@ui-tars-test/agent-sdk 0.3.0

package/dist/ToolCallEngine.js

@@ -0,0 +1,190 @@
+/**
+* Copyright (c) 2025 Bytedance, Inc. and its affiliates.
+* SPDX-License-Identifier: Apache-2.0
+*/
+"use strict";
+var __webpack_require__ = {};
+(()=>{
+    __webpack_require__.d = (exports1, definition)=>{
+        for(var key in definition)if (__webpack_require__.o(definition, key) && !__webpack_require__.o(exports1, key)) Object.defineProperty(exports1, key, {
+            enumerable: true,
+            get: definition[key]
+        });
+    };
+})();
+(()=>{
+    __webpack_require__.o = (obj, prop)=>Object.prototype.hasOwnProperty.call(obj, prop);
+})();
+(()=>{
+    __webpack_require__.r = (exports1)=>{
+        if ('undefined' != typeof Symbol && Symbol.toStringTag) Object.defineProperty(exports1, Symbol.toStringTag, {
+            value: 'Module'
+        });
+        Object.defineProperty(exports1, '__esModule', {
+            value: true
+        });
+    };
+})();
+var __webpack_exports__ = {};
+__webpack_require__.r(__webpack_exports__);
+__webpack_require__.d(__webpack_exports__, {
+    GUIAgentToolCallEngine: ()=>GUIAgentToolCallEngine
+});
+const agent_interface_namespaceObject = require("@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");
+const utils_namespaceObject = require("@ui-tars-test/shared/utils");
+function _define_property(obj, key, value) {
+    if (key in obj) Object.defineProperty(obj, key, {
+        value: value,
+        enumerable: true,
+        configurable: true,
+        writable: true
+    });
+    else obj[key] = value;
+    return obj;
+}
+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 {
+    preparePrompt(instructions, tools) {
+        return instructions;
+    }
+    prepareRequest(context) {
+        var _context_messages_find;
+        defaultLogger.log("\u3010New Sys Prompt'\u3011 System Prompt:", JSON.stringify((null == (_context_messages_find = context.messages.find((m)=>'system' === m.role)) ? void 0 : _context_messages_find.content) || ''));
+        return {
+            model: context.model,
+            messages: context.messages,
+            temperature: context.temperature || 0.7,
+            stream: true
+        };
+    }
+    initStreamProcessingState() {
+        return {
+            contentBuffer: '',
+            toolCalls: [],
+            reasoningBuffer: '',
+            finishReason: null
+        };
+    }
+    processStreamingChunk(chunk, state) {
+        var _chunk_choices_, _chunk_choices_1;
+        const delta = null == (_chunk_choices_ = chunk.choices[0]) ? void 0 : _chunk_choices_.delta;
+        if (null == delta ? void 0 : delta.content) state.contentBuffer += delta.content;
+        if (null == (_chunk_choices_1 = chunk.choices[0]) ? void 0 : _chunk_choices_1.finish_reason) state.finishReason = chunk.choices[0].finish_reason;
+        return {
+            content: (null == delta ? void 0 : delta.content) || '',
+            reasoningContent: '',
+            hasToolCallUpdate: false,
+            toolCalls: []
+        };
+    }
+    generateToolCallId() {
+        return `call_${Date.now()}_${Math.random().toString(36).substring(2, 11)}`;
+    }
+    finalizeStreamProcessing(state) {
+        const fullContent = state.contentBuffer;
+        defaultLogger.log("\u3010New Sys Prompt'\u3011 Model Response:", fullContent);
+        defaultLogger.log('[finalizeStreamProcessing] fullContent', fullContent);
+        console.log('[CLI DEBUG] [ToolCallEngine] Full model response received:', fullContent);
+        let parsedGUIResponse = null;
+        if (this.customActionParser) {
+            parsedGUIResponse = this.customActionParser(fullContent);
+            defaultLogger.log('[finalizeStreamProcessing] Using custom action parser');
+            console.log('[CLI DEBUG] [ToolCallEngine] Using custom action parser');
+        }
+        if (!parsedGUIResponse) {
+            console.log('[CLI DEBUG] [ToolCallEngine] Using default action parser (XML parser)');
+            parsedGUIResponse = defaultParser.parsePrediction(fullContent);
