRequest(id: string, payload: LLMRequestHookPayload): Promise<void> {\n    try {\n      const safeStringify = (obj: unknown, max = 800): string => {\n        try {\n          const s = JSON.stringify(obj);\n          return s.length > max ? s.slice(0, max) + '…(truncated)' : s;\n        } catch {\n          return '[unserializable]';\n        }\n      };\n\n      const req = payload?.request;\n      const messages = req && Array.isArray(req.messages) ? req.messages : [];\n      const model = req && typeof req.model === 'string' ? req.model : 'unknown';\n\n      const hasImages = (() => {\n        try {\n          let cnt = 0;\n          for (const m of messages as Array<{ content?: unknown }>) {\n            const content = m?.content;\n            if (Array.isArray(content)) {\n              for (const part of content as Array<{ type?: string }>) {\n                if (part?.type === 'image_url') cnt++;\n              }\n            }\n          }\n          return cnt;\n        } catch {\n          return undefined as unknown as number | undefined;\n        }\n      })();\n\n      const summary = {\n        id,\n        model,\n        messagesCount: messages.length,\n        hasImages,\n      };\n\n      this.logger.info('[GUIAgent] onLLMRequest summary:', safeStringify(summary));\n\n      const firstMsg = messages[0];\n      if (firstMsg) {\n        this.logger.debug('[GUIAgent] onLLMRequest first message:', safeStringify(firstMsg, 1200));\n      }\n    } catch (e) {\n      this.logger.error('[GUIAgent] onLLMRequest logging failed:', e);\n    }\n  }\n\n  async onEachAgentLoopStart(sessionId: string) {\n    this.logger.info('onEachAgentLoopStart', sessionId);\n  }\n\n  async onAgentLoopEnd(id: string): Promise<void> {\n    // await this.browserOperator.cleanup();\n  }\n\n  async onBeforeToolCall(\n    id: string,\n    toolCall: { toolCallId: string; name: string },\n    args: unknown,\n  ) {\n    return args;\n  }\n\n  async onAfterToolCall(\n    id: string,\n    toolCall: { toolCallId: string; name: string },\n    result: unknown,\n  ): Promise<unknown> {\n    this.logger.info('onAfterToolCall toolCall', JSON.stringify(toolCall));\n\n    if (toolCall.name !== GUI_ADAPTED_TOOL_NAME) {\n      this.logger.info('onAfterToolCall: skipping screenshot');\n      return;\n    }\n\n    await sleep(this.loopIntervalInMs);\n\n    const output = await this.operator!.doScreenshot();\n    if (!output) {\n      this.logger.error('Failed to get screenshot');\n      return;\n    }\n\n    const base64Tool = new Base64ImageParser(output.base64);\n    const base64Uri = base64Tool.getDataUri();\n    if (!base64Uri) {\n      this.logger.error('Failed to get base64 image uri');\n      return;\n    }\n\n    const { width: imageWidth, height: imageHeight } = base64Tool.getDimensions() || {\n      width: -1,\n      height: -1,\n    };\n\n    const content: ChatCompletionContentPart[] = [\n      {\n        type: 'image_url',\n        image_url: {\n          url: base64Uri,\n          detail: this.detailCalculator(imageWidth, imageHeight),\n        },\n      },\n    ];\n\n    if (output?.url) {\n      content.push({\n        type: 'text',\n        text: `The current page's url: ${output.url}`,\n      });\n    }\n\n    const eventStream = this.getEventStream();\n    const events = eventStream.getEvents();\n    this.logger.info('onAfterToolCall events length:', events.length);\n\n    const event = eventStream.createEvent('environment_input', {\n      description: 'Browser Screenshot',\n      content,\n      metadata: {\n        type: 'screenshot',\n        url: output?.url,\n      },\n    });\n    eventStream.sendEvent(event);\n    return result;\n  }\n}\n"],"names":["__webpack_require__","definition","key","Object","obj","prop","Symbol","defaultLogger","ConsoleLogger","LogLevel","GUIAgent","BaseGUIAgent","Tool","GUI_ADAPTED_TOOL_NAME","input","createGUIErrorResponse","normalizeActionCoords","JSON","result","convertToAgentUIAction","undefined","id","payload","safeStringify","max","s","req","messages","Array","model","hasImages","cnt","m","content","part","summary","firstMsg","e","sessionId","toolCall","args","sleep","output","base64Tool","Base64ImageParser","base64Uri","imageWidth","imageHeight","eventStream","events","event","config","operator","systemPrompt","customeActionParser","normalizeCoordinates","detailCalculator","maxLoopCount","loopIntervalInMs","finalSystemPrompt","SYSTEM_PROMPT","p","isSystemPromptTemplate","assembleSystemPrompt","AdaptedToolCallEngine","GUIAgentToolCallEngine","defaultNormalizeCoords","defaultDetailCalculator"],"mappings":";;;;;;;IAAAA,oBAAoB,CAAC,GAAG,CAAC,UAASC;QACjC,IAAI,IAAIC,OAAOD,WACR,IAAGD,oBAAoB,CAAC,CAACC,YAAYC,QAAQ,CAACF,oBAAoB,CAAC,CAAC,UAASE,MACzEC,OAAO,cAAc,CAAC,UAASD,KAAK;YAAE,YAAY;YAAM,KAAKD,UAAU,CAACC,IAAI;QAAC;IAGzF;;;ICNAF,oBAAoB,CAAC,GAAG,CAACI,KAAKC,OAAUF,OAAO,SAAS,CAAC,cAAc,CAAC,IAAI,CAACC,KAAKC;;;ICClFL,oBAAoB,CAAC,GAAG,CAAC;QACxB,IAAG,AAAkB,eAAlB,OAAOM,UAA0BA,OAAO,WAAW,EACrDH,OAAO,cAAc,CAAC,UAASG,OAAO,WAAW,EAAE;YAAE,OAAO;QAAS;QAEtEH,OAAO,cAAc,CAAC,UAAS,cAAc;YAAE,OAAO;QAAK;IAC5D;;;;;;;;;;;;;;;;ACHC;;;;;;;;;;AA2BD,MAAMI,gBAAgB,IAAIC,4BAAAA,aAAaA,CAAC,cAAcC,4BAAAA,QAAAA,CAAAA,KAAc;AAE7D,MAAMC,iBAAqCC,qBAAAA,YAAYA;IAuD5D,MAAM,aAAa;QAEjB,IAAI,CAAC,YAAY,CACf,IAAIC,4BAAAA,IAAIA,CAAC;YACP,IAAIC,sCAAAA,qBAAqBA;YACzB,aAAa;YACb,YAAY,CAAC;YACb,UAAU,OAAOC;gBACf,IAAI,CAAC,MAAM,CAAC,GAAG,CAAC,GAAGD,sCAAAA,qBAAqBA,CAAC,OAAO,CAAC,EAAEC;gBACnD,IAAI,CAAC,IAAI,CAAC,QAAQ,EAChB,OAAOC,AAAAA,IAAAA,kCAAAA,sBAAAA,AAAAA,EAAuBD,MAAM,MAAM,EAAE;gBAE9C,IAAIA,MAAM,YAAY,EACpB,OAAOC,AAAAA,IAAAA,kCAAAA,sBAAAA,AAAAA,EAAuBD,MAAM,MAAM,EAAEA,MAAM,YAAY;gBAGhE,IAAIA,MAAM,eAAe,EACvBA,MAAM,eAAe,GAAGE,AAAAA,IAAAA,sBAAAA,qBAAAA,AAAAA,EACtBF,MAAM,eAAe,EACrB,IAAI,CAAC,oBAAoB;gBAG7B,IAAI,CAAC,MAAM,CAAC,IAAI,CAAC,sBAAsBG,KAAK,SAAS,CAACH,MAAM,eAAe;gBAC3E,MAAMI,SAAS,MAAM,IAAI,CAAC,QAAQ,CAAE,SAAS,CAAC;oBAC5C,SAAS;wBAACJ,MAAM,eAAe;qBAAC;gBAClC;gBACA,IAAII,OAAO,YAAY,EACrB,OAAOH,AAAAA,IAAAA,kCAAAA,sBAAAA,AAAAA,EAAuBD,MAAM,MAAM,EAAEI,OAAO,YAAY;gBAGjE,OAAO;oBACL,SAAS;oBACT,QAAQJ,MAAM,MAAM;oBACpB,kBAAkBK,AAAAA,IAAAA,kCAAAA,sBAAAA,AAAAA,EAAuBL,MAAM,eAAe;oBAC9D,aAAaM;gBACf;YACF;QACF;QAEF,KAAK,CAAC;IACR;IAEA,MAAM,aAAaC,EAAU,EAAEC,OAA8B,EAAiB;QAC5E,IAAI;YACF,MAAMC,gBAAgB,CAACnB,KAAcoB,MAAM,GAAG;gBAC5C,IAAI;oBACF,MAAMC,IAAIR,KAAK,SAAS,CAACb;oBACzB,OAAOqB,EAAE,MAAM,GAAGD,MAAMC,EAAE,KAAK,CAAC,GAAGD,OAAO,sBAAiBC;gBAC7D,EAAE,OAAM;oBACN,OAAO;gBACT;YACF;YAEA,MAAMC,MAAMJ,QAAAA,UAAAA,KAAAA,IAAAA,QAAS,OAAO;YAC5B,MAAMK,WAAWD,OAAOE,MAAM,OAAO,CAACF,IAAI,QAAQ,IAAIA,IAAI,QAAQ,GAAG,EAAE;YACvE,MAAMG,QAAQH,OAAO,AAAqB,YAArB,OAAOA,IAAI,KAAK,GAAgBA,IAAI,KAAK,GAAG;YAEjE,MAAMI,YAAa,AAAC;gBAClB,IAAI;oBACF,IAAIC,MAAM;oBACV,KAAK,MAAMC,KAAKL,SAA0C;wBACxD,MAAMM,UAAUD,QAAAA,IAAAA,KAAAA,IAAAA,EAAG,OAAO;wBAC1B,IAAIJ,MAAM,OAAO,CAACK,UAChB;4BAAA,KAAK,MAAMC,QAAQD,QACjB,IAAIC,AAAAA,CAAAA,QAAAA,OAAAA,KAAAA,IAAAA,KAAM,IAAI,AAAD,MAAM,aAAaH;wBAClC;oBAEJ;oBACA,OAAOA;gBACT,EAAE,OAAM;oBACN;gBACF;YACF;YAEA,MAAMI,UAAU;gBACdd;gBACAQ;gBACA,eAAeF,SAAS,MAAM;gBAC9BG;YACF;YAEA,IAAI,CAAC,MAAM,CAAC,IAAI,CAAC,oCAAoCP,cAAcY;YAEnE,MAAMC,WAAWT,QAAQ,CAAC,EAAE;YAC5B,IAAIS,UACF,IAAI,CAAC,MAAM,CAAC,KAAK,CAAC,0CAA0Cb,cAAca,UAAU;QAExF,EAAE,OAAOC,GAAG;YACV,IAAI,CAAC,MAAM,CAAC,KAAK,CAAC,2CAA2CA;QAC/D;IACF;IAEA,MAAM,qBAAqBC,SAAiB,EAAE;QAC5C,IAAI,CAAC,MAAM,CAAC,IAAI,CAAC,wBAAwBA;IAC3C;IAEA,MAAM,eAAejB,EAAU,EAAiB,CAEhD;IAEA,MAAM,iBACJA,EAAU,EACVkB,QAA8C,EAC9CC,IAAa,EACb;QACA,OAAOA;IACT;IAEA,MAAM,gBACJnB,EAAU,EACVkB,QAA8C,EAC9CrB,MAAe,EACG;QAClB,IAAI,CAAC,MAAM,CAAC,IAAI,CAAC,4BAA4BD,KAAK,SAAS,CAACsB;QAE5D,IAAIA,SAAS,IAAI,KAAK1B,sCAAAA,qBAAqBA,EAAE,YAC3C,IAAI,CAAC,MAAM,CAAC,IAAI,CAAC;QAInB,MAAM4B,AAAAA,IAAAA,sBAAAA,KAAAA,AAAAA,EAAM,IAAI,CAAC,gBAAgB;QAEjC,MAAMC,SAAS,MAAM,IAAI,CAAC,QAAQ,CAAE,YAAY;QAChD,IAAI,CAACA,QAAQ,YACX,IAAI,CAAC,MAAM,CAAC,KAAK,CAAC;QAIpB,MAAMC,aAAa,IAAIC,4BAAAA,iBAAiBA,CAACF,OAAO,MAAM;QACtD,MAAMG,YAAYF,WAAW,UAAU;QACvC,IAAI,CAACE,WAAW,YACd,IAAI,CAAC,MAAM,CAAC,KAAK,CAAC;QAIpB,MAAM,EAAE,OAAOC,UAAU,EAAE,QAAQC,WAAW,EAAE,GAAGJ,WAAW,aAAa,MAAM;YAC/E,OAAO;YACP,QAAQ;QACV;QAEA,MAAMV,UAAuC;YAC3C;gBACE,MAAM;gBACN,WAAW;oBACT,KAAKY;oBACL,QAAQ,IAAI,CAAC,gBAAgB,CAACC,YAAYC;gBAC5C;YACF;SACD;QAED,IAAIL,QAAAA,SAAAA,KAAAA,IAAAA,OAAQ,GAAG,EACbT,QAAQ,IAAI,CAAC;YACX,MAAM;YACN,MAAM,CAAC,wBAAwB,EAAES,OAAO,GAAG,EAAE;QAC/C;QAGF,MAAMM,cAAc,IAAI,CAAC,cAAc;QACvC,MAAMC,SAASD,YAAY,SAAS;QACpC,IAAI,CAAC,MAAM,CAAC,IAAI,CAAC,kCAAkCC,OAAO,MAAM;QAEhE,MAAMC,QAAQF,YAAY,WAAW,CAAC,qBAAqB;YACzD,aAAa;YACbf;YACA,UAAU;gBACR,MAAM;gBACN,KAAKS,QAAAA,SAAAA,KAAAA,IAAAA,OAAQ,GAAG;YAClB;QACF;QACAM,YAAY,SAAS,CAACE;QACtB,OAAOhC;IACT;IA1NA,YAAYiC,MAAyB,CAAE;QACrC,MAAM,EACJC,QAAQ,EACRvB,KAAK,EACLwB,YAAY,EACZC,mBAAmB,EACnBC,oBAAoB,EACpBC,gBAAgB,EAChBC,YAAY,EACZC,mBAAmB,GAAG,EACvB,GAAGP;QACJ,IAAIQ,oBAAoBC,oCAAAA,aAAaA;QACrC,IAAI,AAAwB,YAAxB,OAAOP,cACTM,oBAAoBN;aACf,IAAIzB,MAAM,OAAO,CAACyB,eACvBM,oBAAoBN,aACjB,GAAG,CAAC,CAACQ,IAAO,AAAa,YAAb,OAAOA,IAAiBA,IAAIA,EAAE,OAAO,EACjD,IAAI,CAAC;aACH,IAAIR,gBAAgBS,AAAAA,IAAAA,sBAAAA,sBAAAA,AAAAA,EAAuBT,eAChDM,oBAAoBI,AAAAA,IAAAA,sBAAAA,oBAAAA,AAAAA,EAAqBV,cAAcD,SAAS,mBAAmB;QAErF7C,cAAc,KAAK,CAAC,8BAA8BoD;QAGlD,MAAMK,wBAAwB,cAAcC,2CAAAA,sBAAsBA;YAChE,aAAc;gBACZ,KAAK,CAACX;YACR;QACF;QAEA,KAAK,CAAC;YACJ,MAAM5C,SAAS,KAAK;YACpB,cAAciD;YACd,OAAO,EAAE;YACT,gBAAgBK;YAChB,OAAOnC;YACP,GAAI4B,gBAAgB;gBAAE,eAAeA;YAAa,CAAC;YACnD,UAAUhD,4BAAAA,QAAAA,CAAAA,KAAc;QAC1B,IA3CF,uBAAQ,YAAR,SACA,uBAAQ,wBAAR,SACA,uBAAQ,oBAAR,SACA,uBAAQ,oBAAR;QAyCE,IAAI,CAAC,QAAQ,GAAG2C;QAChB,IAAI,CAAC,oBAAoB,GAAGG,wBAAwBW,yCAAAA,sBAAsBA;QAE1E,IAAI,CAAC,gBAAgB,GAAGV,oBAAoBW,yCAAAA,uBAAuBA;QACnE,IAAI,CAAC,gBAAgB,GAAGT;QACxB,IAAI,CAAC,MAAM,GAAG,IAAI,CAAC,MAAM,CAAC,KAAK,CAAC;IAClC;AA8KF;AAlOE,iBADWhD,UACJ,SAAQ"}

package/dist/GUIAgent.mjs

@@ -2,7 +2,7 @@
 * Copyright (c) 2025 Bytedance, Inc. and its affiliates.
 * SPDX-License-Identifier: Apache-2.0
 */
-import { ConsoleLogger, LogLevel, Tool } from "@tarko/agent";
+import { ConsoleLogger, LogLevel, Tool } from "@ui-tars-test/tarko-agent";
 import { GUIAgentToolCallEngine } from "./ToolCallEngine.mjs";
 import { SYSTEM_PROMPT } from "./prompts.mjs";
 import { Base64ImageParser } from "@agent-infra/media-utils";

