@midscene/core 1.1.1-beta-20260106151323.0 → 1.1.1-beta-20260108085624.0

package/dist/es/ai-model/prompt/playwright-generator.mjs

@@ -1,5 +1,5 @@
 import { PLAYWRIGHT_EXAMPLE_CODE } from "@midscene/shared/constants";
-import { AIActionType, callAI, callAIWithStringResponse } from "../index.mjs";
+import { callAI, callAIWithStringResponse } from "../index.mjs";
 import { createEventCounts, createMessageContent, extractInputDescriptions, filterEventsByType, getScreenshotsForLLM, prepareEventSummary, processEventsForLLM, validateEvents } from "./yaml-generator.mjs";
 const generatePlaywrightTest = async (events, options, modelConfig)=>{
     validateEvents(events);
@@ -47,7 +47,7 @@ ${PLAYWRIGHT_EXAMPLE_CODE}`;
             content: messageContent
         }
     ];
-    const response = await callAIWithStringResponse(prompt, AIActionType.TEXT, modelConfig);
+    const response = await callAIWithStringResponse(prompt, modelConfig);
     if (response?.content && 'string' == typeof response.content) return response.content;
     throw new Error('Failed to generate Playwright test code');
 };
@@ -98,12 +98,12 @@ ${PLAYWRIGHT_EXAMPLE_CODE}`;
             content: messageContent
         }
     ];
-    if (options.stream && options.onChunk) return await callAI(prompt, AIActionType.TEXT, modelConfig, {
+    if (options.stream && options.onChunk) return await callAI(prompt, modelConfig, {
         stream: true,
         onChunk: options.onChunk
     });
     {
-        const response = await callAIWithStringResponse(prompt, AIActionType.TEXT, modelConfig);
+        const response = await callAIWithStringResponse(prompt, modelConfig);
         if (response?.content && 'string' == typeof response.content) return {
             content: response.content,
             usage: response.usage,

package/dist/es/ai-model/prompt/playwright-generator.mjs.map

@@ -1 +1 @@
-{"version":3,"file":"ai-model/prompt/playwright-generator.mjs","sources":["../../../../src/ai-model/prompt/playwright-generator.ts"],"sourcesContent":["import type {\n  StreamingAIResponse,\n  StreamingCodeGenerationOptions,\n} from '@/types';\nimport { PLAYWRIGHT_EXAMPLE_CODE } from '@midscene/shared/constants';\nimport type { IModelConfig } from '@midscene/shared/env';\nimport type { ChatCompletionMessageParam } from 'openai/resources/index';\nimport { AIActionType, callAI, callAIWithStringResponse } from '../index';\n// Import shared utilities and types from yaml-generator\nimport {\n  type ChromeRecordedEvent,\n  type EventCounts,\n  type EventSummary,\n  type InputDescription,\n  type ProcessedEvent,\n  createEventCounts,\n  createMessageContent,\n  extractInputDescriptions,\n  filterEventsByType,\n  getScreenshotsForLLM,\n  prepareEventSummary,\n  processEventsForLLM,\n  validateEvents,\n} from './yaml-generator';\n\n// Playwright-specific interfaces\nexport interface PlaywrightGenerationOptions {\n  testName?: string;\n  includeScreenshots?: boolean;\n  includeTimestamps?: boolean;\n  maxScreenshots?: number;\n  description?: string;\n  viewportSize?: { width: number; height: number };\n  waitForNetworkIdle?: boolean;\n  waitForNetworkIdleTimeout?: number;\n}\n\n// Re-export shared types for backward compatibility\nexport type {\n  ChromeRecordedEvent,\n  EventCounts,\n  InputDescription,\n  ProcessedEvent,\n  EventSummary,\n};\n\n// Re-export shared utilities for backward compatibility\nexport {\n  getScreenshotsForLLM,\n  filterEventsByType,\n  createEventCounts,\n  extractInputDescriptions,\n  processEventsForLLM,\n  prepareEventSummary,\n  createMessageContent,\n  validateEvents,\n};\n\n/**\n * Generates Playwright test code from recorded events\n */\nexport const generatePlaywrightTest = async (\n  events: ChromeRecordedEvent[],\n  options: PlaywrightGenerationOptions,\n  modelConfig: IModelConfig,\n): Promise<string> => {\n  // Validate input\n  validateEvents(events);\n\n  // Prepare event summary using shared utilities\n  const summary = prepareEventSummary(events, {\n    testName: options.testName,\n    maxScreenshots: options.maxScreenshots || 3,\n  });\n\n  // Add Playwright-specific options to summary\n  const playwrightSummary = {\n    ...summary,\n    waitForNetworkIdle: options.waitForNetworkIdle !== false,\n    waitForNetworkIdleTimeout: options.waitForNetworkIdleTimeout || 2000,\n    viewportSize: options.viewportSize || { width: 1280, height: 800 },\n  };\n\n  // Get screenshots for visual context\n  const screenshots = getScreenshotsForLLM(events, options.maxScreenshots || 3);\n\n  // Create prompt text\n  const promptText = `Generate a Playwright test using @midscene/web/playwright that reproduces this recorded browser session. The test should be based on the following events and follow the structure of the example provided. Make the test descriptive with appropriate assertions and validations.\n\nEvent Summary:\n${JSON.stringify(playwrightSummary, null, 2)}\n\nGenerated code should:\n1. Import required dependencies\n2. Set up the test with proper configuration\n3. Include a beforeEach hook to navigate to the starting URL\n4. Implement a test that uses Midscene AI methods (aiTap, aiInput, aiAssert, etc.)\n5. Include appropriate assertions and validations\n6. Follow best practices for Playwright tests\n7. Be ready to execute without further modification\n\nImportant: Return ONLY the raw Playwright test code. Do NOT wrap the response in markdown code blocks (no \\`\\`\\`typescript, \\`\\`\\`javascript or \\`\\`\\`). Start directly with the code content.`;\n\n  // Create message content with screenshots\n  const messageContent = createMessageContent(\n    promptText,\n    screenshots,\n    options.includeScreenshots !== false,\n  );\n\n  // Create system prompt\n  const systemPrompt = `You are an expert test automation engineer specializing in Playwright and Midscene. \nYour task is to generate a complete, executable Playwright test using @midscene/web/playwright that reproduces a recorded browser session.\n\n${PLAYWRIGHT_EXAMPLE_CODE}`;\n\n  // Use LLM to generate the Playwright test code\n  const prompt: ChatCompletionMessageParam[] = [\n    {\n      role: 'system',\n      content: systemPrompt,\n    },\n    {\n      role: 'user',\n      content: messageContent,\n    },\n  ];\n\n  const response = await callAIWithStringResponse(\n    prompt,\n    AIActionType.TEXT,\n    modelConfig,\n  );\n\n  if (response?.content && typeof response.content === 'string') {\n    return response.content;\n  }\n\n  throw new Error('Failed to generate Playwright test code');\n};\n\n/**\n * Generates Playwright test code from recorded events with streaming support\n */\nexport const generatePlaywrightTestStream = async (\n  events: ChromeRecordedEvent[],\n  options: PlaywrightGenerationOptions & StreamingCodeGenerationOptions,\n  modelConfig: IModelConfig,\n): Promise<StreamingAIResponse> => {\n  // Validate input\n  validateEvents(events);\n\n  // Prepare event summary using shared utilities\n  const summary = prepareEventSummary(events, {\n    testName: options.testName,\n    maxScreenshots: options.maxScreenshots || 3,\n  });\n\n  // Add Playwright-specific options to summary\n  const playwrightSummary = {\n    ...summary,\n    waitForNetworkIdle: options.waitForNetworkIdle !== false,\n    waitForNetworkIdleTimeout: options.waitForNetworkIdleTimeout || 2000,\n    viewportSize: options.viewportSize || { width: 1280, height: 800 },\n  };\n\n  // Get screenshots for visual context\n  const screenshots = getScreenshotsForLLM(events, options.maxScreenshots || 3);\n\n  // Create prompt text\n  const promptText = `Generate a Playwright test using @midscene/web/playwright that reproduces this recorded browser session. The test should be based on the following events and follow the structure of the example provided. Make the test descriptive with appropriate assertions and validations.\n\nEvent Summary:\n${JSON.stringify(playwrightSummary, null, 2)}\n\nGenerated code should:\n1. Import required dependencies\n2. Set up the test with proper configuration\n3. Include a beforeEach hook to navigate to the starting URL\n4. Implement a test that uses Midscene AI methods (aiTap, aiInput, aiAssert, etc.)\n5. Include appropriate assertions and validations\n6. Follow best practices for Playwright tests\n7. Be ready to execute without further modification\n8. can't wrap this test code in markdown code block\n\nImportant: Return ONLY the raw Playwright test code. Do NOT wrap the response in markdown code blocks (no \\`\\`\\`typescript, \\`\\`\\`javascript or \\`\\`\\`). Start directly with the code content.`;\n\n  // Create message content with screenshots\n  const messageContent = createMessageContent(\n    promptText,\n    screenshots,\n    options.includeScreenshots !== false,\n  );\n\n  // Create system prompt\n  const systemPrompt = `You are an expert test automation engineer specializing in Playwright and Midscene. \nYour task is to generate a complete, executable Playwright test using @midscene/web/playwright that reproduces a recorded browser session.\n\n${PLAYWRIGHT_EXAMPLE_CODE}`;\n\n  // Use LLM to generate the Playwright test code with streaming\n  const prompt: ChatCompletionMessageParam[] = [\n    {\n      role: 'system',\n      content: systemPrompt,\n    },\n    {\n      role: 'user',\n      content: messageContent,\n    },\n  ];\n\n  if (options.stream && options.onChunk) {\n    // Use streaming\n    return await callAI(prompt, AIActionType.TEXT, modelConfig, {\n      stream: true,\n      onChunk: options.onChunk,\n    });\n  } else {\n    // Fallback to non-streaming\n    const response = await callAIWithStringResponse(\n      prompt,\n      AIActionType.TEXT,\n      modelConfig,\n    );\n\n    if (response?.content && typeof response.content === 'string') {\n      return {\n        content: response.content,\n        usage: response.usage,\n        isStreamed: false,\n      };\n    }\n\n    throw new Error('Failed to generate Playwright test code');\n  }\n};\n"],"names":["generatePlaywrightTest","events","options","modelConfig","validateEvents","summary","prepareEventSummary","playwrightSummary","screenshots","getScreenshotsForLLM","promptText","JSON","messageContent","createMessageContent","systemPrompt","PLAYWRIGHT_EXAMPLE_CODE","prompt","response","callAIWithStringResponse","AIActionType","Error","generatePlaywrightTestStream","callAI"],"mappings":";;;AA6DO,MAAMA,yBAAyB,OACpCC,QACAC,SACAC;IAGAC,eAAeH;IAGf,MAAMI,UAAUC,oBAAoBL,QAAQ;QAC1C,UAAUC,QAAQ,QAAQ;QAC1B,gBAAgBA,QAAQ,cAAc,IAAI;IAC5C;IAGA,MAAMK,oBAAoB;QACxB,GAAGF,OAAO;QACV,oBAAoBH,AAA+B,UAA/BA,QAAQ,kBAAkB;QAC9C,2BAA2BA,QAAQ,yBAAyB,IAAI;QAChE,cAAcA,QAAQ,YAAY,IAAI;YAAE,OAAO;YAAM,QAAQ;QAAI;IACnE;IAGA,MAAMM,cAAcC,qBAAqBR,QAAQC,QAAQ,cAAc,IAAI;IAG3E,MAAMQ,aAAa,CAAC;;;AAGtB,EAAEC,KAAK,SAAS,CAACJ,mBAAmB,MAAM,GAAG;;;;;;;;;;;8LAWiJ,CAAC;IAG7L,MAAMK,iBAAiBC,qBACrBH,YACAF,aACAN,AAA+B,UAA/BA,QAAQ,kBAAkB;IAI5B,MAAMY,eAAe,CAAC;;;AAGxB,EAAEC,yBAAyB;IAGzB,MAAMC,SAAuC;QAC3C;YACE,MAAM;YACN,SAASF;QACX;QACA;YACE,MAAM;YACN,SAASF;QACX;KACD;IAED,MAAMK,WAAW,MAAMC,yBACrBF,QACAG,aAAa,IAAI,EACjBhB;IAGF,IAAIc,UAAU,WAAW,AAA4B,YAA5B,OAAOA,SAAS,OAAO,EAC9C,OAAOA,SAAS,OAAO;IAGzB,MAAM,IAAIG,MAAM;AAClB;AAKO,MAAMC,+BAA+B,OAC1CpB,QACAC,SACAC;IAGAC,eAAeH;IAGf,MAAMI,UAAUC,oBAAoBL,QAAQ;QAC1C,UAAUC,QAAQ,QAAQ;QAC1B,gBAAgBA,QAAQ,cAAc,IAAI;IAC5C;IAGA,MAAMK,oBAAoB;QACxB,GAAGF,OAAO;QACV,oBAAoBH,AAA+B,UAA/BA,QAAQ,kBAAkB;QAC9C,2BAA2BA,QAAQ,yBAAyB,IAAI;QAChE,cAAcA,QAAQ,YAAY,IAAI;YAAE,OAAO;YAAM,QAAQ;QAAI;IACnE;IAGA,MAAMM,cAAcC,qBAAqBR,QAAQC,QAAQ,cAAc,IAAI;IAG3E,MAAMQ,aAAa,CAAC;;;AAGtB,EAAEC,KAAK,SAAS,CAACJ,mBAAmB,MAAM,GAAG;;;;;;;;;;;;8LAYiJ,CAAC;IAG7L,MAAMK,iBAAiBC,qBACrBH,YACAF,aACAN,AAA+B,UAA/BA,QAAQ,kBAAkB;IAI5B,MAAMY,eAAe,CAAC;;;AAGxB,EAAEC,yBAAyB;IAGzB,MAAMC,SAAuC;QAC3C;YACE,MAAM;YACN,SAASF;QACX;QACA;YACE,MAAM;YACN,SAASF;QACX;KACD;IAED,IAAIV,QAAQ,MAAM,IAAIA,QAAQ,OAAO,EAEnC,OAAO,MAAMoB,OAAON,QAAQG,aAAa,IAAI,EAAEhB,aAAa;QAC1D,QAAQ;QACR,SAASD,QAAQ,OAAO;IAC1B;IACK;QAEL,MAAMe,WAAW,MAAMC,yBACrBF,QACAG,aAAa,IAAI,EACjBhB;QAGF,IAAIc,UAAU,WAAW,AAA4B,YAA5B,OAAOA,SAAS,OAAO,EAC9C,OAAO;YACL,SAASA,SAAS,OAAO;YACzB,OAAOA,SAAS,KAAK;YACrB,YAAY;QACd;QAGF,MAAM,IAAIG,MAAM;IAClB;AACF"}
+{"version":3,"file":"ai-model/prompt/playwright-generator.mjs","sources":["../../../../src/ai-model/prompt/playwright-generator.ts"],"sourcesContent":["import type {\n  StreamingAIResponse,\n  StreamingCodeGenerationOptions,\n} from '@/types';\nimport { PLAYWRIGHT_EXAMPLE_CODE } from '@midscene/shared/constants';\nimport type { IModelConfig } from '@midscene/shared/env';\nimport type { ChatCompletionMessageParam } from 'openai/resources/index';\nimport { callAI, callAIWithStringResponse } from '../index';\n// Import shared utilities and types from yaml-generator\nimport {\n  type ChromeRecordedEvent,\n  type EventCounts,\n  type EventSummary,\n  type InputDescription,\n  type ProcessedEvent,\n  createEventCounts,\n  createMessageContent,\n  extractInputDescriptions,\n  filterEventsByType,\n  getScreenshotsForLLM,\n  prepareEventSummary,\n  processEventsForLLM,\n  validateEvents,\n} from './yaml-generator';\n\n// Playwright-specific interfaces\nexport interface PlaywrightGenerationOptions {\n  testName?: string;\n  includeScreenshots?: boolean;\n  includeTimestamps?: boolean;\n  maxScreenshots?: number;\n  description?: string;\n  viewportSize?: { width: number; height: number };\n  waitForNetworkIdle?: boolean;\n  waitForNetworkIdleTimeout?: number;\n}\n\n// Re-export shared types for backward compatibility\nexport type {\n  ChromeRecordedEvent,\n  EventCounts,\n  InputDescription,\n  ProcessedEvent,\n  EventSummary,\n};\n\n// Re-export shared utilities for backward compatibility\nexport {\n  getScreenshotsForLLM,\n  filterEventsByType,\n  createEventCounts,\n  extractInputDescriptions,\n  processEventsForLLM,\n  prepareEventSummary,\n  createMessageContent,\n  validateEvents,\n};\n\n/**\n * Generates Playwright test code from recorded events\n */\nexport const generatePlaywrightTest = async (\n  events: ChromeRecordedEvent[],\n  options: PlaywrightGenerationOptions,\n  modelConfig: IModelConfig,\n): Promise<string> => {\n  // Validate input\n  validateEvents(events);\n\n  // Prepare event summary using shared utilities\n  const summary = prepareEventSummary(events, {\n    testName: options.testName,\n    maxScreenshots: options.maxScreenshots || 3,\n  });\n\n  // Add Playwright-specific options to summary\n  const playwrightSummary = {\n    ...summary,\n    waitForNetworkIdle: options.waitForNetworkIdle !== false,\n    waitForNetworkIdleTimeout: options.waitForNetworkIdleTimeout || 2000,\n    viewportSize: options.viewportSize || { width: 1280, height: 800 },\n  };\n\n  // Get screenshots for visual context\n  const screenshots = getScreenshotsForLLM(events, options.maxScreenshots || 3);\n\n  // Create prompt text\n  const promptText = `Generate a Playwright test using @midscene/web/playwright that reproduces this recorded browser session. The test should be based on the following events and follow the structure of the example provided. Make the test descriptive with appropriate assertions and validations.\n\nEvent Summary:\n${JSON.stringify(playwrightSummary, null, 2)}\n\nGenerated code should:\n1. Import required dependencies\n2. Set up the test with proper configuration\n3. Include a beforeEach hook to navigate to the starting URL\n4. Implement a test that uses Midscene AI methods (aiTap, aiInput, aiAssert, etc.)\n5. Include appropriate assertions and validations\n6. Follow best practices for Playwright tests\n7. Be ready to execute without further modification\n\nImportant: Return ONLY the raw Playwright test code. Do NOT wrap the response in markdown code blocks (no \\`\\`\\`typescript, \\`\\`\\`javascript or \\`\\`\\`). Start directly with the code content.`;\n\n  // Create message content with screenshots\n  const messageContent = createMessageContent(\n    promptText,\n    screenshots,\n    options.includeScreenshots !== false,\n  );\n\n  // Create system prompt\n  const systemPrompt = `You are an expert test automation engineer specializing in Playwright and Midscene. \nYour task is to generate a complete, executable Playwright test using @midscene/web/playwright that reproduces a recorded browser session.\n\n${PLAYWRIGHT_EXAMPLE_CODE}`;\n\n  // Use LLM to generate the Playwright test code\n  const prompt: ChatCompletionMessageParam[] = [\n    {\n      role: 'system',\n      content: systemPrompt,\n    },\n    {\n      role: 'user',\n      content: messageContent,\n    },\n  ];\n\n  const response = await callAIWithStringResponse(prompt, modelConfig);\n\n  if (response?.content && typeof response.content === 'string') {\n    return response.content;\n  }\n\n  throw new Error('Failed to generate Playwright test code');\n};\n\n/**\n * Generates Playwright test code from recorded events with streaming support\n */\nexport const generatePlaywrightTestStream = async (\n  events: ChromeRecordedEvent[],\n  options: PlaywrightGenerationOptions & StreamingCodeGenerationOptions,\n  modelConfig: IModelConfig,\n): Promise<StreamingAIResponse> => {\n  // Validate input\n  validateEvents(events);\n\n  // Prepare event summary using shared utilities\n  const summary = prepareEventSummary(events, {\n    testName: options.testName,\n    maxScreenshots: options.maxScreenshots || 3,\n  });\n\n  // Add Playwright-specific options to summary\n  const playwrightSummary = {\n    ...summary,\n    waitForNetworkIdle: options.waitForNetworkIdle !== false,\n    waitForNetworkIdleTimeout: options.waitForNetworkIdleTimeout || 2000,\n    viewportSize: options.viewportSize || { width: 1280, height: 800 },\n  };\n\n  // Get screenshots for visual context\n  const screenshots = getScreenshotsForLLM(events, options.maxScreenshots || 3);\n\n  // Create prompt text\n  const promptText = `Generate a Playwright test using @midscene/web/playwright that reproduces this recorded browser session. The test should be based on the following events and follow the structure of the example provided. Make the test descriptive with appropriate assertions and validations.\n\nEvent Summary:\n${JSON.stringify(playwrightSummary, null, 2)}\n\nGenerated code should:\n1. Import required dependencies\n2. Set up the test with proper configuration\n3. Include a beforeEach hook to navigate to the starting URL\n4. Implement a test that uses Midscene AI methods (aiTap, aiInput, aiAssert, etc.)\n5. Include appropriate assertions and validations\n6. Follow best practices for Playwright tests\n7. Be ready to execute without further modification\n8. can't wrap this test code in markdown code block\n\nImportant: Return ONLY the raw Playwright test code. Do NOT wrap the response in markdown code blocks (no \\`\\`\\`typescript, \\`\\`\\`javascript or \\`\\`\\`). Start directly with the code content.`;\n\n  // Create message content with screenshots\n  const messageContent = createMessageContent(\n    promptText,\n    screenshots,\n    options.includeScreenshots !== false,\n  );\n\n  // Create system prompt\n  const systemPrompt = `You are an expert test automation engineer specializing in Playwright and Midscene. \nYour task is to generate a complete, executable Playwright test using @midscene/web/playwright that reproduces a recorded browser session.\n\n${PLAYWRIGHT_EXAMPLE_CODE}`;\n\n  // Use LLM to generate the Playwright test code with streaming\n  const prompt: ChatCompletionMessageParam[] = [\n    {\n      role: 'system',\n      content: systemPrompt,\n    },\n    {\n      role: 'user',\n      content: messageContent,\n    },\n  ];\n\n  if (options.stream && options.onChunk) {\n    // Use streaming\n    return await callAI(prompt, modelConfig, {\n      stream: true,\n      onChunk: options.onChunk,\n    });\n  } else {\n    // Fallback to non-streaming\n    const response = await callAIWithStringResponse(prompt, modelConfig);\n\n    if (response?.content && typeof response.content === 'string') {\n      return {\n        content: response.content,\n        usage: response.usage,\n        isStreamed: false,\n      };\n    }\n\n    throw new Error('Failed to generate Playwright test code');\n  }\n};\n"],"names":["generatePlaywrightTest","events","options","modelConfig","validateEvents","summary","prepareEventSummary","playwrightSummary","screenshots","getScreenshotsForLLM","promptText","JSON","messageContent","createMessageContent","systemPrompt","PLAYWRIGHT_EXAMPLE_CODE","prompt","response","callAIWithStringResponse","Error","generatePlaywrightTestStream","callAI"],"mappings":";;;AA6DO,MAAMA,yBAAyB,OACpCC,QACAC,SACAC;IAGAC,eAAeH;IAGf,MAAMI,UAAUC,oBAAoBL,QAAQ;QAC1C,UAAUC,QAAQ,QAAQ;QAC1B,gBAAgBA,QAAQ,cAAc,IAAI;IAC5C;IAGA,MAAMK,oBAAoB;QACxB,GAAGF,OAAO;QACV,oBAAoBH,AAA+B,UAA/BA,QAAQ,kBAAkB;QAC9C,2BAA2BA,QAAQ,yBAAyB,IAAI;QAChE,cAAcA,QAAQ,YAAY,IAAI;YAAE,OAAO;YAAM,QAAQ;QAAI;IACnE;IAGA,MAAMM,cAAcC,qBAAqBR,QAAQC,QAAQ,cAAc,IAAI;IAG3E,MAAMQ,aAAa,CAAC;;;AAGtB,EAAEC,KAAK,SAAS,CAACJ,mBAAmB,MAAM,GAAG;;;;;;;;;;;8LAWiJ,CAAC;IAG7L,MAAMK,iBAAiBC,qBACrBH,YACAF,aACAN,AAA+B,UAA/BA,QAAQ,kBAAkB;IAI5B,MAAMY,eAAe,CAAC;;;AAGxB,EAAEC,yBAAyB;IAGzB,MAAMC,SAAuC;QAC3C;YACE,MAAM;YACN,SAASF;QACX;QACA;YACE,MAAM;YACN,SAASF;QACX;KACD;IAED,MAAMK,WAAW,MAAMC,yBAAyBF,QAAQb;IAExD,IAAIc,UAAU,WAAW,AAA4B,YAA5B,OAAOA,SAAS,OAAO,EAC9C,OAAOA,SAAS,OAAO;IAGzB,MAAM,IAAIE,MAAM;AAClB;AAKO,MAAMC,+BAA+B,OAC1CnB,QACAC,SACAC;IAGAC,eAAeH;IAGf,MAAMI,UAAUC,oBAAoBL,QAAQ;QAC1C,UAAUC,QAAQ,QAAQ;QAC1B,gBAAgBA,QAAQ,cAAc,IAAI;IAC5C;IAGA,MAAMK,oBAAoB;QACxB,GAAGF,OAAO;QACV,oBAAoBH,AAA+B,UAA/BA,QAAQ,kBAAkB;QAC9C,2BAA2BA,QAAQ,yBAAyB,IAAI;QAChE,cAAcA,QAAQ,YAAY,IAAI;YAAE,OAAO;YAAM,QAAQ;QAAI;IACnE;IAGA,MAAMM,cAAcC,qBAAqBR,QAAQC,QAAQ,cAAc,IAAI;IAG3E,MAAMQ,aAAa,CAAC;;;AAGtB,EAAEC,KAAK,SAAS,CAACJ,mBAAmB,MAAM,GAAG;;;;;;;;;;;;8LAYiJ,CAAC;IAG7L,MAAMK,iBAAiBC,qBACrBH,YACAF,aACAN,AAA+B,UAA/BA,QAAQ,kBAAkB;IAI5B,MAAMY,eAAe,CAAC;;;AAGxB,EAAEC,yBAAyB;IAGzB,MAAMC,SAAuC;QAC3C;YACE,MAAM;YACN,SAASF;QACX;QACA;YACE,MAAM;YACN,SAASF;QACX;KACD;IAED,IAAIV,QAAQ,MAAM,IAAIA,QAAQ,OAAO,EAEnC,OAAO,MAAMmB,OAAOL,QAAQb,aAAa;QACvC,QAAQ;QACR,SAASD,QAAQ,OAAO;IAC1B;IACK;QAEL,MAAMe,WAAW,MAAMC,yBAAyBF,QAAQb;QAExD,IAAIc,UAAU,WAAW,AAA4B,YAA5B,OAAOA,SAAS,OAAO,EAC9C,OAAO;YACL,SAASA,SAAS,OAAO;YACzB,OAAOA,SAAS,KAAK;YACrB,YAAY;QACd;QAGF,MAAM,IAAIE,MAAM;IAClB;AACF"}