+            defaultLogger.log('[finalizeStreamProcessing] Using default action parser');
+        }
+        if (parsedGUIResponse) console.log('[CLI DEBUG] [ToolCallEngine] Parsed response:', JSON.stringify(parsedGUIResponse, null, 2));
+        else console.log('[CLI DEBUG] [ToolCallEngine] Parsing failed or returned null');
+        if (!parsedGUIResponse || parsedGUIResponse.errorMessage) return {
+            content: '',
+            rawContent: fullContent,
+            toolCalls: [
+                {
+                    id: this.generateToolCallId(),
+                    type: 'function',
+                    function: {
+                        name: external_constants_js_namespaceObject.GUI_ADAPTED_TOOL_NAME,
+                        arguments: JSON.stringify({
+                            action: '',
+                            step: '',
+                            thought: '',
+                            operator_action: null,
+                            errorMessage: (null == parsedGUIResponse ? void 0 : parsedGUIResponse.errorMessage) ?? 'Failed to parse GUI Action from output'
+                        })
+                    }
+                }
+            ],
+            finishReason: 'tool_calls'
+        };
+        const toolCalls = [];
+        let finished = false;
+        let finishMessage = null;
+        for (const action of parsedGUIResponse.actions){
+            if ('finished' === action.type) {
+                var _action_inputs;
+                finished = true;
+                finishMessage = (null == (_action_inputs = action.inputs) ? void 0 : _action_inputs.content) ?? null;
+                continue;
+            }
+            toolCalls.push({
+                id: this.generateToolCallId(),
+                type: 'function',
+                function: {
+                    name: external_constants_js_namespaceObject.GUI_ADAPTED_TOOL_NAME,
+                    arguments: JSON.stringify({
+                        action: (0, utils_namespaceObject.serializeAction)(action),
+                        step: '',
+                        thought: parsedGUIResponse.reasoningContent ?? '',
+                        operator_action: action
+                    })
+                }
+            });
+        }
+        const content = finishMessage ?? '';
+        const reasoningContent = parsedGUIResponse.reasoningContent ?? '';
+        const contentForWebUI = content.replace(/\\n|\n/g, '<br>');
+        const reasoningContentForWebUI = reasoningContent.replace(/\\n|\n/g, '<br>');
+        return {
+            content: contentForWebUI,
+            rawContent: fullContent,
+            reasoningContent: reasoningContentForWebUI,
+            toolCalls,
+            finishReason: toolCalls.length > 0 && !finished ? 'tool_calls' : 'stop'
+        };
+    }
+    buildHistoricalAssistantMessage(currentLoopAssistantEvent) {
+        return {
+            role: 'assistant',
+            content: currentLoopAssistantEvent.rawContent || currentLoopAssistantEvent.content
+        };
+    }
+    buildHistoricalToolCallResultMessages(toolCallResults) {
+        return toolCallResults.map((result)=>{
+            const textContent = result.content.filter((part)=>'text' === part.type).map((part)=>part.text).join('');
+            return {
+                role: 'user',
+                content: `Tool "${result.toolName}" result:\n${textContent}`
+            };
+        });
+    }
+    constructor(customActionParser){
+        super(), _define_property(this, "customActionParser", void 0);
+        this.customActionParser = customActionParser;
+    }
+}
+exports.GUIAgentToolCallEngine = __webpack_exports__.GUIAgentToolCallEngine;
+for(var __webpack_i__ in __webpack_exports__)if (-1 === [
+    "GUIAgentToolCallEngine"
+].indexOf(__webpack_i__)) exports[__webpack_i__] = __webpack_exports__[__webpack_i__];
+Object.defineProperty(exports, '__esModule', {
+    value: true
+});
+
+//# sourceMappingURL=ToolCallEngine.js.map