package/dist/GUIAgent.mjs.map

@@ -1 +1 @@
-{"version":3,"file":"GUIAgent.mjs","sources":["webpack://@ui-tars-test/agent-sdk/./src/GUIAgent.ts"],"sourcesContent":["/*\n * Copyright (c) 2025 Bytedance, Inc. and its affiliates.\n * SPDX-License-Identifier: Apache-2.0\n */\nimport {\n  LLMRequestHookPayload,\n  ChatCompletionContentPart,\n  LogLevel,\n  Tool,\n  ConsoleLogger,\n} from '@tarko/agent';\nimport { GUIAgentToolCallEngine } from './ToolCallEngine';\nimport { SYSTEM_PROMPT } from './prompts';\nimport { Base64ImageParser } from '@agent-infra/media-utils';\nimport { Operator, BaseGUIAgent } from '@ui-tars-test/shared/base';\nimport {\n  GUIAgentConfig,\n  NormalizeCoordinates,\n  ImageDetailCalculator,\n} from '@ui-tars-test/shared/types';\nimport {\n  assembleSystemPrompt,\n  isSystemPromptTemplate,\n  normalizeActionCoords,\n  sleep,\n} from '@ui-tars-test/shared/utils';\nimport { GUI_ADAPTED_TOOL_NAME } from './constants';\nimport { convertToAgentUIAction, createGUIErrorResponse } from './utils';\nimport { defaultNormalizeCoords, defaultDetailCalculator } from './defaultImpls';\n\nconst defaultLogger = new ConsoleLogger('[GUIAgent]', LogLevel.DEBUG);\n\nexport class GUIAgent<T extends Operator> extends BaseGUIAgent {\n  static label = 'GUI Agent';\n\n  private operator: Operator | undefined;\n  private normalizeCoordinates: NormalizeCoordinates;\n  private detailCalculator: ImageDetailCalculator;\n  private loopIntervalInMs: number;\n\n  constructor(config: GUIAgentConfig<T>) {\n    const {\n      operator,\n      model,\n      systemPrompt,\n      customeActionParser,\n      normalizeCoordinates,\n      detailCalculator,\n      maxLoopCount,\n      loopIntervalInMs = 500,\n    } = config;\n    let finalSystemPrompt = SYSTEM_PROMPT;\n    if (typeof systemPrompt === 'string') {\n      finalSystemPrompt = systemPrompt;\n    } else if (Array.isArray(systemPrompt)) {\n      finalSystemPrompt = systemPrompt\n        .map((p) => (typeof p === 'string' ? p : p.content))\n        .join('\\n\\n');\n    } else if (systemPrompt && isSystemPromptTemplate(systemPrompt)) {\n      finalSystemPrompt = assembleSystemPrompt(systemPrompt, operator.getSupportedActions());\n    }\n    defaultLogger.debug('final instructions for sp:', finalSystemPrompt);\n\n    // Create a adapted ToolCallEngine constructor that captures the customeActionParser\n    const AdaptedToolCallEngine = class extends GUIAgentToolCallEngine {\n      constructor() {\n        super(customeActionParser);\n      }\n    };\n\n    super({\n      name: GUIAgent.label,\n      instructions: finalSystemPrompt,\n      tools: [],\n      toolCallEngine: AdaptedToolCallEngine,\n      model: model,\n      ...(maxLoopCount && { maxIterations: maxLoopCount }),\n      logLevel: LogLevel.DEBUG,\n    });\n    this.operator = operator;\n    this.normalizeCoordinates = normalizeCoordinates ?? defaultNormalizeCoords;\n    // Default detail calculator implementation\n    this.detailCalculator = detailCalculator ?? defaultDetailCalculator;\n    this.loopIntervalInMs = loopIntervalInMs;\n    this.logger = this.logger.spawn('[GUIAgent]');\n  }\n\n  async initialize() {\n    // Register the GUI tool\n    this.registerTool(\n      new Tool({\n        id: GUI_ADAPTED_TOOL_NAME,\n        description: 'operator tool',\n        parameters: {}, // no need to pass parameters\n        function: async (input) => {\n          this.logger.log(`${GUI_ADAPTED_TOOL_NAME} input:`, input);\n          if (!this.operator) {\n            return createGUIErrorResponse(input.action, 'Operator not initialized');\n          }\n          if (input.errorMessage) {\n            return createGUIErrorResponse(input.action, input.errorMessage);\n          }\n          // normalize coordinates\n          if (input.operator_action) {\n            input.operator_action = normalizeActionCoords(\n              input.operator_action,\n              this.normalizeCoordinates,\n            );\n          }\n          this.logger.info('action to execute:', JSON.stringify(input.operator_action));\n          const result = await this.operator!.doExecute({\n            actions: [input.operator_action],\n          });\n          if (result.errorMessage) {\n            return createGUIErrorResponse(input.action, result.errorMessage);\n          }\n          // return { action: input.action, status: 'success', result };\n          return {\n            success: true,\n            action: input.action,\n            normalizedAction: convertToAgentUIAction(input.operator_action),\n            observation: undefined, // Reserved for future implementation\n          };\n        },\n      }),\n    );\n    super.initialize();\n  }\n\n  async onLLMRequest(id: string, payload: LLMRequestHookPayload): Promise<void> {\n    try {\n      const safeStringify = (obj: unknown, max = 800): string => {\n        try {\n          const s = JSON.stringify(obj);\n          return s.length > max ? s.slice(0, max) + '…(truncated)' : s;\n        } catch {\n          return '[unserializable]';\n        }\n      };\n\n      const req = payload?.request;\n      const messages = req && Array.isArray(req.messages) ? req.messages : [];\n      const model = req && typeof req.model === 'string' ? req.model : 'unknown';\n\n      const hasImages = (() => {\n        try {\n          let cnt = 0;\n          for (const m of messages as Array<{ content?: unknown }>) {\n            const content = m?.content;\n            if (Array.isArray(content)) {\n              for (const part of content as Array<{ type?: string }>) {\n                if (part?.type === 'image_url') cnt++;\n              }\n            }\n          }\n          return cnt;\n        } catch {\n          return undefined as unknown as number | undefined;\n        }\n      })();\n\n      const summary = {\n        id,\n        model,\n        messagesCount: messages.length,\n        hasImages,\n      };\n\n      this.logger.info('[GUIAgent] onLLMRequest summary:', safeStringify(summary));\n\n      const firstMsg = messages[0];\n      if (firstMsg) {\n        this.logger.debug('[GUIAgent] onLLMRequest first message:', safeStringify(firstMsg, 1200));\n      }\n    } catch (e) {\n      this.logger.error('[GUIAgent] onLLMRequest logging failed:', e);\n    }\n  }\n\n  async onEachAgentLoopStart(sessionId: string) {\n    this.logger.info('onEachAgentLoopStart', sessionId);\n  }\n\n  async onAgentLoopEnd(id: string): Promise<void> {\n    // await this.browserOperator.cleanup();\n  }\n\n  async onBeforeToolCall(\n    id: string,\n    toolCall: { toolCallId: string; name: string },\n    args: unknown,\n  ) {\n    return args;\n  }\n\n  async onAfterToolCall(\n    id: string,\n    toolCall: { toolCallId: string; name: string },\n    result: unknown,\n  ): Promise<unknown> {\n    this.logger.info('onAfterToolCall toolCall', JSON.stringify(toolCall));\n\n    if (toolCall.name !== GUI_ADAPTED_TOOL_NAME) {\n      this.logger.info('onAfterToolCall: skipping screenshot');\n      return;\n    }\n\n    await sleep(this.loopIntervalInMs);\n\n    const output = await this.operator!.doScreenshot();\n    if (!output) {\n      this.logger.error('Failed to get screenshot');\n      return;\n    }\n\n    const base64Tool = new Base64ImageParser(output.base64);\n    const base64Uri = base64Tool.getDataUri();\n    if (!base64Uri) {\n      this.logger.error('Failed to get base64 image uri');\n      return;\n    }\n\n    const { width: imageWidth, height: imageHeight } = base64Tool.getDimensions() || {\n      width: -1,\n      height: -1,\n    };\n\n    const content: ChatCompletionContentPart[] = [\n      {\n        type: 'image_url',\n        image_url: {\n          url: base64Uri,\n          detail: this.detailCalculator(imageWidth, imageHeight),\n        },\n      },\n    ];\n\n    if (output?.url) {\n      content.push({\n        type: 'text',\n        text: `The current page's url: ${output.url}`,\n      });\n    }\n\n    const eventStream = this.getEventStream();\n    const events = eventStream.getEvents();\n    this.logger.info('onAfterToolCall events length:', events.length);\n\n    const event = eventStream.createEvent('environment_input', {\n      description: 'Browser Screenshot',\n      content,\n      metadata: {\n        type: 'screenshot',\n        url: output?.url,\n      },\n    });\n    eventStream.sendEvent(event);\n    return result;\n  }\n}\n"],"names":["defaultLogger","ConsoleLogger","LogLevel","GUIAgent","BaseGUIAgent","Tool","GUI_ADAPTED_TOOL_NAME","input","createGUIErrorResponse","normalizeActionCoords","JSON","result","convertToAgentUIAction","undefined","id","payload","safeStringify","obj","max","s","req","messages","Array","model","hasImages","cnt","m","content","part","summary","firstMsg","e","sessionId","toolCall","args","sleep","output","base64Tool","Base64ImageParser","base64Uri","imageWidth","imageHeight","eventStream","events","event","config","operator","systemPrompt","customeActionParser","normalizeCoordinates","detailCalculator","maxLoopCount","loopIntervalInMs","finalSystemPrompt","SYSTEM_PROMPT","p","isSystemPromptTemplate","assembleSystemPrompt","AdaptedToolCallEngine","GUIAgentToolCallEngine","defaultNormalizeCoords","defaultDetailCalculator"],"mappings":";;;;;;;;;;;;;AAGC;;;;;;;;;;AA2BD,MAAMA,gBAAgB,IAAIC,cAAc,cAAcC,SAAS,KAAK;AAE7D,MAAMC,iBAAqCC;IAuDhD,MAAM,aAAa;QAEjB,IAAI,CAAC,YAAY,CACf,IAAIC,KAAK;YACP,IAAIC;YACJ,aAAa;YACb,YAAY,CAAC;YACb,UAAU,OAAOC;gBACf,IAAI,CAAC,MAAM,CAAC,GAAG,CAAC,GAAGD,sBAAsB,OAAO,CAAC,EAAEC;gBACnD,IAAI,CAAC,IAAI,CAAC,QAAQ,EAChB,OAAOC,uBAAuBD,MAAM,MAAM,EAAE;gBAE9C,IAAIA,MAAM,YAAY,EACpB,OAAOC,uBAAuBD,MAAM,MAAM,EAAEA,MAAM,YAAY;gBAGhE,IAAIA,MAAM,eAAe,EACvBA,MAAM,eAAe,GAAGE,sBACtBF,MAAM,eAAe,EACrB,IAAI,CAAC,oBAAoB;gBAG7B,IAAI,CAAC,MAAM,CAAC,IAAI,CAAC,sBAAsBG,KAAK,SAAS,CAACH,MAAM,eAAe;gBAC3E,MAAMI,SAAS,MAAM,IAAI,CAAC,QAAQ,CAAE,SAAS,CAAC;oBAC5C,SAAS;wBAACJ,MAAM,eAAe;qBAAC;gBAClC;gBACA,IAAII,OAAO,YAAY,EACrB,OAAOH,uBAAuBD,MAAM,MAAM,EAAEI,OAAO,YAAY;gBAGjE,OAAO;oBACL,SAAS;oBACT,QAAQJ,MAAM,MAAM;oBACpB,kBAAkBK,uBAAuBL,MAAM,eAAe;oBAC9D,aAAaM;gBACf;YACF;QACF;QAEF,KAAK,CAAC;IACR;IAEA,MAAM,aAAaC,EAAU,EAAEC,OAA8B,EAAiB;QAC5E,IAAI;YACF,MAAMC,gBAAgB,CAACC,KAAcC,MAAM,GAAG;gBAC5C,IAAI;oBACF,MAAMC,IAAIT,KAAK,SAAS,CAACO;oBACzB,OAAOE,EAAE,MAAM,GAAGD,MAAMC,EAAE,KAAK,CAAC,GAAGD,OAAO,sBAAiBC;gBAC7D,EAAE,OAAM;oBACN,OAAO;gBACT;YACF;YAEA,MAAMC,MAAML,QAAAA,UAAAA,KAAAA,IAAAA,QAAS,OAAO;YAC5B,MAAMM,WAAWD,OAAOE,MAAM,OAAO,CAACF,IAAI,QAAQ,IAAIA,IAAI,QAAQ,GAAG,EAAE;YACvE,MAAMG,QAAQH,OAAO,AAAqB,YAArB,OAAOA,IAAI,KAAK,GAAgBA,IAAI,KAAK,GAAG;YAEjE,MAAMI,YAAa,AAAC;gBAClB,IAAI;oBACF,IAAIC,MAAM;oBACV,KAAK,MAAMC,KAAKL,SAA0C;wBACxD,MAAMM,UAAUD,QAAAA,IAAAA,KAAAA,IAAAA,EAAG,OAAO;wBAC1B,IAAIJ,MAAM,OAAO,CAACK,UAChB;4BAAA,KAAK,MAAMC,QAAQD,QACjB,IAAIC,AAAAA,CAAAA,QAAAA,OAAAA,KAAAA,IAAAA,KAAM,IAAI,AAAD,MAAM,aAAaH;wBAClC;oBAEJ;oBACA,OAAOA;gBACT,EAAE,OAAM;oBACN;gBACF;YACF;YAEA,MAAMI,UAAU;gBACdf;gBACAS;gBACA,eAAeF,SAAS,MAAM;gBAC9BG;YACF;YAEA,IAAI,CAAC,MAAM,CAAC,IAAI,CAAC,oCAAoCR,cAAca;YAEnE,MAAMC,WAAWT,QAAQ,CAAC,EAAE;YAC5B,IAAIS,UACF,IAAI,CAAC,MAAM,CAAC,KAAK,CAAC,0CAA0Cd,cAAcc,UAAU;QAExF,EAAE,OAAOC,GAAG;YACV,IAAI,CAAC,MAAM,CAAC,KAAK,CAAC,2CAA2CA;QAC/D;IACF;IAEA,MAAM,qBAAqBC,SAAiB,EAAE;QAC5C,IAAI,CAAC,MAAM,CAAC,IAAI,CAAC,wBAAwBA;IAC3C;IAEA,MAAM,eAAelB,EAAU,EAAiB,CAEhD;IAEA,MAAM,iBACJA,EAAU,EACVmB,QAA8C,EAC9CC,IAAa,EACb;QACA,OAAOA;IACT;IAEA,MAAM,gBACJpB,EAAU,EACVmB,QAA8C,EAC9CtB,MAAe,EACG;QAClB,IAAI,CAAC,MAAM,CAAC,IAAI,CAAC,4BAA4BD,KAAK,SAAS,CAACuB;QAE5D,IAAIA,SAAS,IAAI,KAAK3B,uBAAuB,YAC3C,IAAI,CAAC,MAAM,CAAC,IAAI,CAAC;QAInB,MAAM6B,MAAM,IAAI,CAAC,gBAAgB;QAEjC,MAAMC,SAAS,MAAM,IAAI,CAAC,QAAQ,CAAE,YAAY;QAChD,IAAI,CAACA,QAAQ,YACX,IAAI,CAAC,MAAM,CAAC,KAAK,CAAC;QAIpB,MAAMC,aAAa,IAAIC,kBAAkBF,OAAO,MAAM;QACtD,MAAMG,YAAYF,WAAW,UAAU;QACvC,IAAI,CAACE,WAAW,YACd,IAAI,CAAC,MAAM,CAAC,KAAK,CAAC;QAIpB,MAAM,EAAE,OAAOC,UAAU,EAAE,QAAQC,WAAW,EAAE,GAAGJ,WAAW,aAAa,MAAM;YAC/E,OAAO;YACP,QAAQ;QACV;QAEA,MAAMV,UAAuC;YAC3C;gBACE,MAAM;gBACN,WAAW;oBACT,KAAKY;oBACL,QAAQ,IAAI,CAAC,gBAAgB,CAACC,YAAYC;gBAC5C;YACF;SACD;QAED,IAAIL,QAAAA,SAAAA,KAAAA,IAAAA,OAAQ,GAAG,EACbT,QAAQ,IAAI,CAAC;YACX,MAAM;YACN,MAAM,CAAC,wBAAwB,EAAES,OAAO,GAAG,EAAE;QAC/C;QAGF,MAAMM,cAAc,IAAI,CAAC,cAAc;QACvC,MAAMC,SAASD,YAAY,SAAS;QACpC,IAAI,CAAC,MAAM,CAAC,IAAI,CAAC,kCAAkCC,OAAO,MAAM;QAEhE,MAAMC,QAAQF,YAAY,WAAW,CAAC,qBAAqB;YACzD,aAAa;YACbf;YACA,UAAU;gBACR,MAAM;gBACN,KAAKS,QAAAA,SAAAA,KAAAA,IAAAA,OAAQ,GAAG;YAClB;QACF;QACAM,YAAY,SAAS,CAACE;QACtB,OAAOjC;IACT;IA1NA,YAAYkC,MAAyB,CAAE;QACrC,MAAM,EACJC,QAAQ,EACRvB,KAAK,EACLwB,YAAY,EACZC,mBAAmB,EACnBC,oBAAoB,EACpBC,gBAAgB,EAChBC,YAAY,EACZC,mBAAmB,GAAG,EACvB,GAAGP;QACJ,IAAIQ,oBAAoBC;QACxB,IAAI,AAAwB,YAAxB,OAAOP,cACTM,oBAAoBN;aACf,IAAIzB,MAAM,OAAO,CAACyB,eACvBM,oBAAoBN,aACjB,GAAG,CAAC,CAACQ,IAAO,AAAa,YAAb,OAAOA,IAAiBA,IAAIA,EAAE,OAAO,EACjD,IAAI,CAAC;aACH,IAAIR,gBAAgBS,uBAAuBT,eAChDM,oBAAoBI,qBAAqBV,cAAcD,SAAS,mBAAmB;QAErF9C,cAAc,KAAK,CAAC,8BAA8BqD;QAGlD,MAAMK,wBAAwB,cAAcC;YAC1C,aAAc;gBACZ,KAAK,CAACX;YACR;QACF;QAEA,KAAK,CAAC;YACJ,MAAM7C,SAAS,KAAK;YACpB,cAAckD;YACd,OAAO,EAAE;YACT,gBAAgBK;YAChB,OAAOnC;YACP,GAAI4B,gBAAgB;gBAAE,eAAeA;YAAa,CAAC;YACnD,UAAUjD,SAAS,KAAK;QAC1B,IA3CF,uBAAQ,YAAR,SACA,uBAAQ,wBAAR,SACA,uBAAQ,oBAAR,SACA,uBAAQ,oBAAR;QAyCE,IAAI,CAAC,QAAQ,GAAG4C;QAChB,IAAI,CAAC,oBAAoB,GAAGG,wBAAwBW;QAEpD,IAAI,CAAC,gBAAgB,GAAGV,oBAAoBW;QAC5C,IAAI,CAAC,gBAAgB,GAAGT;QACxB,IAAI,CAAC,MAAM,GAAG,IAAI,CAAC,MAAM,CAAC,KAAK,CAAC;IAClC;AA8KF;AAlOE,iBADWjD,UACJ,SAAQ"}