package/dist/es/ai-model/prompt/yaml-generator.mjs

@@ -1,5 +1,5 @@
 import { YAML_EXAMPLE_CODE } from "@midscene/shared/constants";
-import { AIActionType, callAI, callAIWithStringResponse } from "../index.mjs";
+import { callAI, callAIWithStringResponse } from "../index.mjs";
 const getScreenshotsForLLM = (events, maxScreenshots = 1)=>{
     const eventsWithScreenshots = events.filter((event)=>event.screenshotBefore || event.screenshotAfter || event.screenshotWithBox);
     const sortedEvents = [
@@ -141,7 +141,7 @@ Important: Return ONLY the raw YAML content. Do NOT wrap the response in markdow
                     }))
             });
         }
-        const response = await callAIWithStringResponse(prompt, AIActionType.TEXT, modelConfig);
+        const response = await callAIWithStringResponse(prompt, modelConfig);
         if (response?.content && 'string' == typeof response.content) return response.content;
         throw new Error('Failed to generate YAML test configuration');
     } catch (error) {
@@ -197,12 +197,12 @@ Important: Return ONLY the raw YAML content. Do NOT wrap the response in markdow
                     }))
             });
         }
-        if (options.stream && options.onChunk) return await callAI(prompt, AIActionType.TEXT, modelConfig, {
+        if (options.stream && options.onChunk) return await callAI(prompt, modelConfig, {
             stream: true,
             onChunk: options.onChunk
         });
         {
-            const response = await callAIWithStringResponse(prompt, AIActionType.TEXT, modelConfig);
+            const response = await callAIWithStringResponse(prompt, modelConfig);
             if (response?.content && 'string' == typeof response.content) return {
                 content: response.content,
                 usage: response.usage,