package/dist/ToolCallEngine.js.map

@@ -0,0 +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"}

package/dist/ToolCallEngine.mjs

@@ -0,0 +1,156 @@
+/**
+* Copyright (c) 2025 Bytedance, Inc. and its affiliates.
+* SPDX-License-Identifier: Apache-2.0
+*/
+import { ToolCallEngine } from "@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";
+import { serializeAction } from "@ui-tars-test/shared/utils";
+function _define_property(obj, key, value) {
+    if (key in obj) Object.defineProperty(obj, key, {
+        value: value,
+        enumerable: true,
+        configurable: true,
+        writable: true
+    });
+    else obj[key] = value;
+    return obj;
+}
+const defaultParser = new DefaultActionParser();
+const defaultLogger = new ConsoleLogger('[GUIAgent:ToolCallEngine]', LogLevel.DEBUG);
+class GUIAgentToolCallEngine extends ToolCallEngine {
+    preparePrompt(instructions, tools) {
+        return instructions;
+    }
+    prepareRequest(context) {
+        var _context_messages_find;
+        defaultLogger.log("\u3010New Sys Prompt'\u3011 System Prompt:", JSON.stringify((null == (_context_messages_find = context.messages.find((m)=>'system' === m.role)) ? void 0 : _context_messages_find.content) || ''));
+        return {
+            model: context.model,
+            messages: context.messages,
+            temperature: context.temperature || 0.7,
+            stream: true
+        };
+    }
+    initStreamProcessingState() {
+        return {
+            contentBuffer: '',
+            toolCalls: [],
+            reasoningBuffer: '',
+            finishReason: null
+        };
+    }
+    processStreamingChunk(chunk, state) {
+        var _chunk_choices_, _chunk_choices_1;
+        const delta = null == (_chunk_choices_ = chunk.choices[0]) ? void 0 : _chunk_choices_.delta;
+        if (null == delta ? void 0 : delta.content) state.contentBuffer += delta.content;
+        if (null == (_chunk_choices_1 = chunk.choices[0]) ? void 0 : _chunk_choices_1.finish_reason) state.finishReason = chunk.choices[0].finish_reason;
+        return {
+            content: (null == delta ? void 0 : delta.content) || '',
+            reasoningContent: '',
+            hasToolCallUpdate: false,
+            toolCalls: []
+        };
+    }
+    generateToolCallId() {
+        return `call_${Date.now()}_${Math.random().toString(36).substring(2, 11)}`;
+    }
+    finalizeStreamProcessing(state) {
+        const fullContent = state.contentBuffer;
+        defaultLogger.log("\u3010New Sys Prompt'\u3011 Model Response:", fullContent);
+        defaultLogger.log('[finalizeStreamProcessing] fullContent', fullContent);
+        console.log('[CLI DEBUG] [ToolCallEngine] Full model response received:', fullContent);
+        let parsedGUIResponse = null;
+        if (this.customActionParser) {
+            parsedGUIResponse = this.customActionParser(fullContent);
+            defaultLogger.log('[finalizeStreamProcessing] Using custom action parser');
+            console.log('[CLI DEBUG] [ToolCallEngine] Using custom action parser');
+        }
+        if (!parsedGUIResponse) {
+            console.log('[CLI DEBUG] [ToolCallEngine] Using default action parser (XML parser)');
+            parsedGUIResponse = defaultParser.parsePrediction(fullContent);
+            defaultLogger.log('[finalizeStreamProcessing] Using default action parser');
+        }