+{"version":3,"file":"GUIAgent.mjs","sources":["webpack://@ui-tars-test/agent-sdk/./src/GUIAgent.ts"],"sourcesContent":["/*\n * Copyright (c) 2025 Bytedance, Inc. and its affiliates.\n * SPDX-License-Identifier: Apache-2.0\n */\nimport {\n  LLMRequestHookPayload,\n  ChatCompletionContentPart,\n  LogLevel,\n  Tool,\n  ConsoleLogger,\n} from '@ui-tars-test/tarko-agent';\nimport { GUIAgentToolCallEngine } from './ToolCallEngine';\nimport { SYSTEM_PROMPT } from './prompts';\nimport { Base64ImageParser } from '@agent-infra/media-utils';\nimport { Operator, BaseGUIAgent } from '@ui-tars-test/shared/base';\nimport {\n  GUIAgentConfig,\n  NormalizeCoordinates,\n  ImageDetailCalculator,\n} from '@ui-tars-test/shared/types';\nimport {\n  assembleSystemPrompt,\n  isSystemPromptTemplate,\n  normalizeActionCoords,\n  sleep,\n} from '@ui-tars-test/shared/utils';\nimport { GUI_ADAPTED_TOOL_NAME } from './constants';\nimport { convertToAgentUIAction, createGUIErrorResponse } from './utils';\nimport { defaultNormalizeCoords, defaultDetailCalculator } from './defaultImpls';\n\nconst defaultLogger = new ConsoleLogger('[GUIAgent]', LogLevel.DEBUG);\n\nexport class GUIAgent<T extends Operator> extends BaseGUIAgent {\n  static label = 'GUI Agent';\n\n  private operator: Operator | undefined;\n  private normalizeCoordinates: NormalizeCoordinates;\n  private detailCalculator: ImageDetailCalculator;\n  private loopIntervalInMs: number;\n\n  constructor(config: GUIAgentConfig<T>) {\n    const {\n      operator,\n      model,\n      systemPrompt,\n      customeActionParser,\n      normalizeCoordinates,\n      detailCalculator,\n      maxLoopCount,\n      loopIntervalInMs = 500,\n    } = config;\n    let finalSystemPrompt = SYSTEM_PROMPT;\n    if (typeof systemPrompt === 'string') {\n      finalSystemPrompt = systemPrompt;\n    } else if (Array.isArray(systemPrompt)) {\n      finalSystemPrompt = systemPrompt\n        .map((p) => (typeof p === 'string' ? p : p.content))\n        .join('\\n\\n');\n    } else if (systemPrompt && isSystemPromptTemplate(systemPrompt)) {\n      finalSystemPrompt = assembleSystemPrompt(systemPrompt, operator.getSupportedActions());\n    }\n    defaultLogger.debug('final instructions for sp:', finalSystemPrompt);\n\n    // Create a adapted ToolCallEngine constructor that captures the customeActionParser\n    const AdaptedToolCallEngine = class extends GUIAgentToolCallEngine {\n      constructor() {\n        super(customeActionParser);\n      }\n    };\n\n    super({\n      name: GUIAgent.label,\n      instructions: finalSystemPrompt,\n      tools: [],\n      toolCallEngine: AdaptedToolCallEngine,\n      model: model,\n      ...(maxLoopCount && { maxIterations: maxLoopCount }),\n      logLevel: LogLevel.DEBUG,\n    });\n    this.operator = operator;\n    this.normalizeCoordinates = normalizeCoordinates ?? defaultNormalizeCoords;\n    // Default detail calculator implementation\n    this.detailCalculator = detailCalculator ?? defaultDetailCalculator;\n    this.loopIntervalInMs = loopIntervalInMs;\n    this.logger = this.logger.spawn('[GUIAgent]');\n  }\n\n  async initialize() {\n    // Register the GUI tool\n    this.registerTool(\n      new Tool({\n        id: GUI_ADAPTED_TOOL_NAME,\n        description: 'operator tool',\n        parameters: {}, // no need to pass parameters\n        function: async (input) => {\n          this.logger.log(`${GUI_ADAPTED_TOOL_NAME} input:`, input);\n          if (!this.operator) {\n            return createGUIErrorResponse(input.action, 'Operator not initialized');\n          }\n          if (input.errorMessage) {\n            return createGUIErrorResponse(input.action, input.errorMessage);\n          }\n          // normalize coordinates\n          if (input.operator_action) {\n            input.operator_action = normalizeActionCoords(\n              input.operator_action,\n              this.normalizeCoordinates,\n            );\n          }\n          this.logger.info('action to execute:', JSON.stringify(input.operator_action));\n          const result = await this.operator!.doExecute({\n            actions: [input.operator_action],\n          });\n          if (result.errorMessage) {\n            return createGUIErrorResponse(input.action, result.errorMessage);\n          }\n          // return { action: input.action, status: 'success', result };\n          return {\n            success: true,\n            action: input.action,\n            normalizedAction: convertToAgentUIAction(input.operator_action),\n            observation: undefined, // Reserved for future implementation\n          };\n        },\n      }),\n    );\n    super.initialize();\n  }\n\n  async onLLMRequest(id: string, payload: LLMRequestHookPayload): Promise<void> {\n    try {\n      const safeStringify = (obj: unknown, max = 800): string => {\n        try {\n          const s = JSON.stringify(obj);\n          return s.length > max ? s.slice(0, max) + '…(truncated)' : s;\n        } catch {\n          return '[unserializable]';\n        }\n      };\n\n      const req = payload?.request;\n      const messages = req && Array.isArray(req.messages) ? req.messages : [];\n      const model = req && typeof req.model === 'string' ? req.model : 'unknown';\n\n      const hasImages = (() => {\n        try {\n          let cnt = 0;\n          for (const m of messages as Array<{ content?: unknown }>) {\n            const content = m?.content;\n            if (Array.isArray(content)) {\n              for (const part of content as Array<{ type?: string }>) {\n                if (part?.type === 'image_url') cnt++;\n              }\n            }\n          }\n          return cnt;\n        } catch {\n          return undefined as unknown as number | undefined;\n        }\n      })();\n\n      const summary = {\n        id,\n        model,\n        messagesCount: messages.length,\n        hasImages,\n      };\n\n      this.logger.info('[GUIAgent] onLLMRequest summary:', safeStringify(summary));\n\n      const firstMsg = messages[0];\n      if (firstMsg) {\n        this.logger.debug('[GUIAgent] onLLMRequest first message:', safeStringify(firstMsg, 1200));\n      }\n    } catch (e) {\n      this.logger.error('[GUIAgent] onLLMRequest logging failed:', e);\n    }\n  }\n\n  async onEachAgentLoopStart(sessionId: string) {\n    this.logger.info('onEachAgentLoopStart', sessionId);\n  }\n\n  async onAgentLoopEnd(id: string): Promise<void> {\n    // await this.browserOperator.cleanup();\n  }\n\n  async onBeforeToolCall(\n    id: string,\n    toolCall: { toolCallId: string; name: string },\n    args: unknown,\n  ) {\n    return args;\n  }\n\n  async onAfterToolCall(\n    id: string,\n    toolCall: { toolCallId: string; name: string },\n    result: unknown,\n  ): Promise<unknown> {\n    this.logger.info('onAfterToolCall toolCall', JSON.stringify(toolCall));\n\n    if (toolCall.name !== GUI_ADAPTED_TOOL_NAME) {\n      this.logger.info('onAfterToolCall: skipping screenshot');\n      return;\n    }\n\n    await sleep(this.loopIntervalInMs);\n\n    const output = await this.operator!.doScreenshot();\n    if (!output) {\n      this.logger.error('Failed to get screenshot');\n      return;\n    }\n\n    const base64Tool = new Base64ImageParser(output.base64);\n    const base64Uri = base64Tool.getDataUri();\n    if (!base64Uri) {\n      this.logger.error('Failed to get base64 image uri');\n      return;\n    }\n\n    const { width: imageWidth, height: imageHeight } = base64Tool.getDimensions() || {\n      width: -1,\n      height: -1,\n    };\n\n    const content: ChatCompletionContentPart[] = [\n      {\n        type: 'image_url',\n        image_url: {\n          url: base64Uri,\n          detail: this.detailCalculator(imageWidth, imageHeight),\n        },\n      },\n    ];\n\n    if (output?.url) {\n      content.push({\n        type: 'text',\n        text: `The current page's url: ${output.url}`,\n      });\n    }\n\n    const eventStream = this.getEventStream();\n    const events = eventStream.getEvents();\n    this.logger.info('onAfterToolCall events length:', events.length);\n\n    const event = eventStream.createEvent('environment_input', {\n      description: 'Browser Screenshot',\n      content,\n      metadata: {\n        type: 'screenshot',\n        url: output?.url,\n      },\n    });\n    eventStream.sendEvent(event);\n    return result;\n  }\n}\n"],"names":["defaultLogger","ConsoleLogger","LogLevel","GUIAgent","BaseGUIAgent","Tool","GUI_ADAPTED_TOOL_NAME","input","createGUIErrorResponse","normalizeActionCoords","JSON","result","convertToAgentUIAction","undefined","id","payload","safeStringify","obj","max","s","req","messages","Array","model","hasImages","cnt","m","content","part","summary","firstMsg","e","sessionId","toolCall","args","sleep","output","base64Tool","Base64ImageParser","base64Uri","imageWidth","imageHeight","eventStream","events","event","config","operator","systemPrompt","customeActionParser","normalizeCoordinates","detailCalculator","maxLoopCount","loopIntervalInMs","finalSystemPrompt","SYSTEM_PROMPT","p","isSystemPromptTemplate","assembleSystemPrompt","AdaptedToolCallEngine","GUIAgentToolCallEngine","defaultNormalizeCoords","defaultDetailCalculator"],"mappings":";;;;;;;;;;;;;AAGC;;;;;;;;;;AA2BD,MAAMA,gBAAgB,IAAIC,cAAc,cAAcC,SAAS,KAAK;AAE7D,MAAMC,iBAAqCC;IAuDhD,MAAM,aAAa;QAEjB,IAAI,CAAC,YAAY,CACf,IAAIC,KAAK;YACP,IAAIC;YACJ,aAAa;YACb,YAAY,CAAC;YACb,UAAU,OAAOC;gBACf,IAAI,CAAC,MAAM,CAAC,GAAG,CAAC,GAAGD,sBAAsB,OAAO,CAAC,EAAEC;gBACnD,IAAI,CAAC,IAAI,CAAC,QAAQ,EAChB,OAAOC,uBAAuBD,MAAM,MAAM,EAAE;gBAE9C,IAAIA,MAAM,YAAY,EACpB,OAAOC,uBAAuBD,MAAM,MAAM,EAAEA,MAAM,YAAY;gBAGhE,IAAIA,MAAM,eAAe,EACvBA,MAAM,eAAe,GAAGE,sBACtBF,MAAM,eAAe,EACrB,IAAI,CAAC,oBAAoB;gBAG7B,IAAI,CAAC,MAAM,CAAC,IAAI,CAAC,sBAAsBG,KAAK,SAAS,CAACH,MAAM,eAAe;gBAC3E,MAAMI,SAAS,MAAM,IAAI,CAAC,QAAQ,CAAE,SAAS,CAAC;oBAC5C,SAAS;wBAACJ,MAAM,eAAe;qBAAC;gBAClC;gBACA,IAAII,OAAO,YAAY,EACrB,OAAOH,uBAAuBD,MAAM,MAAM,EAAEI,OAAO,YAAY;gBAGjE,OAAO;oBACL,SAAS;oBACT,QAAQJ,MAAM,MAAM;oBACpB,kBAAkBK,uBAAuBL,MAAM,eAAe;oBAC9D,aAAaM;gBACf;YACF;QACF;QAEF,KAAK,CAAC;IACR;IAEA,MAAM,aAAaC,EAAU,EAAEC,OAA8B,EAAiB;QAC5E,IAAI;YACF,MAAMC,gBAAgB,CAACC,KAAcC,MAAM,GAAG;gBAC5C,IAAI;oBACF,MAAMC,IAAIT,KAAK,SAAS,CAACO;oBACzB,OAAOE,EAAE,MAAM,GAAGD,MAAMC,EAAE,KAAK,CAAC,GAAGD,OAAO,sBAAiBC;gBAC7D,EAAE,OAAM;oBACN,OAAO;gBACT;YACF;YAEA,MAAMC,MAAML,QAAAA,UAAAA,KAAAA,IAAAA,QAAS,OAAO;YAC5B,MAAMM,WAAWD,OAAOE,MAAM,OAAO,CAACF,IAAI,QAAQ,IAAIA,IAAI,QAAQ,GAAG,EAAE;YACvE,MAAMG,QAAQH,OAAO,AAAqB,YAArB,OAAOA,IAAI,KAAK,GAAgBA,IAAI,KAAK,GAAG;YAEjE,MAAMI,YAAa,AAAC;gBAClB,IAAI;oBACF,IAAIC,MAAM;oBACV,KAAK,MAAMC,KAAKL,SAA0C;wBACxD,MAAMM,UAAUD,QAAAA,IAAAA,KAAAA,IAAAA,EAAG,OAAO;wBAC1B,IAAIJ,MAAM,OAAO,CAACK,UAChB;4BAAA,KAAK,MAAMC,QAAQD,QACjB,IAAIC,AAAAA,CAAAA,QAAAA,OAAAA,KAAAA,IAAAA,KAAM,IAAI,AAAD,MAAM,aAAaH;wBAClC;oBAEJ;oBACA,OAAOA;gBACT,EAAE,OAAM;oBACN;gBACF;YACF;YAEA,MAAMI,UAAU;gBACdf;gBACAS;gBACA,eAAeF,SAAS,MAAM;gBAC9BG;YACF;YAEA,IAAI,CAAC,MAAM,CAAC,IAAI,CAAC,oCAAoCR,cAAca;YAEnE,MAAMC,WAAWT,QAAQ,CAAC,EAAE;YAC5B,IAAIS,UACF,IAAI,CAAC,MAAM,CAAC,KAAK,CAAC,0CAA0Cd,cAAcc,UAAU;QAExF,EAAE,OAAOC,GAAG;YACV,IAAI,CAAC,MAAM,CAAC,KAAK,CAAC,2CAA2CA;QAC/D;IACF;IAEA,MAAM,qBAAqBC,SAAiB,EAAE;QAC5C,IAAI,CAAC,MAAM,CAAC,IAAI,CAAC,wBAAwBA;IAC3C;IAEA,MAAM,eAAelB,EAAU,EAAiB,CAEhD;IAEA,MAAM,iBACJA,EAAU,EACVmB,QAA8C,EAC9CC,IAAa,EACb;QACA,OAAOA;IACT;IAEA,MAAM,gBACJpB,EAAU,EACVmB,QAA8C,EAC9CtB,MAAe,EACG;QAClB,IAAI,CAAC,MAAM,CAAC,IAAI,CAAC,4BAA4BD,KAAK,SAAS,CAACuB;QAE5D,IAAIA,SAAS,IAAI,KAAK3B,uBAAuB,YAC3C,IAAI,CAAC,MAAM,CAAC,IAAI,CAAC;QAInB,MAAM6B,MAAM,IAAI,CAAC,gBAAgB;QAEjC,MAAMC,SAAS,MAAM,IAAI,CAAC,QAAQ,CAAE,YAAY;QAChD,IAAI,CAACA,QAAQ,YACX,IAAI,CAAC,MAAM,CAAC,KAAK,CAAC;QAIpB,MAAMC,aAAa,IAAIC,kBAAkBF,OAAO,MAAM;QACtD,MAAMG,YAAYF,WAAW,UAAU;QACvC,IAAI,CAACE,WAAW,YACd,IAAI,CAAC,MAAM,CAAC,KAAK,CAAC;QAIpB,MAAM,EAAE,OAAOC,UAAU,EAAE,QAAQC,WAAW,EAAE,GAAGJ,WAAW,aAAa,MAAM;YAC/E,OAAO;YACP,QAAQ;QACV;QAEA,MAAMV,UAAuC;YAC3C;gBACE,MAAM;gBACN,WAAW;oBACT,KAAKY;oBACL,QAAQ,IAAI,CAAC,gBAAgB,CAACC,YAAYC;gBAC5C;YACF;SACD;QAED,IAAIL,QAAAA,SAAAA,KAAAA,IAAAA,OAAQ,GAAG,EACbT,QAAQ,IAAI,CAAC;YACX,MAAM;YACN,MAAM,CAAC,wBAAwB,EAAES,OAAO,GAAG,EAAE;QAC/C;QAGF,MAAMM,cAAc,IAAI,CAAC,cAAc;QACvC,MAAMC,SAASD,YAAY,SAAS;QACpC,IAAI,CAAC,MAAM,CAAC,IAAI,CAAC,kCAAkCC,OAAO,MAAM;QAEhE,MAAMC,QAAQF,YAAY,WAAW,CAAC,qBAAqB;YACzD,aAAa;YACbf;YACA,UAAU;gBACR,MAAM;gBACN,KAAKS,QAAAA,SAAAA,KAAAA,IAAAA,OAAQ,GAAG;YAClB;QACF;QACAM,YAAY,SAAS,CAACE;QACtB,OAAOjC;IACT;IA1NA,YAAYkC,MAAyB,CAAE;QACrC,MAAM,EACJC,QAAQ,EACRvB,KAAK,EACLwB,YAAY,EACZC,mBAAmB,EACnBC,oBAAoB,EACpBC,gBAAgB,EAChBC,YAAY,EACZC,mBAAmB,GAAG,EACvB,GAAGP;QACJ,IAAIQ,oBAAoBC;QACxB,IAAI,AAAwB,YAAxB,OAAOP,cACTM,oBAAoBN;aACf,IAAIzB,MAAM,OAAO,CAACyB,eACvBM,oBAAoBN,aACjB,GAAG,CAAC,CAACQ,IAAO,AAAa,YAAb,OAAOA,IAAiBA,IAAIA,EAAE,OAAO,EACjD,IAAI,CAAC;aACH,IAAIR,gBAAgBS,uBAAuBT,eAChDM,oBAAoBI,qBAAqBV,cAAcD,SAAS,mBAAmB;QAErF9C,cAAc,KAAK,CAAC,8BAA8BqD;QAGlD,MAAMK,wBAAwB,cAAcC;YAC1C,aAAc;gBACZ,KAAK,CAACX;YACR;QACF;QAEA,KAAK,CAAC;YACJ,MAAM7C,SAAS,KAAK;YACpB,cAAckD;YACd,OAAO,EAAE;YACT,gBAAgBK;YAChB,OAAOnC;YACP,GAAI4B,gBAAgB;gBAAE,eAAeA;YAAa,CAAC;YACnD,UAAUjD,SAAS,KAAK;QAC1B,IA3CF,uBAAQ,YAAR,SACA,uBAAQ,wBAAR,SACA,uBAAQ,oBAAR,SACA,uBAAQ,oBAAR;QAyCE,IAAI,CAAC,QAAQ,GAAG4C;QAChB,IAAI,CAAC,oBAAoB,GAAGG,wBAAwBW;QAEpD,IAAI,CAAC,gBAAgB,GAAGV,oBAAoBW;QAC5C,IAAI,CAAC,gBAAgB,GAAGT;QACxB,IAAI,CAAC,MAAM,GAAG,IAAI,CAAC,MAAM,CAAC,KAAK,CAAC;IAClC;AA8KF;AAlOE,iBADWjD,UACJ,SAAQ"}