package/dist/es/ai-model/prompt/yaml-generator.mjs.map

@@ -1 +1 @@
-{"version":3,"file":"ai-model/prompt/yaml-generator.mjs","sources":["../../../../src/ai-model/prompt/yaml-generator.ts"],"sourcesContent":["import type {\n  StreamingAIResponse,\n  StreamingCodeGenerationOptions,\n} from '@/types';\nimport { YAML_EXAMPLE_CODE } from '@midscene/shared/constants';\nimport type { IModelConfig } from '@midscene/shared/env';\nimport {\n  AIActionType,\n  type ChatCompletionMessageParam,\n  callAI,\n  callAIWithStringResponse,\n} from '../index';\n\n// Common interfaces for test generation (shared between YAML and Playwright)\nexport interface EventCounts {\n  navigation: number;\n  click: number;\n  input: number;\n  scroll: number;\n  total: number;\n}\n\nexport interface InputDescription {\n  description: string;\n  value: string;\n}\n\nexport interface ProcessedEvent {\n  type: string;\n  timestamp: number;\n  url?: string;\n  title?: string;\n  elementDescription?: string;\n  value?: string;\n  pageInfo?: any;\n  elementRect?: any;\n}\n\nexport interface EventSummary {\n  testName: string;\n  startUrl: string;\n  eventCounts: EventCounts;\n  urls: string[];\n  clickDescriptions: string[];\n  inputDescriptions: InputDescription[];\n  events: ProcessedEvent[];\n}\n\n// Common ChromeRecordedEvent interface\nexport interface ChromeRecordedEvent {\n  type: string;\n  timestamp: number;\n  url?: string;\n  title?: string;\n  elementDescription?: string;\n  value?: string;\n  pageInfo?: any;\n  elementRect?: any;\n  screenshotBefore?: string;\n  screenshotAfter?: string;\n  screenshotWithBox?: string;\n}\n\nexport interface YamlGenerationOptions {\n  testName?: string;\n  includeTimestamps?: boolean;\n  maxScreenshots?: number;\n  description?: string;\n}\n\nexport interface FilteredEvents {\n  navigationEvents: ChromeRecordedEvent[];\n  clickEvents: ChromeRecordedEvent[];\n  inputEvents: ChromeRecordedEvent[];\n  scrollEvents: ChromeRecordedEvent[];\n}\n\n// Common utility functions (shared between YAML and Playwright generators)\n\n/**\n * Get screenshots from events for LLM context\n */\nexport const getScreenshotsForLLM = (\n  events: ChromeRecordedEvent[],\n  maxScreenshots = 1,\n): string[] => {\n  // Find events with screenshots, prioritizing navigation and click events\n  const eventsWithScreenshots = events.filter(\n    (event) =>\n      event.screenshotBefore ||\n      event.screenshotAfter ||\n      event.screenshotWithBox,\n  );\n\n  // Sort them by priority (navigation first, then clicks, then others)\n  const sortedEvents = [...eventsWithScreenshots].sort((a, b) => {\n    if (a.type === 'navigation' && b.type !== 'navigation') return -1;\n    if (a.type !== 'navigation' && b.type === 'navigation') return 1;\n    if (a.type === 'click' && b.type !== 'click') return -1;\n    if (a.type !== 'click' && b.type === 'click') return 1;\n    return 0;\n  });\n\n  // Extract up to maxScreenshots screenshots\n  const screenshots: string[] = [];\n  for (const event of sortedEvents) {\n    // Prefer the most informative screenshot\n    const screenshot =\n      event.screenshotWithBox ||\n      event.screenshotAfter ||\n      event.screenshotBefore;\n    if (screenshot && !screenshots.includes(screenshot)) {\n      screenshots.push(screenshot);\n      if (screenshots.length >= maxScreenshots) break;\n    }\n  }\n\n  return screenshots;\n};\n\n/**\n * Filter events by type for easier processing\n */\nexport const filterEventsByType = (\n  events: ChromeRecordedEvent[],\n): FilteredEvents => {\n  return {\n    navigationEvents: events.filter((event) => event.type === 'navigation'),\n    clickEvents: events.filter((event) => event.type === 'click'),\n    inputEvents: events.filter((event) => event.type === 'input'),\n    scrollEvents: events.filter((event) => event.type === 'scroll'),\n  };\n};\n\n/**\n * Create event counts summary\n */\nexport const createEventCounts = (\n  filteredEvents: FilteredEvents,\n  totalEvents: number,\n): EventCounts => {\n  return {\n    navigation: filteredEvents.navigationEvents.length,\n    click: filteredEvents.clickEvents.length,\n    input: filteredEvents.inputEvents.length,\n    scroll: filteredEvents.scrollEvents.length,\n    total: totalEvents,\n  };\n};\n\n/**\n * Extract input descriptions from input events\n */\nexport const extractInputDescriptions = (\n  inputEvents: ChromeRecordedEvent[],\n): InputDescription[] => {\n  return inputEvents\n    .map((event) => ({\n      description: event.elementDescription || '',\n      value: event.value || '',\n    }))\n    .filter((item) => item.description && item.value);\n};\n\n/**\n * Process events for LLM consumption\n */\nexport const processEventsForLLM = (\n  events: ChromeRecordedEvent[],\n): ProcessedEvent[] => {\n  return events.map((event) => ({\n    type: event.type,\n    timestamp: event.timestamp,\n    url: event.url,\n    title: event.title,\n    elementDescription: event.elementDescription,\n    value: event.value,\n    pageInfo: event.pageInfo,\n    elementRect: event.elementRect,\n  }));\n};\n\n/**\n * Prepare comprehensive event summary for LLM\n */\nexport const prepareEventSummary = (\n  events: ChromeRecordedEvent[],\n  options: { testName?: string; maxScreenshots?: number } = {},\n): EventSummary => {\n  const filteredEvents = filterEventsByType(events);\n  const eventCounts = createEventCounts(filteredEvents, events.length);\n\n  // Extract useful information from events\n  const startUrl =\n    filteredEvents.navigationEvents.length > 0\n      ? filteredEvents.navigationEvents[0].url || ''\n      : '';\n\n  const clickDescriptions = filteredEvents.clickEvents\n    .map((event) => event.elementDescription)\n    .filter((desc): desc is string => Boolean(desc))\n    .slice(0, 10);\n\n  const inputDescriptions = extractInputDescriptions(\n    filteredEvents.inputEvents,\n  ).slice(0, 10);\n\n  const urls = filteredEvents.navigationEvents\n    .map((e) => e.url)\n    .filter((url): url is string => Boolean(url))\n    .slice(0, 5);\n\n  const processedEvents = processEventsForLLM(events);\n\n  return {\n    testName: options.testName || 'Automated test from recorded events',\n    startUrl,\n    eventCounts,\n    urls,\n    clickDescriptions,\n    inputDescriptions,\n    events: processedEvents,\n  };\n};\n\n/**\n * Create message content for LLM with optional screenshots\n */\nexport const createMessageContent = (\n  promptText: string,\n  screenshots: string[] = [],\n  includeScreenshots = true,\n) => {\n  const messageContent: any[] = [\n    {\n      type: 'text',\n      text: promptText,\n    },\n  ];\n\n  // Add screenshots if available and requested\n  if (includeScreenshots && screenshots.length > 0) {\n    messageContent.unshift({\n      type: 'text',\n      text: 'Here are screenshots from the recording session to help you understand the context:',\n    });\n\n    screenshots.forEach((screenshot) => {\n      messageContent.push({\n        type: 'image_url',\n        image_url: {\n          url: screenshot,\n        },\n      });\n    });\n  }\n\n  return messageContent;\n};\n\n/**\n * Validate events before processing\n */\nexport const validateEvents = (events: ChromeRecordedEvent[]): void => {\n  if (!events.length) {\n    throw new Error('No events provided for test generation');\n  }\n};\n\n// YAML-specific generation functions\n\n/**\n * Generates YAML test configuration from recorded events using AI\n */\nexport const generateYamlTest = async (\n  events: ChromeRecordedEvent[],\n  options: YamlGenerationOptions,\n  modelConfig: IModelConfig,\n): Promise<string> => {\n  try {\n    // Validate input\n    validateEvents(events);\n\n    // Prepare event summary using shared utilities\n    const summary = prepareEventSummary(events, {\n      testName: options.testName,\n      maxScreenshots: options.maxScreenshots || 3,\n    });\n\n    // Add YAML-specific options to summary\n    const yamlSummary = {\n      ...summary,\n      includeTimestamps: options.includeTimestamps || false,\n    };\n\n    // Get screenshots for visual context\n    const screenshots = getScreenshotsForLLM(\n      events,\n      options.maxScreenshots || 3,\n    );\n\n    // Use LLM to generate the YAML test configuration\n    const prompt: ChatCompletionMessageParam[] = [\n      {\n        role: 'system',\n        content: `You are an expert in Midscene.js YAML test generation. Generate clean, accurate YAML following these rules: ${YAML_EXAMPLE_CODE}`,\n      },\n      {\n        role: 'user',\n        content: `Generate YAML test for Midscene.js automation from recorded browser events.\n\nEvent Summary:\n${JSON.stringify(yamlSummary, null, 2)}\n\nConvert events:\n- navigation → target.url\n- click → aiTap with element description\n- input → aiInput with value and locate\n- scroll → aiScroll with appropriate direction\n- Add aiAssert for important state changes\n\nImportant: Return ONLY the raw YAML content. Do NOT wrap the response in markdown code blocks (no \\`\\`\\`yaml or \\`\\`\\`). Start directly with the YAML content.`,\n      },\n    ];\n\n    // Add screenshots if available and requested\n    if (screenshots.length > 0) {\n      prompt.push({\n        role: 'user',\n        content:\n          'Here are screenshots from the recording session to help you understand the context:',\n      });\n\n      prompt.push({\n        role: 'user',\n        content: screenshots.map((screenshot) => ({\n          type: 'image_url',\n          image_url: {\n            url: screenshot,\n          },\n        })),\n      });\n    }\n\n    const response = await callAIWithStringResponse(\n      prompt,\n      AIActionType.TEXT,\n      modelConfig,\n    );\n\n    if (response?.content && typeof response.content === 'string') {\n      return response.content;\n    }\n\n    throw new Error('Failed to generate YAML test configuration');\n  } catch (error) {\n    throw new Error(`Failed to generate YAML test: ${error}`);\n  }\n};\n\n/**\n * Generates YAML test configuration from recorded events using AI with streaming support\n */\nexport const generateYamlTestStream = async (\n  events: ChromeRecordedEvent[],\n  options: YamlGenerationOptions & StreamingCodeGenerationOptions,\n  modelConfig: IModelConfig,\n): Promise<StreamingAIResponse> => {\n  try {\n    // Validate input\n    validateEvents(events);\n\n    // Prepare event summary using shared utilities\n    const summary = prepareEventSummary(events, {\n      testName: options.testName,\n      maxScreenshots: options.maxScreenshots || 3,\n    });\n\n    // Add YAML-specific options to summary\n    const yamlSummary = {\n      ...summary,\n      includeTimestamps: options.includeTimestamps || false,\n    };\n\n    // Get screenshots for visual context\n    const screenshots = getScreenshotsForLLM(\n      events,\n      options.maxScreenshots || 3,\n    );\n\n    // Use LLM to generate the YAML test configuration\n    const prompt: ChatCompletionMessageParam[] = [\n      {\n        role: 'system',\n        content: `You are an expert in Midscene.js YAML test generation. Generate clean, accurate YAML following these rules: ${YAML_EXAMPLE_CODE}`,\n      },\n      {\n        role: 'user',\n        content: `Generate YAML test for Midscene.js automation from recorded browser events.\n\nEvent Summary:\n${JSON.stringify(yamlSummary, null, 2)}\n\nConvert events:\n- navigation → target.url\n- click → aiTap with element description\n- input → aiInput with value and locate\n- scroll → aiScroll with appropriate direction\n- Add aiAssert for important state changes\n\nImportant: Return ONLY the raw YAML content. Do NOT wrap the response in markdown code blocks (no \\`\\`\\`yaml or \\`\\`\\`). Start directly with the YAML content.`,\n      },\n    ];\n\n    // Add screenshots if available and requested\n    if (screenshots.length > 0) {\n      prompt.push({\n        role: 'user',\n        content:\n          'Here are screenshots from the recording session to help you understand the context:',\n      });\n\n      prompt.push({\n        role: 'user',\n        content: screenshots.map((screenshot) => ({\n          type: 'image_url',\n          image_url: {\n            url: screenshot,\n          },\n        })),\n      });\n    }\n\n    if (options.stream && options.onChunk) {\n      // Use streaming\n      return await callAI(prompt, AIActionType.TEXT, modelConfig, {\n        stream: true,\n        onChunk: options.onChunk,\n      });\n    } else {\n      // Fallback to non-streaming\n      const response = await callAIWithStringResponse(\n        prompt,\n        AIActionType.TEXT,\n        modelConfig,\n      );\n\n      if (response?.content && typeof response.content === 'string') {\n        return {\n          content: response.content,\n          usage: response.usage,\n          isStreamed: false,\n        };\n      }\n\n      throw new Error('Failed to generate YAML test configuration');\n    }\n  } catch (error) {\n    throw new Error(`Failed to generate YAML test: ${error}`);\n  }\n};\n"],"names":["getScreenshotsForLLM","events","maxScreenshots","eventsWithScreenshots","event","sortedEvents","a","b","screenshots","screenshot","filterEventsByType","createEventCounts","filteredEvents","totalEvents","extractInputDescriptions","inputEvents","item","processEventsForLLM","prepareEventSummary","options","eventCounts","startUrl","clickDescriptions","desc","Boolean","inputDescriptions","urls","e","url","processedEvents","createMessageContent","promptText","includeScreenshots","messageContent","validateEvents","Error","generateYamlTest","modelConfig","summary","yamlSummary","prompt","YAML_EXAMPLE_CODE","JSON","response","callAIWithStringResponse","AIActionType","error","generateYamlTestStream","callAI"],"mappings":";;AAkFO,MAAMA,uBAAuB,CAClCC,QACAC,iBAAiB,CAAC;IAGlB,MAAMC,wBAAwBF,OAAO,MAAM,CACzC,CAACG,QACCA,MAAM,gBAAgB,IACtBA,MAAM,eAAe,IACrBA,MAAM,iBAAiB;IAI3B,MAAMC,eAAe;WAAIF;KAAsB,CAAC,IAAI,CAAC,CAACG,GAAGC;QACvD,IAAID,AAAW,iBAAXA,EAAE,IAAI,IAAqBC,AAAW,iBAAXA,EAAE,IAAI,EAAmB,OAAO;QAC/D,IAAID,AAAW,iBAAXA,EAAE,IAAI,IAAqBC,AAAW,iBAAXA,EAAE,IAAI,EAAmB,OAAO;QAC/D,IAAID,AAAW,YAAXA,EAAE,IAAI,IAAgBC,AAAW,YAAXA,EAAE,IAAI,EAAc,OAAO;QACrD,IAAID,AAAW,YAAXA,EAAE,IAAI,IAAgBC,AAAW,YAAXA,EAAE,IAAI,EAAc,OAAO;QACrD,OAAO;IACT;IAGA,MAAMC,cAAwB,EAAE;IAChC,KAAK,MAAMJ,SAASC,aAAc;QAEhC,MAAMI,aACJL,MAAM,iBAAiB,IACvBA,MAAM,eAAe,IACrBA,MAAM,gBAAgB;QACxB,IAAIK,cAAc,CAACD,YAAY,QAAQ,CAACC,aAAa;YACnDD,YAAY,IAAI,CAACC;YACjB,IAAID,YAAY,MAAM,IAAIN,gBAAgB;QAC5C;IACF;IAEA,OAAOM;AACT;AAKO,MAAME,qBAAqB,CAChCT,SAEO;QACL,kBAAkBA,OAAO,MAAM,CAAC,CAACG,QAAUA,AAAe,iBAAfA,MAAM,IAAI;QACrD,aAAaH,OAAO,MAAM,CAAC,CAACG,QAAUA,AAAe,YAAfA,MAAM,IAAI;QAChD,aAAaH,OAAO,MAAM,CAAC,CAACG,QAAUA,AAAe,YAAfA,MAAM,IAAI;QAChD,cAAcH,OAAO,MAAM,CAAC,CAACG,QAAUA,AAAe,aAAfA,MAAM,IAAI;IACnD;AAMK,MAAMO,oBAAoB,CAC/BC,gBACAC,cAEO;QACL,YAAYD,eAAe,gBAAgB,CAAC,MAAM;QAClD,OAAOA,eAAe,WAAW,CAAC,MAAM;QACxC,OAAOA,eAAe,WAAW,CAAC,MAAM;QACxC,QAAQA,eAAe,YAAY,CAAC,MAAM;QAC1C,OAAOC;IACT;AAMK,MAAMC,2BAA2B,CACtCC,cAEOA,YACJ,GAAG,CAAC,CAACX,QAAW;YACf,aAAaA,MAAM,kBAAkB,IAAI;YACzC,OAAOA,MAAM,KAAK,IAAI;QACxB,IACC,MAAM,CAAC,CAACY,OAASA,KAAK,WAAW,IAAIA,KAAK,KAAK;AAM7C,MAAMC,sBAAsB,CACjChB,SAEOA,OAAO,GAAG,CAAC,CAACG,QAAW;YAC5B,MAAMA,MAAM,IAAI;YAChB,WAAWA,MAAM,SAAS;YAC1B,KAAKA,MAAM,GAAG;YACd,OAAOA,MAAM,KAAK;YAClB,oBAAoBA,MAAM,kBAAkB;YAC5C,OAAOA,MAAM,KAAK;YAClB,UAAUA,MAAM,QAAQ;YACxB,aAAaA,MAAM,WAAW;QAChC;AAMK,MAAMc,sBAAsB,CACjCjB,QACAkB,UAA0D,CAAC,CAAC;IAE5D,MAAMP,iBAAiBF,mBAAmBT;IAC1C,MAAMmB,cAAcT,kBAAkBC,gBAAgBX,OAAO,MAAM;IAGnE,MAAMoB,WACJT,eAAe,gBAAgB,CAAC,MAAM,GAAG,IACrCA,eAAe,gBAAgB,CAAC,EAAE,CAAC,GAAG,IAAI,KAC1C;IAEN,MAAMU,oBAAoBV,eAAe,WAAW,CACjD,GAAG,CAAC,CAACR,QAAUA,MAAM,kBAAkB,EACvC,MAAM,CAAC,CAACmB,OAAyBC,QAAQD,OACzC,KAAK,CAAC,GAAG;IAEZ,MAAME,oBAAoBX,yBACxBF,eAAe,WAAW,EAC1B,KAAK,CAAC,GAAG;IAEX,MAAMc,OAAOd,eAAe,gBAAgB,CACzC,GAAG,CAAC,CAACe,IAAMA,EAAE,GAAG,EAChB,MAAM,CAAC,CAACC,MAAuBJ,QAAQI,MACvC,KAAK,CAAC,GAAG;IAEZ,MAAMC,kBAAkBZ,oBAAoBhB;IAE5C,OAAO;QACL,UAAUkB,QAAQ,QAAQ,IAAI;QAC9BE;QACAD;QACAM;QACAJ;QACAG;QACA,QAAQI;IACV;AACF;AAKO,MAAMC,uBAAuB,CAClCC,YACAvB,cAAwB,EAAE,EAC1BwB,qBAAqB,IAAI;IAEzB,MAAMC,iBAAwB;QAC5B;YACE,MAAM;YACN,MAAMF;QACR;KACD;IAGD,IAAIC,sBAAsBxB,YAAY,MAAM,GAAG,GAAG;QAChDyB,eAAe,OAAO,CAAC;YACrB,MAAM;YACN,MAAM;QACR;QAEAzB,YAAY,OAAO,CAAC,CAACC;YACnBwB,eAAe,IAAI,CAAC;gBAClB,MAAM;gBACN,WAAW;oBACT,KAAKxB;gBACP;YACF;QACF;IACF;IAEA,OAAOwB;AACT;AAKO,MAAMC,iBAAiB,CAACjC;IAC7B,IAAI,CAACA,OAAO,MAAM,EAChB,MAAM,IAAIkC,MAAM;AAEpB;AAOO,MAAMC,mBAAmB,OAC9BnC,QACAkB,SACAkB;IAEA,IAAI;QAEFH,eAAejC;QAGf,MAAMqC,UAAUpB,oBAAoBjB,QAAQ;YAC1C,UAAUkB,QAAQ,QAAQ;YAC1B,gBAAgBA,QAAQ,cAAc,IAAI;QAC5C;QAGA,MAAMoB,cAAc;YAClB,GAAGD,OAAO;YACV,mBAAmBnB,QAAQ,iBAAiB,IAAI;QAClD;QAGA,MAAMX,cAAcR,qBAClBC,QACAkB,QAAQ,cAAc,IAAI;QAI5B,MAAMqB,SAAuC;YAC3C;gBACE,MAAM;gBACN,SAAS,CAAC,4GAA4G,EAAEC,mBAAmB;YAC7I;YACA;gBACE,MAAM;gBACN,SAAS,CAAC;;;AAGlB,EAAEC,KAAK,SAAS,CAACH,aAAa,MAAM,GAAG;;;;;;;;;8JASuH,CAAC;YACzJ;SACD;QAGD,IAAI/B,YAAY,MAAM,GAAG,GAAG;YAC1BgC,OAAO,IAAI,CAAC;gBACV,MAAM;gBACN,SACE;YACJ;YAEAA,OAAO,IAAI,CAAC;gBACV,MAAM;gBACN,SAAShC,YAAY,GAAG,CAAC,CAACC,aAAgB;wBACxC,MAAM;wBACN,WAAW;4BACT,KAAKA;wBACP;oBACF;YACF;QACF;QAEA,MAAMkC,WAAW,MAAMC,yBACrBJ,QACAK,aAAa,IAAI,EACjBR;QAGF,IAAIM,UAAU,WAAW,AAA4B,YAA5B,OAAOA,SAAS,OAAO,EAC9C,OAAOA,SAAS,OAAO;QAGzB,MAAM,IAAIR,MAAM;IAClB,EAAE,OAAOW,OAAO;QACd,MAAM,IAAIX,MAAM,CAAC,8BAA8B,EAAEW,OAAO;IAC1D;AACF;AAKO,MAAMC,yBAAyB,OACpC9C,QACAkB,SACAkB;IAEA,IAAI;QAEFH,eAAejC;QAGf,MAAMqC,UAAUpB,oBAAoBjB,QAAQ;YAC1C,UAAUkB,QAAQ,QAAQ;YAC1B,gBAAgBA,QAAQ,cAAc,IAAI;QAC5C;QAGA,MAAMoB,cAAc;YAClB,GAAGD,OAAO;YACV,mBAAmBnB,QAAQ,iBAAiB,IAAI;QAClD;QAGA,MAAMX,cAAcR,qBAClBC,QACAkB,QAAQ,cAAc,IAAI;QAI5B,MAAMqB,SAAuC;YAC3C;gBACE,MAAM;gBACN,SAAS,CAAC,4GAA4G,EAAEC,mBAAmB;YAC7I;YACA;gBACE,MAAM;gBACN,SAAS,CAAC;;;AAGlB,EAAEC,KAAK,SAAS,CAACH,aAAa,MAAM,GAAG;;;;;;;;;8JASuH,CAAC;YACzJ;SACD;QAGD,IAAI/B,YAAY,MAAM,GAAG,GAAG;YAC1BgC,OAAO,IAAI,CAAC;gBACV,MAAM;gBACN,SACE;YACJ;YAEAA,OAAO,IAAI,CAAC;gBACV,MAAM;gBACN,SAAShC,YAAY,GAAG,CAAC,CAACC,aAAgB;wBACxC,MAAM;wBACN,WAAW;4BACT,KAAKA;wBACP;oBACF;YACF;QACF;QAEA,IAAIU,QAAQ,MAAM,IAAIA,QAAQ,OAAO,EAEnC,OAAO,MAAM6B,OAAOR,QAAQK,aAAa,IAAI,EAAER,aAAa;YAC1D,QAAQ;YACR,SAASlB,QAAQ,OAAO;QAC1B;QACK;YAEL,MAAMwB,WAAW,MAAMC,yBACrBJ,QACAK,aAAa,IAAI,EACjBR;YAGF,IAAIM,UAAU,WAAW,AAA4B,YAA5B,OAAOA,SAAS,OAAO,EAC9C,OAAO;gBACL,SAASA,SAAS,OAAO;gBACzB,OAAOA,SAAS,KAAK;gBACrB,YAAY;YACd;YAGF,MAAM,IAAIR,MAAM;QAClB;IACF,EAAE,OAAOW,OAAO;QACd,MAAM,IAAIX,MAAM,CAAC,8BAA8B,EAAEW,OAAO;IAC1D;AACF"}
+{"version":3,"file":"ai-model/prompt/yaml-generator.mjs","sources":["../../../../src/ai-model/prompt/yaml-generator.ts"],"sourcesContent":["import type {\n  StreamingAIResponse,\n  StreamingCodeGenerationOptions,\n} from '@/types';\nimport { YAML_EXAMPLE_CODE } from '@midscene/shared/constants';\nimport type { IModelConfig } from '@midscene/shared/env';\nimport {\n  type ChatCompletionMessageParam,\n  callAI,\n  callAIWithStringResponse,\n} from '../index';\n\n// Common interfaces for test generation (shared between YAML and Playwright)\nexport interface EventCounts {\n  navigation: number;\n  click: number;\n  input: number;\n  scroll: number;\n  total: number;\n}\n\nexport interface InputDescription {\n  description: string;\n  value: string;\n}\n\nexport interface ProcessedEvent {\n  type: string;\n  timestamp: number;\n  url?: string;\n  title?: string;\n  elementDescription?: string;\n  value?: string;\n  pageInfo?: any;\n  elementRect?: any;\n}\n\nexport interface EventSummary {\n  testName: string;\n  startUrl: string;\n  eventCounts: EventCounts;\n  urls: string[];\n  clickDescriptions: string[];\n  inputDescriptions: InputDescription[];\n  events: ProcessedEvent[];\n}\n\n// Common ChromeRecordedEvent interface\nexport interface ChromeRecordedEvent {\n  type: string;\n  timestamp: number;\n  url?: string;\n  title?: string;\n  elementDescription?: string;\n  value?: string;\n  pageInfo?: any;\n  elementRect?: any;\n  screenshotBefore?: string;\n  screenshotAfter?: string;\n  screenshotWithBox?: string;\n}\n\nexport interface YamlGenerationOptions {\n  testName?: string;\n  includeTimestamps?: boolean;\n  maxScreenshots?: number;\n  description?: string;\n}\n\nexport interface FilteredEvents {\n  navigationEvents: ChromeRecordedEvent[];\n  clickEvents: ChromeRecordedEvent[];\n  inputEvents: ChromeRecordedEvent[];\n  scrollEvents: ChromeRecordedEvent[];\n}\n\n// Common utility functions (shared between YAML and Playwright generators)\n\n/**\n * Get screenshots from events for LLM context\n */\nexport const getScreenshotsForLLM = (\n  events: ChromeRecordedEvent[],\n  maxScreenshots = 1,\n): string[] => {\n  // Find events with screenshots, prioritizing navigation and click events\n  const eventsWithScreenshots = events.filter(\n    (event) =>\n      event.screenshotBefore ||\n      event.screenshotAfter ||\n      event.screenshotWithBox,\n  );\n\n  // Sort them by priority (navigation first, then clicks, then others)\n  const sortedEvents = [...eventsWithScreenshots].sort((a, b) => {\n    if (a.type === 'navigation' && b.type !== 'navigation') return -1;\n    if (a.type !== 'navigation' && b.type === 'navigation') return 1;\n    if (a.type === 'click' && b.type !== 'click') return -1;\n    if (a.type !== 'click' && b.type === 'click') return 1;\n    return 0;\n  });\n\n  // Extract up to maxScreenshots screenshots\n  const screenshots: string[] = [];\n  for (const event of sortedEvents) {\n    // Prefer the most informative screenshot\n    const screenshot =\n      event.screenshotWithBox ||\n      event.screenshotAfter ||\n      event.screenshotBefore;\n    if (screenshot && !screenshots.includes(screenshot)) {\n      screenshots.push(screenshot);\n      if (screenshots.length >= maxScreenshots) break;\n    }\n  }\n\n  return screenshots;\n};\n\n/**\n * Filter events by type for easier processing\n */\nexport const filterEventsByType = (\n  events: ChromeRecordedEvent[],\n): FilteredEvents => {\n  return {\n    navigationEvents: events.filter((event) => event.type === 'navigation'),\n    clickEvents: events.filter((event) => event.type === 'click'),\n    inputEvents: events.filter((event) => event.type === 'input'),\n    scrollEvents: events.filter((event) => event.type === 'scroll'),\n  };\n};\n\n/**\n * Create event counts summary\n */\nexport const createEventCounts = (\n  filteredEvents: FilteredEvents,\n  totalEvents: number,\n): EventCounts => {\n  return {\n    navigation: filteredEvents.navigationEvents.length,\n    click: filteredEvents.clickEvents.length,\n    input: filteredEvents.inputEvents.length,\n    scroll: filteredEvents.scrollEvents.length,\n    total: totalEvents,\n  };\n};\n\n/**\n * Extract input descriptions from input events\n */\nexport const extractInputDescriptions = (\n  inputEvents: ChromeRecordedEvent[],\n): InputDescription[] => {\n  return inputEvents\n    .map((event) => ({\n      description: event.elementDescription || '',\n      value: event.value || '',\n    }))\n    .filter((item) => item.description && item.value);\n};\n\n/**\n * Process events for LLM consumption\n */\nexport const processEventsForLLM = (\n  events: ChromeRecordedEvent[],\n): ProcessedEvent[] => {\n  return events.map((event) => ({\n    type: event.type,\n    timestamp: event.timestamp,\n    url: event.url,\n    title: event.title,\n    elementDescription: event.elementDescription,\n    value: event.value,\n    pageInfo: event.pageInfo,\n    elementRect: event.elementRect,\n  }));\n};\n\n/**\n * Prepare comprehensive event summary for LLM\n */\nexport const prepareEventSummary = (\n  events: ChromeRecordedEvent[],\n  options: { testName?: string; maxScreenshots?: number } = {},\n): EventSummary => {\n  const filteredEvents = filterEventsByType(events);\n  const eventCounts = createEventCounts(filteredEvents, events.length);\n\n  // Extract useful information from events\n  const startUrl =\n    filteredEvents.navigationEvents.length > 0\n      ? filteredEvents.navigationEvents[0].url || ''\n      : '';\n\n  const clickDescriptions = filteredEvents.clickEvents\n    .map((event) => event.elementDescription)\n    .filter((desc): desc is string => Boolean(desc))\n    .slice(0, 10);\n\n  const inputDescriptions = extractInputDescriptions(\n    filteredEvents.inputEvents,\n  ).slice(0, 10);\n\n  const urls = filteredEvents.navigationEvents\n    .map((e) => e.url)\n    .filter((url): url is string => Boolean(url))\n    .slice(0, 5);\n\n  const processedEvents = processEventsForLLM(events);\n\n  return {\n    testName: options.testName || 'Automated test from recorded events',\n    startUrl,\n    eventCounts,\n    urls,\n    clickDescriptions,\n    inputDescriptions,\n    events: processedEvents,\n  };\n};\n\n/**\n * Create message content for LLM with optional screenshots\n */\nexport const createMessageContent = (\n  promptText: string,\n  screenshots: string[] = [],\n  includeScreenshots = true,\n) => {\n  const messageContent: any[] = [\n    {\n      type: 'text',\n      text: promptText,\n    },\n  ];\n\n  // Add screenshots if available and requested\n  if (includeScreenshots && screenshots.length > 0) {\n    messageContent.unshift({\n      type: 'text',\n      text: 'Here are screenshots from the recording session to help you understand the context:',\n    });\n\n    screenshots.forEach((screenshot) => {\n      messageContent.push({\n        type: 'image_url',\n        image_url: {\n          url: screenshot,\n        },\n      });\n    });\n  }\n\n  return messageContent;\n};\n\n/**\n * Validate events before processing\n */\nexport const validateEvents = (events: ChromeRecordedEvent[]): void => {\n  if (!events.length) {\n    throw new Error('No events provided for test generation');\n  }\n};\n\n// YAML-specific generation functions\n\n/**\n * Generates YAML test configuration from recorded events using AI\n */\nexport const generateYamlTest = async (\n  events: ChromeRecordedEvent[],\n  options: YamlGenerationOptions,\n  modelConfig: IModelConfig,\n): Promise<string> => {\n  try {\n    // Validate input\n    validateEvents(events);\n\n    // Prepare event summary using shared utilities\n    const summary = prepareEventSummary(events, {\n      testName: options.testName,\n      maxScreenshots: options.maxScreenshots || 3,\n    });\n\n    // Add YAML-specific options to summary\n    const yamlSummary = {\n      ...summary,\n      includeTimestamps: options.includeTimestamps || false,\n    };\n\n    // Get screenshots for visual context\n    const screenshots = getScreenshotsForLLM(\n      events,\n      options.maxScreenshots || 3,\n    );\n\n    // Use LLM to generate the YAML test configuration\n    const prompt: ChatCompletionMessageParam[] = [\n      {\n        role: 'system',\n        content: `You are an expert in Midscene.js YAML test generation. Generate clean, accurate YAML following these rules: ${YAML_EXAMPLE_CODE}`,\n      },\n      {\n        role: 'user',\n        content: `Generate YAML test for Midscene.js automation from recorded browser events.\n\nEvent Summary:\n${JSON.stringify(yamlSummary, null, 2)}\n\nConvert events:\n- navigation → target.url\n- click → aiTap with element description\n- input → aiInput with value and locate\n- scroll → aiScroll with appropriate direction\n- Add aiAssert for important state changes\n\nImportant: Return ONLY the raw YAML content. Do NOT wrap the response in markdown code blocks (no \\`\\`\\`yaml or \\`\\`\\`). Start directly with the YAML content.`,\n      },\n    ];\n\n    // Add screenshots if available and requested\n    if (screenshots.length > 0) {\n      prompt.push({\n        role: 'user',\n        content:\n          'Here are screenshots from the recording session to help you understand the context:',\n      });\n\n      prompt.push({\n        role: 'user',\n        content: screenshots.map((screenshot) => ({\n          type: 'image_url',\n          image_url: {\n            url: screenshot,\n          },\n        })),\n      });\n    }\n\n    const response = await callAIWithStringResponse(prompt, modelConfig);\n\n    if (response?.content && typeof response.content === 'string') {\n      return response.content;\n    }\n\n    throw new Error('Failed to generate YAML test configuration');\n  } catch (error) {\n    throw new Error(`Failed to generate YAML test: ${error}`);\n  }\n};\n\n/**\n * Generates YAML test configuration from recorded events using AI with streaming support\n */\nexport const generateYamlTestStream = async (\n  events: ChromeRecordedEvent[],\n  options: YamlGenerationOptions & StreamingCodeGenerationOptions,\n  modelConfig: IModelConfig,\n): Promise<StreamingAIResponse> => {\n  try {\n    // Validate input\n    validateEvents(events);\n\n    // Prepare event summary using shared utilities\n    const summary = prepareEventSummary(events, {\n      testName: options.testName,\n      maxScreenshots: options.maxScreenshots || 3,\n    });\n\n    // Add YAML-specific options to summary\n    const yamlSummary = {\n      ...summary,\n      includeTimestamps: options.includeTimestamps || false,\n    };\n\n    // Get screenshots for visual context\n    const screenshots = getScreenshotsForLLM(\n      events,\n      options.maxScreenshots || 3,\n    );\n\n    // Use LLM to generate the YAML test configuration\n    const prompt: ChatCompletionMessageParam[] = [\n      {\n        role: 'system',\n        content: `You are an expert in Midscene.js YAML test generation. Generate clean, accurate YAML following these rules: ${YAML_EXAMPLE_CODE}`,\n      },\n      {\n        role: 'user',\n        content: `Generate YAML test for Midscene.js automation from recorded browser events.\n\nEvent Summary:\n${JSON.stringify(yamlSummary, null, 2)}\n\nConvert events:\n- navigation → target.url\n- click → aiTap with element description\n- input → aiInput with value and locate\n- scroll → aiScroll with appropriate direction\n- Add aiAssert for important state changes\n\nImportant: Return ONLY the raw YAML content. Do NOT wrap the response in markdown code blocks (no \\`\\`\\`yaml or \\`\\`\\`). Start directly with the YAML content.`,\n      },\n    ];\n\n    // Add screenshots if available and requested\n    if (screenshots.length > 0) {\n      prompt.push({\n        role: 'user',\n        content:\n          'Here are screenshots from the recording session to help you understand the context:',\n      });\n\n      prompt.push({\n        role: 'user',\n        content: screenshots.map((screenshot) => ({\n          type: 'image_url',\n          image_url: {\n            url: screenshot,\n          },\n        })),\n      });\n    }\n\n    if (options.stream && options.onChunk) {\n      // Use streaming\n      return await callAI(prompt, modelConfig, {\n        stream: true,\n        onChunk: options.onChunk,\n      });\n    } else {\n      // Fallback to non-streaming\n      const response = await callAIWithStringResponse(prompt, modelConfig);\n\n      if (response?.content && typeof response.content === 'string') {\n        return {\n          content: response.content,\n          usage: response.usage,\n          isStreamed: false,\n        };\n      }\n\n      throw new Error('Failed to generate YAML test configuration');\n    }\n  } catch (error) {\n    throw new Error(`Failed to generate YAML test: ${error}`);\n  }\n};\n"],"names":["getScreenshotsForLLM","events","maxScreenshots","eventsWithScreenshots","event","sortedEvents","a","b","screenshots","screenshot","filterEventsByType","createEventCounts","filteredEvents","totalEvents","extractInputDescriptions","inputEvents","item","processEventsForLLM","prepareEventSummary","options","eventCounts","startUrl","clickDescriptions","desc","Boolean","inputDescriptions","urls","e","url","processedEvents","createMessageContent","promptText","includeScreenshots","messageContent","validateEvents","Error","generateYamlTest","modelConfig","summary","yamlSummary","prompt","YAML_EXAMPLE_CODE","JSON","response","callAIWithStringResponse","error","generateYamlTestStream","callAI"],"mappings":";;AAiFO,MAAMA,uBAAuB,CAClCC,QACAC,iBAAiB,CAAC;IAGlB,MAAMC,wBAAwBF,OAAO,MAAM,CACzC,CAACG,QACCA,MAAM,gBAAgB,IACtBA,MAAM,eAAe,IACrBA,MAAM,iBAAiB;IAI3B,MAAMC,eAAe;WAAIF;KAAsB,CAAC,IAAI,CAAC,CAACG,GAAGC;QACvD,IAAID,AAAW,iBAAXA,EAAE,IAAI,IAAqBC,AAAW,iBAAXA,EAAE,IAAI,EAAmB,OAAO;QAC/D,IAAID,AAAW,iBAAXA,EAAE,IAAI,IAAqBC,AAAW,iBAAXA,EAAE,IAAI,EAAmB,OAAO;QAC/D,IAAID,AAAW,YAAXA,EAAE,IAAI,IAAgBC,AAAW,YAAXA,EAAE,IAAI,EAAc,OAAO;QACrD,IAAID,AAAW,YAAXA,EAAE,IAAI,IAAgBC,AAAW,YAAXA,EAAE,IAAI,EAAc,OAAO;QACrD,OAAO;IACT;IAGA,MAAMC,cAAwB,EAAE;IAChC,KAAK,MAAMJ,SAASC,aAAc;QAEhC,MAAMI,aACJL,MAAM,iBAAiB,IACvBA,MAAM,eAAe,IACrBA,MAAM,gBAAgB;QACxB,IAAIK,cAAc,CAACD,YAAY,QAAQ,CAACC,aAAa;YACnDD,YAAY,IAAI,CAACC;YACjB,IAAID,YAAY,MAAM,IAAIN,gBAAgB;QAC5C;IACF;IAEA,OAAOM;AACT;AAKO,MAAME,qBAAqB,CAChCT,SAEO;QACL,kBAAkBA,OAAO,MAAM,CAAC,CAACG,QAAUA,AAAe,iBAAfA,MAAM,IAAI;QACrD,aAAaH,OAAO,MAAM,CAAC,CAACG,QAAUA,AAAe,YAAfA,MAAM,IAAI;QAChD,aAAaH,OAAO,MAAM,CAAC,CAACG,QAAUA,AAAe,YAAfA,MAAM,IAAI;QAChD,cAAcH,OAAO,MAAM,CAAC,CAACG,QAAUA,AAAe,aAAfA,MAAM,IAAI;IACnD;AAMK,MAAMO,oBAAoB,CAC/BC,gBACAC,cAEO;QACL,YAAYD,eAAe,gBAAgB,CAAC,MAAM;QAClD,OAAOA,eAAe,WAAW,CAAC,MAAM;QACxC,OAAOA,eAAe,WAAW,CAAC,MAAM;QACxC,QAAQA,eAAe,YAAY,CAAC,MAAM;QAC1C,OAAOC;IACT;AAMK,MAAMC,2BAA2B,CACtCC,cAEOA,YACJ,GAAG,CAAC,CAACX,QAAW;YACf,aAAaA,MAAM,kBAAkB,IAAI;YACzC,OAAOA,MAAM,KAAK,IAAI;QACxB,IACC,MAAM,CAAC,CAACY,OAASA,KAAK,WAAW,IAAIA,KAAK,KAAK;AAM7C,MAAMC,sBAAsB,CACjChB,SAEOA,OAAO,GAAG,CAAC,CAACG,QAAW;YAC5B,MAAMA,MAAM,IAAI;YAChB,WAAWA,MAAM,SAAS;YAC1B,KAAKA,MAAM,GAAG;YACd,OAAOA,MAAM,KAAK;YAClB,oBAAoBA,MAAM,kBAAkB;YAC5C,OAAOA,MAAM,KAAK;YAClB,UAAUA,MAAM,QAAQ;YACxB,aAAaA,MAAM,WAAW;QAChC;AAMK,MAAMc,sBAAsB,CACjCjB,QACAkB,UAA0D,CAAC,CAAC;IAE5D,MAAMP,iBAAiBF,mBAAmBT;IAC1C,MAAMmB,cAAcT,kBAAkBC,gBAAgBX,OAAO,MAAM;IAGnE,MAAMoB,WACJT,eAAe,gBAAgB,CAAC,MAAM,GAAG,IACrCA,eAAe,gBAAgB,CAAC,EAAE,CAAC,GAAG,IAAI,KAC1C;IAEN,MAAMU,oBAAoBV,eAAe,WAAW,CACjD,GAAG,CAAC,CAACR,QAAUA,MAAM,kBAAkB,EACvC,MAAM,CAAC,CAACmB,OAAyBC,QAAQD,OACzC,KAAK,CAAC,GAAG;IAEZ,MAAME,oBAAoBX,yBACxBF,eAAe,WAAW,EAC1B,KAAK,CAAC,GAAG;IAEX,MAAMc,OAAOd,eAAe,gBAAgB,CACzC,GAAG,CAAC,CAACe,IAAMA,EAAE,GAAG,EAChB,MAAM,CAAC,CAACC,MAAuBJ,QAAQI,MACvC,KAAK,CAAC,GAAG;IAEZ,MAAMC,kBAAkBZ,oBAAoBhB;IAE5C,OAAO;QACL,UAAUkB,QAAQ,QAAQ,IAAI;QAC9BE;QACAD;QACAM;QACAJ;QACAG;QACA,QAAQI;IACV;AACF;AAKO,MAAMC,uBAAuB,CAClCC,YACAvB,cAAwB,EAAE,EAC1BwB,qBAAqB,IAAI;IAEzB,MAAMC,iBAAwB;QAC5B;YACE,MAAM;YACN,MAAMF;QACR;KACD;IAGD,IAAIC,sBAAsBxB,YAAY,MAAM,GAAG,GAAG;QAChDyB,eAAe,OAAO,CAAC;YACrB,MAAM;YACN,MAAM;QACR;QAEAzB,YAAY,OAAO,CAAC,CAACC;YACnBwB,eAAe,IAAI,CAAC;gBAClB,MAAM;gBACN,WAAW;oBACT,KAAKxB;gBACP;YACF;QACF;IACF;IAEA,OAAOwB;AACT;AAKO,MAAMC,iBAAiB,CAACjC;IAC7B,IAAI,CAACA,OAAO,MAAM,EAChB,MAAM,IAAIkC,MAAM;AAEpB;AAOO,MAAMC,mBAAmB,OAC9BnC,QACAkB,SACAkB;IAEA,IAAI;QAEFH,eAAejC;QAGf,MAAMqC,UAAUpB,oBAAoBjB,QAAQ;YAC1C,UAAUkB,QAAQ,QAAQ;YAC1B,gBAAgBA,QAAQ,cAAc,IAAI;QAC5C;QAGA,MAAMoB,cAAc;YAClB,GAAGD,OAAO;YACV,mBAAmBnB,QAAQ,iBAAiB,IAAI;QAClD;QAGA,MAAMX,cAAcR,qBAClBC,QACAkB,QAAQ,cAAc,IAAI;QAI5B,MAAMqB,SAAuC;YAC3C;gBACE,MAAM;gBACN,SAAS,CAAC,4GAA4G,EAAEC,mBAAmB;YAC7I;YACA;gBACE,MAAM;gBACN,SAAS,CAAC;;;AAGlB,EAAEC,KAAK,SAAS,CAACH,aAAa,MAAM,GAAG;;;;;;;;;8JASuH,CAAC;YACzJ;SACD;QAGD,IAAI/B,YAAY,MAAM,GAAG,GAAG;YAC1BgC,OAAO,IAAI,CAAC;gBACV,MAAM;gBACN,SACE;YACJ;YAEAA,OAAO,IAAI,CAAC;gBACV,MAAM;gBACN,SAAShC,YAAY,GAAG,CAAC,CAACC,aAAgB;wBACxC,MAAM;wBACN,WAAW;4BACT,KAAKA;wBACP;oBACF;YACF;QACF;QAEA,MAAMkC,WAAW,MAAMC,yBAAyBJ,QAAQH;QAExD,IAAIM,UAAU,WAAW,AAA4B,YAA5B,OAAOA,SAAS,OAAO,EAC9C,OAAOA,SAAS,OAAO;QAGzB,MAAM,IAAIR,MAAM;IAClB,EAAE,OAAOU,OAAO;QACd,MAAM,IAAIV,MAAM,CAAC,8BAA8B,EAAEU,OAAO;IAC1D;AACF;AAKO,MAAMC,yBAAyB,OACpC7C,QACAkB,SACAkB;IAEA,IAAI;QAEFH,eAAejC;QAGf,MAAMqC,UAAUpB,oBAAoBjB,QAAQ;YAC1C,UAAUkB,QAAQ,QAAQ;YAC1B,gBAAgBA,QAAQ,cAAc,IAAI;QAC5C;QAGA,MAAMoB,cAAc;YAClB,GAAGD,OAAO;YACV,mBAAmBnB,QAAQ,iBAAiB,IAAI;QAClD;QAGA,MAAMX,cAAcR,qBAClBC,QACAkB,QAAQ,cAAc,IAAI;QAI5B,MAAMqB,SAAuC;YAC3C;gBACE,MAAM;gBACN,SAAS,CAAC,4GAA4G,EAAEC,mBAAmB;YAC7I;YACA;gBACE,MAAM;gBACN,SAAS,CAAC;;;AAGlB,EAAEC,KAAK,SAAS,CAACH,aAAa,MAAM,GAAG;;;;;;;;;8JASuH,CAAC;YACzJ;SACD;QAGD,IAAI/B,YAAY,MAAM,GAAG,GAAG;YAC1BgC,OAAO,IAAI,CAAC;gBACV,MAAM;gBACN,SACE;YACJ;YAEAA,OAAO,IAAI,CAAC;gBACV,MAAM;gBACN,SAAShC,YAAY,GAAG,CAAC,CAACC,aAAgB;wBACxC,MAAM;wBACN,WAAW;4BACT,KAAKA;wBACP;oBACF;YACF;QACF;QAEA,IAAIU,QAAQ,MAAM,IAAIA,QAAQ,OAAO,EAEnC,OAAO,MAAM4B,OAAOP,QAAQH,aAAa;YACvC,QAAQ;YACR,SAASlB,QAAQ,OAAO;QAC1B;QACK;YAEL,MAAMwB,WAAW,MAAMC,yBAAyBJ,QAAQH;YAExD,IAAIM,UAAU,WAAW,AAA4B,YAA5B,OAAOA,SAAS,OAAO,EAC9C,OAAO;gBACL,SAASA,SAAS,OAAO;gBACzB,OAAOA,SAAS,KAAK;gBACrB,YAAY;YACd;YAGF,MAAM,IAAIR,MAAM;QAClB;IACF,EAAE,OAAOU,OAAO;QACd,MAAM,IAAIV,MAAM,CAAC,8BAA8B,EAAEU,OAAO;IAC1D;AACF"}