+        if (parsedGUIResponse) console.log('[CLI DEBUG] [ToolCallEngine] Parsed response:', JSON.stringify(parsedGUIResponse, null, 2));
+        else console.log('[CLI DEBUG] [ToolCallEngine] Parsing failed or returned null');
+        if (!parsedGUIResponse || parsedGUIResponse.errorMessage) return {
+            content: '',
+            rawContent: fullContent,
+            toolCalls: [
+                {
+                    id: this.generateToolCallId(),
+                    type: 'function',
+                    function: {
+                        name: GUI_ADAPTED_TOOL_NAME,
+                        arguments: JSON.stringify({
+                            action: '',
+                            step: '',
+                            thought: '',
+                            operator_action: null,
+                            errorMessage: (null == parsedGUIResponse ? void 0 : parsedGUIResponse.errorMessage) ?? 'Failed to parse GUI Action from output'
+                        })
+                    }
+                }
+            ],
+            finishReason: 'tool_calls'
+        };
+        const toolCalls = [];
+        let finished = false;
+        let finishMessage = null;
+        for (const action of parsedGUIResponse.actions){
+            if ('finished' === action.type) {
+                var _action_inputs;
+                finished = true;
+                finishMessage = (null == (_action_inputs = action.inputs) ? void 0 : _action_inputs.content) ?? null;
+                continue;
+            }
+            toolCalls.push({
+                id: this.generateToolCallId(),
+                type: 'function',
+                function: {
+                    name: GUI_ADAPTED_TOOL_NAME,
+                    arguments: JSON.stringify({
+                        action: serializeAction(action),
+                        step: '',
+                        thought: parsedGUIResponse.reasoningContent ?? '',
+                        operator_action: action
+                    })
+                }
+            });
+        }
+        const content = finishMessage ?? '';
+        const reasoningContent = parsedGUIResponse.reasoningContent ?? '';
+        const contentForWebUI = content.replace(/\\n|\n/g, '<br>');
+        const reasoningContentForWebUI = reasoningContent.replace(/\\n|\n/g, '<br>');
+        return {
+            content: contentForWebUI,
+            rawContent: fullContent,
+            reasoningContent: reasoningContentForWebUI,
+            toolCalls,
+            finishReason: toolCalls.length > 0 && !finished ? 'tool_calls' : 'stop'
+        };
+    }
+    buildHistoricalAssistantMessage(currentLoopAssistantEvent) {
+        return {
+            role: 'assistant',
+            content: currentLoopAssistantEvent.rawContent || currentLoopAssistantEvent.content
+        };
+    }
+    buildHistoricalToolCallResultMessages(toolCallResults) {
+        return toolCallResults.map((result)=>{
+            const textContent = result.content.filter((part)=>'text' === part.type).map((part)=>part.text).join('');
+            return {
+                role: 'user',
+                content: `Tool "${result.toolName}" result:\n${textContent}`
+            };
+        });
+    }
+    constructor(customActionParser){
+        super(), _define_property(this, "customActionParser", void 0);
+        this.customActionParser = customActionParser;
+    }
+}
+export { GUIAgentToolCallEngine };
+
+//# sourceMappingURL=ToolCallEngine.mjs.map