package/dist/ToolCallEngine.d.ts

@@ -1,4 +1,4 @@
-import { ToolCallEngine, Tool, ToolCallEnginePrepareRequestContext, ChatCompletionCreateParams, ChatCompletionAssistantMessageParam, ChatCompletionChunk, MultimodalToolCallResult, AgentEventStream, ChatCompletionMessageParam, ParsedModelResponse, StreamProcessingState, StreamChunkResult } from '@tarko/agent-interface';
+import { ToolCallEngine, Tool, ToolCallEnginePrepareRequestContext, ChatCompletionCreateParams, ChatCompletionAssistantMessageParam, ChatCompletionChunk, MultimodalToolCallResult, AgentEventStream, ChatCompletionMessageParam, ParsedModelResponse, StreamProcessingState, StreamChunkResult } from '@ui-tars-test/tarko-agent-interface';
 import { CustomActionParser } from '@ui-tars-test/shared/types';
 /**
  * GUIAgentToolCallEngine - Minimal prompt engineering tool call engine

package/dist/ToolCallEngine.d.ts.map

@@ -1 +1 @@
-{"version":3,"file":"ToolCallEngine.d.ts","sourceRoot":"","sources":["../src/ToolCallEngine.ts"],"names":[],"mappings":"AAAA,OAAO,EACL,cAAc,EACd,IAAI,EACJ,mCAAmC,EACnC,0BAA0B,EAC1B,mCAAmC,EACnC,mBAAmB,EACnB,wBAAwB,EACxB,gBAAgB,EAChB,0BAA0B,EAE1B,mBAAmB,EACnB,qBAAqB,EACrB,iBAAiB,EAClB,MAAM,wBAAwB,CAAC;AAKhC,OAAO,EAAE,kBAAkB,EAAE,MAAM,4BAA4B,CAAC;AAKhE;;;;;;;;;GASG;AACH,qBAAa,sBAAuB,SAAQ,cAAc;IACxD,OAAO,CAAC,kBAAkB,CAAC,CAAqB;gBAEpC,kBAAkB,CAAC,EAAE,kBAAkB;IAKnD;;OAEG;IACH,aAAa,CAAC,YAAY,EAAE,MAAM,EAAE,KAAK,EAAE,IAAI,EAAE,GAAG,MAAM;IAI1D;;;;OAIG;IACH,cAAc,CAAC,OAAO,EAAE,mCAAmC,GAAG,0BAA0B;IAaxF;;;;OAIG;IACH,yBAAyB,IAAI,qBAAqB;IASlD;;;;OAIG;IACH,qBAAqB,CACnB,KAAK,EAAE,mBAAmB,EAC1B,KAAK,EAAE,qBAAqB,GAC3B,iBAAiB;IAsBpB;;OAEG;IACH,OAAO,CAAC,kBAAkB;IAI1B;;OAEG;IACH,wBAAwB,CAAC,KAAK,EAAE,qBAAqB,GAAG,mBAAmB;IAiG3E;;;;;OAKG;IACH,+BAA+B,CAC7B,yBAAyB,EAAE,gBAAgB,CAAC,qBAAqB,GAChE,mCAAmC;IAOtC;;;;;OAKG;IACH,qCAAqC,CACnC,eAAe,EAAE,wBAAwB,EAAE,GAC1C,0BAA0B,EAAE;CAchC"}
+{"version":3,"file":"ToolCallEngine.d.ts","sourceRoot":"","sources":["../src/ToolCallEngine.ts"],"names":[],"mappings":"AAAA,OAAO,EACL,cAAc,EACd,IAAI,EACJ,mCAAmC,EACnC,0BAA0B,EAC1B,mCAAmC,EACnC,mBAAmB,EACnB,wBAAwB,EACxB,gBAAgB,EAChB,0BAA0B,EAE1B,mBAAmB,EACnB,qBAAqB,EACrB,iBAAiB,EAClB,MAAM,qCAAqC,CAAC;AAK7C,OAAO,EAAE,kBAAkB,EAAE,MAAM,4BAA4B,CAAC;AAKhE;;;;;;;;;GASG;AACH,qBAAa,sBAAuB,SAAQ,cAAc;IACxD,OAAO,CAAC,kBAAkB,CAAC,CAAqB;gBAEpC,kBAAkB,CAAC,EAAE,kBAAkB;IAKnD;;OAEG;IACH,aAAa,CAAC,YAAY,EAAE,MAAM,EAAE,KAAK,EAAE,IAAI,EAAE,GAAG,MAAM;IAI1D;;;;OAIG;IACH,cAAc,CAAC,OAAO,EAAE,mCAAmC,GAAG,0BAA0B;IAaxF;;;;OAIG;IACH,yBAAyB,IAAI,qBAAqB;IASlD;;;;OAIG;IACH,qBAAqB,CACnB,KAAK,EAAE,mBAAmB,EAC1B,KAAK,EAAE,qBAAqB,GAC3B,iBAAiB;IAsBpB;;OAEG;IACH,OAAO,CAAC,kBAAkB;IAI1B;;OAEG;IACH,wBAAwB,CAAC,KAAK,EAAE,qBAAqB,GAAG,mBAAmB;IAiG3E;;;;;OAKG;IACH,+BAA+B,CAC7B,yBAAyB,EAAE,gBAAgB,CAAC,qBAAqB,GAChE,mCAAmC;IAOtC;;;;;OAKG;IACH,qCAAqC,CACnC,eAAe,EAAE,wBAAwB,EAAE,GAC1C,0BAA0B,EAAE;CAchC"}

package/dist/ToolCallEngine.js

@@ -30,7 +30,7 @@ __webpack_require__.r(__webpack_exports__);
 __webpack_require__.d(__webpack_exports__, {
     GUIAgentToolCallEngine: ()=>GUIAgentToolCallEngine
 });
-const agent_interface_namespaceObject = require("@tarko/agent-interface");
+const tarko_agent_interface_namespaceObject = require("@ui-tars-test/tarko-agent-interface");
 const action_parser_namespaceObject = require("@ui-tars-test/action-parser");
 const external_constants_js_namespaceObject = require("./constants.js");
 const logger_namespaceObject = require("@agent-infra/logger");
@@ -47,7 +47,7 @@ function _define_property(obj, key, value) {
 }
 const defaultParser = new action_parser_namespaceObject.DefaultActionParser();
 const defaultLogger = new logger_namespaceObject.ConsoleLogger('[GUIAgent:ToolCallEngine]', logger_namespaceObject.LogLevel.DEBUG);
-class GUIAgentToolCallEngine extends agent_interface_namespaceObject.ToolCallEngine {
+class GUIAgentToolCallEngine extends tarko_agent_interface_namespaceObject.ToolCallEngine {
     preparePrompt(instructions, tools) {
         return instructions;
     }