package/dist/es/ai-model/service-caller/index.mjs

@@ -3,7 +3,7 @@ import { getDebug } from "@midscene/shared/logger";
 import { assert, ifInBrowser } from "@midscene/shared/utils";
 import { jsonrepair } from "jsonrepair";
 import openai_0 from "openai";
-async function createChatClient({ AIActionTypeValue, modelConfig }) {
+async function createChatClient({ modelConfig }) {
     const { socksProxy, httpProxy, modelName, openaiBaseURL, openaiApiKey, openaiExtraConfig, modelDescription, uiTarsModelVersion: uiTarsVersion, vlMode, createOpenAIClient, timeout } = modelConfig;
     let proxyAgent;
     const debugProxy = getDebug('ai:call:proxy');
@@ -102,9 +102,8 @@ async function createChatClient({ AIActionTypeValue, modelConfig }) {
         vlMode
     };
 }
-async function callAI(messages, AIActionTypeValue, modelConfig, options) {
+async function callAI(messages, modelConfig, options) {
     const { completion, modelName, modelDescription, uiTarsVersion, vlMode } = await createChatClient({
-        AIActionTypeValue,
         modelConfig
     });
     const maxTokens = globalConfigManager.getEnvConfigValue(MIDSCENE_MODEL_MAX_TOKENS) ?? globalConfigManager.getEnvConfigValue(OPENAI_MAX_TOKENS);
@@ -240,8 +239,8 @@ async function callAI(messages, AIActionTypeValue, modelConfig, options) {
         throw newError;
     }
 }
-async function callAIWithObjectResponse(messages, AIActionTypeValue, modelConfig, options) {
-    const response = await callAI(messages, AIActionTypeValue, modelConfig, {
+async function callAIWithObjectResponse(messages, modelConfig, options) {
+    const response = await callAI(messages, modelConfig, {
         deepThink: options?.deepThink
     });
     assert(response, 'empty response');
@@ -255,8 +254,8 @@ async function callAIWithObjectResponse(messages, AIActionTypeValue, modelConfig
         reasoning_content: response.reasoning_content
     };
 }
-async function callAIWithStringResponse(msgs, AIActionTypeValue, modelConfig) {
-    const { content, usage } = await callAI(msgs, AIActionTypeValue, modelConfig);
+async function callAIWithStringResponse(msgs, modelConfig) {
+    const { content, usage } = await callAI(msgs, modelConfig);
     return {
         content,
         usage