package/dist/ToolCallEngine.mjs.map

@@ -0,0 +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"}

package/dist/constants.d.ts

@@ -0,0 +1,6 @@
+/**
+ * Copyright (c) 2025 Bytedance, Inc. and its affiliates.
+ * SPDX-License-Identifier: Apache-2.0
+ */
+export declare const GUI_ADAPTED_TOOL_NAME = "browser_vision_control";
+//# sourceMappingURL=constants.d.ts.map

package/dist/constants.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"constants.d.ts","sourceRoot":"","sources":["../src/constants.ts"],"names":[],"mappings":"AAAA;;;GAGG;AACH,eAAO,MAAM,qBAAqB,2BAA2B,CAAC"}

package/dist/constants.js

@@ -0,0 +1,42 @@
+/**
+* Copyright (c) 2025 Bytedance, Inc. and its affiliates.
+* SPDX-License-Identifier: Apache-2.0
+*/
+"use strict";
+var __webpack_require__ = {};
+(()=>{
+    __webpack_require__.d = (exports1, definition)=>{
+        for(var key in definition)if (__webpack_require__.o(definition, key) && !__webpack_require__.o(exports1, key)) Object.defineProperty(exports1, key, {
+            enumerable: true,
+            get: definition[key]
+        });
+    };
+})();
+(()=>{
+    __webpack_require__.o = (obj, prop)=>Object.prototype.hasOwnProperty.call(obj, prop);
+})();
+(()=>{
+    __webpack_require__.r = (exports1)=>{
+        if ('undefined' != typeof Symbol && Symbol.toStringTag) Object.defineProperty(exports1, Symbol.toStringTag, {
+            value: 'Module'
+        });
+        Object.defineProperty(exports1, '__esModule', {
+            value: true
+        });
+    };
+})();
+var __webpack_exports__ = {};
+__webpack_require__.r(__webpack_exports__);
+__webpack_require__.d(__webpack_exports__, {
+    GUI_ADAPTED_TOOL_NAME: ()=>GUI_ADAPTED_TOOL_NAME
+});
+const GUI_ADAPTED_TOOL_NAME = 'browser_vision_control';
+exports.GUI_ADAPTED_TOOL_NAME = __webpack_exports__.GUI_ADAPTED_TOOL_NAME;
+for(var __webpack_i__ in __webpack_exports__)if (-1 === [
+    "GUI_ADAPTED_TOOL_NAME"
+].indexOf(__webpack_i__)) exports[__webpack_i__] = __webpack_exports__[__webpack_i__];
+Object.defineProperty(exports, '__esModule', {
+    value: true
+});
+
+//# sourceMappingURL=constants.js.map

package/dist/constants.js.map

@@ -0,0 +1 @@
+{"version":3,"file":"constants.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/constants.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 */\nexport const GUI_ADAPTED_TOOL_NAME = 'browser_vision_control';\n"],"names":["__webpack_require__","definition","key","Object","obj","prop","Symbol","GUI_ADAPTED_TOOL_NAME"],"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;;;;;;;ACFO,MAAMI,wBAAwB"}

package/dist/constants.mjs

@@ -0,0 +1,8 @@
+/**
+* Copyright (c) 2025 Bytedance, Inc. and its affiliates.
+* SPDX-License-Identifier: Apache-2.0
+*/
+const GUI_ADAPTED_TOOL_NAME = 'browser_vision_control';
+export { GUI_ADAPTED_TOOL_NAME };
+
+//# sourceMappingURL=constants.mjs.map

package/dist/constants.mjs.map

@@ -0,0 +1 @@
+{"version":3,"file":"constants.mjs","sources":["webpack://@ui-tars-test/agent-sdk/./src/constants.ts"],"sourcesContent":["/**\n * Copyright (c) 2025 Bytedance, Inc. and its affiliates.\n * SPDX-License-Identifier: Apache-2.0\n */\nexport const GUI_ADAPTED_TOOL_NAME = 'browser_vision_control';\n"],"names":["GUI_ADAPTED_TOOL_NAME"],"mappings":";;;;AAIO,MAAMA,wBAAwB"}

package/dist/defaultImpls.d.ts

@@ -0,0 +1,15 @@
+import { ImageDetailCalculator, NormalizeCoordinates } from '@ui-tars-test/shared/types';
+/**
+ * Default coordinate normalization function
+ * Normalizes raw coordinates by dividing by 1000 (simple scaling)
+ * @param rawCoords - The raw coordinates to normalize
+ * @returns Object containing normalized coordinates
+ */
+export declare const defaultNormalizeCoords: NormalizeCoordinates;
+/**
+ * Default implementation for detail calculation based on pixel count
+ * detail:low mode: 1,048,576 px (1024×1024)
+ * detail:high mode: 4,014,080 px (2048×1960)
+ */
+export declare const defaultDetailCalculator: ImageDetailCalculator;
+//# sourceMappingURL=defaultImpls.d.ts.map