package/dist/ToolCallEngine.js.map

@@ -1 +1 @@
-{"version":3,"file":"ToolCallEngine.js","sources":["webpack://@ui-tars-test/agent-sdk/webpack/runtime/define_property_getters","webpack://@ui-tars-test/agent-sdk/webpack/runtime/has_own_property","webpack://@ui-tars-test/agent-sdk/webpack/runtime/make_namespace_object","webpack://@ui-tars-test/agent-sdk/./src/ToolCallEngine.ts"],"sourcesContent":["__webpack_require__.d = (exports, definition) => {\n\tfor(var key in definition) {\n        if(__webpack_require__.o(definition, key) && !__webpack_require__.o(exports, key)) {\n            Object.defineProperty(exports, key, { enumerable: true, get: definition[key] });\n        }\n    }\n};","__webpack_require__.o = (obj, prop) => (Object.prototype.hasOwnProperty.call(obj, prop))","// define __esModule on exports\n__webpack_require__.r = (exports) => {\n\tif(typeof Symbol !== 'undefined' && Symbol.toStringTag) {\n\t\tObject.defineProperty(exports, Symbol.toStringTag, { value: 'Module' });\n\t}\n\tObject.defineProperty(exports, '__esModule', { value: true });\n};","import {\n  ToolCallEngine,\n  Tool,\n  ToolCallEnginePrepareRequestContext,\n  ChatCompletionCreateParams,\n  ChatCompletionAssistantMessageParam,\n  ChatCompletionChunk,\n  MultimodalToolCallResult,\n  AgentEventStream,\n  ChatCompletionMessageParam,\n  ChatCompletionMessageToolCall,\n  ParsedModelResponse,\n  StreamProcessingState,\n  StreamChunkResult,\n} from '@tarko/agent-interface';\nimport { DefaultActionParser } from '@ui-tars-test/action-parser';\nimport { GUI_ADAPTED_TOOL_NAME } from './constants';\nimport { ConsoleLogger, LogLevel } from '@agent-infra/logger';\nimport { serializeAction } from '@ui-tars-test/shared/utils';\nimport { CustomActionParser } from '@ui-tars-test/shared/types';\n\nconst defaultParser = new DefaultActionParser();\nconst defaultLogger = new ConsoleLogger('[GUIAgent:ToolCallEngine]', LogLevel.DEBUG);\n\n/**\n * GUIAgentToolCallEngine - Minimal prompt engineering tool call engine\n *\n * This is the simplest possible implementation of a tool call engine that:\n * 1. Uses prompt engineering to instruct the LLM to output tool calls in a specific format\n * 2. Parses tool calls from LLM response text using simple regex matching\n * 3. Does not support streaming (focuses on core functionality only)\n *\n * Format used: <tool_call>{\"name\": \"tool_name\", \"arguments\": {...}}</tool_call>\n */\nexport class GUIAgentToolCallEngine extends ToolCallEngine {\n  private customActionParser?: CustomActionParser;\n\n  constructor(customActionParser?: CustomActionParser) {\n    super();\n    this.customActionParser = customActionParser;\n  }\n\n  /**\n   * Prepare system prompt with tool information and instructions\n   */\n  preparePrompt(instructions: string, tools: Tool[]): string {\n    return instructions;\n  }\n\n  /**\n   * Prepare request parameters for the LLM\n   *\n   * FIXME: move to base tool call engine.\n   */\n  prepareRequest(context: ToolCallEnginePrepareRequestContext): ChatCompletionCreateParams {\n    defaultLogger.log(\n      \"【New Sys Prompt'】 System Prompt:\",\n      JSON.stringify(context.messages.find((m) => m.role === 'system')?.content || ''),\n    );\n    return {\n      model: context.model,\n      messages: context.messages,\n      temperature: context.temperature || 0.7,\n      stream: true,\n    };\n  }\n\n  /**\n   * Initialize processing state (minimal implementation)\n   *\n   * FIXME: move to base tool call engine.\n   */\n  initStreamProcessingState(): StreamProcessingState {\n    return {\n      contentBuffer: '',\n      toolCalls: [],\n      reasoningBuffer: '',\n      finishReason: null,\n    };\n  }\n\n  /**\n   * Process streaming chunks - simply accumulate content\n   *\n   * FIXME: make it optional\n   */\n  processStreamingChunk(\n    chunk: ChatCompletionChunk,\n    state: StreamProcessingState,\n  ): StreamChunkResult {\n    const delta = chunk.choices[0]?.delta;\n\n    // Accumulate content\n    if (delta?.content) {\n      state.contentBuffer += delta.content;\n    }\n\n    // Record finish reason\n    if (chunk.choices[0]?.finish_reason) {\n      state.finishReason = chunk.choices[0].finish_reason;\n    }\n\n    // Return incremental content without tool call detection during streaming\n    return {\n      content: delta?.content || '',\n      reasoningContent: '',\n      hasToolCallUpdate: false,\n      toolCalls: [],\n    };\n  }\n\n  /**\n   * Generate a tool call ID\n   */\n  private generateToolCallId(): string {\n    return `call_${Date.now()}_${Math.random().toString(36).substring(2, 11)}`;\n  }\n\n  /**\n   * Extract tool calls from complete response text\n   */\n  finalizeStreamProcessing(state: StreamProcessingState): ParsedModelResponse {\n    const fullContent = state.contentBuffer;\n    defaultLogger.log(\"【New Sys Prompt'】 Model Response:\", fullContent);\n    defaultLogger.log('[finalizeStreamProcessing] fullContent', fullContent);\n\n    // Add explicit log to confirm XML parsing intent\n    console.log('[CLI DEBUG] [ToolCallEngine] Full model response received:', fullContent);\n\n    // Try custom action parser first if available\n    let parsedGUIResponse = null;\n    if (this.customActionParser) {\n      parsedGUIResponse = this.customActionParser(fullContent);\n      defaultLogger.log('[finalizeStreamProcessing] Using custom action parser');\n      console.log('[CLI DEBUG] [ToolCallEngine] Using custom action parser');\n    }\n\n    // Fall back to default parser if custom parser is not available or returns null\n    if (!parsedGUIResponse) {\n      console.log('[CLI DEBUG] [ToolCallEngine] Using default action parser (XML parser)');\n      parsedGUIResponse = defaultParser.parsePrediction(fullContent);\n      defaultLogger.log('[finalizeStreamProcessing] Using default action parser');\n    }\n\n    if (parsedGUIResponse) {\n      console.log(\n        '[CLI DEBUG] [ToolCallEngine] Parsed response:',\n        JSON.stringify(parsedGUIResponse, null, 2),\n      );\n    } else {\n      console.log('[CLI DEBUG] [ToolCallEngine] Parsing failed or returned null');\n    }\n\n    if (!parsedGUIResponse || parsedGUIResponse.errorMessage) {\n      return {\n        content: '',\n        rawContent: fullContent,\n        toolCalls: [\n          {\n            id: this.generateToolCallId(),\n            type: 'function',\n            function: {\n              name: GUI_ADAPTED_TOOL_NAME,\n              arguments: JSON.stringify({\n                action: '',\n                step: '',\n                thought: '',\n                operator_action: null,\n                errorMessage:\n                  parsedGUIResponse?.errorMessage ?? 'Failed to parse GUI Action from output',\n              }),\n            },\n          },\n        ],\n        finishReason: 'tool_calls',\n      };\n    }\n\n    const toolCalls: ChatCompletionMessageToolCall[] = [];\n\n    let finished = false;\n    let finishMessage: string | null = null;\n    for (const action of parsedGUIResponse.actions) {\n      if (action.type === 'finished') {\n        finished = true;\n        finishMessage = action.inputs?.content ?? null;\n        continue;\n      }\n      toolCalls.push({\n        id: this.generateToolCallId(),\n        type: 'function',\n        function: {\n          name: GUI_ADAPTED_TOOL_NAME,\n          arguments: JSON.stringify({\n            action: serializeAction(action),\n            step: '',\n            thought: parsedGUIResponse.reasoningContent ?? '',\n            operator_action: action,\n          }),\n        },\n      });\n    }\n\n    const content = finishMessage ?? '';\n    const reasoningContent = parsedGUIResponse.reasoningContent ?? '';\n    const contentForWebUI = content.replace(/\\\\n|\\n/g, '<br>');\n    const reasoningContentForWebUI = reasoningContent.replace(/\\\\n|\\n/g, '<br>');\n\n    // No tool calls found - return regular response\n    return {\n      content: contentForWebUI,\n      rawContent: fullContent,\n      reasoningContent: reasoningContentForWebUI,\n      toolCalls,\n      finishReason: toolCalls.length > 0 && !finished ? 'tool_calls' : 'stop',\n    };\n  }\n\n  /**\n   * Build assistant message for conversation history\n   * For PE engines, we preserve the raw content including tool call markup\n   *\n   * FIXME: move to base tool call engine.\n   */\n  buildHistoricalAssistantMessage(\n    currentLoopAssistantEvent: AgentEventStream.AssistantMessageEvent,\n  ): ChatCompletionAssistantMessageParam {\n    return {\n      role: 'assistant',\n      content: currentLoopAssistantEvent.rawContent || currentLoopAssistantEvent.content,\n    };\n  }\n\n  /**\n   * Build tool result messages as user messages\n   * PE engines format tool results as user input for next iteration\n   *\n   * FIXME: move to base tool call engine.\n   */\n  buildHistoricalToolCallResultMessages(\n    toolCallResults: MultimodalToolCallResult[],\n  ): ChatCompletionMessageParam[] {\n    return toolCallResults.map((result) => {\n      // Extract text content from multimodal result\n      const textContent = result.content\n        .filter((part) => part.type === 'text')\n        .map((part) => (part as { text: string }).text)\n        .join('');\n\n      return {\n        role: 'user',\n        content: `Tool \"${result.toolName}\" result:\\n${textContent}`,\n      };\n    });\n  }\n}\n"],"names":["__webpack_require__","definition","key","Object","obj","prop","Symbol","defaultParser","DefaultActionParser","defaultLogger","ConsoleLogger","LogLevel","GUIAgentToolCallEngine","ToolCallEngine","instructions","tools","context","_context_messages_find","JSON","m","chunk","state","_chunk_choices_","_chunk_choices_1","delta","Date","Math","fullContent","console","parsedGUIResponse","GUI_ADAPTED_TOOL_NAME","toolCalls","finished","finishMessage","action","_action_inputs","serializeAction","content","reasoningContent","contentForWebUI","reasoningContentForWebUI","currentLoopAssistantEvent","toolCallResults","result","textContent","part","customActionParser"],"mappings":";;;;;;;IAAAA,oBAAoB,CAAC,GAAG,CAAC,UAASC;QACjC,IAAI,IAAIC,OAAOD,WACR,IAAGD,oBAAoB,CAAC,CAACC,YAAYC,QAAQ,CAACF,oBAAoB,CAAC,CAAC,UAASE,MACzEC,OAAO,cAAc,CAAC,UAASD,KAAK;YAAE,YAAY;YAAM,KAAKD,UAAU,CAACC,IAAI;QAAC;IAGzF;;;ICNAF,oBAAoB,CAAC,GAAG,CAACI,KAAKC,OAAUF,OAAO,SAAS,CAAC,cAAc,CAAC,IAAI,CAACC,KAAKC;;;ICClFL,oBAAoB,CAAC,GAAG,CAAC;QACxB,IAAG,AAAkB,eAAlB,OAAOM,UAA0BA,OAAO,WAAW,EACrDH,OAAO,cAAc,CAAC,UAASG,OAAO,WAAW,EAAE;YAAE,OAAO;QAAS;QAEtEH,OAAO,cAAc,CAAC,UAAS,cAAc;YAAE,OAAO;QAAK;IAC5D;;;;;;;;;;;;;;;;;;;;;;ACeA,MAAMI,gBAAgB,IAAIC,8BAAAA,mBAAmBA;AAC7C,MAAMC,gBAAgB,IAAIC,uBAAAA,aAAaA,CAAC,6BAA6BC,uBAAAA,QAAAA,CAAAA,KAAc;AAY5E,MAAMC,+BAA+BC,gCAAAA,cAAcA;IAWxD,cAAcC,YAAoB,EAAEC,KAAa,EAAU;QACzD,OAAOD;IACT;IAOA,eAAeE,OAA4C,EAA8B;YAGtEC;QAFjBR,cAAc,GAAG,CACf,8CACAS,KAAK,SAAS,CAACD,AAAAA,SAAAA,CAAAA,yBAAAA,QAAQ,QAAQ,CAAC,IAAI,CAAC,CAACE,IAAMA,AAAW,aAAXA,EAAE,IAAI,CAAa,IAAhDF,KAAAA,IAAAA,uBAAmD,OAAO,AAAD,KAAK;QAE/E,OAAO;YACL,OAAOD,QAAQ,KAAK;YACpB,UAAUA,QAAQ,QAAQ;YAC1B,aAAaA,QAAQ,WAAW,IAAI;YACpC,QAAQ;QACV;IACF;IAOA,4BAAmD;QACjD,OAAO;YACL,eAAe;YACf,WAAW,EAAE;YACb,iBAAiB;YACjB,cAAc;QAChB;IACF;IAOA,sBACEI,KAA0B,EAC1BC,KAA4B,EACT;YACLC,iBAQVC;QARJ,MAAMC,QAAQ,QAAAF,CAAAA,kBAAAA,MAAM,OAAO,CAAC,EAAE,AAAD,IAAfA,KAAAA,IAAAA,gBAAkB,KAAK;QAGrC,IAAIE,QAAAA,QAAAA,KAAAA,IAAAA,MAAO,OAAO,EAChBH,MAAM,aAAa,IAAIG,MAAM,OAAO;QAItC,IAAI,QAAAD,CAAAA,mBAAAA,MAAM,OAAO,CAAC,EAAE,AAAD,IAAfA,KAAAA,IAAAA,iBAAkB,aAAa,EACjCF,MAAM,YAAY,GAAGD,MAAM,OAAO,CAAC,EAAE,CAAC,aAAa;QAIrD,OAAO;YACL,SAASI,AAAAA,CAAAA,QAAAA,QAAAA,KAAAA,IAAAA,MAAO,OAAO,AAAD,KAAK;YAC3B,kBAAkB;YAClB,mBAAmB;YACnB,WAAW,EAAE;QACf;IACF;IAKQ,qBAA6B;QACnC,OAAO,CAAC,KAAK,EAAEC,KAAK,GAAG,GAAG,CAAC,EAAEC,KAAK,MAAM,GAAG,QAAQ,CAAC,IAAI,SAAS,CAAC,GAAG,KAAK;IAC5E;IAKA,yBAAyBL,KAA4B,EAAuB;QAC1E,MAAMM,cAAcN,MAAM,aAAa;QACvCZ,cAAc,GAAG,CAAC,+CAAqCkB;QACvDlB,cAAc,GAAG,CAAC,0CAA0CkB;QAG5DC,QAAQ,GAAG,CAAC,8DAA8DD;QAG1E,IAAIE,oBAAoB;QACxB,IAAI,IAAI,CAAC,kBAAkB,EAAE;YAC3BA,oBAAoB,IAAI,CAAC,kBAAkB,CAACF;YAC5ClB,cAAc,GAAG,CAAC;YAClBmB,QAAQ,GAAG,CAAC;QACd;QAGA,IAAI,CAACC,mBAAmB;YACtBD,QAAQ,GAAG,CAAC;YACZC,oBAAoBtB,cAAc,eAAe,CAACoB;YAClDlB,cAAc,GAAG,CAAC;QACpB;QAEA,IAAIoB,mBACFD,QAAQ,GAAG,CACT,iDACAV,KAAK,SAAS,CAACW,mBAAmB,MAAM;aAG1CD,QAAQ,GAAG,CAAC;QAGd,IAAI,CAACC,qBAAqBA,kBAAkB,YAAY,EACtD,OAAO;YACL,SAAS;YACT,YAAYF;YACZ,WAAW;gBACT;oBACE,IAAI,IAAI,CAAC,kBAAkB;oBAC3B,MAAM;oBACN,UAAU;wBACR,MAAMG,sCAAAA,qBAAqBA;wBAC3B,WAAWZ,KAAK,SAAS,CAAC;4BACxB,QAAQ;4BACR,MAAM;4BACN,SAAS;4BACT,iBAAiB;4BACjB,cACEW,AAAAA,CAAAA,QAAAA,oBAAAA,KAAAA,IAAAA,kBAAmB,YAAY,AAAD,KAAK;wBACvC;oBACF;gBACF;aACD;YACD,cAAc;QAChB;QAGF,MAAME,YAA6C,EAAE;QAErD,IAAIC,WAAW;QACf,IAAIC,gBAA+B;QACnC,KAAK,MAAMC,UAAUL,kBAAkB,OAAO,CAAE;YAC9C,IAAIK,AAAgB,eAAhBA,OAAO,IAAI,EAAiB;oBAEdC;gBADhBH,WAAW;gBACXC,gBAAgBE,AAAAA,SAAAA,CAAAA,iBAAAA,OAAO,MAAM,AAAD,IAAZA,KAAAA,IAAAA,eAAe,OAAO,AAAD,KAAK;gBAC1C;YACF;YACAJ,UAAU,IAAI,CAAC;gBACb,IAAI,IAAI,CAAC,kBAAkB;gBAC3B,MAAM;gBACN,UAAU;oBACR,MAAMD,sCAAAA,qBAAqBA;oBAC3B,WAAWZ,KAAK,SAAS,CAAC;wBACxB,QAAQkB,AAAAA,IAAAA,sBAAAA,eAAAA,AAAAA,EAAgBF;wBACxB,MAAM;wBACN,SAASL,kBAAkB,gBAAgB,IAAI;wBAC/C,iBAAiBK;oBACnB;gBACF;YACF;QACF;QAEA,MAAMG,UAAUJ,iBAAiB;QACjC,MAAMK,mBAAmBT,kBAAkB,gBAAgB,IAAI;QAC/D,MAAMU,kBAAkBF,QAAQ,OAAO,CAAC,WAAW;QACnD,MAAMG,2BAA2BF,iBAAiB,OAAO,CAAC,WAAW;QAGrE,OAAO;YACL,SAASC;YACT,YAAYZ;YACZ,kBAAkBa;YAClBT;YACA,cAAcA,UAAU,MAAM,GAAG,KAAK,CAACC,WAAW,eAAe;QACnE;IACF;IAQA,gCACES,yBAAiE,EAC5B;QACrC,OAAO;YACL,MAAM;YACN,SAASA,0BAA0B,UAAU,IAAIA,0BAA0B,OAAO;QACpF;IACF;IAQA,sCACEC,eAA2C,EACb;QAC9B,OAAOA,gBAAgB,GAAG,CAAC,CAACC;YAE1B,MAAMC,cAAcD,OAAO,OAAO,CAC/B,MAAM,CAAC,CAACE,OAASA,AAAc,WAAdA,KAAK,IAAI,EAC1B,GAAG,CAAC,CAACA,OAAUA,KAA0B,IAAI,EAC7C,IAAI,CAAC;YAER,OAAO;gBACL,MAAM;gBACN,SAAS,CAAC,MAAM,EAAEF,OAAO,QAAQ,CAAC,WAAW,EAAEC,aAAa;YAC9D;QACF;IACF;IAzNA,YAAYE,kBAAuC,CAAE;QACnD,KAAK,IAHP,uBAAQ,sBAAR;QAIE,IAAI,CAAC,kBAAkB,GAAGA;IAC5B;AAuNF"}