package/dist/es/ai-model/service-caller/index.mjs.map

@@ -1 +1 @@
-{"version":3,"file":"ai-model/service-caller/index.mjs","sources":["../../../../src/ai-model/service-caller/index.ts"],"sourcesContent":["import {\n  AIResponseFormat,\n  type AIUsageInfo,\n  type DeepThinkOption,\n} from '@/types';\nimport type { CodeGenerationChunk, StreamingCallback } from '@/types';\nimport {\n  type IModelConfig,\n  MIDSCENE_LANGFUSE_DEBUG,\n  MIDSCENE_LANGSMITH_DEBUG,\n  MIDSCENE_MODEL_MAX_TOKENS,\n  OPENAI_MAX_TOKENS,\n  type TVlModeTypes,\n  type UITarsModelVersion,\n  globalConfigManager,\n} from '@midscene/shared/env';\n\nimport { getDebug } from '@midscene/shared/logger';\nimport { assert, ifInBrowser } from '@midscene/shared/utils';\nimport { jsonrepair } from 'jsonrepair';\nimport OpenAI from 'openai';\nimport type { ChatCompletionMessageParam } from 'openai/resources/index';\nimport type { Stream } from 'openai/streaming';\nimport type { AIActionType, AIArgs } from '../../common';\n\nasync function createChatClient({\n  AIActionTypeValue,\n  modelConfig,\n}: {\n  AIActionTypeValue: AIActionType;\n  modelConfig: IModelConfig;\n}): Promise<{\n  completion: OpenAI.Chat.Completions;\n  modelName: string;\n  modelDescription: string;\n  uiTarsVersion?: UITarsModelVersion;\n  vlMode: TVlModeTypes | undefined;\n}> {\n  const {\n    socksProxy,\n    httpProxy,\n    modelName,\n    openaiBaseURL,\n    openaiApiKey,\n    openaiExtraConfig,\n    modelDescription,\n    uiTarsModelVersion: uiTarsVersion,\n    vlMode,\n    createOpenAIClient,\n    timeout,\n  } = modelConfig;\n\n  let proxyAgent: any = undefined;\n  const debugProxy = getDebug('ai:call:proxy');\n\n  // Helper function to sanitize proxy URL for logging (remove credentials)\n  // Uses URL API instead of regex to avoid ReDoS vulnerabilities\n  const sanitizeProxyUrl = (url: string): string => {\n    try {\n      const parsed = new URL(url);\n      if (parsed.username) {\n        // Keep username for debugging, hide password for security\n        parsed.password = '****';\n        return parsed.href;\n      }\n      return url;\n    } catch {\n      // If URL parsing fails, return original URL (will be caught later)\n      return url;\n    }\n  };\n\n  if (httpProxy) {\n    debugProxy('using http proxy', sanitizeProxyUrl(httpProxy));\n    if (ifInBrowser) {\n      console.warn(\n        'HTTP proxy is configured but not supported in browser environment',\n      );\n    } else {\n      // Dynamic import with variable to avoid bundler static analysis\n      const moduleName = 'undici';\n      const { ProxyAgent } = await import(moduleName);\n      proxyAgent = new ProxyAgent({\n        uri: httpProxy,\n        // Note: authentication is handled via the URI (e.g., http://user:pass@proxy.com:8080)\n      });\n    }\n  } else if (socksProxy) {\n    debugProxy('using socks proxy', sanitizeProxyUrl(socksProxy));\n    if (ifInBrowser) {\n      console.warn(\n        'SOCKS proxy is configured but not supported in browser environment',\n      );\n    } else {\n      try {\n        // Dynamic import with variable to avoid bundler static analysis\n        const moduleName = 'fetch-socks';\n        const { socksDispatcher } = await import(moduleName);\n        // Parse SOCKS proxy URL (e.g., socks5://127.0.0.1:1080)\n        const proxyUrl = new URL(socksProxy);\n\n        // Validate hostname\n        if (!proxyUrl.hostname) {\n          throw new Error('SOCKS proxy URL must include a valid hostname');\n        }\n\n        // Validate and parse port\n        const port = Number.parseInt(proxyUrl.port, 10);\n        if (!proxyUrl.port || Number.isNaN(port)) {\n          throw new Error('SOCKS proxy URL must include a valid port');\n        }\n\n        // Parse SOCKS version from protocol\n        const protocol = proxyUrl.protocol.replace(':', '');\n        const socksType =\n          protocol === 'socks4' ? 4 : protocol === 'socks5' ? 5 : 5;\n\n        proxyAgent = socksDispatcher({\n          type: socksType,\n          host: proxyUrl.hostname,\n          port,\n          ...(proxyUrl.username\n            ? {\n                userId: decodeURIComponent(proxyUrl.username),\n                password: decodeURIComponent(proxyUrl.password || ''),\n              }\n            : {}),\n        });\n        debugProxy('socks proxy configured successfully', {\n          type: socksType,\n          host: proxyUrl.hostname,\n          port: port,\n        });\n      } catch (error) {\n        console.error('Failed to configure SOCKS proxy:', error);\n        throw new Error(\n          `Invalid SOCKS proxy URL: ${socksProxy}. Expected format: socks4://host:port, socks5://host:port, or with authentication: socks5://user:pass@host:port`,\n        );\n      }\n    }\n  }\n\n  const openAIOptions = {\n    baseURL: openaiBaseURL,\n    apiKey: openaiApiKey,\n    // Use fetchOptions.dispatcher for fetch-based SDK instead of httpAgent\n    // Note: Type assertion needed due to undici version mismatch between dependencies\n    ...(proxyAgent ? { fetchOptions: { dispatcher: proxyAgent as any } } : {}),\n    ...openaiExtraConfig,\n    ...(typeof timeout === 'number' ? { timeout } : {}),\n    dangerouslyAllowBrowser: true,\n  };\n\n  const baseOpenAI = new OpenAI(openAIOptions);\n\n  let openai: OpenAI = baseOpenAI;\n\n  // LangSmith wrapper\n  if (\n    openai &&\n    globalConfigManager.getEnvConfigInBoolean(MIDSCENE_LANGSMITH_DEBUG)\n  ) {\n    if (ifInBrowser) {\n      throw new Error('langsmith is not supported in browser');\n    }\n    console.log('DEBUGGING MODE: langsmith wrapper enabled');\n    // Use variable to prevent static analysis by bundlers\n    const langsmithModule = 'langsmith/wrappers';\n    const { wrapOpenAI } = await import(langsmithModule);\n    openai = wrapOpenAI(openai);\n  }\n\n  // Langfuse wrapper\n  if (\n    openai &&\n    globalConfigManager.getEnvConfigInBoolean(MIDSCENE_LANGFUSE_DEBUG)\n  ) {\n    if (ifInBrowser) {\n      throw new Error('langfuse is not supported in browser');\n    }\n    console.log('DEBUGGING MODE: langfuse wrapper enabled');\n    // Use variable to prevent static analysis by bundlers\n    const langfuseModule = 'langfuse';\n    const { observeOpenAI } = await import(langfuseModule);\n    openai = observeOpenAI(openai);\n  }\n\n  if (createOpenAIClient) {\n    const wrappedClient = await createOpenAIClient(baseOpenAI, openAIOptions);\n\n    if (wrappedClient) {\n      openai = wrappedClient as OpenAI;\n    }\n  }\n\n  return {\n    completion: openai.chat.completions,\n    modelName,\n    modelDescription,\n    uiTarsVersion,\n    vlMode,\n  };\n}\n\nexport async function callAI(\n  messages: ChatCompletionMessageParam[],\n  AIActionTypeValue: AIActionType,\n  modelConfig: IModelConfig,\n  options?: {\n    stream?: boolean;\n    onChunk?: StreamingCallback;\n    deepThink?: DeepThinkOption;\n  },\n): Promise<{\n  content: string;\n  reasoning_content?: string;\n  usage?: AIUsageInfo;\n  isStreamed: boolean;\n}> {\n  const { completion, modelName, modelDescription, uiTarsVersion, vlMode } =\n    await createChatClient({\n      AIActionTypeValue,\n      modelConfig,\n    });\n\n  const maxTokens =\n    globalConfigManager.getEnvConfigValue(MIDSCENE_MODEL_MAX_TOKENS) ??\n    globalConfigManager.getEnvConfigValue(OPENAI_MAX_TOKENS);\n  const debugCall = getDebug('ai:call');\n  const debugProfileStats = getDebug('ai:profile:stats');\n  const debugProfileDetail = getDebug('ai:profile:detail');\n\n  const startTime = Date.now();\n  const temperature = modelConfig.temperature ?? 0;\n\n  const isStreaming = options?.stream && options?.onChunk;\n  let content: string | undefined;\n  let accumulated = '';\n  let accumulatedReasoning = '';\n  let usage: OpenAI.CompletionUsage | undefined;\n  let timeCost: number | undefined;\n\n  const buildUsageInfo = (usageData?: OpenAI.CompletionUsage) => {\n    if (!usageData) return undefined;\n\n    const cachedInputTokens = (\n      usageData as { prompt_tokens_details?: { cached_tokens?: number } }\n    )?.prompt_tokens_details?.cached_tokens;\n\n    return {\n      prompt_tokens: usageData.prompt_tokens ?? 0,\n      completion_tokens: usageData.completion_tokens ?? 0,\n      total_tokens: usageData.total_tokens ?? 0,\n      cached_input: cachedInputTokens ?? 0,\n      time_cost: timeCost ?? 0,\n      model_name: modelName,\n      model_description: modelDescription,\n      intent: modelConfig.intent,\n    } satisfies AIUsageInfo;\n  };\n\n  const commonConfig = {\n    temperature,\n    stream: !!isStreaming,\n    max_tokens: typeof maxTokens === 'number' ? maxTokens : undefined,\n    ...(vlMode === 'qwen2.5-vl' // qwen vl v2 specific config\n      ? {\n          vl_high_resolution_images: true,\n        }\n      : {}),\n  };\n  const {\n    config: deepThinkConfig,\n    debugMessage,\n    warningMessage,\n  } = resolveDeepThinkConfig({\n    deepThink: options?.deepThink,\n    vlMode,\n  });\n  if (debugMessage) {\n    debugCall(debugMessage);\n  }\n  if (warningMessage) {\n    debugCall(warningMessage);\n    console.warn(warningMessage);\n  }\n\n  try {\n    debugCall(\n      `sending ${isStreaming ? 'streaming ' : ''}request to ${modelName}`,\n    );\n\n    if (isStreaming) {\n      const stream = (await completion.create(\n        {\n          model: modelName,\n          messages,\n          ...commonConfig,\n          ...deepThinkConfig,\n        },\n        {\n          stream: true,\n        },\n      )) as Stream<OpenAI.Chat.Completions.ChatCompletionChunk> & {\n        _request_id?: string | null;\n      };\n\n      for await (const chunk of stream) {\n        const content = chunk.choices?.[0]?.delta?.content || '';\n        const reasoning_content =\n          (chunk.choices?.[0]?.delta as any)?.reasoning_content || '';\n\n        // Check for usage info in any chunk (OpenAI provides usage in separate chunks)\n        if (chunk.usage) {\n          usage = chunk.usage;\n        }\n\n        if (content || reasoning_content) {\n          accumulated += content;\n          accumulatedReasoning += reasoning_content;\n          const chunkData: CodeGenerationChunk = {\n            content,\n            reasoning_content,\n            accumulated,\n            isComplete: false,\n            usage: undefined,\n          };\n          options.onChunk!(chunkData);\n        }\n\n        // Check if stream is complete\n        if (chunk.choices?.[0]?.finish_reason) {\n          timeCost = Date.now() - startTime;\n\n          // If usage is not available from the stream, provide a basic usage info\n          if (!usage) {\n            // Estimate token counts based on content length (rough approximation)\n            const estimatedTokens = Math.max(\n              1,\n              Math.floor(accumulated.length / 4),\n            );\n            usage = {\n              prompt_tokens: estimatedTokens,\n              completion_tokens: estimatedTokens,\n              total_tokens: estimatedTokens * 2,\n            };\n          }\n\n          // Send final chunk\n          const finalChunk: CodeGenerationChunk = {\n            content: '',\n            accumulated,\n            reasoning_content: '',\n            isComplete: true,\n            usage: buildUsageInfo(usage),\n          };\n          options.onChunk!(finalChunk);\n          break;\n        }\n      }\n      content = accumulated;\n      debugProfileStats(\n        `streaming model, ${modelName}, mode, ${vlMode || 'default'}, cost-ms, ${timeCost}, temperature, ${temperature ?? ''}`,\n      );\n    } else {\n      const result = await completion.create({\n        model: modelName,\n        messages,\n        ...commonConfig,\n        ...deepThinkConfig,\n      } as any);\n      timeCost = Date.now() - startTime;\n\n      debugProfileStats(\n        `model, ${modelName}, mode, ${vlMode || 'default'}, ui-tars-version, ${uiTarsVersion}, prompt-tokens, ${result.usage?.prompt_tokens || ''}, completion-tokens, ${result.usage?.completion_tokens || ''}, total-tokens, ${result.usage?.total_tokens || ''}, cost-ms, ${timeCost}, requestId, ${result._request_id || ''}, temperature, ${temperature ?? ''}`,\n      );\n\n      debugProfileDetail(`model usage detail: ${JSON.stringify(result.usage)}`);\n\n      assert(\n        result.choices,\n        `invalid response from LLM service: ${JSON.stringify(result)}`,\n      );\n      content = result.choices[0].message.content!;\n      accumulatedReasoning =\n        (result.choices[0].message as any)?.reasoning_content || '';\n      usage = result.usage;\n    }\n\n    debugCall(`response reasoning content: ${accumulatedReasoning}`);\n    debugCall(`response content: ${content}`);\n    assert(content, 'empty content');\n\n    // Ensure we always have usage info for streaming responses\n    if (isStreaming && !usage) {\n      // Estimate token counts based on content length (rough approximation)\n      const estimatedTokens = Math.max(\n        1,\n        Math.floor((content || '').length / 4),\n      );\n      usage = {\n        prompt_tokens: estimatedTokens,\n        completion_tokens: estimatedTokens,\n        total_tokens: estimatedTokens * 2,\n      } as OpenAI.CompletionUsage;\n    }\n\n    return {\n      content: content || '',\n      reasoning_content: accumulatedReasoning || undefined,\n      usage: buildUsageInfo(usage),\n      isStreamed: !!isStreaming,\n    };\n  } catch (e: any) {\n    console.error(' call AI error', e);\n    const newError = new Error(\n      `failed to call ${isStreaming ? 'streaming ' : ''}AI model service (${modelName}): ${e.message}\\nTrouble shooting: https://midscenejs.com/model-provider.html`,\n      {\n        cause: e,\n      },\n    );\n    throw newError;\n  }\n}\n\nexport async function callAIWithObjectResponse<T>(\n  messages: ChatCompletionMessageParam[],\n  AIActionTypeValue: AIActionType,\n  modelConfig: IModelConfig,\n  options?: {\n    deepThink?: DeepThinkOption;\n  },\n): Promise<{\n  content: T;\n  contentString: string;\n  usage?: AIUsageInfo;\n  reasoning_content?: string;\n}> {\n  const response = await callAI(messages, AIActionTypeValue, modelConfig, {\n    deepThink: options?.deepThink,\n  });\n  assert(response, 'empty response');\n  const vlMode = modelConfig.vlMode;\n  const jsonContent = safeParseJson(response.content, vlMode);\n  assert(\n    typeof jsonContent === 'object',\n    `failed to parse json response from model (${modelConfig.modelName}): ${response.content}`,\n  );\n  return {\n    content: jsonContent,\n    contentString: response.content,\n    usage: response.usage,\n    reasoning_content: response.reasoning_content,\n  };\n}\n\nexport async function callAIWithStringResponse(\n  msgs: AIArgs,\n  AIActionTypeValue: AIActionType,\n  modelConfig: IModelConfig,\n): Promise<{ content: string; usage?: AIUsageInfo }> {\n  const { content, usage } = await callAI(msgs, AIActionTypeValue, modelConfig);\n  return { content, usage };\n}\n\nexport function extractJSONFromCodeBlock(response: string) {\n  try {\n    // First, try to match a JSON object directly in the response\n    const jsonMatch = response.match(/^\\s*(\\{[\\s\\S]*\\})\\s*$/);\n    if (jsonMatch) {\n      return jsonMatch[1];\n    }\n\n    // If no direct JSON object is found, try to extract JSON from a code block\n    const codeBlockMatch = response.match(\n      /```(?:json)?\\s*(\\{[\\s\\S]*?\\})\\s*```/,\n    );\n    if (codeBlockMatch) {\n      return codeBlockMatch[1];\n    }\n\n    // If no code block is found, try to find a JSON-like structure in the text\n    const jsonLikeMatch = response.match(/\\{[\\s\\S]*\\}/);\n    if (jsonLikeMatch) {\n      return jsonLikeMatch[0];\n    }\n  } catch {}\n  // If no JSON-like structure is found, return the original response\n  return response;\n}\n\nexport function preprocessDoubaoBboxJson(input: string) {\n  if (input.includes('bbox')) {\n    // when its values like 940 445 969 490, replace all /\\d+\\s+\\d+/g with /$1,$2/g\n    while (/\\d+\\s+\\d+/.test(input)) {\n      input = input.replace(/(\\d+)\\s+(\\d+)/g, '$1,$2');\n    }\n  }\n  return input;\n}\n\nexport function resolveDeepThinkConfig({\n  deepThink,\n  vlMode,\n}: {\n  deepThink?: DeepThinkOption;\n  vlMode?: TVlModeTypes;\n}): {\n  config: Record<string, unknown>;\n  debugMessage?: string;\n  warningMessage?: string;\n} {\n  const normalizedDeepThink = deepThink === 'unset' ? undefined : deepThink;\n\n  if (normalizedDeepThink === undefined) {\n    return { config: {}, debugMessage: undefined };\n  }\n\n  if (vlMode === 'qwen3-vl') {\n    return {\n      config: { enable_thinking: normalizedDeepThink },\n      debugMessage: `deepThink mapped to enable_thinking=${normalizedDeepThink} for qwen3-vl`,\n    };\n  }\n\n  if (vlMode === 'doubao-vision') {\n    return {\n      config: {\n        thinking: { type: normalizedDeepThink ? 'enabled' : 'disabled' },\n      },\n      debugMessage: `deepThink mapped to thinking.type=${normalizedDeepThink ? 'enabled' : 'disabled'} for doubao-vision`,\n    };\n  }\n\n  return {\n    config: {},\n    debugMessage: `deepThink ignored: unsupported model_family \"${vlMode ?? 'default'}\"`,\n    warningMessage: `The \"deepThink\" option is not supported for model_family \"${vlMode ?? 'default'}\".`,\n  };\n}\n\n/**\n * Normalize a parsed JSON object by trimming whitespace from:\n * 1. All object keys (e.g., \" prompt \" -> \"prompt\")\n * 2. All string values (e.g., \" Tap \" -> \"Tap\")\n * This handles LLM output that may include leading/trailing spaces.\n */\nfunction normalizeJsonObject(obj: any): any {\n  // Handle null and undefined\n  if (obj === null || obj === undefined) {\n    return obj;\n  }\n\n  // Handle arrays - recursively normalize each element\n  if (Array.isArray(obj)) {\n    return obj.map((item) => normalizeJsonObject(item));\n  }\n\n  // Handle objects\n  if (typeof obj === 'object') {\n    const normalized: any = {};\n\n    for (const [key, value] of Object.entries(obj)) {\n      // Trim the key to remove leading/trailing spaces\n      const trimmedKey = key.trim();\n\n      // Recursively normalize the value\n      let normalizedValue = normalizeJsonObject(value);\n\n      // Trim all string values\n      if (typeof normalizedValue === 'string') {\n        normalizedValue = normalizedValue.trim();\n      }\n\n      normalized[trimmedKey] = normalizedValue;\n    }\n\n    return normalized;\n  }\n\n  // Handle primitive strings\n  if (typeof obj === 'string') {\n    return obj.trim();\n  }\n\n  // Return other primitives as-is\n  return obj;\n}\n\nexport function safeParseJson(input: string, vlMode: TVlModeTypes | undefined) {\n  const cleanJsonString = extractJSONFromCodeBlock(input);\n  // match the point\n  if (cleanJsonString?.match(/\\((\\d+),(\\d+)\\)/)) {\n    return cleanJsonString\n      .match(/\\((\\d+),(\\d+)\\)/)\n      ?.slice(1)\n      .map(Number);\n  }\n\n  let parsed: any;\n  let lastError: unknown;\n  try {\n    parsed = JSON.parse(cleanJsonString);\n    return normalizeJsonObject(parsed);\n  } catch (error) {\n    lastError = error;\n  }\n  try {\n    parsed = JSON.parse(jsonrepair(cleanJsonString));\n    return normalizeJsonObject(parsed);\n  } catch (error) {\n    lastError = error;\n  }\n\n  if (vlMode === 'doubao-vision' || vlMode === 'vlm-ui-tars') {\n    const jsonString = preprocessDoubaoBboxJson(cleanJsonString);\n    try {\n      parsed = JSON.parse(jsonrepair(jsonString));\n      return normalizeJsonObject(parsed);\n    } catch (error) {\n      lastError = error;\n    }\n  }\n  throw Error(\n    `failed to parse LLM response into JSON. Error - ${String(\n      lastError ?? 