package/dist/defaultImpls.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"defaultImpls.d.ts","sourceRoot":"","sources":["../src/defaultImpls.ts"],"names":[],"mappings":"AAIA,OAAO,EAAe,qBAAqB,EAAE,oBAAoB,EAAE,MAAM,4BAA4B,CAAC;AAEtG;;;;;GAKG;AACH,eAAO,MAAM,sBAAsB,EAAE,oBAYpC,CAAC;AAEF;;;;GAIG;AACH,eAAO,MAAM,uBAAuB,EAAE,qBAgBrC,CAAC"}

package/dist/defaultImpls.js

@@ -0,0 +1,67 @@
+/**
+* Copyright (c) 2025 Bytedance, Inc. and its affiliates.
+* SPDX-License-Identifier: Apache-2.0
+*/
+"use strict";
+var __webpack_require__ = {};
+(()=>{
+    __webpack_require__.d = (exports1, definition)=>{
+        for(var key in definition)if (__webpack_require__.o(definition, key) && !__webpack_require__.o(exports1, key)) Object.defineProperty(exports1, key, {
+            enumerable: true,
+            get: definition[key]
+        });
+    };
+})();
+(()=>{
+    __webpack_require__.o = (obj, prop)=>Object.prototype.hasOwnProperty.call(obj, prop);
+})();
+(()=>{
+    __webpack_require__.r = (exports1)=>{
+        if ('undefined' != typeof Symbol && Symbol.toStringTag) Object.defineProperty(exports1, Symbol.toStringTag, {
+            value: 'Module'
+        });
+        Object.defineProperty(exports1, '__esModule', {
+            value: true
+        });
+    };
+})();
+var __webpack_exports__ = {};
+__webpack_require__.r(__webpack_exports__);
+__webpack_require__.d(__webpack_exports__, {
+    defaultDetailCalculator: ()=>defaultDetailCalculator,
+    defaultNormalizeCoords: ()=>defaultNormalizeCoords
+});
+const defaultNormalizeCoords = (rawCoords)=>{
+    if (!rawCoords.raw) return {
+        normalized: rawCoords
+    };
+    const normalizedCoords = {
+        ...rawCoords,
+        normalized: {
+            x: rawCoords.raw.x / 1000,
+            y: rawCoords.raw.y / 1000
+        }
+    };
+    return {
+        normalized: normalizedCoords
+    };
+};
+const defaultDetailCalculator = (width, height)=>{
+    const LOW_DETAIL_THRESHOLD = 1048576;
+    const HIGH_DETAIL_THRESHOLD = 4014080;
+    const pixelCount = width * height;
+    if (pixelCount <= LOW_DETAIL_THRESHOLD) return 'low';
+    if (pixelCount <= HIGH_DETAIL_THRESHOLD) return 'high';
+    return 'auto';
+};
+exports.defaultDetailCalculator = __webpack_exports__.defaultDetailCalculator;
+exports.defaultNormalizeCoords = __webpack_exports__.defaultNormalizeCoords;
+for(var __webpack_i__ in __webpack_exports__)if (-1 === [
+    "defaultDetailCalculator",
+    "defaultNormalizeCoords"
+].indexOf(__webpack_i__)) exports[__webpack_i__] = __webpack_exports__[__webpack_i__];
+Object.defineProperty(exports, '__esModule', {
+    value: true
+});
+
+//# sourceMappingURL=defaultImpls.js.map