+{"version":3,"file":"ToolCallEngine.js","sources":["webpack://@ui-tars-test/agent-sdk/webpack/runtime/define_property_getters","webpack://@ui-tars-test/agent-sdk/webpack/runtime/has_own_property","webpack://@ui-tars-test/agent-sdk/webpack/runtime/make_namespace_object","webpack://@ui-tars-test/agent-sdk/./src/ToolCallEngine.ts"],"sourcesContent":["__webpack_require__.d = (exports, definition) => {\n\tfor(var key in definition) {\n        if(__webpack_require__.o(definition, key) && !__webpack_require__.o(exports, key)) {\n            Object.defineProperty(exports, key, { enumerable: true, get: definition[key] });\n        }\n    }\n};","__webpack_require__.o = (obj, prop) => (Object.prototype.hasOwnProperty.call(obj, prop))","// define __esModule on exports\n__webpack_require__.r = (exports) => {\n\tif(typeof Symbol !== 'undefined' && Symbol.toStringTag) {\n\t\tObject.defineProperty(exports, Symbol.toStringTag, { value: 'Module' });\n\t}\n\tObject.defineProperty(exports, '__esModule', { value: true });\n};","import {\n  ToolCallEngine,\n  Tool,\n  ToolCallEnginePrepareRequestContext,\n  ChatCompletionCreateParams,\n  ChatCompletionAssistantMessageParam,\n  ChatCompletionChunk,\n  MultimodalToolCallResult,\n  AgentEventStream,\n  ChatCompletionMessageParam,\n  ChatCompletionMessageToolCall,\n  ParsedModelResponse,\n  StreamProcessingState,\n  StreamChunkResult,\n} from '@ui-tars-test/tarko-agent-interface';\nimport { DefaultActionParser } from '@ui-tars-test/action-parser';\nimport { GUI_ADAPTED_TOOL_NAME } from './constants';\nimport { ConsoleLogger, LogLevel } from '@agent-infra/logger';\nimport { serializeAction } from '@ui-tars-test/shared/utils';\nimport { CustomActionParser } from '@ui-tars-test/shared/types';\n\nconst defaultParser = new DefaultActionParser();\nconst defaultLogger = new ConsoleLogger('[GUIAgent:ToolCallEngine]', LogLevel.DEBUG);\n\n/**\n * GUIAgentToolCallEngine - Minimal prompt engineering tool call engine\n *\n * This is the simplest possible implementation of a tool call engine that:\n * 1. Uses prompt engineering to instruct the LLM to output tool calls in a specific format\n * 2. Parses tool calls from LLM response text using simple regex matching\n * 3. Does not support streaming (focuses on core functionality only)\n *\n * Format used: <tool_call>{\"name\": \"tool_name\", \"arguments\": {...}}</tool_call>\n */\nexport class GUIAgentToolCallEngine extends ToolCallEngine {\n  private customActionParser?: CustomActionParser;\n\n  constructor(customActionParser?: CustomActionParser) {\n    super();\n    this.customActionParser = customActionParser;\n  }\n\n  /**\n   * Prepare system prompt with tool information and instructions\n   */\n  preparePrompt(instructions: string, tools: Tool[]): string {\n    return instructions;\n  }\n\n  /**\n   * Prepare request parameters for the LLM\n   *\n   * FIXME: move to base tool call engine.\n   */\n  prepareRequest(context: ToolCallEnginePrepareRequestContext): ChatCompletionCreateParams {\n    defaultLogger.log(\n      \"【New Sys Prompt'】 System Prompt:\",\n      JSON.stringify(context.messages.find((m) => m.role === 'system')?.content || ''),\n    );\n    return {\n      model: context.model,\n      messages: context.messages,\n      temperature: context.temperature || 0.7,\n      stream: true,\n    };\n  }\n\n  /**\n   * Initialize processing state (minimal implementation)\n   *\n   * FIXME: move to base tool call engine.\n   */\n  initStreamProcessingState(): StreamProcessingState {\n    return {\n      contentBuffer: '',\n      toolCalls: [],\n      reasoningBuffer: '',\n      finishReason: null,\n    };\n  }\n\n  /**\n   * Process streaming chunks - simply accumulate content\n   *\n   * FIXME: make it optional\n   */\n  processStreamingChunk(\n    chunk: ChatCompletionChunk,\n    state: StreamProcessingState,\n  ): StreamChunkResult {\n    const delta = chunk.choices[0]?.delta;\n\n    // Accumulate content\n    if (delta?.content) {\n      state.contentBuffer += delta.content;\n    }\n\n    // Record finish reason\n    if (chunk.choices[0]?.finish_reason) {\n      state.finishReason = chunk.choices[0].finish_reason;\n    }\n\n    // Return incremental content without tool call detection during streaming\n    return {\n      content: delta?.content || '',\n      reasoningContent: '',\n      hasToolCallUpdate: false,\n      toolCalls: [],\n    };\n  }\n\n  /**\n   * Generate a tool call ID\n   */\n  private generateToolCallId(): string {\n    return `call_${Date.now()}_${Math.random().toString(36).substring(2, 11)}`;\n  }\n\n  /**\n   * Extract tool calls from complete response text\n   */\n  finalizeStreamProcessing(state: StreamProcessingState): ParsedModelResponse {\n    const fullContent = state.contentBuffer;\n    defaultLogger.log(\"【New Sys Prompt'】 Model Response:\", fullContent);\n    defaultLogger.log('[finalizeStreamProcessing] fullContent', fullContent);\n\n    // Add explicit log to confirm XML parsing intent\n    console.log('[CLI DEBUG] [ToolCallEngine] Full model response received:', fullContent);\n\n    // Try custom action parser first if available\n    let parsedGUIResponse = null;\n    if (this.customActionParser) {\n      parsedGUIResponse = this.customActionParser(fullContent);\n      defaultLogger.log('[finalizeStreamProcessing] Using custom action parser');\n      console.log('[CLI DEBUG] [ToolCallEngine] Using custom action parser');\n    }\n\n    // Fall back to default parser if custom parser is not available or returns null\n    if (!parsedGUIResponse) {\n      console.log('[CLI DEBUG] [ToolCallEngine] Using default action parser (XML parser)');\n      parsedGUIResponse = defaultParser.parsePrediction(fullContent);\n      defaultLogger.log('[finalizeStreamProcessing] Using default action parser');\n    }\n\n    if (parsedGUIResponse) {\n      console.log(\n        '[CLI DEBUG] [ToolCallEngine] Parsed response:',\n        JSON.stringify(parsedGUIResponse, null, 2),\n      );\n    } else {\n      console.log('[CLI DEBUG] [ToolCallEngine] Parsing failed or returned null');\n    }\n\n    if (!parsedGUIResponse || parsedGUIResponse.errorMessage) {\n      return {\n        content: '',\n        rawContent: fullContent,\n        toolCalls: [\n          {\n            id: this.generateToolCallId(),\n            type: 'function',\n            function: {\n              name: GUI_ADAPTED_TOOL_NAME,\n              arguments: JSON.stringify({\n                action: '',\n                step: '',\n                thought: '',\n                operator_action: null,\n                errorMessage:\n                  parsedGUIResponse?.errorMessage ?? 'Failed to parse GUI Action from output',\n              }),\n            },\n          },\n        ],\n        finishReason: 'tool_calls',\n      };\n    }\n\n    const toolCalls: ChatCompletionMessageToolCall[] = [];\n\n    let finished = false;\n    let finishMessage: string | null = null;\n    for (const action of parsedGUIResponse.actions) {\n      if (action.type === 'finished') {\n        finished = true;\n        finishMessage = action.inputs?.content ?? null;\n        continue;\n      }\n      toolCalls.push({\n        id: this.generateToolCallId(),\n        type: 'function',\n        function: {\n          name: GUI_ADAPTED_TOOL_NAME,\n          arguments: JSON.stringify({\n            action: serializeAction(action),\n            step: '',\n            thought: parsedGUIResponse.reasoningContent ?? '',\n            operator_action: action,\n          }),\n        },\n      });\n    }\n\n    const content = finishMessage ?? '';\n    const reasoningContent = parsedGUIResponse.reasoningContent ?? '';\n    const contentForWebUI = content.replace(/\\\\n|\\n/g, '<br>');\n    const reasoningContentForWebUI = reasoningContent.replace(/\\\\n|\\n/g, '<br>');\n\n    // No tool calls found - return regular response\n    return {\n      content: contentForWebUI,\n      rawContent: fullContent,\n      reasoningContent: reasoningContentForWebUI,\n      toolCalls,\n      finishReason: toolCalls.length > 0 && !finished ? 'tool_calls' : 'stop',\n    };\n  }\n\n  /**\n   * Build assistant message for conversation history\n   * For PE engines, we preserve the raw content including tool call markup\n   *\n   * FIXME: move to base tool call engine.\n   */\n  buildHistoricalAssistantMessage(\n    currentLoopAssistantEvent: AgentEventStream.AssistantMessageEvent,\n  ): ChatCompletionAssistantMessageParam {\n    return {\n      role: 'assistant',\n      content: currentLoopAssistantEvent.rawContent || currentLoopAssistantEvent.content,\n    };\n  }\n\n  /**\n   * Build tool result messages as user messages\n   * PE engines format tool results as user input for next iteration\n   *\n   * FIXME: move to base tool call engine.\n   */\n  buildHistoricalToolCallResultMessages(\n    toolCallResults: MultimodalToolCallResult[],\n  ): ChatCompletionMessageParam[] {\n    return toolCallResults.map((result) => {\n      // Extract text content from multimodal result\n      const textContent = result.content\n        .filter((part) => part.type === 'text')\n        .map((part) => (part as { text: string }).text)\n        .join('');\n\n      return {\n        role: 'user',\n        content: `Tool \"${result.toolName}\" result:\\n${textContent}`,\n      };\n    });\n  }\n}\n"],"names":["__webpack_require__","definition","key","Object","obj","prop","Symbol","defaultParser","DefaultActionParser","defaultLogger","ConsoleLogger","LogLevel","GUIAgentToolCallEngine","ToolCallEngine","instructions","tools","context","_context_messages_find","JSON","m","chunk","state","_chunk_choices_","_chunk_choices_1","delta","Date","Math","fullContent","console","parsedGUIResponse","GUI_ADAPTED_TOOL_NAME","toolCalls","finished","finishMessage","action","_action_inputs","serializeAction","content","reasoningContent","contentForWebUI","reasoningContentForWebUI","currentLoopAssistantEvent","toolCallResults","result","textContent","part","customActionParser"],"mappings":";;;;;;;IAAAA,oBAAoB,CAAC,GAAG,CAAC,UAASC;QACjC,IAAI,IAAIC,OAAOD,WACR,IAAGD,oBAAoB,CAAC,CAACC,YAAYC,QAAQ,CAACF,oBAAoB,CAAC,CAAC,UAASE,MACzEC,OAAO,cAAc,CAAC,UAASD,KAAK;YAAE,YAAY;YAAM,KAAKD,UAAU,CAACC,IAAI;QAAC;IAGzF;;;ICNAF,oBAAoB,CAAC,GAAG,CAACI,KAAKC,OAAUF,OAAO,SAAS,CAAC,cAAc,CAAC,IAAI,CAACC,KAAKC;;;ICClFL,oBAAoB,CAAC,GAAG,CAAC;QACxB,IAAG,AAAkB,eAAlB,OAAOM,UAA0BA,OAAO,WAAW,EACrDH,OAAO,cAAc,CAAC,UAASG,OAAO,WAAW,EAAE;YAAE,OAAO;QAAS;QAEtEH,OAAO,cAAc,CAAC,UAAS,cAAc;YAAE,OAAO;QAAK;IAC5D;;;;;;;;;;;;;;;;;;;;;;ACeA,MAAMI,gBAAgB,IAAIC,8BAAAA,mBAAmBA;AAC7C,MAAMC,gBAAgB,IAAIC,uBAAAA,aAAaA,CAAC,6BAA6BC,uBAAAA,QAAAA,CAAAA,KAAc;AAY5E,MAAMC,+BAA+BC,sCAAAA,cAAcA;IAWxD,cAAcC,YAAoB,EAAEC,KAAa,EAAU;QACzD,OAAOD;IACT;IAOA,eAAeE,OAA4C,EAA8B;YAGtEC;QAFjBR,cAAc,GAAG,CACf,8CACAS,KAAK,SAAS,CAACD,AAAAA,SAAAA,CAAAA,yBAAAA,QAAQ,QAAQ,CAAC,IAAI,CAAC,CAACE,IAAMA,AAAW,aAAXA,EAAE,IAAI,CAAa,IAAhDF,KAAAA,IAAAA,uBAAmD,OAAO,AAAD,KAAK;QAE/E,OAAO;YACL,OAAOD,QAAQ,KAAK;YACpB,UAAUA,QAAQ,QAAQ;YAC1B,aAAaA,QAAQ,WAAW,IAAI;YACpC,QAAQ;QACV;IACF;IAOA,4BAAmD;QACjD,OAAO;YACL,eAAe;YACf,WAAW,EAAE;YACb,iBAAiB;YACjB,cAAc;QAChB;IACF;IAOA,sBACEI,KAA0B,EAC1BC,KAA4B,EACT;YACLC,iBAQVC;QARJ,MAAMC,QAAQ,QAAAF,CAAAA,kBAAAA,MAAM,OAAO,CAAC,EAAE,AAAD,IAAfA,KAAAA,IAAAA,gBAAkB,KAAK;QAGrC,IAAIE,QAAAA,QAAAA,KAAAA,IAAAA,MAAO,OAAO,EAChBH,MAAM,aAAa,IAAIG,MAAM,OAAO;QAItC,IAAI,QAAAD,CAAAA,mBAAAA,MAAM,OAAO,CAAC,EAAE,AAAD,IAAfA,KAAAA,IAAAA,iBAAkB,aAAa,EACjCF,MAAM,YAAY,GAAGD,MAAM,OAAO,CAAC,EAAE,CAAC,aAAa;QAIrD,OAAO;YACL,SAASI,AAAAA,CAAAA,QAAAA,QAAAA,KAAAA,IAAAA,MAAO,OAAO,AAAD,KAAK;YAC3B,kBAAkB;YAClB,mBAAmB;YACnB,WAAW,EAAE;QACf;IACF;IAKQ,qBAA6B;QACnC,OAAO,CAAC,KAAK,EAAEC,KAAK,GAAG,GAAG,CAAC,EAAEC,KAAK,MAAM,GAAG,QAAQ,CAAC,IAAI,SAAS,CAAC,GAAG,KAAK;IAC5E;IAKA,yBAAyBL,KAA4B,EAAuB;QAC1E,MAAMM,cAAcN,MAAM,aAAa;QACvCZ,cAAc,GAAG,CAAC,+CAAqCkB;QACvDlB,cAAc,GAAG,CAAC,0CAA0CkB;QAG5DC,QAAQ,GAAG,CAAC,8DAA8DD;QAG1E,IAAIE,oBAAoB;QACxB,IAAI,IAAI,CAAC,kBAAkB,EAAE;YAC3BA,oBAAoB,IAAI,CAAC,kBAAkB,CAACF;YAC5ClB,cAAc,GAAG,CAAC;YAClBmB,QAAQ,GAAG,CAAC;QACd;QAGA,IAAI,CAACC,mBAAmB;YACtBD,QAAQ,GAAG,CAAC;YACZC,oBAAoBtB,cAAc,eAAe,CAACoB;YAClDlB,cAAc,GAAG,CAAC;QACpB;QAEA,IAAIoB,mBACFD,QAAQ,GAAG,CACT,iDACAV,KAAK,SAAS,CAACW,mBAAmB,MAAM;aAG1CD,QAAQ,GAAG,CAAC;QAGd,IAAI,CAACC,qBAAqBA,kBAAkB,YAAY,EACtD,OAAO;YACL,SAAS;YACT,YAAYF;YACZ,WAAW;gBACT;oBACE,IAAI,IAAI,CAAC,kBAAkB;oBAC3B,MAAM;oBACN,UAAU;wBACR,MAAMG,sCAAAA,qBAAqBA;wBAC3B,WAAWZ,KAAK,SAAS,CAAC;4BACxB,QAAQ;4BACR,MAAM;4BACN,SAAS;4BACT,iBAAiB;4BACjB,cACEW,AAAAA,CAAAA,QAAAA,oBAAAA,KAAAA,IAAAA,kBAAmB,YAAY,AAAD,KAAK;wBACvC;oBACF;gBACF;aACD;YACD,cAAc;QAChB;QAGF,MAAME,YAA6C,EAAE;QAErD,IAAIC,WAAW;QACf,IAAIC,gBAA+B;QACnC,KAAK,MAAMC,UAAUL,kBAAkB,OAAO,CAAE;YAC9C,IAAIK,AAAgB,eAAhBA,OAAO,IAAI,EAAiB;oBAEdC;gBADhBH,WAAW;gBACXC,gBAAgBE,AAAAA,SAAAA,CAAAA,iBAAAA,OAAO,MAAM,AAAD,IAAZA,KAAAA,IAAAA,eAAe,OAAO,AAAD,KAAK;gBAC1C;YACF;YACAJ,UAAU,IAAI,CAAC;gBACb,IAAI,IAAI,CAAC,kBAAkB;gBAC3B,MAAM;gBACN,UAAU;oBACR,MAAMD,sCAAAA,qBAAqBA;oBAC3B,WAAWZ,KAAK,SAAS,CAAC;wBACxB,QAAQkB,AAAAA,IAAAA,sBAAAA,eAAAA,AAAAA,EAAgBF;wBACxB,MAAM;wBACN,SAASL,kBAAkB,gBAAgB,IAAI;wBAC/C,iBAAiBK;oBACnB;gBACF;YACF;QACF;QAEA,MAAMG,UAAUJ,iBAAiB;QACjC,MAAMK,mBAAmBT,kBAAkB,gBAAgB,IAAI;QAC/D,MAAMU,kBAAkBF,QAAQ,OAAO,CAAC,WAAW;QACnD,MAAMG,2BAA2BF,iBAAiB,OAAO,CAAC,WAAW;QAGrE,OAAO;YACL,SAASC;YACT,YAAYZ;YACZ,kBAAkBa;YAClBT;YACA,cAAcA,UAAU,MAAM,GAAG,KAAK,CAACC,WAAW,eAAe;QACnE;IACF;IAQA,gCACES,yBAAiE,EAC5B;QACrC,OAAO;YACL,MAAM;YACN,SAASA,0BAA0B,UAAU,IAAIA,0BAA0B,OAAO;QACpF;IACF;IAQA,sCACEC,eAA2C,EACb;QAC9B,OAAOA,gBAAgB,GAAG,CAAC,CAACC;YAE1B,MAAMC,cAAcD,OAAO,OAAO,CAC/B,MAAM,CAAC,CAACE,OAASA,AAAc,WAAdA,KAAK,IAAI,EAC1B,GAAG,CAAC,CAACA,OAAUA,KAA0B,IAAI,EAC7C,IAAI,CAAC;YAER,OAAO;gBACL,MAAM;gBACN,SAAS,CAAC,MAAM,EAAEF,OAAO,QAAQ,CAAC,WAAW,EAAEC,aAAa;YAC9D;QACF;IACF;IAzNA,YAAYE,kBAAuC,CAAE;QACnD,KAAK,IAHP,uBAAQ,sBAAR;QAIE,IAAI,CAAC,kBAAkB,GAAGA;IAC5B;AAuNF"}