'unknown error',\n    )}. Response - \\n ${input}`,\n  );\n}\n"],"names":["createChatClient","AIActionTypeValue","modelConfig","socksProxy","httpProxy","modelName","openaiBaseURL","openaiApiKey","openaiExtraConfig","modelDescription","uiTarsVersion","vlMode","createOpenAIClient","timeout","proxyAgent","debugProxy","getDebug","sanitizeProxyUrl","url","parsed","URL","ifInBrowser","console","moduleName","ProxyAgent","socksDispatcher","proxyUrl","Error","port","Number","protocol","socksType","decodeURIComponent","error","openAIOptions","baseOpenAI","OpenAI","openai","globalConfigManager","MIDSCENE_LANGSMITH_DEBUG","langsmithModule","wrapOpenAI","MIDSCENE_LANGFUSE_DEBUG","langfuseModule","observeOpenAI","wrappedClient","callAI","messages","options","completion","maxTokens","MIDSCENE_MODEL_MAX_TOKENS","OPENAI_MAX_TOKENS","debugCall","debugProfileStats","debugProfileDetail","startTime","Date","temperature","isStreaming","content","accumulated","accumulatedReasoning","usage","timeCost","buildUsageInfo","usageData","cachedInputTokens","commonConfig","undefined","deepThinkConfig","debugMessage","warningMessage","resolveDeepThinkConfig","stream","chunk","reasoning_content","chunkData","estimatedTokens","Math","finalChunk","result","JSON","assert","e","newError","callAIWithObjectResponse","response","jsonContent","safeParseJson","callAIWithStringResponse","msgs","extractJSONFromCodeBlock","jsonMatch","codeBlockMatch","jsonLikeMatch","preprocessDoubaoBboxJson","input","deepThink","normalizedDeepThink","normalizeJsonObject","obj","Array","item","normalized","key","value","Object","trimmedKey","normalizedValue","cleanJsonString","lastError","jsonrepair","jsonString","String"],"mappings":";;;;;AAyBA,eAAeA,iBAAiB,EAC9BC,iBAAiB,EACjBC,WAAW,EAIZ;IAOC,MAAM,EACJC,UAAU,EACVC,SAAS,EACTC,SAAS,EACTC,aAAa,EACbC,YAAY,EACZC,iBAAiB,EACjBC,gBAAgB,EAChB,oBAAoBC,aAAa,EACjCC,MAAM,EACNC,kBAAkB,EAClBC,OAAO,EACR,GAAGX;IAEJ,IAAIY;IACJ,MAAMC,aAAaC,SAAS;IAI5B,MAAMC,mBAAmB,CAACC;QACxB,IAAI;YACF,MAAMC,SAAS,IAAIC,IAAIF;YACvB,IAAIC,OAAO,QAAQ,EAAE;gBAEnBA,OAAO,QAAQ,GAAG;gBAClB,OAAOA,OAAO,IAAI;YACpB;YACA,OAAOD;QACT,EAAE,OAAM;YAEN,OAAOA;QACT;IACF;IAEA,IAAId,WAAW;QACbW,WAAW,oBAAoBE,iBAAiBb;QAChD,IAAIiB,aACFC,QAAQ,IAAI,CACV;aAEG;YAEL,MAAMC,aAAa;YACnB,MAAM,EAAEC,UAAU,EAAE,GAAG,MAAM,MAAM,CAACD;YACpCT,aAAa,IAAIU,WAAW;gBAC1B,KAAKpB;YAEP;QACF;IACF,OAAO,IAAID,YAAY;QACrBY,WAAW,qBAAqBE,iBAAiBd;QACjD,IAAIkB,aACFC,QAAQ,IAAI,CACV;aAGF,IAAI;YAEF,MAAMC,aAAa;YACnB,MAAM,EAAEE,eAAe,EAAE,GAAG,MAAM,MAAM,CAACF;YAEzC,MAAMG,WAAW,IAAIN,IAAIjB;YAGzB,IAAI,CAACuB,SAAS,QAAQ,EACpB,MAAM,IAAIC,MAAM;YAIlB,MAAMC,OAAOC,OAAO,QAAQ,CAACH,SAAS,IAAI,EAAE;YAC5C,IAAI,CAACA,SAAS,IAAI,IAAIG,OAAO,KAAK,CAACD,OACjC,MAAM,IAAID,MAAM;YAIlB,MAAMG,WAAWJ,SAAS,QAAQ,CAAC,OAAO,CAAC,KAAK;YAChD,MAAMK,YACJD,AAAa,aAAbA,WAAwB,IAAIA,AAAa,aAAbA,WAAwB,IAAI;YAE1DhB,aAAaW,gBAAgB;gBAC3B,MAAMM;gBACN,MAAML,SAAS,QAAQ;gBACvBE;gBACA,GAAIF,SAAS,QAAQ,GACjB;oBACE,QAAQM,mBAAmBN,SAAS,QAAQ;oBAC5C,UAAUM,mBAAmBN,SAAS,QAAQ,IAAI;gBACpD,IACA,CAAC,CAAC;YACR;YACAX,WAAW,uCAAuC;gBAChD,MAAMgB;gBACN,MAAML,SAAS,QAAQ;gBACvB,MAAME;YACR;QACF,EAAE,OAAOK,OAAO;YACdX,QAAQ,KAAK,CAAC,oCAAoCW;YAClD,MAAM,IAAIN,MACR,CAAC,yBAAyB,EAAExB,WAAW,+GAA+G,CAAC;QAE3J;IAEJ;IAEA,MAAM+B,gBAAgB;QACpB,SAAS5B;QACT,QAAQC;QAGR,GAAIO,aAAa;YAAE,cAAc;gBAAE,YAAYA;YAAkB;QAAE,IAAI,CAAC,CAAC;QACzE,GAAGN,iBAAiB;QACpB,GAAI,AAAmB,YAAnB,OAAOK,UAAuB;YAAEA;QAAQ,IAAI,CAAC,CAAC;QAClD,yBAAyB;IAC3B;IAEA,MAAMsB,aAAa,IAAIC,SAAOF;IAE9B,IAAIG,SAAiBF;IAGrB,IACEE,UACAC,oBAAoB,qBAAqB,CAACC,2BAC1C;QACA,IAAIlB,aACF,MAAM,IAAIM,MAAM;QAElBL,QAAQ,GAAG,CAAC;QAEZ,MAAMkB,kBAAkB;QACxB,MAAM,EAAEC,UAAU,EAAE,GAAG,MAAM,MAAM,CAACD;QACpCH,SAASI,WAAWJ;IACtB;IAGA,IACEA,UACAC,oBAAoB,qBAAqB,CAACI,0BAC1C;QACA,IAAIrB,aACF,MAAM,IAAIM,MAAM;QAElBL,QAAQ,GAAG,CAAC;QAEZ,MAAMqB,iBAAiB;QACvB,MAAM,EAAEC,aAAa,EAAE,GAAG,MAAM,MAAM,CAACD;QACvCN,SAASO,cAAcP;IACzB;IAEA,IAAIzB,oBAAoB;QACtB,MAAMiC,gBAAgB,MAAMjC,mBAAmBuB,YAAYD;QAE3D,IAAIW,eACFR,SAASQ;IAEb;IAEA,OAAO;QACL,YAAYR,OAAO,IAAI,CAAC,WAAW;QACnChC;QACAI;QACAC;QACAC;IACF;AACF;AAEO,eAAemC,OACpBC,QAAsC,EACtC9C,iBAA+B,EAC/BC,WAAyB,EACzB8C,OAIC;IAOD,MAAM,EAAEC,UAAU,EAAE5C,SAAS,EAAEI,gBAAgB,EAAEC,aAAa,EAAEC,MAAM,EAAE,GACtE,MAAMX,iBAAiB;QACrBC;QACAC;IACF;IAEF,MAAMgD,YACJZ,oBAAoB,iBAAiB,CAACa,8BACtCb,oBAAoB,iBAAiB,CAACc;IACxC,MAAMC,YAAYrC,SAAS;IAC3B,MAAMsC,oBAAoBtC,SAAS;IACnC,MAAMuC,qBAAqBvC,SAAS;IAEpC,MAAMwC,YAAYC,KAAK,GAAG;IAC1B,MAAMC,cAAcxD,YAAY,WAAW,IAAI;IAE/C,MAAMyD,cAAcX,SAAS,UAAUA,SAAS;IAChD,IAAIY;IACJ,IAAIC,cAAc;IAClB,IAAIC,uBAAuB;IAC3B,IAAIC;IACJ,IAAIC;IAEJ,MAAMC,iBAAiB,CAACC;QACtB,IAAI,CAACA,WAAW;QAEhB,MAAMC,oBACJD,WACC,uBAAuB;QAE1B,OAAO;YACL,eAAeA,UAAU,aAAa,IAAI;YAC1C,mBAAmBA,UAAU,iBAAiB,IAAI;YAClD,cAAcA,UAAU,YAAY,IAAI;YACxC,cAAcC,qBAAqB;YACnC,WAAWH,YAAY;YACvB,YAAY3D;YACZ,mBAAmBI;YACnB,QAAQP,YAAY,MAAM;QAC5B;IACF;IAEA,MAAMkE,eAAe;QACnBV;QACA,QAAQ,CAAC,CAACC;QACV,YAAY,AAAqB,YAArB,OAAOT,YAAyBA,YAAYmB;QACxD,GAAI1D,AAAW,iBAAXA,SACA;YACE,2BAA2B;QAC7B,IACA,CAAC,CAAC;IACR;IACA,MAAM,EACJ,QAAQ2D,eAAe,EACvBC,YAAY,EACZC,cAAc,EACf,GAAGC,uBAAuB;QACzB,WAAWzB,SAAS;QACpBrC;IACF;IACA,IAAI4D,cACFlB,UAAUkB;IAEZ,IAAIC,gBAAgB;QAClBnB,UAAUmB;QACVlD,QAAQ,IAAI,CAACkD;IACf;IAEA,IAAI;QACFnB,UACE,CAAC,QAAQ,EAAEM,cAAc,eAAe,GAAG,WAAW,EAAEtD,WAAW;QAGrE,IAAIsD,aAAa;YACf,MAAMe,SAAU,MAAMzB,WAAW,MAAM,CACrC;gBACE,OAAO5C;gBACP0C;gBACA,GAAGqB,YAAY;gBACf,GAAGE,eAAe;YACpB,GACA;gBACE,QAAQ;YACV;YAKF,WAAW,MAAMK,SAASD,OAAQ;gBAChC,MAAMd,UAAUe,MAAM,OAAO,EAAE,CAAC,EAAE,EAAE,OAAO,WAAW;gBACtD,MAAMC,oBACHD,MAAM,OAAO,EAAE,CAAC,EAAE,EAAE,OAAe,qBAAqB;gBAG3D,IAAIA,MAAM,KAAK,EACbZ,QAAQY,MAAM,KAAK;gBAGrB,IAAIf,WAAWgB,mBAAmB;oBAChCf,eAAeD;oBACfE,wBAAwBc;oBACxB,MAAMC,YAAiC;wBACrCjB;wBACAgB;wBACAf;wBACA,YAAY;wBACZ,OAAOQ;oBACT;oBACArB,QAAQ,OAAO,CAAE6B;gBACnB;gBAGA,IAAIF,MAAM,OAAO,EAAE,CAAC,EAAE,EAAE,eAAe;oBACrCX,WAAWP,KAAK,GAAG,KAAKD;oBAGxB,IAAI,CAACO,OAAO;wBAEV,MAAMe,kBAAkBC,KAAK,GAAG,CAC9B,GACAA,KAAK,KAAK,CAAClB,YAAY,MAAM,GAAG;wBAElCE,QAAQ;4BACN,eAAee;4BACf,mBAAmBA;4BACnB,cAAcA,AAAkB,IAAlBA;wBAChB;oBACF;oBAGA,MAAME,aAAkC;wBACtC,SAAS;wBACTnB;wBACA,mBAAmB;wBACnB,YAAY;wBACZ,OAAOI,eAAeF;oBACxB;oBACAf,QAAQ,OAAO,CAAEgC;oBACjB;gBACF;YACF;YACApB,UAAUC;YACVP,kBACE,CAAC,iBAAiB,EAAEjD,UAAU,QAAQ,EAAEM,UAAU,UAAU,WAAW,EAAEqD,SAAS,eAAe,EAAEN,eAAe,IAAI;QAE1H,OAAO;YACL,MAAMuB,SAAS,MAAMhC,WAAW,MAAM,CAAC;gBACrC,OAAO5C;gBACP0C;gBACA,GAAGqB,YAAY;gBACf,GAAGE,eAAe;YACpB;YACAN,WAAWP,KAAK,GAAG,KAAKD;YAExBF,kBACE,CAAC,OAAO,EAAEjD,UAAU,QAAQ,EAAEM,UAAU,UAAU,mBAAmB,EAAED,cAAc,iBAAiB,EAAEuE,OAAO,KAAK,EAAE,iBAAiB,GAAG,qBAAqB,EAAEA,OAAO,KAAK,EAAE,qBAAqB,GAAG,gBAAgB,EAAEA,OAAO,KAAK,EAAE,gBAAgB,GAAG,WAAW,EAAEjB,SAAS,aAAa,EAAEiB,OAAO,WAAW,IAAI,GAAG,eAAe,EAAEvB,eAAe,IAAI;YAG9VH,mBAAmB,CAAC,oBAAoB,EAAE2B,KAAK,SAAS,CAACD,OAAO,KAAK,GAAG;YAExEE,OACEF,OAAO,OAAO,EACd,CAAC,mCAAmC,EAAEC,KAAK,SAAS,CAACD,SAAS;YAEhErB,UAAUqB,OAAO,OAAO,CAAC,EAAE,CAAC,OAAO,CAAC,OAAO;YAC3CnB,uBACGmB,OAAO,OAAO,CAAC,EAAE,CAAC,OAAO,EAAU,qBAAqB;YAC3DlB,QAAQkB,OAAO,KAAK;QACtB;QAEA5B,UAAU,CAAC,4BAA4B,EAAES,sBAAsB;QAC/DT,UAAU,CAAC,kBAAkB,EAAEO,SAAS;QACxCuB,OAAOvB,SAAS;QAGhB,IAAID,eAAe,CAACI,OAAO;YAEzB,MAAMe,kBAAkBC,KAAK,GAAG,CAC9B,GACAA,KAAK,KAAK,CAAEnB,AAAAA,CAAAA,WAAW,EAAC,EAAG,MAAM,GAAG;YAEtCG,QAAQ;gBACN,eAAee;gBACf,mBAAmBA;gBACnB,cAAcA,AAAkB,IAAlBA;YAChB;QACF;QAEA,OAAO;YACL,SAASlB,WAAW;YACpB,mBAAmBE,wBAAwBO;YAC3C,OAAOJ,eAAeF;YACtB,YAAY,CAAC,CAACJ;QAChB;IACF,EAAE,OAAOyB,GAAQ;QACf9D,QAAQ,KAAK,CAAC,kBAAkB8D;QAChC,MAAMC,WAAW,IAAI1D,MACnB,CAAC,eAAe,EAAEgC,cAAc,eAAe,GAAG,kBAAkB,EAAEtD,UAAU,GAAG,EAAE+E,EAAE,OAAO,CAAC,8DAA8D,CAAC,EAC9J;YACE,OAAOA;QACT;QAEF,MAAMC;IACR;AACF;AAEO,eAAeC,yBACpBvC,QAAsC,EACtC9C,iBAA+B,EAC/BC,WAAyB,EACzB8C,OAEC;IAOD,MAAMuC,WAAW,MAAMzC,OAAOC,UAAU9C,mBAAmBC,aAAa;QACtE,WAAW8C,SAAS;IACtB;IACAmC,OAAOI,UAAU;IACjB,MAAM5E,SAAST,YAAY,MAAM;IACjC,MAAMsF,cAAcC,cAAcF,SAAS,OAAO,EAAE5E;IACpDwE,OACE,AAAuB,YAAvB,OAAOK,aACP,CAAC,0CAA0C,EAAEtF,YAAY,SAAS,CAAC,GAAG,EAAEqF,SAAS,OAAO,EAAE;IAE5F,OAAO;QACL,SAASC;QACT,eAAeD,SAAS,OAAO;QAC/B,OAAOA,SAAS,KAAK;QACrB,mBAAmBA,SAAS,iBAAiB;IAC/C;AACF;AAEO,eAAeG,yBACpBC,IAAY,EACZ1F,iBAA+B,EAC/BC,WAAyB;IAEzB,MAAM,EAAE0D,OAAO,EAAEG,KAAK,EAAE,GAAG,MAAMjB,OAAO6C,MAAM1F,mBAAmBC;IACjE,OAAO;QAAE0D;QAASG;IAAM;AAC1B;AAEO,SAAS6B,yBAAyBL,QAAgB;IACvD,IAAI;QAEF,MAAMM,YAAYN,SAAS,KAAK,CAAC;QACjC,IAAIM,WACF,OAAOA,SAAS,CAAC,EAAE;QAIrB,MAAMC,iBAAiBP,SAAS,KAAK,CACnC;QAEF,IAAIO,gBACF,OAAOA,cAAc,CAAC,EAAE;QAI1B,MAAMC,gBAAgBR,SAAS,KAAK,CAAC;QACrC,IAAIQ,eACF,OAAOA,aAAa,CAAC,EAAE;IAE3B,EAAE,OAAM,CAAC;IAET,OAAOR;AACT;AAEO,SAASS,yBAAyBC,KAAa;IACpD,IAAIA,MAAM,QAAQ,CAAC,SAEjB,MAAO,YAAY,IAAI,CAACA,OACtBA,QAAQA,MAAM,OAAO,CAAC,kBAAkB;IAG5C,OAAOA;AACT;AAEO,SAASxB,uBAAuB,EACrCyB,SAAS,EACTvF,MAAM,EAIP;IAKC,MAAMwF,sBAAsBD,AAAc,YAAdA,YAAwB7B,SAAY6B;IAEhE,IAAIC,AAAwB9B,WAAxB8B,qBACF,OAAO;QAAE,QAAQ,CAAC;QAAG,cAAc9B;IAAU;IAG/C,IAAI1D,AAAW,eAAXA,QACF,OAAO;QACL,QAAQ;YAAE,iBAAiBwF;QAAoB;QAC/C,cAAc,CAAC,oCAAoC,EAAEA,oBAAoB,aAAa,CAAC;IACzF;IAGF,IAAIxF,AAAW,oBAAXA,QACF,OAAO;QACL,QAAQ;YACN,UAAU;gBAAE,MAAMwF,sBAAsB,YAAY;YAAW;QACjE;QACA,cAAc,CAAC,kCAAkC,EAAEA,sBAAsB,YAAY,WAAW,kBAAkB,CAAC;IACrH;IAGF,OAAO;QACL,QAAQ,CAAC;QACT,cAAc,CAAC,6CAA6C,EAAExF,UAAU,UAAU,CAAC,CAAC;QACpF,gBAAgB,CAAC,0DAA0D,EAAEA,UAAU,UAAU,EAAE,CAAC;IACtG;AACF;AAQA,SAASyF,oBAAoBC,GAAQ;IAEnC,IAAIA,QAAAA,KACF,OAAOA;IAIT,IAAIC,MAAM,OAAO,CAACD,MAChB,OAAOA,IAAI,GAAG,CAAC,CAACE,OAASH,oBAAoBG;IAI/C,IAAI,AAAe,YAAf,OAAOF,KAAkB;QAC3B,MAAMG,aAAkB,CAAC;QAEzB,KAAK,MAAM,CAACC,KAAKC,MAAM,IAAIC,OAAO,OAAO,CAACN,KAAM;YAE9C,MAAMO,aAAaH,IAAI,IAAI;YAG3B,IAAII,kBAAkBT,oBAAoBM;YAG1C,IAAI,AAA2B,YAA3B,OAAOG,iBACTA,kBAAkBA,gBAAgB,IAAI;YAGxCL,UAAU,CAACI,WAAW,GAAGC;QAC3B;QAEA,OAAOL;IACT;IAGA,IAAI,AAAe,YAAf,OAAOH,KACT,OAAOA,IAAI,IAAI;IAIjB,OAAOA;AACT;AAEO,SAASZ,cAAcQ,KAAa,EAAEtF,MAAgC;IAC3E,MAAMmG,kBAAkBlB,yBAAyBK;IAEjD,IAAIa,iBAAiB,MAAM,oBACzB,OAAOA,gBACJ,KAAK,CAAC,oBACL,MAAM,GACP,IAAIjF;IAGT,IAAIV;IACJ,IAAI4F;IACJ,IAAI;QACF5F,SAAS+D,KAAK,KAAK,CAAC4B;QACpB,OAAOV,oBAAoBjF;IAC7B,EAAE,OAAOc,OAAO;QACd8E,YAAY9E;IACd;IACA,IAAI;QACFd,SAAS+D,KAAK,KAAK,CAAC8B,WAAWF;QAC/B,OAAOV,oBAAoBjF;IAC7B,EAAE,OAAOc,OAAO;QACd8E,YAAY9E;IACd;IAEA,IAAItB,AAAW,oBAAXA,UAA8BA,AAAW,kBAAXA,QAA0B;QAC1D,MAAMsG,aAAajB,yBAAyBc;QAC5C,IAAI;YACF3F,SAAS+D,KAAK,KAAK,CAAC8B,WAAWC;YAC/B,OAAOb,oBAAoBjF;QAC7B,EAAE,OAAOc,OAAO;YACd8E,YAAY9E;QACd;IACF;IACA,MAAMN,MACJ,CAAC,gDAAgD,EAAEuF,OACjDH,aAAa,iBACb,gBAAgB,EAAEd,OAAO;AAE/B"}
+{"version":3,"file":"ai-model/service-caller/index.mjs","sources":["../../../../src/ai-model/service-caller/index.ts"],"sourcesContent":["import {\n  AIResponseFormat,\n  type AIUsageInfo,\n  type DeepThinkOption,\n} from '@/types';\nimport type { CodeGenerationChunk, StreamingCallback } from '@/types';\nimport {\n  type IModelConfig,\n  MIDSCENE_LANGFUSE_DEBUG,\n  MIDSCENE_LANGSMITH_DEBUG,\n  MIDSCENE_MODEL_MAX_TOKENS,\n  OPENAI_MAX_TOKENS,\n  type TVlModeTypes,\n  type UITarsModelVersion,\n  globalConfigManager,\n} from '@midscene/shared/env';\n\nimport { getDebug } from '@midscene/shared/logger';\nimport { assert, ifInBrowser } from '@midscene/shared/utils';\nimport { jsonrepair } from 'jsonrepair';\nimport OpenAI from 'openai';\nimport type { ChatCompletionMessageParam } from 'openai/resources/index';\nimport type { Stream } from 'openai/streaming';\nimport type { AIArgs } from '../../common';\n\nasync function createChatClient({\n  modelConfig,\n}: {\n  modelConfig: IModelConfig;\n}): Promise<{\n  completion: OpenAI.Chat.Completions;\n  modelName: string;\n  modelDescription: string;\n  uiTarsVersion?: UITarsModelVersion;\n  vlMode: TVlModeTypes | undefined;\n}> {\n  const {\n    socksProxy,\n    httpProxy,\n    modelName,\n    openaiBaseURL,\n    openaiApiKey,\n    openaiExtraConfig,\n    modelDescription,\n    uiTarsModelVersion: uiTarsVersion,\n    vlMode,\n    createOpenAIClient,\n    timeout,\n  } = modelConfig;\n\n  let proxyAgent: any = undefined;\n  const debugProxy = getDebug('ai:call:proxy');\n\n  // Helper function to sanitize proxy URL for logging (remove credentials)\n  // Uses URL API instead of regex to avoid ReDoS vulnerabilities\n  const sanitizeProxyUrl = (url: string): string => {\n    try {\n      const parsed = new URL(url);\n      if (parsed.username) {\n        // Keep username for debugging, hide password for security\n        parsed.password = '****';\n        return parsed.href;\n      }\n      return url;\n    } catch {\n      // If URL parsing fails, return original URL (will be caught later)\n      return url;\n    }\n  };\n\n  if (httpProxy) {\n    debugProxy('using http proxy', sanitizeProxyUrl(httpProxy));\n    if (ifInBrowser) {\n      console.warn(\n        'HTTP proxy is configured but not supported in browser environment',\n      );\n    } else {\n      // Dynamic import with variable to avoid bundler static analysis\n      const moduleName = 'undici';\n      const { ProxyAgent } = await import(moduleName);\n      proxyAgent = new ProxyAgent({\n        uri: httpProxy,\n        // Note: authentication is handled via the URI (e.g., http://user:pass@proxy.com:8080)\n      });\n    }\n  } else if (socksProxy) {\n    debugProxy('using socks proxy', sanitizeProxyUrl(socksProxy));\n    if (ifInBrowser) {\n      console.warn(\n        'SOCKS proxy is configured but not supported in browser environment',\n      );\n    } else {\n      try {\n        // Dynamic import with variable to avoid bundler static analysis\n        const moduleName = 'fetch-socks';\n        const { socksDispatcher } = await import(moduleName);\n        // Parse SOCKS proxy URL (e.g., socks5://127.0.0.1:1080)\n        const proxyUrl = new URL(socksProxy);\n\n        // Validate hostname\n        if (!proxyUrl.hostname) {\n          throw new Error('SOCKS proxy URL must include a valid hostname');\n        }\n\n        // Validate and parse port\n        const port = Number.parseInt(proxyUrl.port, 10);\n        if (!proxyUrl.port || Number.isNaN(port)) {\n          throw new Error('SOCKS proxy URL must include a valid port');\n        }\n\n        // Parse SOCKS version from protocol\n        const protocol = proxyUrl.protocol.replace(':', '');\n        const socksType =\n          protocol === 'socks4' ? 4 : protocol === 'socks5' ? 5 : 5;\n\n        proxyAgent = socksDispatcher({\n          type: socksType,\n          host: proxyUrl.hostname,\n          port,\n          ...(proxyUrl.username\n            ? {\n                userId: decodeURIComponent(proxyUrl.username),\n                password: decodeURIComponent(proxyUrl.password || ''),\n              }\n            : {}),\n        });\n        debugProxy('socks proxy configured successfully', {\n          type: socksType,\n          host: proxyUrl.hostname,\n          port: port,\n        });\n      } catch (error) {\n        console.error('Failed to configure SOCKS proxy:', error);\n        throw new Error(\n          `Invalid SOCKS proxy URL: ${socksProxy}. Expected format: socks4://host:port, socks5://host:port, or with authentication: socks5://user:pass@host:port`,\n        );\n      }\n    }\n  }\n\n  const openAIOptions = {\n    baseURL: openaiBaseURL,\n    apiKey: openaiApiKey,\n    // Use fetchOptions.dispatcher for fetch-based SDK instead of httpAgent\n    // Note: Type assertion needed due to undici version mismatch between dependencies\n    ...(proxyAgent ? { fetchOptions: { dispatcher: proxyAgent as any } } : {}),\n    ...openaiExtraConfig,\n    ...(typeof timeout === 'number' ? { timeout } : {}),\n    dangerouslyAllowBrowser: true,\n  };\n\n  const baseOpenAI = new OpenAI(openAIOptions);\n\n  let openai: OpenAI = baseOpenAI;\n\n  // LangSmith wrapper\n  if (\n    openai &&\n    globalConfigManager.getEnvConfigInBoolean(MIDSCENE_LANGSMITH_DEBUG)\n  ) {\n    if (ifInBrowser) {\n      throw new Error('langsmith is not supported in browser');\n    }\n    console.log('DEBUGGING MODE: langsmith wrapper enabled');\n    // Use variable to prevent static analysis by bundlers\n    const langsmithModule = 'langsmith/wrappers';\n    const { wrapOpenAI } = await import(langsmithModule);\n    openai = wrapOpenAI(openai);\n  }\n\n  // Langfuse wrapper\n  if (\n    openai &&\n    globalConfigManager.getEnvConfigInBoolean(MIDSCENE_LANGFUSE_DEBUG)\n  ) {\n    if (ifInBrowser) {\n      throw new Error('langfuse is not supported in browser');\n    }\n    console.log('DEBUGGING MODE: langfuse wrapper enabled');\n    // Use variable to prevent static analysis by bundlers\n    const langfuseModule = 'langfuse';\n    const { observeOpenAI } = await import(langfuseModule);\n    openai = observeOpenAI(openai);\n  }\n\n  if (createOpenAIClient) {\n    const wrappedClient = await createOpenAIClient(baseOpenAI, openAIOptions);\n\n    if (wrappedClient) {\n      openai = wrappedClient as OpenAI;\n    }\n  }\n\n  return {\n    completion: openai.chat.completions,\n    modelName,\n    modelDescription,\n    uiTarsVersion,\n    vlMode,\n  };\n}\n\nexport async function callAI(\n  messages: ChatCompletionMessageParam[],\n  modelConfig: IModelConfig,\n  options?: {\n    stream?: boolean;\n    onChunk?: StreamingCallback;\n    deepThink?: DeepThinkOption;\n  },\n): Promise<{\n  content: string;\n  reasoning_content?: string;\n  usage?: AIUsageInfo;\n  isStreamed: boolean;\n}> {\n  const { completion, modelName, modelDescription, uiTarsVersion, vlMode } =\n    await createChatClient({\n      modelConfig,\n    });\n\n  const maxTokens =\n    globalConfigManager.getEnvConfigValue(MIDSCENE_MODEL_MAX_TOKENS) ??\n    globalConfigManager.getEnvConfigValue(OPENAI_MAX_TOKENS);\n  const debugCall = getDebug('ai:call');\n  const debugProfileStats = getDebug('ai:profile:stats');\n  const debugProfileDetail = getDebug('ai:profile:detail');\n\n  const startTime = Date.now();\n  const temperature = modelConfig.temperature ?? 0;\n\n  const isStreaming = options?.stream && options?.onChunk;\n  let content: string | undefined;\n  let accumulated = '';\n  let accumulatedReasoning = '';\n  let usage: OpenAI.CompletionUsage | undefined;\n  let timeCost: number | undefined;\n\n  const buildUsageInfo = (usageData?: OpenAI.CompletionUsage) => {\n    if (!usageData) return undefined;\n\n    const cachedInputTokens = (\n      usageData as { prompt_tokens_details?: { cached_tokens?: number } }\n    )?.prompt_tokens_details?.cached_tokens;\n\n    return {\n      prompt_tokens: usageData.prompt_tokens ?? 0,\n      completion_tokens: usageData.completion_tokens ?? 0,\n      total_tokens: usageData.total_tokens ?? 0,\n      cached_input: cachedInputTokens ?? 0,\n      time_cost: timeCost ?? 0,\n      model_name: modelName,\n      model_description: modelDescription,\n      intent: modelConfig.intent,\n    } satisfies AIUsageInfo;\n  };\n\n  const commonConfig = {\n    temperature,\n    stream: !!isStreaming,\n    max_tokens: typeof maxTokens === 'number' ? maxTokens : undefined,\n    ...(vlMode === 'qwen2.5-vl' // qwen vl v2 specific config\n      ? {\n          vl_high_resolution_images: true,\n        }\n      : {}),\n  };\n  const {\n    config: deepThinkConfig,\n    debugMessage,\n    warningMessage,\n  } = resolveDeepThinkConfig({\n    deepThink: options?.deepThink,\n    vlMode,\n  });\n  if (debugMessage) {\n    debugCall(debugMessage);\n  }\n  if (warningMessage) {\n    debugCall(warningMessage);\n    console.warn(warningMessage);\n  }\n\n  try {\n    debugCall(\n      `sending ${isStreaming ? 'streaming ' : ''}request to ${modelName}`,\n    );\n\n    if (isStreaming) {\n      const stream = (await completion.create(\n        {\n          model: modelName,\n          messages,\n          ...commonConfig,\n          ...deepThinkConfig,\n        },\n        {\n          stream: true,\n        },\n      )) as Stream<OpenAI.Chat.Completions.ChatCompletionChunk> & {\n        _request_id?: string | null;\n      };\n\n      for await (const chunk of stream) {\n        const content = chunk.choices?.[0]?.delta?.content || '';\n        const reasoning_content =\n          (chunk.choices?.