package/dist/defaultImpls.js.map

@@ -0,0 +1 @@
+{"version":3,"file":"defaultImpls.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/defaultImpls.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 { Coordinates, ImageDetailCalculator, NormalizeCoordinates } from '@ui-tars-test/shared/types';\n\n/**\n * Default coordinate normalization function\n * Normalizes raw coordinates by dividing by 1000 (simple scaling)\n * @param rawCoords - The raw coordinates to normalize\n * @returns Object containing normalized coordinates\n */\nexport const defaultNormalizeCoords: NormalizeCoordinates = (rawCoords: Coordinates) => {\n  if (!rawCoords.raw) {\n    return { normalized: rawCoords };\n  }\n  const normalizedCoords = {\n    ...rawCoords,\n    normalized: {\n      x: rawCoords.raw.x / 1000,\n      y: rawCoords.raw.y / 1000,\n    },\n  };\n  return { normalized: normalizedCoords };\n};\n\n/**\n * Default implementation for detail calculation based on pixel count\n * detail:low mode: 1,048,576 px (1024×1024)\n * detail:high mode: 4,014,080 px (2048×1960)\n */\nexport const defaultDetailCalculator: ImageDetailCalculator = (\n  width: number,\n  height: number,\n): 'low' | 'high' | 'auto' => {\n  const LOW_DETAIL_THRESHOLD = 1024 * 1024; // 1,048,576 px\n  const HIGH_DETAIL_THRESHOLD = 2048 * 1960; // 4,014,080 px\n\n  const pixelCount = width * height;\n  if (pixelCount <= LOW_DETAIL_THRESHOLD) {\n    return 'low';\n  } else if (pixelCount <= HIGH_DETAIL_THRESHOLD) {\n    return 'high';\n  } else {\n    // For images larger than high detail threshold, use high detail\n    return 'auto';\n  }\n};\n"],"names":["__webpack_require__","definition","key","Object","obj","prop","Symbol","defaultNormalizeCoords","rawCoords","normalizedCoords","defaultDetailCalculator","width","height","LOW_DETAIL_THRESHOLD","HIGH_DETAIL_THRESHOLD","pixelCount"],"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;;;;;;;;ACMO,MAAMI,yBAA+C,CAACC;IAC3D,IAAI,CAACA,UAAU,GAAG,EAChB,OAAO;QAAE,YAAYA;IAAU;IAEjC,MAAMC,mBAAmB;QACvB,GAAGD,SAAS;QACZ,YAAY;YACV,GAAGA,UAAU,GAAG,CAAC,CAAC,GAAG;YACrB,GAAGA,UAAU,GAAG,CAAC,CAAC,GAAG;QACvB;IACF;IACA,OAAO;QAAE,YAAYC;IAAiB;AACxC;AAOO,MAAMC,0BAAiD,CAC5DC,OACAC;IAEA,MAAMC,uBAAuB;IAC7B,MAAMC,wBAAwB;IAE9B,MAAMC,aAAaJ,QAAQC;IAC3B,IAAIG,cAAcF,sBAChB,OAAO;IACF,IAAIE,cAAcD,uBACvB,OAAO;IAGP,OAAO;AAEX"}

package/dist/defaultImpls.mjs

@@ -0,0 +1,30 @@
+/**
+* Copyright (c) 2025 Bytedance, Inc. and its affiliates.
+* SPDX-License-Identifier: Apache-2.0
+*/
+const defaultNormalizeCoords = (rawCoords)=>{
+    if (!rawCoords.raw) return {
+        normalized: rawCoords
+    };
+    const normalizedCoords = {
+        ...rawCoords,
+        normalized: {
+            x: rawCoords.raw.x / 1000,
+            y: rawCoords.raw.y / 1000
+        }
+    };
+    return {
+        normalized: normalizedCoords
+    };
+};
+const defaultDetailCalculator = (width, height)=>{
+    const LOW_DETAIL_THRESHOLD = 1048576;
+    const HIGH_DETAIL_THRESHOLD = 4014080;
+    const pixelCount = width * height;
+    if (pixelCount <= LOW_DETAIL_THRESHOLD) return 'low';
+    if (pixelCount <= HIGH_DETAIL_THRESHOLD) return 'high';
+    return 'auto';
+};
+export { defaultDetailCalculator, defaultNormalizeCoords };
+
+//# sourceMappingURL=defaultImpls.mjs.map