package/dist/ToolCallEngine.mjs

@@ -2,7 +2,7 @@
 * Copyright (c) 2025 Bytedance, Inc. and its affiliates.
 * SPDX-License-Identifier: Apache-2.0
 */
-import { ToolCallEngine } from "@tarko/agent-interface";
+import { ToolCallEngine } from "@ui-tars-test/tarko-agent-interface";
 import { DefaultActionParser } from "@ui-tars-test/action-parser";
 import { GUI_ADAPTED_TOOL_NAME } from "./constants.mjs";
 import { ConsoleLogger, LogLevel } from "@agent-infra/logger";

package/dist/ToolCallEngine.mjs.map

@@ -1 +1 @@
-{"version":3,"file":"ToolCallEngine.mjs","sources":["webpack://@ui-tars-test/agent-sdk/./src/ToolCallEngine.ts"],"sourcesContent":["import {\n  ToolCallEngine,\n  Tool,\n  ToolCallEnginePrepareRequestContext,\n  ChatCompletionCreateParams,\n  ChatCompletionAssistantMessageParam,\n  ChatCompletionChunk,\n  MultimodalToolCallResult,\n  AgentEventStream,\n  ChatCompletionMessageParam,\n  ChatCompletionMessageToolCall,\n  ParsedModelResponse,\n  StreamProcessingState,\n  StreamChunkResult,\n} from '@tarko/agent-interface';\nimport { DefaultActionParser } from '@ui-tars-test/action-parser';\nimport { GUI_ADAPTED_TOOL_NAME } from './constants';\nimport { ConsoleLogger, LogLevel } from '@agent-infra/logger';\nimport { serializeAction } from '@ui-tars-test/shared/utils';\nimport { CustomActionParser } from '@ui-tars-test/shared/types';\n\nconst defaultParser = new DefaultActionParser();\nconst defaultLogger = new ConsoleLogger('[GUIAgent:ToolCallEngine]', LogLevel.DEBUG);\n\n/**\n * GUIAgentToolCallEngine - Minimal prompt engineering tool call engine\n *\n * This is the simplest possible implementation of a tool call engine that:\n * 1. Uses prompt engineering to instruct the LLM to output tool calls in a specific format\n * 2. Parses tool calls from LLM response text using simple regex matching\n * 3. Does not support streaming (focuses on core functionality only)\n *\n * Format used: <tool_call>{\"name\": \"tool_name\", \"arguments\": {...}}</tool_call>\n */\nexport class GUIAgentToolCallEngine extends ToolCallEngine {\n  private customActionParser?: CustomActionParser;\n\n  constructor(customActionParser?: CustomActionParser) {\n    super();\n    this.customActionParser = customActionParser;\n  }\n\n  /**\n   * Prepare system prompt with tool information and instructions\n   */\n  preparePrompt(instructions: string, tools: Tool[]): string {\n    return instructions;\n  }\n\n  /**\n   * Prepare request parameters for the LLM\n   *\n   * FIXME: move to base tool call engine.\n   */\n  prepareRequest(context: ToolCallEnginePrepareRequestContext): ChatCompletionCreateParams {\n    defaultLogger.log(\n      \"【New Sys Prompt'】 System Prompt:\",\n      JSON.stringify(context.messages.find((m) => m.role === 'system')?.content || ''),\n    );\n    return {\n      model: context.model,\n      messages: context.messages,\n      temperature: context.temperature || 0.7,\n      stream: true,\n    };\n  }\n\n  /**\n   * Initialize processing state (minimal implementation)\n   *\n   * FIXME: move to base tool call engine.\n   */\n  initStreamProcessingState(): StreamProcessingState {\n    return {\n      contentBuffer: '',\n      toolCalls: [],\n      reasoningBuffer: '',\n      finishReason: null,\n    };\n  }\n\n  /**\n   * Process streaming chunks - simply accumulate content\n   *\n   * FIXME: make it optional\n   */\n  processStreamingChunk(\n    chunk: ChatCompletionChunk,\n    state: StreamProcessingState,\n  ): StreamChunkResult {\n    const delta = chunk.choices[0]?.delta;\n\n    // Accumulate content\n    if (delta?.content) {\n      state.contentBuffer += delta.content;\n    }\n\n    // Record finish reason\n    if (chunk.choices[0]?.finish_reason) {\n      state.finishReason = chunk.choices[0].finish_reason;\n    }\n\n    // Return incremental content without tool call detection during streaming\n    return {\n      content: delta?.content || '',\n      reasoningContent: '',\n      hasToolCallUpdate: false,\n      toolCalls: [],\n    };\n  }\n\n  /**\n   * Generate a tool call ID\n   */\n  private generateToolCallId(): string {\n    return `call_${Date.now()}_${Math.random().toString(36).substring(2, 11)}`;\n  }\n\n  /**\n   * Extract tool calls from complete response text\n   */\n  finalizeStreamProcessing(state: StreamProcessingState): ParsedModelResponse {\n    const fullContent = state.contentBuffer;\n    defaultLogger.log(\"【New Sys Prompt'】 Model Response:\", fullContent);\n    defaultLogger.log('[finalizeStreamProcessing] fullContent', fullContent);\n\n    // Add explicit log to confirm XML parsing intent\n    console.log('[CLI DEBUG] [ToolCallEngine] Full model response received:', fullContent);\n\n    // Try custom action parser first if available\n    let parsedGUIResponse = null;\n    if (this.customActionParser) {\n      parsedGUIResponse = this.customActionParser(fullContent);\n      defaultLogger.log('[finalizeStreamProcessing] Using custom action parser');\n      console.log('[CLI DEBUG] [ToolCallEngine] Using custom action parser');\n    }\n\n    // Fall back to default parser if custom parser is not available or returns null\n    if (!parsedGUIResponse) {\n      console.log('[CLI DEBUG] [ToolCallEngine] Using default action parser (XML parser)');\n      parsedGUIResponse = defaultParser.parsePrediction(fullContent);\n      defaultLogger.log('[finalizeStreamProcessing] Using default action parser');\n    }\n\n    if (parsedGUIResponse) {\n      console.log(\n        '[CLI DEBUG] [ToolCallEngine] Parsed response:',\n        JSON.stringify(parsedGUIResponse, null, 2),\n      );\n    } else {\n      console.log('[CLI DEBUG] [ToolCallEngine] Parsing failed or returned null');\n    }\n\n    if (!parsedGUIResponse || parsedGUIResponse.errorMessage) {\n      return {\n        content: '',\n        rawContent: fullContent,\n        toolCalls: [\n          {\n            id: this.generateToolCallId(),\n            type: 'function',\n            function: {\n              name: GUI_ADAPTED_TOOL_NAME,\n              arguments: JSON.stringify({\n                action: '',\n                step: '',\n                thought: '',\n                operator_action: null,\n                errorMessage:\n                  parsedGUIResponse?.errorMessage ?? 'Failed to parse GUI Action from output',\n              }),\n            },\n          },\n        ],\n        finishReason: 'tool_calls',\n      };\n    }\n\n    const toolCalls: ChatCompletionMessageToolCall[] = [];\n\n    let finished = false;\n    let finishMessage: string | null = null;\n    for (const action of parsedGUIResponse.actions) {\n      if (action.type === 'finished') {\n        finished = true;\n        finishMessage = action.inputs?.content ?? null;\n        continue;\n      }\n      toolCalls.push({\n        id: this.generateToolCallId(),\n        type: 'function',\n        function: {\n          name: GUI_ADAPTED_TOOL_NAME,\n          arguments: JSON.stringify({\n            action: serializeAction(action),\n            step: '',\n            thought: parsedGUIResponse.reasoningContent ?? '',\n            operator_action: action,\n          }),\n        },\n      });\n    }\n\n    const content = finishMessage ?? '';\n    const reasoningContent = parsedGUIResponse.reasoningContent ?? '';\n    const contentForWebUI = content.replace(/\\\\n|\\n/g, '<br>');\n    const reasoningContentForWebUI = reasoningContent.replace(/\\\\n|\\n/g, '<br>');\n\n    // No tool calls found - return regular response\n    return {\n      content: contentForWebUI,\n      rawContent: fullContent,\n      reasoningContent: reasoningContentForWebUI,\n      toolCalls,\n      finishReason: toolCalls.length > 0 && !finished ? 'tool_calls' : 'stop',\n    };\n  }\n\n  /**\n   * Build assistant message for conversation history\n   * For PE engines, we preserve the raw content including tool call markup\n   *\n   * FIXME: move to base tool call engine.\n   */\n  buildHistoricalAssistantMessage(\n    currentLoopAssistantEvent: AgentEventStream.AssistantMessageEvent,\n  ): ChatCompletionAssistantMessageParam {\n    return {\n      role: 'assistant',\n      content: currentLoopAssistantEvent.rawContent || currentLoopAssistantEvent.content,\n    };\n  }\n\n  /**\n   * Build tool result messages as user messages\n   * PE engines format tool results as user input for next iteration\n   *\n   * FIXME: move to base tool call engine.\n   */\n  buildHistoricalToolCallResultMessages(\n    toolCallResults: MultimodalToolCallResult[],\n  ): ChatCompletionMessageParam[] {\n    return toolCallResults.map((result) => {\n      // Extract text content from multimodal result\n      const textContent = result.content\n        .filter((part) => part.type === 'text')\n        .map((part) => (part as { text: string }).text)\n        .join('');\n\n      return {\n        role: 'user',\n        content: `Tool \"${result.toolName}\" result:\\n${textContent}`,\n      };\n    });\n  }\n}\n"],"names":["defaultParser","DefaultActionParser","defaultLogger","ConsoleLogger","LogLevel","GUIAgentToolCallEngine","ToolCallEngine","instructions","tools","context","_context_messages_find","JSON","m","chunk","state","_chunk_choices_","_chunk_choices_1","delta","Date","Math","fullContent","console","parsedGUIResponse","GUI_ADAPTED_TOOL_NAME","toolCalls","finished","finishMessage","action","_action_inputs","serializeAction","content","reasoningContent","contentForWebUI","reasoningContentForWebUI","currentLoopAssistantEvent","toolCallResults","result","textContent","part","customActionParser"],"mappings":";;;;;;;;;;;;;;;;;;;AAqBA,MAAMA,gBAAgB,IAAIC;AAC1B,MAAMC,gBAAgB,IAAIC,cAAc,6BAA6BC,SAAS,KAAK;AAY5E,MAAMC,+BAA+BC;IAW1C,cAAcC,YAAoB,EAAEC,KAAa,EAAU;QACzD,OAAOD;IACT;IAOA,eAAeE,OAA4C,EAA8B;YAGtEC;QAFjBR,cAAc,GAAG,CACf,8CACAS,KAAK,SAAS,CAACD,AAAAA,SAAAA,CAAAA,yBAAAA,QAAQ,QAAQ,CAAC,IAAI,CAAC,CAACE,IAAMA,AAAW,aAAXA,EAAE,IAAI,CAAa,IAAhDF,KAAAA,IAAAA,uBAAmD,OAAO,AAAD,KAAK;QAE/E,OAAO;YACL,OAAOD,QAAQ,KAAK;YACpB,UAAUA,QAAQ,QAAQ;YAC1B,aAAaA,QAAQ,WAAW,IAAI;YACpC,QAAQ;QACV;IACF;IAOA,4BAAmD;QACjD,OAAO;YACL,eAAe;YACf,WAAW,EAAE;YACb,iBAAiB;YACjB,cAAc;QAChB;IACF;IAOA,sBACEI,KAA0B,EAC1BC,KAA4B,EACT;YACLC,iBAQVC;QARJ,MAAMC,QAAQ,QAAAF,CAAAA,kBAAAA,MAAM,OAAO,CAAC,EAAE,AAAD,IAAfA,KAAAA,IAAAA,gBAAkB,KAAK;QAGrC,IAAIE,QAAAA,QAAAA,KAAAA,IAAAA,MAAO,OAAO,EAChBH,MAAM,aAAa,IAAIG,MAAM,OAAO;QAItC,IAAI,QAAAD,CAAAA,mBAAAA,MAAM,OAAO,CAAC,EAAE,AAAD,IAAfA,KAAAA,IAAAA,iBAAkB,aAAa,EACjCF,MAAM,YAAY,GAAGD,MAAM,OAAO,CAAC,EAAE,CAAC,aAAa;QAIrD,OAAO;YACL,SAASI,AAAAA,CAAAA,QAAAA,QAAAA,KAAAA,IAAAA,MAAO,OAAO,AAAD,KAAK;YAC3B,kBAAkB;YAClB,mBAAmB;YACnB,WAAW,EAAE;QACf;IACF;IAKQ,qBAA6B;QACnC,OAAO,CAAC,KAAK,EAAEC,KAAK,GAAG,GAAG,CAAC,EAAEC,KAAK,MAAM,GAAG,QAAQ,CAAC,IAAI,SAAS,CAAC,GAAG,KAAK;IAC5E;IAKA,yBAAyBL,KAA4B,EAAuB;QAC1E,MAAMM,cAAcN,MAAM,aAAa;QACvCZ,cAAc,GAAG,CAAC,+CAAqCkB;QACvDlB,cAAc,GAAG,CAAC,0CAA0CkB;QAG5DC,QAAQ,GAAG,CAAC,8DAA8DD;QAG1E,IAAIE,oBAAoB;QACxB,IAAI,IAAI,CAAC,kBAAkB,EAAE;YAC3BA,oBAAoB,IAAI,CAAC,kBAAkB,CAACF;YAC5ClB,cAAc,GAAG,CAAC;YAClBmB,QAAQ,GAAG,CAAC;QACd;QAGA,IAAI,CAACC,mBAAmB;YACtBD,QAAQ,GAAG,CAAC;YACZC,oBAAoBtB,cAAc,eAAe,CAACoB;YAClDlB,cAAc,GAAG,CAAC;QACpB;QAEA,IAAIoB,mBACFD,QAAQ,GAAG,CACT,iDACAV,KAAK,SAAS,CAACW,mBAAmB,MAAM;aAG1CD,QAAQ,GAAG,CAAC;QAGd,IAAI,CAACC,qBAAqBA,kBAAkB,YAAY,EACtD,OAAO;YACL,SAAS;YACT,YAAYF;YACZ,WAAW;gBACT;oBACE,IAAI,IAAI,CAAC,kBAAkB;oBAC3B,MAAM;oBACN,UAAU;wBACR,MAAMG;wBACN,WAAWZ,KAAK,SAAS,CAAC;4BACxB,QAAQ;4BACR,MAAM;4BACN,SAAS;4BACT,iBAAiB;4BACjB,cACEW,AAAAA,CAAAA,QAAAA,oBAAAA,KAAAA,IAAAA,kBAAmB,YAAY,AAAD,KAAK;wBACvC;oBACF;gBACF;aACD;YACD,cAAc;QAChB;QAGF,MAAME,YAA6C,EAAE;QAErD,IAAIC,WAAW;QACf,IAAIC,gBAA+B;QACnC,KAAK,MAAMC,UAAUL,kBAAkB,OAAO,CAAE;YAC9C,IAAIK,AAAgB,eAAhBA,OAAO,IAAI,EAAiB;oBAEdC;gBADhBH,WAAW;gBACXC,gBAAgBE,AAAAA,SAAAA,CAAAA,iBAAAA,OAAO,MAAM,AAAD,IAAZA,KAAAA,IAAAA,eAAe,OAAO,AAAD,KAAK;gBAC1C;YACF;YACAJ,UAAU,IAAI,CAAC;gBACb,IAAI,IAAI,CAAC,kBAAkB;gBAC3B,MAAM;gBACN,UAAU;oBACR,MAAMD;oBACN,WAAWZ,KAAK,SAAS,CAAC;wBACxB,QAAQkB,gBAAgBF;wBACxB,MAAM;wBACN,SAASL,kBAAkB,gBAAgB,IAAI;wBAC/C,iBAAiBK;oBACnB;gBACF;YACF;QACF;QAEA,MAAMG,UAAUJ,iBAAiB;QACjC,MAAMK,mBAAmBT,kBAAkB,gBAAgB,IAAI;QAC/D,MAAMU,kBAAkBF,QAAQ,OAAO,CAAC,WAAW;QACnD,MAAMG,2BAA2BF,iBAAiB,OAAO,CAAC,WAAW;QAGrE,OAAO;YACL,SAASC;YACT,YAAYZ;YACZ,kBAAkBa;YAClBT;YACA,cAAcA,UAAU,MAAM,GAAG,KAAK,CAACC,WAAW,eAAe;QACnE;IACF;IAQA,gCACES,yBAAiE,EAC5B;QACrC,OAAO;YACL,MAAM;YACN,SAASA,0BAA0B,UAAU,IAAIA,0BAA0B,OAAO;QACpF;IACF;IAQA,sCACEC,eAA2C,EACb;QAC9B,OAAOA,gBAAgB,GAAG,CAAC,CAACC;YAE1B,MAAMC,cAAcD,OAAO,OAAO,CAC/B,MAAM,CAAC,CAACE,OAASA,AAAc,WAAdA,KAAK,IAAI,EAC1B,GAAG,CAAC,CAACA,OAAUA,KAA0B,IAAI,EAC7C,IAAI,CAAC;YAER,OAAO;gBACL,MAAM;gBACN,SAAS,CAAC,MAAM,EAAEF,OAAO,QAAQ,CAAC,WAAW,EAAEC,aAAa;YAC9D;QACF;IACF;IAzNA,YAAYE,kBAAuC,CAAE;QACnD,KAAK,IAHP,uBAAQ,sBAAR;QAIE,IAAI,CAAC,kBAAkB,GAAGA;IAC5B;AAuNF"}
+{"version":3,"file":"ToolCallEngine.mjs","sources":["webpack://@ui-tars-test/agent-sdk/./src/ToolCallEngine.ts"],"sourcesContent":["import {\n  ToolCallEngine,\n  Tool,\n  ToolCallEnginePrepareRequestContext,\n  ChatCompletionCreateParams,\n  ChatCompletionAssistantMessageParam,\n  ChatCompletionChunk,\n  MultimodalToolCallResult,\n  AgentEventStream,\n  ChatCompletionMessageParam,\n  ChatCompletionMessageToolCall,\n  ParsedModelResponse,\n  StreamProcessingState,\n  StreamChunkResult,\n} from '@ui-tars-test/tarko-agent-interface';\nimport { DefaultActionParser } from '@ui-tars-test/action-parser';\nimport { GUI_ADAPTED_TOOL_NAME } from './constants';\nimport { ConsoleLogger, LogLevel } from '@agent-infra/logger';\nimport { serializeAction } from '@ui-tars-test/shared/utils';\nimport { CustomActionParser } from '@ui-tars-test/shared/types';\n\nconst defaultParser = new DefaultActionParser();\nconst defaultLogger = new ConsoleLogger('[GUIAgent:ToolCallEngine]', LogLevel.DEBUG);\n\n/**\n * GUIAgentToolCallEngine - Minimal prompt engineering tool call engine\n *\n * This is the simplest possible implementation of a tool call engine that:\n * 1. Uses prompt engineering to instruct the LLM to output tool calls in a specific format\n * 2. Parses tool calls from LLM response text using simple regex matching\n * 3. Does not support streaming (focuses on core functionality only)\n *\n * Format used: <tool_call>{\"name\": \"tool_name\", \"arguments\": {...}}</tool_call>\n */\nexport class GUIAgentToolCallEngine extends ToolCallEngine {\n  private customActionParser?: CustomActionParser;\n\n  constructor(customActionParser?: CustomActionParser) {\n    super();\n    this.customActionParser = customActionParser;\n  }\n\n  /**\n   * Prepare system prompt with tool information and instructions\n   */\n  preparePrompt(instructions: string, tools: Tool[]): string {\n    return instructions;\n  }\n\n  /**\n   * Prepare request parameters for the LLM\n   *\n   * FIXME: move to base tool call engine.\n   */\n  prepareRequest(context: ToolCallEnginePrepareRequestContext): ChatCompletionCreateParams {\n    defaultLogger.log(\n      \"【New Sys Prompt'】 System Prompt:\",\n      JSON.stringify(context.messages.find((m) => m.role === 'system')?.content || ''),\n    );\n    return {\n      model: context.model,\n      messages: context.messages,\n      temperature: context.temperature || 0.7,\n      stream: true,\n    };\n  }\n\n  /**\n   * Initialize processing state (minimal implementation)\n   *\n   * FIXME: move to base tool call engine.\n   */\n  initStreamProcessingState(): StreamProcessingState {\n    return {\n      contentBuffer: '',\n      toolCalls: [],\n      reasoningBuffer: '',\n      finishReason: null,\n    };\n  }\n\n  /**\n   * Process streaming chunks - simply accumulate content\n   *\n   * FIXME: make it optional\n   */\n  processStreamingChunk(\n    chunk: ChatCompletionChunk,\n    state: StreamProcessingState,\n  ): StreamChunkResult {\n    const delta = chunk.choices[0]?.delta;\n\n    // Accumulate content\n    if (delta?.content) {\n      state.contentBuffer += delta.content;\n    }\n\n    // Record finish reason\n    if (chunk.choices[0]?.finish_reason) {\n      state.finishReason = chunk.choices[0].finish_reason;\n    }\n\n    // Return incremental content without tool call detection during streaming\n    return {\n      content: delta?.content || '',\n      reasoningContent: '',\n      hasToolCallUpdate: false,\n      toolCalls: [],\n    };\n  }\n\n  /**\n   * Generate a tool call ID\n   */\n  private generateToolCallId(): string {\n    return `call_${Date.now()}_${Math.random().toString(36).substring(2, 11)}`;\n  }\n\n  /**\n   * Extract tool calls from complete response text\n   */\n  finalizeStreamProcessing(state: StreamProcessingState): ParsedModelResponse {\n    const fullContent = state.contentBuffer;\n    defaultLogger.log(\"【New Sys Prompt'】 Model Response:\", fullContent);\n    defaultLogger.log('[finalizeStreamProcessing] fullContent', fullContent);\n\n    // Add explicit log to confirm XML parsing intent\n    console.log('[CLI DEBUG] [ToolCallEngine] Full model response received:', fullContent);\n\n    // Try custom action parser first if available\n    let parsedGUIResponse = null;\n    if (this.customActionParser) {\n      parsedGUIResponse = this.customActionParser(fullContent);\n      defaultLogger.log('[finalizeStreamProcessing] Using custom action parser');\n      console.log('[CLI DEBUG] [ToolCallEngine] Using custom action parser');\n    }\n\n    // Fall back to default parser if custom parser is not available or returns null\n    if (!parsedGUIResponse) {\n      console.log('[CLI DEBUG] [ToolCallEngine] Using default action parser (XML parser)');\n      parsedGUIResponse = defaultParser.parsePrediction(fullContent);\n      defaultLogger.log('[finalizeStreamProcessing] Using default action parser');\n    }\n\n    if (parsedGUIResponse) {\n      console.log(\n        '[CLI DEBUG] [ToolCallEngine] Parsed response:',\n        JSON.stringify(parsedGUIResponse, null, 2),\n      );\n    } else {\n      console.log('[CLI DEBUG] [ToolCallEngine] Parsing failed or returned null');\n    }\n\n    if (!parsedGUIResponse || parsedGUIResponse.errorMessage) {\n      return {\n        content: '',\n        rawContent: fullContent,\n        toolCalls: [\n          {\n            id: this.generateToolCallId(),\n            type: 'function',\n            function: {\n              name: GUI_ADAPTED_TOOL_NAME,\n              arguments: JSON.stringify({\n                action: '',\n                step: '',\n                thought: '',\n                operator_action: null,\n                errorMessage:\n                  parsedGUIResponse?.errorMessage ?? 'Failed to parse GUI Action from output',\n              }),\n            },\n          },\n        ],\n        finishReason: 'tool_calls',\n      };\n    }\n\n    const toolCalls: ChatCompletionMessageToolCall[] = [];\n\n    let finished = false;\n    let finishMessage: string | null = null;\n    for (const action of parsedGUIResponse.actions) {\n      if (action.type === 'finished') {\n        finished = true;\n        finishMessage = action.inputs?.content ?? null;\n        continue;\n      }\n      toolCalls.push({\n        id: this.generateToolCallId(),\n        type: 'function',\n        function: {\n          name: GUI_ADAPTED_TOOL_NAME,\n          arguments: JSON.stringify({\n            action: serializeAction(action),\n            step: '',\n            thought: parsedGUIResponse.reasoningContent ?? '',\n            operator_action: action,\n          }),\n        },\n      });\n    }\n\n    const content = finishMessage ?? '';\n    const reasoningContent = parsedGUIResponse.reasoningContent ?? '';\n    const contentForWebUI = content.replace(/\\\\n|\\n/g, '<br>');\n    const reasoningContentForWebUI = reasoningContent.replace(/\\\\n|\\n/g, '<br>');\n\n    // No tool calls found - return regular response\n    return {\n      content: contentForWebUI,\n      rawContent: fullContent,\n      reasoningContent: reasoningContentForWebUI,\n      toolCalls,\n      finishReason: toolCalls.length > 0 && !finished ? 'tool_calls' : 'stop',\n    };\n  }\n\n  /**\n   * Build assistant message for conversation history\n   * For PE engines, we preserve the raw content including tool call markup\n   *\n   * FIXME: move to base tool call engine.\n   */\n  buildHistoricalAssistantMessage(\n    currentLoopAssistantEvent: AgentEventStream.AssistantMessageEvent,\n  ): ChatCompletionAssistantMessageParam {\n    return {\n      role: 'assistant',\n      content: currentLoopAssistantEvent.rawContent || currentLoopAssistantEvent.content,\n    };\n  }\n\n  /**\n   * Build tool result messages as user messages\n   * PE engines format tool results as user input for next iteration\n   *\n   * FIXME: move to base tool call engine.\n   */\n  buildHistoricalToolCallResultMessages(\n    toolCallResults: MultimodalToolCallResult[],\n  ): ChatCompletionMessageParam[] {\n    return toolCallResults.map((result) => {\n      // Extract text content from multimodal result\n      const textContent = result.content\n        .filter((part) => part.type === 'text')\n        .map((part) => (part as { text: string }).text)\n        .join('');\n\n      return {\n        role: 'user',\n        content: `Tool \"${result.toolName}\" result:\\n${textContent}`,\n      };\n    });\n  }\n}\n"],"names":["defaultParser","DefaultActionParser","defaultLogger","ConsoleLogger","LogLevel","GUIAgentToolCallEngine","ToolCallEngine","instructions","tools","context","_context_messages_find","JSON","m","chunk","state","_chunk_choices_","_chunk_choices_1","delta","Date","Math","fullContent","console","parsedGUIResponse","GUI_ADAPTED_TOOL_NAME","toolCalls","finished","finishMessage","action","_action_inputs","serializeAction","content","reasoningContent","contentForWebUI","reasoningContentForWebUI","currentLoopAssistantEvent","toolCallResults","result","textContent","part","customActionParser"],"mappings":";;;;;;;;;;;;;;;;;;;AAqBA,MAAMA,gBAAgB,IAAIC;AAC1B,MAAMC,gBAAgB,IAAIC,cAAc,6BAA6BC,SAAS,KAAK;AAY5E,MAAMC,+BAA+BC;IAW1C,cAAcC,YAAoB,EAAEC,KAAa,EAAU;QACzD,OAAOD;IACT;IAOA,eAAeE,OAA4C,EAA8B;YAGtEC;QAFjBR,cAAc,GAAG,CACf,8CACAS,KAAK,SAAS,CAACD,AAAAA,SAAAA,CAAAA,yBAAAA,QAAQ,QAAQ,CAAC,IAAI,CAAC,CAACE,IAAMA,AAAW,aAAXA,EAAE,IAAI,CAAa,IAAhDF,KAAAA,IAAAA,uBAAmD,OAAO,AAAD,KAAK;QAE/E,OAAO;YACL,OAAOD,QAAQ,KAAK;YACpB,UAAUA,QAAQ,QAAQ;YAC1B,aAAaA,QAAQ,WAAW,IAAI;YACpC,QAAQ;QACV;IACF;IAOA,4BAAmD;QACjD,OAAO;YACL,eAAe;YACf,WAAW,EAAE;YACb,iBAAiB;YACjB,cAAc;QAChB;IACF;IAOA,sBACEI,KAA0B,EAC1BC,KAA4B,EACT;YACLC,iBAQVC;QARJ,MAAMC,QAAQ,QAAAF,CAAAA,kBAAAA,MAAM,OAAO,CAAC,EAAE,AAAD,IAAfA,KAAAA,IAAAA,gBAAkB,KAAK;QAGrC,IAAIE,QAAAA,QAAAA,KAAAA,IAAAA,MAAO,OAAO,EAChBH,MAAM,aAAa,IAAIG,MAAM,OAAO;QAItC,IAAI,QAAAD,CAAAA,mBAAAA,MAAM,OAAO,CAAC,EAAE,AAAD,IAAfA,KAAAA,IAAAA,iBAAkB,aAAa,EACjCF,MAAM,YAAY,GAAGD,MAAM,OAAO,CAAC,EAAE,CAAC,aAAa;QAIrD,OAAO;YACL,SAASI,AAAAA,CAAAA,QAAAA,QAAAA,KAAAA,IAAAA,MAAO,OAAO,AAAD,KAAK;YAC3B,kBAAkB;YAClB,mBAAmB;YACnB,WAAW,EAAE;QACf;IACF;IAKQ,qBAA6B;QACnC,OAAO,CAAC,KAAK,EAAEC,KAAK,GAAG,GAAG,CAAC,EAAEC,KAAK,MAAM,GAAG,QAAQ,CAAC,IAAI,SAAS,CAAC,GAAG,KAAK;IAC5E;IAKA,yBAAyBL,KAA4B,EAAuB;QAC1E,MAAMM,cAAcN,MAAM,aAAa;QACvCZ,cAAc,GAAG,CAAC,+CAAqCkB;QACvDlB,cAAc,GAAG,CAAC,0CAA0CkB;QAG5DC,QAAQ,GAAG,CAAC,8DAA8DD;QAG1E,IAAIE,oBAAoB;QACxB,IAAI,IAAI,CAAC,kBAAkB,EAAE;YAC3BA,oBAAoB,IAAI,CAAC,kBAAkB,CAACF;YAC5ClB,cAAc,GAAG,CAAC;YAClBmB,QAAQ,GAAG,CAAC;QACd;QAGA,IAAI,CAACC,mBAAmB;YACtBD,QAAQ,GAAG,CAAC;YACZC,oBAAoBtB,cAAc,eAAe,CAACoB;YAClDlB,cAAc,GAAG,CAAC;QACpB;QAEA,IAAIoB,mBACFD,QAAQ,GAAG,CACT,iDACAV,KAAK,SAAS,CAACW,mBAAmB,MAAM;aAG1CD,QAAQ,GAAG,CAAC;QAGd,IAAI,CAACC,qBAAqBA,kBAAkB,YAAY,EACtD,OAAO;YACL,SAAS;YACT,YAAYF;YACZ,WAAW;gBACT;oBACE,IAAI,IAAI,CAAC,kBAAkB;oBAC3B,MAAM;oBACN,UAAU;wBACR,MAAMG;wBACN,WAAWZ,KAAK,SAAS,CAAC;4BACxB,QAAQ;4BACR,MAAM;4BACN,SAAS;4BACT,iBAAiB;4BACjB,cACEW,AAAAA,CAAAA,QAAAA,oBAAAA,KAAAA,IAAAA,kBAAmB,YAAY,AAAD,KAAK;wBACvC;oBACF;gBACF;aACD;YACD,cAAc;QAChB;QAGF,MAAME,YAA6C,EAAE;QAErD,IAAIC,WAAW;QACf,IAAIC,gBAA+B;QACnC,KAAK,MAAMC,UAAUL,kBAAkB,OAAO,CAAE;YAC9C,IAAIK,AAAgB,eAAhBA,OAAO,IAAI,EAAiB;oBAEdC;gBADhBH,WAAW;gBACXC,gBAAgBE,AAAAA,SAAAA,CAAAA,iBAAAA,OAAO,MAAM,AAAD,IAAZA,KAAAA,IAAAA,eAAe,OAAO,AAAD,KAAK;gBAC1C;YACF;YACAJ,UAAU,IAAI,CAAC;gBACb,IAAI,IAAI,CAAC,kBAAkB;gBAC3B,MAAM;gBACN,UAAU;oBACR,MAAMD;oBACN,WAAWZ,KAAK,SAAS,CAAC;wBACxB,QAAQkB,gBAAgBF;wBACxB,MAAM;wBACN,SAASL,kBAAkB,gBAAgB,IAAI;wBAC/C,iBAAiBK;oBACnB;gBACF;YACF;QACF;QAEA,MAAMG,UAAUJ,iBAAiB;QACjC,MAAMK,mBAAmBT,kBAAkB,gBAAgB,IAAI;QAC/D,MAAMU,kBAAkBF,QAAQ,OAAO,CAAC,WAAW;QACnD,MAAMG,2BAA2BF,iBAAiB,OAAO,CAAC,WAAW;QAGrE,OAAO;YACL,SAASC;YACT,YAAYZ;YACZ,kBAAkBa;YAClBT;YACA,cAAcA,UAAU,MAAM,GAAG,KAAK,CAACC,WAAW,eAAe;QACnE;IACF;IAQA,gCACES,yBAAiE,EAC5B;QACrC,OAAO;YACL,MAAM;YACN,SAASA,0BAA0B,UAAU,IAAIA,0BAA0B,OAAO;QACpF;IACF;IAQA,sCACEC,eAA2C,EACb;QAC9B,OAAOA,gBAAgB,GAAG,CAAC,CAACC;YAE1B,MAAMC,cAAcD,OAAO,OAAO,CAC/B,MAAM,CAAC,CAACE,OAASA,AAAc,WAAdA,KAAK,IAAI,EAC1B,GAAG,CAAC,CAACA,OAAUA,KAA0B,IAAI,EAC7C,IAAI,CAAC;YAER,OAAO;gBACL,MAAM;gBACN,SAAS,CAAC,MAAM,EAAEF,OAAO,QAAQ,CAAC,WAAW,EAAEC,aAAa;YAC9D;QACF;IACF;IAzNA,YAAYE,kBAAuC,CAAE;QACnD,KAAK,IAHP,uBAAQ,sBAAR;QAIE,IAAI,CAAC,kBAAkB,GAAGA;IAC5B;AAuNF"}