[0]?.delta as any)?.reasoning_content || '';\n\n        // Check for usage info in any chunk (OpenAI provides usage in separate chunks)\n        if (chunk.usage) {\n          usage = chunk.usage;\n        }\n\n        if (content || reasoning_content) {\n          accumulated += content;\n          accumulatedReasoning += reasoning_content;\n          const chunkData: CodeGenerationChunk = {\n            content,\n            reasoning_content,\n            accumulated,\n            isComplete: false,\n            usage: undefined,\n          };\n          options.onChunk!(chunkData);\n        }\n\n        // Check if stream is complete\n        if (chunk.choices?.[0]?.finish_reason) {\n          timeCost = Date.now() - startTime;\n\n          // If usage is not available from the stream, provide a basic usage info\n          if (!usage) {\n            // Estimate token counts based on content length (rough approximation)\n            const estimatedTokens = Math.max(\n              1,\n              Math.floor(accumulated.length / 4),\n            );\n            usage = {\n              prompt_tokens: estimatedTokens,\n              completion_tokens: estimatedTokens,\n              total_tokens: estimatedTokens * 2,\n            };\n          }\n\n          // Send final chunk\n          const finalChunk: CodeGenerationChunk = {\n            content: '',\n            accumulated,\n            reasoning_content: '',\n            isComplete: true,\n            usage: buildUsageInfo(usage),\n          };\n          options.onChunk!(finalChunk);\n          break;\n        }\n      }\n      content = accumulated;\n      debugProfileStats(\n        `streaming model, ${modelName}, mode, ${vlMode || 'default'}, cost-ms, ${timeCost}, temperature, ${temperature ?? ''}`,\n      );\n    } else {\n      const result = await completion.create({\n        model: modelName,\n        messages,\n        ...commonConfig,\n        ...deepThinkConfig,\n      } as any);\n      timeCost = Date.now() - startTime;\n\n      debugProfileStats(\n        `model, ${modelName}, mode, ${vlMode || 'default'}, ui-tars-version, ${uiTarsVersion}, prompt-tokens, ${result.usage?.prompt_tokens || ''}, completion-tokens, ${result.usage?.completion_tokens || ''}, total-tokens, ${result.usage?.total_tokens || ''}, cost-ms, ${timeCost}, requestId, ${result._request_id || ''}, temperature, ${temperature ?? ''}`,\n      );\n\n      debugProfileDetail(`model usage detail: ${JSON.stringify(result.usage)}`);\n\n      assert(\n        result.choices,\n        `invalid response from LLM service: ${JSON.stringify(result)}`,\n      );\n      content = result.choices[0].message.content!;\n      accumulatedReasoning =\n        (result.choices[0].message as any)?.reasoning_content || '';\n      usage = result.usage;\n    }\n\n    debugCall(`response reasoning content: ${accumulatedReasoning}`);\n    debugCall(`response content: ${content}`);\n    assert(content, 'empty content');\n\n    // Ensure we always have usage info for streaming responses\n    if (isStreaming && !usage) {\n      // Estimate token counts based on content length (rough approximation)\n      const estimatedTokens = Math.max(\n        1,\n        Math.floor((content || '').length / 4),\n      );\n      usage = {\n        prompt_tokens: estimatedTokens,\n        completion_tokens: estimatedTokens,\n        total_tokens: estimatedTokens * 2,\n      } as OpenAI.CompletionUsage;\n    }\n\n    return {\n      content: content || '',\n      reasoning_content: accumulatedReasoning || undefined,\n      usage: buildUsageInfo(usage),\n      isStreamed: !!isStreaming,\n    };\n  } catch (e: any) {\n    console.error(' call AI error', e);\n    const newError = new Error(\n      `failed to call ${isStreaming ? 'streaming ' : ''}AI model service (${modelName}): ${e.message}\\nTrouble shooting: https://midscenejs.com/model-provider.html`,\n      {\n        cause: e,\n      },\n    );\n    throw newError;\n  }\n}\n\nexport async function callAIWithObjectResponse<T>(\n  messages: ChatCompletionMessageParam[],\n  modelConfig: IModelConfig,\n  options?: {\n    deepThink?: DeepThinkOption;\n  },\n): Promise<{\n  content: T;\n  contentString: string;\n  usage?: AIUsageInfo;\n  reasoning_content?: string;\n}> {\n  const response = await callAI(messages, modelConfig, {\n    deepThink: options?.deepThink,\n  });\n  assert(response, 'empty response');\n  const vlMode = modelConfig.vlMode;\n  const jsonContent = safeParseJson(response.content, vlMode);\n  assert(\n    typeof jsonContent === 'object',\n    `failed to parse json response from model (${modelConfig.modelName}): ${response.content}`,\n  );\n  return {\n    content: jsonContent,\n    contentString: response.content,\n    usage: response.usage,\n    reasoning_content: response.reasoning_content,\n  };\n}\n\nexport async function callAIWithStringResponse(\n  msgs: AIArgs,\n  modelConfig: IModelConfig,\n): Promise<{ content: string; usage?: AIUsageInfo }> {\n  const { content, usage } = await callAI(msgs, modelConfig);\n  return { content, usage };\n}\n\nexport function extractJSONFromCodeBlock(response: string) {\n  try {\n    // First, try to match a JSON object directly in the response\n    const jsonMatch = response.match(/^\\s*(\\{[\\s\\S]*\\})\\s*$/);\n    if (jsonMatch) {\n      return jsonMatch[1];\n    }\n\n    // If no direct JSON object is found, try to extract JSON from a code block\n    const codeBlockMatch = response.match(\n      /```(?:json)?\\s*(\\{[\\s\\S]*?\\})\\s*```/,\n    );\n    if (codeBlockMatch) {\n      return codeBlockMatch[1];\n    }\n\n    // If no code block is found, try to find a JSON-like structure in the text\n    const jsonLikeMatch = response.match(/\\{[\\s\\S]*\\}/);\n    if (jsonLikeMatch) {\n      return jsonLikeMatch[0];\n    }\n  } catch {}\n  // If no JSON-like structure is found, return the original response\n  return response;\n}\n\nexport function preprocessDoubaoBboxJson(input: string) {\n  if (input.includes('bbox')) {\n    // when its values like 940 445 969 490, replace all /\\d+\\s+\\d+/g with /$1,$2/g\n    while (/\\d+\\s+\\d+/.test(input)) {\n      input = input.replace(/(\\d+)\\s+(\\d+)/g, '$1,$2');\n    }\n  }\n  return input;\n}\n\nexport function resolveDeepThinkConfig({\n  deepThink,\n  vlMode,\n}: {\n  deepThink?: DeepThinkOption;\n  vlMode?: TVlModeTypes;\n}): {\n  config: Record<string, unknown>;\n  debugMessage?: string;\n  warningMessage?: string;\n} {\n  const normalizedDeepThink = deepThink === 'unset' ? undefined : deepThink;\n\n  if (normalizedDeepThink === undefined) {\n    return { config: {}, debugMessage: undefined };\n  }\n\n  if (vlMode === 'qwen3-vl') {\n    return {\n      config: { enable_thinking: normalizedDeepThink },\n      debugMessage: `deepThink mapped to enable_thinking=${normalizedDeepThink} for qwen3-vl`,\n    };\n  }\n\n  if (vlMode === 'doubao-vision') {\n    return {\n      config: {\n        thinking: { type: normalizedDeepThink ? 'enabled' : 'disabled' },\n      },\n      debugMessage: `deepThink mapped to thinking.type=${normalizedDeepThink ? 'enabled' : 'disabled'} for doubao-vision`,\n    };\n  }\n\n  return {\n    config: {},\n    debugMessage: `deepThink ignored: unsupported model_family \"${vlMode ?? 'default'}\"`,\n    warningMessage: `The \"deepThink\" option is not supported for model_family \"${vlMode ?? 'default'}\".`,\n  };\n}\n\n/**\n * Normalize a parsed JSON object by trimming whitespace from:\n * 1. All object keys (e.g., \" prompt \" -> \"prompt\")\n * 2. All string values (e.g., \" Tap \" -> \"Tap\")\n * This handles LLM output that may include leading/trailing spaces.\n */\nfunction normalizeJsonObject(obj: any): any {\n  // Handle null and undefined\n  if (obj === null || obj === undefined) {\n    return obj;\n  }\n\n  // Handle arrays - recursively normalize each element\n  if (Array.isArray(obj)) {\n    return obj.map((item) => normalizeJsonObject(item));\n  }\n\n  // Handle objects\n  if (typeof obj === 'object') {\n    const normalized: any = {};\n\n    for (const [key, value] of Object.entries(obj)) {\n      // Trim the key to remove leading/trailing spaces\n      const trimmedKey = key.trim();\n\n      // Recursively normalize the value\n      let normalizedValue = normalizeJsonObject(value);\n\n      // Trim all string values\n      if (typeof normalizedValue === 'string') {\n        normalizedValue = normalizedValue.trim();\n      }\n\n      normalized[trimmedKey] = normalizedValue;\n    }\n\n    return normalized;\n  }\n\n  // Handle primitive strings\n  if (typeof obj === 'string') {\n    return obj.trim();\n  }\n\n  // Return other primitives as-is\n  return obj;\n}\n\nexport function safeParseJson(input: string, vlMode: TVlModeTypes | undefined) {\n  const cleanJsonString = extractJSONFromCodeBlock(input);\n  // match the point\n  if (cleanJsonString?.match(/\\((\\d+),(\\d+)\\)/)) {\n    return cleanJsonString\n      .match(/\\((\\d+),(\\d+)\\)/)\n      ?.slice(1)\n      .map(Number);\n  }\n\n  let parsed: any;\n  let lastError: unknown;\n  try {\n    parsed = JSON.parse(cleanJsonString);\n    return normalizeJsonObject(parsed);\n  } catch (error) {\n    lastError = error;\n  }\n  try {\n    parsed = JSON.parse(jsonrepair(cleanJsonString));\n    return normalizeJsonObject(parsed);\n  } catch (error) {\n    lastError = error;\n  }\n\n  if (vlMode === 'doubao-vision' || vlMode === 'vlm-ui-tars') {\n    const jsonString = preprocessDoubaoBboxJson(cleanJsonString);\n    try {\n      parsed = JSON.parse(jsonrepair(jsonString));\n      return normalizeJsonObject(parsed);\n    } catch (error) {\n      lastError = error;\n    }\n  }\n  throw Error(\n    `failed to parse LLM response into JSON. Error - ${String(\n      lastError ?? 'unknown error',\n    )}. Response - \\n ${input}`,\n  );\n}\n"],"names":["createChatClient","modelConfig","socksProxy","httpProxy","modelName","openaiBaseURL","openaiApiKey","openaiExtraConfig","modelDescription","uiTarsVersion","vlMode","createOpenAIClient","timeout","proxyAgent","debugProxy","getDebug","sanitizeProxyUrl","url","parsed","URL","ifInBrowser","console","moduleName","ProxyAgent","socksDispatcher","proxyUrl","Error","port","Number","protocol","socksType","decodeURIComponent","error","openAIOptions","baseOpenAI","OpenAI","openai","globalConfigManager","MIDSCENE_LANGSMITH_DEBUG","langsmithModule","wrapOpenAI","MIDSCENE_LANGFUSE_DEBUG","langfuseModule","observeOpenAI","wrappedClient","callAI","messages","options","completion","maxTokens","MIDSCENE_MODEL_MAX_TOKENS","OPENAI_MAX_TOKENS","debugCall","debugProfileStats","debugProfileDetail","startTime","Date","temperature","isStreaming","content","accumulated","accumulatedReasoning","usage","timeCost","buildUsageInfo","usageData","cachedInputTokens","commonConfig","undefined","deepThinkConfig","debugMessage","warningMessage","resolveDeepThinkConfig","stream","chunk","reasoning_content","chunkData","estimatedTokens","Math","finalChunk","result","JSON","assert","e","newError","callAIWithObjectResponse","response","jsonContent","safeParseJson","callAIWithStringResponse","msgs","extractJSONFromCodeBlock","jsonMatch","codeBlockMatch","jsonLikeMatch","preprocessDoubaoBboxJson","input","deepThink","normalizedDeepThink","normalizeJsonObject","obj","Array","item","normalized","key","value","Object","trimmedKey","normalizedValue","cleanJsonString","lastError","jsonrepair","jsonString","String"],"mappings":";;;;;AAyBA,eAAeA,iBAAiB,EAC9BC,WAAW,EAGZ;IAOC,MAAM,EACJC,UAAU,EACVC,SAAS,EACTC,SAAS,EACTC,aAAa,EACbC,YAAY,EACZC,iBAAiB,EACjBC,gBAAgB,EAChB,oBAAoBC,aAAa,EACjCC,MAAM,EACNC,kBAAkB,EAClBC,OAAO,EACR,GAAGX;IAEJ,IAAIY;IACJ,MAAMC,aAAaC,SAAS;IAI5B,MAAMC,mBAAmB,CAACC;QACxB,IAAI;YACF,MAAMC,SAAS,IAAIC,IAAIF;YACvB,IAAIC,OAAO,QAAQ,EAAE;gBAEnBA,OAAO,QAAQ,GAAG;gBAClB,OAAOA,OAAO,IAAI;YACpB;YACA,OAAOD;QACT,EAAE,OAAM;YAEN,OAAOA;QACT;IACF;IAEA,IAAId,WAAW;QACbW,WAAW,oBAAoBE,iBAAiBb;QAChD,IAAIiB,aACFC,QAAQ,IAAI,CACV;aAEG;YAEL,MAAMC,aAAa;YACnB,MAAM,EAAEC,UAAU,EAAE,GAAG,MAAM,MAAM,CAACD;YACpCT,aAAa,IAAIU,WAAW;gBAC1B,KAAKpB;YAEP;QACF;IACF,OAAO,IAAID,YAAY;QACrBY,WAAW,qBAAqBE,iBAAiBd;QACjD,IAAIkB,aACFC,QAAQ,IAAI,CACV;aAGF,IAAI;YAEF,MAAMC,aAAa;YACnB,MAAM,EAAEE,eAAe,EAAE,GAAG,MAAM,MAAM,CAACF;YAEzC,MAAMG,WAAW,IAAIN,IAAIjB;YAGzB,IAAI,CAACuB,SAAS,QAAQ,EACpB,MAAM,IAAIC,MAAM;YAIlB,MAAMC,OAAOC,OAAO,QAAQ,CAACH,SAAS,IAAI,EAAE;YAC5C,IAAI,CAACA,SAAS,IAAI,IAAIG,OAAO,KAAK,CAACD,OACjC,MAAM,IAAID,MAAM;YAIlB,MAAMG,WAAWJ,SAAS,QAAQ,CAAC,OAAO,CAAC,KAAK;YAChD,MAAMK,YACJD,AAAa,aAAbA,WAAwB,IAAIA,AAAa,aAAbA,WAAwB,IAAI;YAE1DhB,aAAaW,gBAAgB;gBAC3B,MAAMM;gBACN,MAAML,SAAS,QAAQ;gBACvBE;gBACA,GAAIF,SAAS,QAAQ,GACjB;oBACE,QAAQM,mBAAmBN,SAAS,QAAQ;oBAC5C,UAAUM,mBAAmBN,SAAS,QAAQ,IAAI;gBACpD,IACA,CAAC,CAAC;YACR;YACAX,WAAW,uCAAuC;gBAChD,MAAMgB;gBACN,MAAML,SAAS,QAAQ;gBACvB,MAAME;YACR;QACF,EAAE,OAAOK,OAAO;YACdX,QAAQ,KAAK,CAAC,oCAAoCW;YAClD,MAAM,IAAIN,MACR,CAAC,yBAAyB,EAAExB,WAAW,+GAA+G,CAAC;QAE3J;IAEJ;IAEA,MAAM+B,gBAAgB;QACpB,SAAS5B;QACT,QAAQC;QAGR,GAAIO,aAAa;YAAE,cAAc;gBAAE,YAAYA;YAAkB;QAAE,IAAI,CAAC,CAAC;QACzE,GAAGN,iBAAiB;QACpB,GAAI,AAAmB,YAAnB,OAAOK,UAAuB;YAAEA;QAAQ,IAAI,CAAC,CAAC;QAClD,yBAAyB;IAC3B;IAEA,MAAMsB,aAAa,IAAIC,SAAOF;IAE9B,IAAIG,SAAiBF;IAGrB,IACEE,UACAC,oBAAoB,qBAAqB,CAACC,2BAC1C;QACA,IAAIlB,aACF,MAAM,IAAIM,MAAM;QAElBL,QAAQ,GAAG,CAAC;QAEZ,MAAMkB,kBAAkB;QACxB,MAAM,EAAEC,UAAU,EAAE,GAAG,MAAM,MAAM,CAACD;QACpCH,SAASI,WAAWJ;IACtB;IAGA,IACEA,UACAC,oBAAoB,qBAAqB,CAACI,0BAC1C;QACA,IAAIrB,aACF,MAAM,IAAIM,MAAM;QAElBL,QAAQ,GAAG,CAAC;QAEZ,MAAMqB,iBAAiB;QACvB,MAAM,EAAEC,aAAa,EAAE,GAAG,MAAM,MAAM,CAACD;QACvCN,SAASO,cAAcP;IACzB;IAEA,IAAIzB,oBAAoB;QACtB,MAAMiC,gBAAgB,MAAMjC,mBAAmBuB,YAAYD;QAE3D,IAAIW,eACFR,SAASQ;IAEb;IAEA,OAAO;QACL,YAAYR,OAAO,IAAI,CAAC,WAAW;QACnChC;QACAI;QACAC;QACAC;IACF;AACF;AAEO,eAAemC,OACpBC,QAAsC,EACtC7C,WAAyB,EACzB8C,OAIC;IAOD,MAAM,EAAEC,UAAU,EAAE5C,SAAS,EAAEI,gBAAgB,EAAEC,aAAa,EAAEC,MAAM,EAAE,GACtE,MAAMV,iBAAiB;QACrBC;IACF;IAEF,MAAMgD,YACJZ,oBAAoB,iBAAiB,CAACa,8BACtCb,oBAAoB,iBAAiB,CAACc;IACxC,MAAMC,YAAYrC,SAAS;IAC3B,MAAMsC,oBAAoBtC,SAAS;IACnC,MAAMuC,qBAAqBvC,SAAS;IAEpC,MAAMwC,YAAYC,KAAK,GAAG;IAC1B,MAAMC,cAAcxD,YAAY,WAAW,IAAI;IAE/C,MAAMyD,cAAcX,SAAS,UAAUA,SAAS;IAChD,IAAIY;IACJ,IAAIC,cAAc;IAClB,IAAIC,uBAAuB;IAC3B,IAAIC;IACJ,IAAIC;IAEJ,MAAMC,iBAAiB,CAACC;QACtB,IAAI,CAACA,WAAW;QAEhB,MAAMC,oBACJD,WACC,uBAAuB;QAE1B,OAAO;YACL,eAAeA,UAAU,aAAa,IAAI;YAC1C,mBAAmBA,UAAU,iBAAiB,IAAI;YAClD,cAAcA,UAAU,YAAY,IAAI;YACxC,cAAcC,qBAAqB;YACnC,WAAWH,YAAY;YACvB,YAAY3D;YACZ,mBAAmBI;YACnB,QAAQP,YAAY,MAAM;QAC5B;IACF;IAEA,MAAMkE,eAAe;QACnBV;QACA,QAAQ,CAAC,CAACC;QACV,YAAY,AAAqB,YAArB,OAAOT,YAAyBA,YAAYmB;QACxD,GAAI1D,AAAW,iBAAXA,SACA;YACE,2BAA2B;QAC7B,IACA,CAAC,CAAC;IACR;IACA,MAAM,EACJ,QAAQ2D,eAAe,EACvBC,YAAY,EACZC,cAAc,EACf,GAAGC,uBAAuB;QACzB,WAAWzB,SAAS;QACpBrC;IACF;IACA,IAAI4D,cACFlB,UAAUkB;IAEZ,IAAIC,gBAAgB;QAClBnB,UAAUmB;QACVlD,QAAQ,IAAI,CAACkD;IACf;IAEA,IAAI;QACFnB,UACE,CAAC,QAAQ,EAAEM,cAAc,eAAe,GAAG,WAAW,EAAEtD,WAAW;QAGrE,IAAIsD,aAAa;YACf,MAAMe,SAAU,MAAMzB,WAAW,MAAM,CACrC;gBACE,OAAO5C;gBACP0C;gBACA,GAAGqB,YAAY;gBACf,GAAGE,eAAe;YACpB,GACA;gBACE,QAAQ;YACV;YAKF,WAAW,MAAMK,SAASD,OAAQ;gBAChC,MAAMd,UAAUe,MAAM,OAAO,EAAE,CAAC,EAAE,EAAE,OAAO,WAAW;gBACtD,MAAMC,oBACHD,MAAM,OAAO,EAAE,CAAC,EAAE,EAAE,OAAe,qBAAqB;gBAG3D,IAAIA,MAAM,KAAK,EACbZ,QAAQY,MAAM,KAAK;gBAGrB,IAAIf,WAAWgB,mBAAmB;oBAChCf,eAAeD;oBACfE,wBAAwBc;oBACxB,MAAMC,YAAiC;wBACrCjB;wBACAgB;wBACAf;wBACA,YAAY;wBACZ,OAAOQ;oBACT;oBACArB,QAAQ,OAAO,CAAE6B;gBACnB;gBAGA,IAAIF,MAAM,OAAO,EAAE,CAAC,EAAE,EAAE,eAAe;oBACrCX,WAAWP,KAAK,GAAG,KAAKD;oBAGxB,IAAI,CAACO,OAAO;wBAEV,MAAMe,kBAAkBC,KAAK,GAAG,CAC9B,GACAA,KAAK,KAAK,CAAClB,YAAY,MAAM,GAAG;wBAElCE,QAAQ;4BACN,eAAee;4BACf,mBAAmBA;4BACnB,cAAcA,AAAkB,IAAlBA;wBAChB;oBACF;oBAGA,MAAME,aAAkC;wBACtC,SAAS;wBACTnB;wBACA,mBAAmB;wBACnB,YAAY;wBACZ,OAAOI,eAAeF;oBACxB;oBACAf,QAAQ,OAAO,CAAEgC;oBACjB;gBACF;YACF;YACApB,UAAUC;YACVP,kBACE,CAAC,iBAAiB,EAAEjD,UAAU,QAAQ,EAAEM,UAAU,UAAU,WAAW,EAAEqD,SAAS,eAAe,EAAEN,eAAe,IAAI;QAE1H,OAAO;YACL,MAAMuB,SAAS,MAAMhC,WAAW,MAAM,CAAC;gBACrC,OAAO5C;gBACP0C;gBACA,GAAGqB,YAAY;gBACf,GAAGE,eAAe;YACpB;YACAN,WAAWP,KAAK,GAAG,KAAKD;YAExBF,kBACE,CAAC,OAAO,EAAEjD,UAAU,QAAQ,EAAEM,UAAU,UAAU,mBAAmB,EAAED,cAAc,iBAAiB,EAAEuE,OAAO,KAAK,EAAE,iBAAiB,GAAG,qBAAqB,EAAEA,OAAO,KAAK,EAAE,qBAAqB,GAAG,gBAAgB,EAAEA,OAAO,KAAK,EAAE,gBAAgB,GAAG,WAAW,EAAEjB,SAAS,aAAa,EAAEiB,OAAO,WAAW,IAAI,GAAG,eAAe,EAAEvB,eAAe,IAAI;YAG9VH,mBAAmB,CAAC,oBAAoB,EAAE2B,KAAK,SAAS,CAACD,OAAO,KAAK,GAAG;YAExEE,OACEF,OAAO,OAAO,EACd,CAAC,mCAAmC,EAAEC,KAAK,SAAS,CAACD,SAAS;YAEhErB,UAAUqB,OAAO,OAAO,CAAC,EAAE,CAAC,OAAO,CAAC,OAAO;YAC3CnB,uBACGmB,OAAO,OAAO,CAAC,EAAE,CAAC,OAAO,EAAU,qBAAqB;YAC3DlB,QAAQkB,OAAO,KAAK;QACtB;QAEA5B,UAAU,CAAC,4BAA4B,EAAES,sBAAsB;QAC/DT,UAAU,CAAC,kBAAkB,EAAEO,SAAS;QACxCuB,OAAOvB,SAAS;QAGhB,IAAID,eAAe,CAACI,OAAO;YAEzB,MAAMe,kBAAkBC,KAAK,GAAG,CAC9B,GACAA,KAAK,KAAK,CAAEnB,AAAAA,CAAAA,WAAW,EAAC,EAAG,MAAM,GAAG;YAEtCG,QAAQ;gBACN,eAAee;gBACf,mBAAmBA;gBACnB,cAAcA,AAAkB,IAAlBA;YAChB;QACF;QAEA,OAAO;YACL,SAASlB,WAAW;YACpB,mBAAmBE,wBAAwBO;YAC3C,OAAOJ,eAAeF;YACtB,YAAY,CAAC,CAACJ;QAChB;IACF,EAAE,OAAOyB,GAAQ;QACf9D,QAAQ,KAAK,CAAC,kBAAkB8D;QAChC,MAAMC,WAAW,IAAI1D,MACnB,CAAC,eAAe,EAAEgC,cAAc,eAAe,GAAG,kBAAkB,EAAEtD,UAAU,GAAG,EAAE+E,EAAE,OAAO,CAAC,8DAA8D,CAAC,EAC9J;YACE,OAAOA;QACT;QAEF,MAAMC;IACR;AACF;AAEO,eAAeC,yBACpBvC,QAAsC,EACtC7C,WAAyB,EACzB8C,OAEC;IAOD,MAAMuC,WAAW,MAAMzC,OAAOC,UAAU7C,aAAa;QACnD,WAAW8C,SAAS;IACtB;IACAmC,OAAOI,UAAU;IACjB,MAAM5E,SAAST,YAAY,MAAM;IACjC,MAAMsF,cAAcC,cAAcF,SAAS,OAAO,EAAE5E;IACpDwE,OACE,AAAuB,YAAvB,OAAOK,aACP,CAAC,0CAA0C,EAAEtF,YAAY,SAAS,CAAC,GAAG,EAAEqF,SAAS,OAAO,EAAE;IAE5F,OAAO;QACL,SAASC;QACT,eAAeD,SAAS,OAAO;QAC/B,OAAOA,SAAS,KAAK;QACrB,mBAAmBA,SAAS,iBAAiB;IAC/C;AACF;AAEO,eAAeG,yBACpBC,IAAY,EACZzF,WAAyB;IAEzB,MAAM,EAAE0D,OAAO,EAAEG,KAAK,EAAE,GAAG,MAAMjB,OAAO6C,MAAMzF;IAC9C,OAAO;QAAE0D;QAASG;IAAM;AAC1B;AAEO,SAAS6B,yBAAyBL,QAAgB;IACvD,IAAI;QAEF,MAAMM,YAAYN,SAAS,KAAK,CAAC;QACjC,IAAIM,WACF,OAAOA,SAAS,CAAC,EAAE;QAIrB,MAAMC,iBAAiBP,SAAS,KAAK,CACnC;QAEF,IAAIO,gBACF,OAAOA,cAAc,CAAC,EAAE;QAI1B,MAAMC,gBAAgBR,SAAS,KAAK,CAAC;QACrC,IAAIQ,eACF,OAAOA,aAAa,CAAC,EAAE;IAE3B,EAAE,OAAM,CAAC;IAET,OAAOR;AACT;AAEO,SAASS,yBAAyBC,KAAa;IACpD,IAAIA,MAAM,QAAQ,CAAC,SAEjB,MAAO,YAAY,IAAI,CAACA,OACtBA,QAAQA,MAAM,OAAO,CAAC,kBAAkB;IAG5C,OAAOA;AACT;AAEO,SAASxB,uBAAuB,EACrCyB,SAAS,EACTvF,MAAM,EAIP;IAKC,MAAMwF,sBAAsBD,AAAc,YAAdA,YAAwB7B,SAAY6B;IAEhE,IAAIC,AAAwB9B,WAAxB8B,qBACF,OAAO;QAAE,QAAQ,CAAC;QAAG,cAAc9B;IAAU;IAG/C,IAAI1D,AAAW,eAAXA,QACF,OAAO;QACL,QAAQ;YAAE,iBAAiBwF;QAAoB;QAC/C,cAAc,CAAC,oCAAoC,EAAEA,oBAAoB,aAAa,CAAC;IACzF;IAGF,IAAIxF,AAAW,oBAAXA,QACF,OAAO;QACL,QAAQ;YACN,UAAU;gBAAE,MAAMwF,sBAAsB,YAAY;YAAW;QACjE;QACA,cAAc,CAAC,kCAAkC,EAAEA,sBAAsB,YAAY,WAAW,kBAAkB,CAAC;IACrH;IAGF,OAAO;QACL,QAAQ,CAAC;QACT,cAAc,CAAC,6CAA6C,EAAExF,UAAU,UAAU,CAAC,CAAC;QACpF,gBAAgB,CAAC,0DAA0D,EAAEA,UAAU,UAAU,EAAE,CAAC;IACtG;AACF;AAQA,SAASyF,oBAAoBC,GAAQ;IAEnC,IAAIA,QAAAA,KACF,OAAOA;IAIT,IAAIC,MAAM,OAAO,CAACD,MAChB,OAAOA,IAAI,GAAG,CAAC,CAACE,OAASH,oBAAoBG;IAI/C,IAAI,AAAe,YAAf,OAAOF,KAAkB;QAC3B,MAAMG,aAAkB,CAAC;QAEzB,KAAK,MAAM,CAACC,KAAKC,MAAM,IAAIC,OAAO,OAAO,CAACN,KAAM;YAE9C,MAAMO,aAAaH,IAAI,IAAI;YAG3B,IAAII,kBAAkBT,oBAAoBM;YAG1C,IAAI,AAA2B,YAA3B,OAAOG,iBACTA,kBAAkBA,gBAAgB,IAAI;YAGxCL,UAAU,CAACI,WAAW,GAAGC;QAC3B;QAEA,OAAOL;IACT;IAGA,IAAI,AAAe,YAAf,OAAOH,KACT,OAAOA,IAAI,IAAI;IAIjB,OAAOA;AACT;AAEO,SAASZ,cAAcQ,KAAa,EAAEtF,MAAgC;IAC3E,MAAMmG,kBAAkBlB,yBAAyBK;IAEjD,IAAIa,iBAAiB,MAAM,oBACzB,OAAOA,gBACJ,KAAK,CAAC,oBACL,MAAM,GACP,IAAIjF;IAGT,IAAIV;IACJ,IAAI4F;IACJ,IAAI;QACF5F,SAAS+D,KAAK,KAAK,CAAC4B;QACpB,OAAOV,oBAAoBjF;IAC7B,EAAE,OAAOc,OAAO;QACd8E,YAAY9E;IACd;IACA,IAAI;QACFd,SAAS+D,KAAK,KAAK,CAAC8B,WAAWF;QAC/B,OAAOV,oBAAoBjF;IAC7B,EAAE,OAAOc,OAAO;QACd8E,YAAY9E;IACd;IAEA,IAAItB,AAAW,oBAAXA,UAA8BA,AAAW,kBAAXA,QAA0B;QAC1D,MAAMsG,aAAajB,yBAAyBc;QAC5C,IAAI;YACF3F,SAAS+D,KAAK,KAAK,CAAC8B,WAAWC;YAC/B,OAAOb,oBAAoBjF;QAC7B,EAAE,OAAOc,OAAO;YACd8E,YAAY9E;QACd;IACF;IACA,MAAMN,MACJ,CAAC,gDAAgD,EAAEuF,OACjDH,aAAa,iBACb,gBAAgB,EAAEd,OAAO;AAE/B"}