package/dist/defaultImpls.mjs.map

@@ -0,0 +1 @@
+{"version":3,"file":"defaultImpls.mjs","sources":["webpack://@ui-tars-test/agent-sdk/./src/defaultImpls.ts"],"sourcesContent":["/*\n * Copyright (c) 2025 Bytedance, Inc. and its affiliates.\n * SPDX-License-Identifier: Apache-2.0\n */\nimport { Coordinates, ImageDetailCalculator, NormalizeCoordinates } from '@ui-tars-test/shared/types';\n\n/**\n * Default coordinate normalization function\n * Normalizes raw coordinates by dividing by 1000 (simple scaling)\n * @param rawCoords - The raw coordinates to normalize\n * @returns Object containing normalized coordinates\n */\nexport const defaultNormalizeCoords: NormalizeCoordinates = (rawCoords: Coordinates) => {\n  if (!rawCoords.raw) {\n    return { normalized: rawCoords };\n  }\n  const normalizedCoords = {\n    ...rawCoords,\n    normalized: {\n      x: rawCoords.raw.x / 1000,\n      y: rawCoords.raw.y / 1000,\n    },\n  };\n  return { normalized: normalizedCoords };\n};\n\n/**\n * Default implementation for detail calculation based on pixel count\n * detail:low mode: 1,048,576 px (1024×1024)\n * detail:high mode: 4,014,080 px (2048×1960)\n */\nexport const defaultDetailCalculator: ImageDetailCalculator = (\n  width: number,\n  height: number,\n): 'low' | 'high' | 'auto' => {\n  const LOW_DETAIL_THRESHOLD = 1024 * 1024; // 1,048,576 px\n  const HIGH_DETAIL_THRESHOLD = 2048 * 1960; // 4,014,080 px\n\n  const pixelCount = width * height;\n  if (pixelCount <= LOW_DETAIL_THRESHOLD) {\n    return 'low';\n  } else if (pixelCount <= HIGH_DETAIL_THRESHOLD) {\n    return 'high';\n  } else {\n    // For images larger than high detail threshold, use high detail\n    return 'auto';\n  }\n};\n"],"names":["defaultNormalizeCoords","rawCoords","normalizedCoords","defaultDetailCalculator","width","height","LOW_DETAIL_THRESHOLD","HIGH_DETAIL_THRESHOLD","pixelCount"],"mappings":";;;;AAYO,MAAMA,yBAA+C,CAACC;IAC3D,IAAI,CAACA,UAAU,GAAG,EAChB,OAAO;QAAE,YAAYA;IAAU;IAEjC,MAAMC,mBAAmB;QACvB,GAAGD,SAAS;QACZ,YAAY;YACV,GAAGA,UAAU,GAAG,CAAC,CAAC,GAAG;YACrB,GAAGA,UAAU,GAAG,CAAC,CAAC,GAAG;QACvB;IACF;IACA,OAAO;QAAE,YAAYC;IAAiB;AACxC;AAOO,MAAMC,0BAAiD,CAC5DC,OACAC;IAEA,MAAMC,uBAAuB;IAC7B,MAAMC,wBAAwB;IAE9B,MAAMC,aAAaJ,QAAQC;IAC3B,IAAIG,cAAcF,sBAChB,OAAO;IACF,IAAIE,cAAcD,uBACvB,OAAO;IAGP,OAAO;AAEX"}

package/dist/index.d.ts

@@ -0,0 +1,3 @@
+export * from './GUIAgent';
+export { GUIAgent as default } from './GUIAgent';
+//# sourceMappingURL=index.d.ts.map

package/dist/index.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"index.d.ts","sourceRoot":"","sources":["../src/index.ts"],"names":[],"mappings":"AAIA,cAAc,YAAY,CAAC;AAC3B,OAAO,EAAE,QAAQ,IAAI,OAAO,EAAE,MAAM,YAAY,CAAC"}