package/dist/utils.d.ts

@@ -1,5 +1,5 @@
 import { BaseAction } from '@ui-tars-test/shared/types';
-import { MultimodalToolCallResult } from '@tarko/agent';
+import { MultimodalToolCallResult } from '@ui-tars-test/tarko-agent';
 export declare const createGUIErrorResponse: (action: BaseAction, errorMessage: string) => MultimodalToolCallResult;
 export declare const convertToAgentUIAction: (action: BaseAction) => {
     action: any;

package/dist/utils.d.ts.map

@@ -1 +1 @@
-{"version":3,"file":"utils.d.ts","sourceRoot":"","sources":["../src/utils.ts"],"names":[],"mappings":"AAMA,OAAO,EAAE,UAAU,EAAE,MAAM,4BAA4B,CAAC;AACxD,OAAO,EAAE,wBAAwB,EAAE,MAAM,cAAc,CAAC;AAExD,eAAO,MAAM,sBAAsB,GACjC,QAAQ,UAAU,EAClB,cAAc,MAAM,KACnB,wBAcF,CAAC;AAEF,eAAO,MAAM,sBAAsB,GAAI,QAAQ,UAAU;;;;;CAUxD,CAAC"}
+{"version":3,"file":"utils.d.ts","sourceRoot":"","sources":["../src/utils.ts"],"names":[],"mappings":"AAMA,OAAO,EAAE,UAAU,EAAE,MAAM,4BAA4B,CAAC;AACxD,OAAO,EAAE,wBAAwB,EAAE,MAAM,2BAA2B,CAAC;AAErE,eAAO,MAAM,sBAAsB,GACjC,QAAQ,UAAU,EAClB,cAAc,MAAM,KACnB,wBAcF,CAAC;AAEF,eAAO,MAAM,sBAAsB,GAAI,QAAQ,UAAU;;;;;CAUxD,CAAC"}

package/dist/utils.js.map

@@ -1 +1 @@
-{"version":3,"file":"utils.js","sources":["webpack://@ui-tars-test/agent-sdk/webpack/runtime/define_property_getters","webpack://@ui-tars-test/agent-sdk/webpack/runtime/has_own_property","webpack://@ui-tars-test/agent-sdk/webpack/runtime/make_namespace_object","webpack://@ui-tars-test/agent-sdk/./src/utils.ts"],"sourcesContent":["__webpack_require__.d = (exports, definition) => {\n\tfor(var key in definition) {\n        if(__webpack_require__.o(definition, key) && !__webpack_require__.o(exports, key)) {\n            Object.defineProperty(exports, key, { enumerable: true, get: definition[key] });\n        }\n    }\n};","__webpack_require__.o = (obj, prop) => (Object.prototype.hasOwnProperty.call(obj, prop))","// define __esModule on exports\n__webpack_require__.r = (exports) => {\n\tif(typeof Symbol !== 'undefined' && Symbol.toStringTag) {\n\t\tObject.defineProperty(exports, Symbol.toStringTag, { value: 'Module' });\n\t}\n\tObject.defineProperty(exports, '__esModule', { value: true });\n};","/* eslint-disable @typescript-eslint/no-explicit-any */\n/*\n * Copyright (c) 2025 Bytedance, Inc. and its affiliates.\n * SPDX-License-Identifier: Apache-2.0\n */\n\nimport { BaseAction } from '@ui-tars-test/shared/types';\nimport { MultimodalToolCallResult } from '@tarko/agent';\n\nexport const createGUIErrorResponse = (\n  action: BaseAction,\n  errorMessage: string,\n): MultimodalToolCallResult => {\n  return {\n    content: [\n      {\n        type: 'text',\n        text: JSON.stringify({\n          action: action,\n          error: errorMessage,\n        }),\n      },\n    ],\n    // @ts-ignore: isError is not in MultimodalToolCallResult type but might be used by internal logic\n    isError: true,\n  };\n};\n\nexport const convertToAgentUIAction = (action: BaseAction) => {\n  // Use type assertion or access properties safely as BaseAction might be generic\n  // or properties might be on specific action types\n  const anyAction = action as any;\n  return {\n    action: anyAction.prediction,\n    thought: anyAction.thought,\n    action_type: anyAction.action_type,\n    action_input: anyAction.action_input,\n  };\n};\n"],"names":["__webpack_require__","definition","key","Object","obj","prop","Symbol","createGUIErrorResponse","action","errorMessage","JSON","convertToAgentUIAction","anyAction"],"mappings":";;;;;;;IAAAA,oBAAoB,CAAC,GAAG,CAAC,UAASC;QACjC,IAAI,IAAIC,OAAOD,WACR,IAAGD,oBAAoB,CAAC,CAACC,YAAYC,QAAQ,CAACF,oBAAoB,CAAC,CAAC,UAASE,MACzEC,OAAO,cAAc,CAAC,UAASD,KAAK;YAAE,YAAY;YAAM,KAAKD,UAAU,CAACC,IAAI;QAAC;IAGzF;;;ICNAF,oBAAoB,CAAC,GAAG,CAACI,KAAKC,OAAUF,OAAO,SAAS,CAAC,cAAc,CAAC,IAAI,CAACC,KAAKC;;;ICClFL,oBAAoB,CAAC,GAAG,CAAC;QACxB,IAAG,AAAkB,eAAlB,OAAOM,UAA0BA,OAAO,WAAW,EACrDH,OAAO,cAAc,CAAC,UAASG,OAAO,WAAW,EAAE;YAAE,OAAO;QAAS;QAEtEH,OAAO,cAAc,CAAC,UAAS,cAAc;YAAE,OAAO;QAAK;IAC5D;;;;;;;;ACGO,MAAMI,yBAAyB,CACpCC,QACAC,eAEO;QACL,SAAS;YACP;gBACE,MAAM;gBACN,MAAMC,KAAK,SAAS,CAAC;oBACnB,QAAQF;oBACR,OAAOC;gBACT;YACF;SACD;QAED,SAAS;IACX;AAGK,MAAME,yBAAyB,CAACH;IAGrC,MAAMI,YAAYJ;IAClB,OAAO;QACL,QAAQI,UAAU,UAAU;QAC5B,SAASA,UAAU,OAAO;QAC1B,aAAaA,UAAU,WAAW;QAClC,cAAcA,UAAU,YAAY;IACtC;AACF"}
+{"version":3,"file":"utils.js","sources":["webpack://@ui-tars-test/agent-sdk/webpack/runtime/define_property_getters","webpack://@ui-tars-test/agent-sdk/webpack/runtime/has_own_property","webpack://@ui-tars-test/agent-sdk/webpack/runtime/make_namespace_object","webpack://@ui-tars-test/agent-sdk/./src/utils.ts"],"sourcesContent":["__webpack_require__.d = (exports, definition) => {\n\tfor(var key in definition) {\n        if(__webpack_require__.o(definition, key) && !__webpack_require__.o(exports, key)) {\n            Object.defineProperty(exports, key, { enumerable: true, get: definition[key] });\n        }\n    }\n};","__webpack_require__.o = (obj, prop) => (Object.prototype.hasOwnProperty.call(obj, prop))","// define __esModule on exports\n__webpack_require__.r = (exports) => {\n\tif(typeof Symbol !== 'undefined' && Symbol.toStringTag) {\n\t\tObject.defineProperty(exports, Symbol.toStringTag, { value: 'Module' });\n\t}\n\tObject.defineProperty(exports, '__esModule', { value: true });\n};","/* eslint-disable @typescript-eslint/no-explicit-any */\n/*\n * Copyright (c) 2025 Bytedance, Inc. and its affiliates.\n * SPDX-License-Identifier: Apache-2.0\n */\n\nimport { BaseAction } from '@ui-tars-test/shared/types';\nimport { MultimodalToolCallResult } from '@ui-tars-test/tarko-agent';\n\nexport const createGUIErrorResponse = (\n  action: BaseAction,\n  errorMessage: string,\n): MultimodalToolCallResult => {\n  return {\n    content: [\n      {\n        type: 'text',\n        text: JSON.stringify({\n          action: action,\n          error: errorMessage,\n        }),\n      },\n    ],\n    // @ts-ignore: isError is not in MultimodalToolCallResult type but might be used by internal logic\n    isError: true,\n  };\n};\n\nexport const convertToAgentUIAction = (action: BaseAction) => {\n  // Use type assertion or access properties safely as BaseAction might be generic\n  // or properties might be on specific action types\n  const anyAction = action as any;\n  return {\n    action: anyAction.prediction,\n    thought: anyAction.thought,\n    action_type: anyAction.action_type,\n    action_input: anyAction.action_input,\n  };\n};\n"],"names":["__webpack_require__","definition","key","Object","obj","prop","Symbol","createGUIErrorResponse","action","errorMessage","JSON","convertToAgentUIAction","anyAction"],"mappings":";;;;;;;IAAAA,oBAAoB,CAAC,GAAG,CAAC,UAASC;QACjC,IAAI,IAAIC,OAAOD,WACR,IAAGD,oBAAoB,CAAC,CAACC,YAAYC,QAAQ,CAACF,oBAAoB,CAAC,CAAC,UAASE,MACzEC,OAAO,cAAc,CAAC,UAASD,KAAK;YAAE,YAAY;YAAM,KAAKD,UAAU,CAACC,IAAI;QAAC;IAGzF;;;ICNAF,oBAAoB,CAAC,GAAG,CAACI,KAAKC,OAAUF,OAAO,SAAS,CAAC,cAAc,CAAC,IAAI,CAACC,KAAKC;;;ICClFL,oBAAoB,CAAC,GAAG,CAAC;QACxB,IAAG,AAAkB,eAAlB,OAAOM,UAA0BA,OAAO,WAAW,EACrDH,OAAO,cAAc,CAAC,UAASG,OAAO,WAAW,EAAE;YAAE,OAAO;QAAS;QAEtEH,OAAO,cAAc,CAAC,UAAS,cAAc;YAAE,OAAO;QAAK;IAC5D;;;;;;;;ACGO,MAAMI,yBAAyB,CACpCC,QACAC,eAEO;QACL,SAAS;YACP;gBACE,MAAM;gBACN,MAAMC,KAAK,SAAS,CAAC;oBACnB,QAAQF;oBACR,OAAOC;gBACT;YACF;SACD;QAED,SAAS;IACX;AAGK,MAAME,yBAAyB,CAACH;IAGrC,MAAMI,YAAYJ;IAClB,OAAO;QACL,QAAQI,UAAU,UAAU;QAC5B,SAASA,UAAU,OAAO;QAC1B,aAAaA,UAAU,WAAW;QAClC,cAAcA,UAAU,YAAY;IACtC;AACF"}

package/dist/utils.mjs.map

@@ -1 +1 @@
-{"version":3,"file":"utils.mjs","sources":["webpack://@ui-tars-test/agent-sdk/./src/utils.ts"],"sourcesContent":["/* eslint-disable @typescript-eslint/no-explicit-any */\n/*\n * Copyright (c) 2025 Bytedance, Inc. and its affiliates.\n * SPDX-License-Identifier: Apache-2.0\n */\n\nimport { BaseAction } from '@ui-tars-test/shared/types';\nimport { MultimodalToolCallResult } from '@tarko/agent';\n\nexport const createGUIErrorResponse = (\n  action: BaseAction,\n  errorMessage: string,\n): MultimodalToolCallResult => {\n  return {\n    content: [\n      {\n        type: 'text',\n        text: JSON.stringify({\n          action: action,\n          error: errorMessage,\n        }),\n      },\n    ],\n    // @ts-ignore: isError is not in MultimodalToolCallResult type but might be used by internal logic\n    isError: true,\n  };\n};\n\nexport const convertToAgentUIAction = (action: BaseAction) => {\n  // Use type assertion or access properties safely as BaseAction might be generic\n  // or properties might be on specific action types\n  const anyAction = action as any;\n  return {\n    action: anyAction.prediction,\n    thought: anyAction.thought,\n    action_type: anyAction.action_type,\n    action_input: anyAction.action_input,\n  };\n};\n"],"names":["createGUIErrorResponse","action","errorMessage","JSON","convertToAgentUIAction","anyAction"],"mappings":";;;;AASO,MAAMA,yBAAyB,CACpCC,QACAC,eAEO;QACL,SAAS;YACP;gBACE,MAAM;gBACN,MAAMC,KAAK,SAAS,CAAC;oBACnB,QAAQF;oBACR,OAAOC;gBACT;YACF;SACD;QAED,SAAS;IACX;AAGK,MAAME,yBAAyB,CAACH;IAGrC,MAAMI,YAAYJ;IAClB,OAAO;QACL,QAAQI,UAAU,UAAU;QAC5B,SAASA,UAAU,OAAO;QAC1B,aAAaA,UAAU,WAAW;QAClC,cAAcA,UAAU,YAAY;IACtC;AACF"}
+{"version":3,"file":"utils.mjs","sources":["webpack://@ui-tars-test/agent-sdk/./src/utils.ts"],"sourcesContent":["/* eslint-disable @typescript-eslint/no-explicit-any */\n/*\n * Copyright (c) 2025 Bytedance, Inc. and its affiliates.\n * SPDX-License-Identifier: Apache-2.0\n */\n\nimport { BaseAction } from '@ui-tars-test/shared/types';\nimport { MultimodalToolCallResult } from '@ui-tars-test/tarko-agent';\n\nexport const createGUIErrorResponse = (\n  action: BaseAction,\n  errorMessage: string,\n): MultimodalToolCallResult => {\n  return {\n    content: [\n      {\n        type: 'text',\n        text: JSON.stringify({\n          action: action,\n          error: errorMessage,\n        }),\n      },\n    ],\n    // @ts-ignore: isError is not in MultimodalToolCallResult type but might be used by internal logic\n    isError: true,\n  };\n};\n\nexport const convertToAgentUIAction = (action: BaseAction) => {\n  // Use type assertion or access properties safely as BaseAction might be generic\n  // or properties might be on specific action types\n  const anyAction = action as any;\n  return {\n    action: anyAction.prediction,\n    thought: anyAction.thought,\n    action_type: anyAction.action_type,\n    action_input: anyAction.action_input,\n  };\n};\n"],"names":["createGUIErrorResponse","action","errorMessage","JSON","convertToAgentUIAction","anyAction"],"mappings":";;;;AASO,MAAMA,yBAAyB,CACpCC,QACAC,eAEO;QACL,SAAS;YACP;gBACE,MAAM;gBACN,MAAMC,KAAK,SAAS,CAAC;oBACnB,QAAQF;oBACR,OAAOC;gBACT;YACF;SACD;QAED,SAAS;IACX;AAGK,MAAME,yBAAyB,CAACH;IAGrC,MAAMI,YAAYJ;IAClB,OAAO;QACL,QAAQI,UAAU,UAAU;QAC5B,SAASA,UAAU,OAAO;QAC1B,aAAaA,UAAU,WAAW;QAClC,cAAcA,UAAU,YAAY;IACtC;AACF"}

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@ui-tars-test/agent-sdk",
-  "version": "0.3.9",
+  "version": "0.3.12",
   "description": "GUI Agent",
   "main": "dist/index.js",
   "module": "dist/index.mjs",
@@ -39,8 +39,9 @@
     "@ui-tars-test/operator-adb": "^0.3.0",
     "@ui-tars-test/operator-browser": "^0.3.0",
     "@ui-tars-test/operator-nutjs": "^0.3.0",
-    "@ui-tars-test/shared": "^0.3.6",
+    "@ui-tars-test/shared": "^0.3.7",
     "@ui-tars-test/tarko-agent": "^0.3.1",
+    "@ui-tars-test/tarko-agent-interface": "^0.3.1",
     "big.js": "^6.2.2",
     "commander": "^14.0.0",
     "dotenv": "^17.2.2",