package/dist/es/ai-model/ui-tars-planning.mjs

@@ -4,7 +4,6 @@ import { getDebug } from "@midscene/shared/logger";
 import { transformHotkeyInput } from "@midscene/shared/us-keyboard-layout";
 import { assert } from "@midscene/shared/utils";
 import { actionParser } from "@ui-tars/action-parser";
-import { AIActionType } from "../common.mjs";
 import { getSummary, getUiTarsPlanningPrompt } from "./prompt/ui-tars-planning.mjs";
 import { callAIWithStringResponse } from "./service-caller/index.mjs";
 const debug = getDebug('ui-tars-planning');
@@ -39,7 +38,7 @@ async function uiTarsPlanning(userInstruction, options) {
             content: systemPrompt
         },
         ...conversationHistory.snapshot()
-    ], AIActionType.INSPECT_ELEMENT, modelConfig);
+    ], modelConfig);
     const convertedText = convertBboxToCoordinates(res.content);
     const { size } = context;
     const { parsed } = actionParser({

package/dist/es/ai-model/ui-tars-planning.mjs.map

@@ -1 +1 @@
-{"version":3,"file":"ai-model/ui-tars-planning.mjs","sources":["../../../src/ai-model/ui-tars-planning.ts"],"sourcesContent":["import type {\n  PlanningAIResponse,\n  PlanningAction,\n  Size,\n  UIContext,\n} from '@/types';\nimport { type IModelConfig, UITarsModelVersion } from '@midscene/shared/env';\nimport { resizeImgBase64 } from '@midscene/shared/img';\nimport { getDebug } from '@midscene/shared/logger';\nimport { transformHotkeyInput } from '@midscene/shared/us-keyboard-layout';\nimport { assert } from '@midscene/shared/utils';\nimport { actionParser } from '@ui-tars/action-parser';\nimport { AIActionType } from '../common';\nimport type { ConversationHistory } from './conversation-history';\nimport { getSummary, getUiTarsPlanningPrompt } from './prompt/ui-tars-planning';\nimport { callAIWithStringResponse } from './service-caller/index';\ntype ActionType =\n  | 'click'\n  | 'left_double'\n  | 'right_single'\n  | 'drag'\n  | 'type'\n  | 'hotkey'\n  | 'finished'\n  | 'scroll'\n  | 'wait';\n\nconst debug = getDebug('ui-tars-planning');\nconst bboxSize = 10;\nconst pointToBbox = (\n  point: { x: number; y: number },\n  width: number,\n  height: number,\n): [number, number, number, number] => {\n  return [\n    Math.round(Math.max(point.x - bboxSize / 2, 0)),\n    Math.round(Math.max(point.y - bboxSize / 2, 0)),\n    Math.round(Math.min(point.x + bboxSize / 2, width)),\n    Math.round(Math.min(point.y + bboxSize / 2, height)),\n  ];\n};\n\nexport async function uiTarsPlanning(\n  userInstruction: string,\n  options: {\n    conversationHistory: ConversationHistory;\n    context: UIContext;\n    modelConfig: IModelConfig;\n    actionContext?: string;\n  },\n): Promise<PlanningAIResponse> {\n  const { conversationHistory, context, modelConfig, actionContext } = options;\n  const { uiTarsModelVersion } = modelConfig;\n\n  let instruction = userInstruction;\n  if (actionContext) {\n    instruction = `<high_priority_knowledge>${actionContext}</high_priority_knowledge>\\n<user_instruction>${userInstruction}</user_instruction>`;\n  }\n\n  const systemPrompt = getUiTarsPlanningPrompt() + instruction;\n\n  const imagePayload = await resizeImageForUiTars(\n    context.screenshotBase64,\n    context.size,\n    uiTarsModelVersion,\n  );\n\n  conversationHistory.append({\n    role: 'user',\n    content: [\n      {\n        type: 'image_url',\n        image_url: {\n          url: imagePayload,\n        },\n      },\n    ],\n  });\n\n  const res = await callAIWithStringResponse(\n    [\n      {\n        role: 'user',\n        content: systemPrompt,\n      },\n      ...conversationHistory.snapshot(),\n    ],\n    AIActionType.INSPECT_ELEMENT,\n    modelConfig,\n  );\n  const convertedText = convertBboxToCoordinates(res.content);\n\n  const { size } = context;\n  const { parsed } = actionParser({\n    prediction: convertedText,\n    factor: [1000, 1000],\n    screenContext: {\n      width: size.width,\n      height: size.height,\n    },\n    modelVer: uiTarsModelVersion,\n  });\n\n  debug(\n    'ui-tars modelVer',\n    uiTarsModelVersion,\n    ', parsed',\n    JSON.stringify(parsed),\n  );\n\n  const transformActions: PlanningAction[] = [];\n  const unhandledActions: Array<{ type: string; thought: string }> = [];\n  let shouldContinue = true;\n  parsed.forEach((action) => {\n    const actionType = (action.action_type || '').toLowerCase();\n    if (actionType === 'click') {\n      assert(action.action_inputs.start_box, 'start_box is required');\n      const point = getPoint(action.action_inputs.start_box, size);\n      transformActions.push({\n        type: 'Tap',\n        param: {\n          locate: {\n            prompt: action.thought || '',\n            bbox: pointToBbox(\n              { x: point[0], y: point[1] },\n              size.width,\n              size.height,\n            ),\n          },\n        },\n      });\n    } else if (actionType === 'left_double') {\n      assert(action.action_inputs.start_box, 'start_box is required');\n      const point = getPoint(action.action_inputs.start_box, size);\n      transformActions.push({\n        type: 'DoubleClick',\n        param: {\n          locate: {\n            prompt: action.thought || '',\n            bbox: pointToBbox(\n              { x: point[0], y: point[1] },\n              size.width,\n              size.height,\n            ),\n          },\n        },\n        thought: action.thought || '',\n      });\n    } else if (actionType === 'right_single') {\n      assert(action.action_inputs.start_box, 'start_box is required');\n      const point = getPoint(action.action_inputs.start_box, size);\n      transformActions.push({\n        type: 'RightClick',\n        param: {\n          locate: {\n            prompt: action.thought || '',\n            bbox: pointToBbox(\n              { x: point[0], y: point[1] },\n              size.width,\n              size.height,\n            ),\n          },\n        },\n        thought: action.thought || '',\n      });\n    } else if (actionType === 'drag') {\n      assert(action.action_inputs.start_box, 'start_box is required');\n      assert(action.action_inputs.end_box, 'end_box is required');\n      const startPoint = getPoint(action.action_inputs.start_box, size);\n      const endPoint = getPoint(action.action_inputs.end_box, size);\n      transformActions.push({\n        type: 'DragAndDrop',\n        param: {\n          from: {\n            prompt: action.thought || '',\n            bbox: pointToBbox(\n              { x: startPoint[0], y: startPoint[1] },\n              size.width,\n              size.height,\n            ),\n          },\n          to: {\n            prompt: action.thought || '',\n            bbox: pointToBbox(\n              { x: endPoint[0], y: endPoint[1] },\n              size.width,\n              size.height,\n            ),\n          },\n        },\n        thought: action.thought || '',\n      });\n    } else if (actionType === 'type') {\n      transformActions.push({\n        type: 'Input',\n        param: {\n          value: action.action_inputs.content,\n        },\n        thought: action.thought || '',\n      });\n    } else if (actionType === 'scroll') {\n      transformActions.push({\n        type: 'Scroll',\n        param: {\n          direction: action.action_inputs.direction,\n        },\n        thought: action.thought || '',\n      });\n    } else if (actionType === 'finished') {\n      shouldContinue = false;\n      transformActions.push({\n        type: 'Finished',\n        param: {},\n        thought: action.thought || '',\n      });\n    } else if (actionType === 'hotkey') {\n      if (!action.action_inputs.key) {\n        console.warn(\n          'No key found in action: hotkey. Will not perform action.',\n        );\n      } else {\n        const keys = transformHotkeyInput(action.action_inputs.key);\n\n        transformActions.push({\n          type: 'KeyboardPress',\n          param: {\n            keyName: keys,\n          },\n          thought: action.thought || '',\n        });\n      }\n    } else if (actionType === 'wait') {\n      transformActions.push({\n        type: 'Sleep',\n        param: {\n          timeMs: 1000,\n        },\n        thought: action.thought || '',\n      });\n    } else if (actionType) {\n      // Track unhandled action types\n      unhandledActions.push({\n        type: actionType,\n        thought: action.thought || '',\n      });\n      debug('Unhandled action type:', actionType, 'thought:', action.thought);\n    }\n  });\n\n  if (transformActions.length === 0) {\n    const errorDetails: string[] = [];\n\n    // Check if parsing failed\n    if (parsed.length === 0) {\n      errorDetails.push('Action parser returned no actions');\n\n      // Check if response has Thought but no Action\n      if (\n        res.content.includes('Thought:') &&\n        !res.content.includes('Action:')\n      ) {\n        errorDetails.push(\n          'Response contains \"Thought:\" but missing \"Action:\" line',\n        );\n      } else {\n        errorDetails.push('Response may be malformed or empty');\n      }\n    }\n\n    // Check if we have unhandled action types\n    if (unhandledActions.length > 0) {\n      const types = unhandledActions.map((a) => a.type).join(', ');\n      errorDetails.push(`Unhandled action types: ${types}`);\n    }\n\n    const errorMessage = [\n      'No actions found in UI-TARS response.',\n      ...errorDetails,\n      `\\nRaw response: ${res.content}`,\n    ].join('\\n');\n\n    throw new Error(errorMessage, {\n      cause: {\n        prediction: res.content,\n        parsed,\n        unhandledActions,\n        convertedText,\n      },\n    });\n  }\n\n  debug('transformActions', JSON.stringify(transformActions, null, 2));\n  const log = getSummary(res.content);\n\n  conversationHistory.append({\n    role: 'assistant',\n    content: log,\n  });\n\n  return {\n    actions: transformActions,\n    log,\n    usage: res.usage,\n    rawResponse: JSON.stringify(res.content, undefined, 2),\n    more_actions_needed_by_instruction: shouldContinue,\n  };\n}\n\n/**\n * Converts bounding box notation to coordinate points\n * @param text - The text containing bbox tags to be converted\n * @returns The text with bbox tags replaced by coordinate points\n */\nfunction convertBboxToCoordinates(text: string): string {\n  // Match the four numbers after <bbox>\n  const pattern = /<bbox>(\\d+)\\s+(\\d+)\\s+(\\d+)\\s+(\\d+)<\\/bbox>/g;\n\n  function replaceMatch(\n    match: string,\n    x1: string,\n    y1: string,\n    x2: string,\n    y2: string,\n  ): string {\n    // Convert strings to numbers and calculate center point\n    const x1Num = Number.parseInt(x1, 10);\n    const y1Num = Number.parseInt(y1, 10);\n    const x2Num = Number.parseInt(x2, 10);\n    const y2Num = Number.parseInt(y2, 10);\n\n    // Use Math.floor to truncate and calculate center point\n    const x = Math.floor((x1Num + x2Num) / 2);\n    const y = Math.floor((y1Num + y2Num) / 2);\n\n    // Return formatted coordinate string\n    return `(${x},${y})`;\n  }\n\n  // Remove [EOS] and replace <bbox> coordinates\n  const cleanedText = text.replace(/\\[EOS\\]/g, '');\n  return cleanedText.replace(pattern, replaceMatch).trim();\n}\n\nfunction getPoint(startBox: string, size: { width: number; height: number }) {\n  const [x, y] = JSON.parse(startBox);\n  return [x * size.width, y * size.height];\n}\n\ninterface BaseAction {\n  action_type: ActionType;\n  action_inputs: Record<string, any>;\n  reflection: string | null;\n  thought: string | null;\n}\n\ninterface ClickAction extends BaseAction {\n  action_type: 'click';\n  action_inputs: {\n    start_box: string; // JSON string of [x, y] coordinates\n  };\n}\n\ninterface DragAction extends BaseAction {\n  action_type: 'drag';\n  action_inputs: {\n    start_box: string; // JSON string of [x, y] coordinates\n    end_box: string; // JSON string of [x, y] coordinates\n  };\n}\n\ninterface WaitAction extends BaseAction {\n  action_type: 'wait';\n  action_inputs: {\n    time: string; // JSON string of [x, y] coordinates\n  };\n}\n\ninterface LeftDoubleAction extends BaseAction {\n  action_type: 'left_double';\n  action_inputs: {\n    start_box: string; // JSON string of [x, y] coordinates\n  };\n}\n\ninterface RightSingleAction extends BaseAction {\n  action_type: 'right_single';\n  action_inputs: {\n    start_box: string; // JSON string of [x, y] coordinates\n  };\n}\n\ninterface TypeAction extends BaseAction {\n  action_type: 'type';\n  action_inputs: {\n    content: string;\n  };\n}\n\ninterface HotkeyAction extends BaseAction {\n  action_type: 'hotkey';\n  action_inputs: {\n    key: string;\n  };\n}\n\ninterface ScrollAction extends BaseAction {\n  action_type: 'scroll';\n  action_inputs: {\n    direction: 'up' | 'down';\n  };\n}\n\ninterface FinishedAction extends BaseAction {\n  action_type: 'finished';\n  action_inputs: Record<string, never>;\n}\n\nexport type Action =\n  | ClickAction\n  | LeftDoubleAction\n  | RightSingleAction\n  | DragAction\n  | TypeAction\n  | HotkeyAction\n  | ScrollAction\n  | FinishedAction\n  | WaitAction;\n\nexport async function resizeImageForUiTars(\n  imageBase64: string,\n  size: Size,\n  uiTarsVersion: UITarsModelVersion | undefined,\n) {\n  if (uiTarsVersion === UITarsModelVersion.V1_5) {\n    debug('ui-tars-v1.5, will check image size', size);\n    const currentPixels = size.width * size.height;\n    const maxPixels = 16384 * 28 * 28; //\n    if (currentPixels > maxPixels) {\n      const resizeFactor = Math.sqrt(maxPixels / currentPixels);\n      const newWidth = Math.floor(size.width * resizeFactor);\n      const newHeight = Math.floor(size.height * resizeFactor);\n      debug(\n        'resize image for ui-tars, new width: %s, new height: %s',\n        newWidth,\n        newHeight,\n      );\n      const resizedImage = await resizeImgBase64(imageBase64, {\n        width: newWidth,\n        height: newHeight,\n      });\n      return resizedImage;\n    }\n  }\n  return imageBase64;\n}\n"],"names":["debug","getDebug","bboxSize","pointToBbox","point","width","height","Math","uiTarsPlanning","userInstruction","options","conversationHistory","context","modelConfig","actionContext","uiTarsModelVersion","instruction","systemPrompt","getUiTarsPlanningPrompt","imagePayload","resizeImageForUiTars","res","callAIWithStringResponse","AIActionType","convertedText","convertBboxToCoordinates","size","parsed","actionParser","JSON","transformActions","unhandledActions","shouldContinue","action","actionType","assert","getPoint","startPoint","endPoint","keys","transformHotkeyInput","console","errorDetails","types","a","errorMessage","Error","log","getSummary","undefined","text","pattern","replaceMatch","match","x1","y1","x2","y2","x1Num","Number","y1Num","x2Num","y2Num","x","y","cleanedText","startBox","imageBase64","uiTarsVersion","UITarsModelVersion","currentPixels","maxPixels","resizeFactor","newWidth","newHeight","resizedImage","resizeImgBase64"],"mappings":";;;;;;;;;AA2BA,MAAMA,QAAQC,SAAS;AACvB,MAAMC,WAAW;AACjB,MAAMC,cAAc,CAClBC,OACAC,OACAC,SAEO;QACLC,KAAK,KAAK,CAACA,KAAK,GAAG,CAACH,MAAM,CAAC,GAAGF,WAAW,GAAG;QAC5CK,KAAK,KAAK,CAACA,KAAK,GAAG,CAACH,MAAM,CAAC,GAAGF,WAAW,GAAG;QAC5CK,KAAK,KAAK,CAACA,KAAK,GAAG,CAACH,MAAM,CAAC,GAAGF,WAAW,GAAGG;QAC5CE,KAAK,KAAK,CAACA,KAAK,GAAG,CAACH,MAAM,CAAC,GAAGF,WAAW,GAAGI;KAC7C;AAGI,eAAeE,eACpBC,eAAuB,EACvBC,OAKC;IAED,MAAM,EAAEC,mBAAmB,EAAEC,OAAO,EAAEC,WAAW,EAAEC,aAAa,EAAE,GAAGJ;IACrE,MAAM,EAAEK,kBAAkB,EAAE,GAAGF;IAE/B,IAAIG,cAAcP;IAClB,IAAIK,eACFE,cAAc,CAAC,yBAAyB,EAAEF,cAAc,8CAA8C,EAAEL,gBAAgB,mBAAmB,CAAC;IAG9I,MAAMQ,eAAeC,4BAA4BF;IAEjD,MAAMG,eAAe,MAAMC,qBACzBR,QAAQ,gBAAgB,EACxBA,QAAQ,IAAI,EACZG;IAGFJ,oBAAoB,MAAM,CAAC;QACzB,MAAM;QACN,SAAS;YACP;gBACE,MAAM;gBACN,WAAW;oBACT,KAAKQ;gBACP;YACF;SACD;IACH;IAEA,MAAME,MAAM,MAAMC,yBAChB;QACE;YACE,MAAM;YACN,SAASL;QACX;WACGN,oBAAoB,QAAQ;KAChC,EACDY,aAAa,eAAe,EAC5BV;IAEF,MAAMW,gBAAgBC,yBAAyBJ,IAAI,OAAO;IAE1D,MAAM,EAAEK,IAAI,EAAE,GAAGd;IACjB,MAAM,EAAEe,MAAM,EAAE,GAAGC,aAAa;QAC9B,YAAYJ;QACZ,QAAQ;YAAC;YAAM;SAAK;QACpB,eAAe;YACb,OAAOE,KAAK,KAAK;YACjB,QAAQA,KAAK,MAAM;QACrB;QACA,UAAUX;IACZ;IAEAf,MACE,oBACAe,oBACA,YACAc,KAAK,SAAS,CAACF;IAGjB,MAAMG,mBAAqC,EAAE;IAC7C,MAAMC,mBAA6D,EAAE;IACrE,IAAIC,iBAAiB;IACrBL,OAAO,OAAO,CAAC,CAACM;QACd,MAAMC,aAAcD,AAAAA,CAAAA,OAAO,WAAW,IAAI,EAAC,EAAG,WAAW;QACzD,IAAIC,AAAe,YAAfA,YAAwB;YAC1BC,OAAOF,OAAO,aAAa,CAAC,SAAS,EAAE;YACvC,MAAM7B,QAAQgC,SAASH,OAAO,aAAa,CAAC,SAAS,EAAEP;YACvDI,iBAAiB,IAAI,CAAC;gBACpB,MAAM;gBACN,OAAO;oBACL,QAAQ;wBACN,QAAQG,OAAO,OAAO,IAAI;wBAC1B,MAAM9B,YACJ;4BAAE,GAAGC,KAAK,CAAC,EAAE;4BAAE,GAAGA,KAAK,CAAC,EAAE;wBAAC,GAC3BsB,KAAK,KAAK,EACVA,KAAK,MAAM;oBAEf;gBACF;YACF;QACF,OAAO,IAAIQ,AAAe,kBAAfA,YAA8B;YACvCC,OAAOF,OAAO,aAAa,CAAC,SAAS,EAAE;YACvC,MAAM7B,QAAQgC,SAASH,OAAO,aAAa,CAAC,SAAS,EAAEP;YACvDI,iBAAiB,IAAI,CAAC;gBACpB,MAAM;gBACN,OAAO;oBACL,QAAQ;wBACN,QAAQG,OAAO,OAAO,IAAI;wBAC1B,MAAM9B,YACJ;4BAAE,GAAGC,KAAK,CAAC,EAAE;4BAAE,GAAGA,KAAK,CAAC,EAAE;wBAAC,GAC3BsB,KAAK,KAAK,EACVA,KAAK,MAAM;oBAEf;gBACF;gBACA,SAASO,OAAO,OAAO,IAAI;YAC7B;QACF,OAAO,IAAIC,AAAe,mBAAfA,YAA+B;YACxCC,OAAOF,OAAO,aAAa,CAAC,SAAS,EAAE;YACvC,MAAM7B,QAAQgC,SAASH,OAAO,aAAa,CAAC,SAAS,EAAEP;YACvDI,iBAAiB,IAAI,CAAC;gBACpB,MAAM;gBACN,OAAO;oBACL,QAAQ;wBACN,QAAQG,OAAO,OAAO,IAAI;wBAC1B,MAAM9B,YACJ;4BAAE,GAAGC,KAAK,CAAC,EAAE;4BAAE,GAAGA,KAAK,CAAC,EAAE;wBAAC,GAC3BsB,KAAK,KAAK,EACVA,KAAK,MAAM;oBAEf;gBACF;gBACA,SAASO,OAAO,OAAO,IAAI;YAC7B;QACF,OAAO,IAAIC,AAAe,WAAfA,YAAuB;YAChCC,OAAOF,OAAO,aAAa,CAAC,SAAS,EAAE;YACvCE,OAAOF,OAAO,aAAa,CAAC,OAAO,EAAE;YACrC,MAAMI,aAAaD,SAASH,OAAO,aAAa,CAAC,SAAS,EAAEP;YAC5D,MAAMY,WAAWF,SAASH,OAAO,aAAa,CAAC,OAAO,EAAEP;YACxDI,iBAAiB,IAAI,CAAC;gBACpB,MAAM;gBACN,OAAO;oBACL,MAAM;wBACJ,QAAQG,OAAO,OAAO,IAAI;wBAC1B,MAAM9B,YACJ;4BAAE,GAAGkC,UAAU,CAAC,EAAE;4BAAE,GAAGA,UAAU,CAAC,EAAE;wBAAC,GACrCX,KAAK,KAAK,EACVA,KAAK,MAAM;oBAEf;oBACA,IAAI;wBACF,QAAQO,OAAO,OAAO,IAAI;wBAC1B,MAAM9B,YACJ;4BAAE,GAAGmC,QAAQ,CAAC,EAAE;4BAAE,GAAGA,QAAQ,CAAC,EAAE;wBAAC,GACjCZ,KAAK,KAAK,EACVA,KAAK,MAAM;oBAEf;gBACF;gBACA,SAASO,OAAO,OAAO,IAAI;YAC7B;QACF,OAAO,IAAIC,AAAe,WAAfA,YACTJ,iBAAiB,IAAI,CAAC;YACpB,MAAM;YACN,OAAO;gBACL,OAAOG,OAAO,aAAa,CAAC,OAAO;YACrC;YACA,SAASA,OAAO,OAAO,IAAI;QAC7B;aACK,IAAIC,AAAe,aAAfA,YACTJ,iBAAiB,IAAI,CAAC;YACpB,MAAM;YACN,OAAO;gBACL,WAAWG,OAAO,aAAa,CAAC,SAAS;YAC3C;YACA,SAASA,OAAO,OAAO,IAAI;QAC7B;aACK,IAAIC,AAAe,eAAfA,YAA2B;YACpCF,iBAAiB;YACjBF,iBAAiB,IAAI,CAAC;gBACpB,MAAM;gBACN,OAAO,CAAC;gBACR,SAASG,OAAO,OAAO,IAAI;YAC7B;QACF,OAAO,IAAIC,AAAe,aAAfA,YACT,IAAKD,OAAO,aAAa,CAAC,GAAG,EAItB;YACL,MAAMM,OAAOC,qBAAqBP,OAAO,aAAa,CAAC,GAAG;YAE1DH,iBAAiB,IAAI,CAAC;gBACpB,MAAM;gBACN,OAAO;oBACL,SAASS;gBACX;gBACA,SAASN,OAAO,OAAO,IAAI;YAC7B;QACF,OAbEQ,QAAQ,IAAI,CACV;aAaC,IAAIP,AAAe,WAAfA,YACTJ,iBAAiB,IAAI,CAAC;YACpB,MAAM;YACN,OAAO;gBACL,QAAQ;YACV;YACA,SAASG,OAAO,OAAO,IAAI;QAC7B;aACK,IAAIC,YAAY;YAErBH,iBAAiB,IAAI,CAAC;gBACpB,MAAMG;gBACN,SAASD,OAAO,OAAO,IAAI;YAC7B;YACAjC,MAAM,0BAA0BkC,YAAY,YAAYD,OAAO,OAAO;QACxE;IACF;IAEA,IAAIH,AAA4B,MAA5BA,iBAAiB,MAAM,EAAQ;QACjC,MAAMY,eAAyB,EAAE;QAGjC,IAAIf,AAAkB,MAAlBA,OAAO,MAAM,EAAQ;YACvBe,aAAa,IAAI,CAAC;YAGlB,IACErB,IAAI,OAAO,CAAC,QAAQ,CAAC,eACrB,CAACA,IAAI,OAAO,CAAC,QAAQ,CAAC,YAEtBqB,aAAa,IAAI,CACf;iBAGFA,aAAa,IAAI,CAAC;QAEtB;QAGA,IAAIX,iBAAiB,MAAM,GAAG,GAAG;YAC/B,MAAMY,QAAQZ,iBAAiB,GAAG,CAAC,CAACa,IAAMA,EAAE,IAAI,EAAE,IAAI,CAAC;YACvDF,aAAa,IAAI,CAAC,CAAC,wBAAwB,EAAEC,OAAO;QACtD;QAEA,MAAME,eAAe;YACnB;eACGH;YACH,CAAC,gBAAgB,EAAErB,IAAI,OAAO,EAAE;SACjC,CAAC,IAAI,CAAC;QAEP,MAAM,IAAIyB,MAAMD,cAAc;YAC5B,OAAO;gBACL,YAAYxB,IAAI,OAAO;gBACvBM;gBACAI;gBACAP;YACF;QACF;IACF;IAEAxB,MAAM,oBAAoB6B,KAAK,SAAS,CAACC,kBAAkB,MAAM;IACjE,MAAMiB,MAAMC,WAAW3B,IAAI,OAAO;IAElCV,oBAAoB,MAAM,CAAC;QACzB,MAAM;QACN,SAASoC;IACX;IAEA,OAAO;QACL,SAASjB;QACTiB;QACA,OAAO1B,IAAI,KAAK;QAChB,aAAaQ,KAAK,SAAS,CAACR,IAAI,OAAO,EAAE4B,QAAW;QACpD,oCAAoCjB;IACtC;AACF;AAOA,SAASP,yBAAyByB,IAAY;IAE5C,MAAMC,UAAU;IAEhB,SAASC,aACPC,KAAa,EACbC,EAAU,EACVC,EAAU,EACVC,EAAU,EACVC,EAAU;QAGV,MAAMC,QAAQC,OAAO,QAAQ,CAACL,IAAI;QAClC,MAAMM,QAAQD,OAAO,QAAQ,CAACJ,IAAI;QAClC,MAAMM,QAAQF,OAAO,QAAQ,CAACH,IAAI;QAClC,MAAMM,QAAQH,OAAO,QAAQ,CAACF,IAAI;QAGlC,MAAMM,IAAIxD,KAAK,KAAK,CAAEmD,AAAAA,CAAAA,QAAQG,KAAI,IAAK;QACvC,MAAMG,IAAIzD,KAAK,KAAK,CAAEqD,AAAAA,CAAAA,QAAQE,KAAI,IAAK;QAGvC,OAAO,CAAC,CAAC,EAAEC,EAAE,CAAC,EAAEC,EAAE,CAAC,CAAC;IACtB;IAGA,MAAMC,cAAcf,KAAK,OAAO,CAAC,YAAY;IAC7C,OAAOe,YAAY,OAAO,CAACd,SAASC,cAAc,IAAI;AACxD;AAEA,SAAShB,SAAS8B,QAAgB,EAAExC,IAAuC;IACzE,MAAM,CAACqC,GAAGC,EAAE,GAAGnC,KAAK,KAAK,CAACqC;IAC1B,OAAO;QAACH,IAAIrC,KAAK,KAAK;QAAEsC,IAAItC,KAAK,MAAM;KAAC;AAC1C;AAkFO,eAAeN,qBACpB+C,WAAmB,EACnBzC,IAAU,EACV0C,aAA6C;IAE7C,IAAIA,kBAAkBC,mBAAmB,IAAI,EAAE;QAC7CrE,MAAM,uCAAuC0B;QAC7C,MAAM4C,gBAAgB5C,KAAK,KAAK,GAAGA,KAAK,MAAM;QAC9C,MAAM6C,YAAY;QAClB,IAAID,gBAAgBC,WAAW;YAC7B,MAAMC,eAAejE,KAAK,IAAI,CAACgE,YAAYD;YAC3C,MAAMG,WAAWlE,KAAK,KAAK,CAACmB,KAAK,KAAK,GAAG8C;YACzC,MAAME,YAAYnE,KAAK,KAAK,CAACmB,KAAK,MAAM,GAAG8C;YAC3CxE,MACE,2DACAyE,UACAC;YAEF,MAAMC,eAAe,MAAMC,gBAAgBT,aAAa;gBACtD,OAAOM;gBACP,QAAQC;YACV;YACA,OAAOC;QACT;IACF;IACA,OAAOR;AACT"}
+{"version":3,"file":"ai-model/ui-tars-planning.mjs","sources":["../../../src/ai-model/ui-tars-planning.ts"],"sourcesContent":["import type {\n  PlanningAIResponse,\n  PlanningAction,\n  Size,\n  UIContext,\n} from '@/types';\nimport { type IModelConfig, UITarsModelVersion } from '@midscene/shared/env';\nimport { resizeImgBase64 } from '@midscene/shared/img';\nimport { getDebug } from '@midscene/shared/logger';\nimport { transformHotkeyInput } from '@midscene/shared/us-keyboard-layout';\nimport { assert } from '@midscene/shared/utils';\nimport { actionParser } from '@ui-tars/action-parser';\nimport type { ConversationHistory } from './conversation-history';\nimport { getSummary, getUiTarsPlanningPrompt } from './prompt/ui-tars-planning';\nimport { callAIWithStringResponse } from './service-caller/index';\ntype ActionType =\n  | 'click'\n  | 'left_double'\n  | 'right_single'\n  | 'drag'\n  | 'type'\n  | 'hotkey'\n  | 'finished'\n  | 'scroll'\n  | 'wait';\n\nconst debug = getDebug('ui-tars-planning');\nconst bboxSize = 10;\nconst pointToBbox = (\n  point: { x: number; y: number },\n  width: number,\n  height: number,\n): [number, number, number, number] => {\n  return [\n    Math.round(Math.max(point.x - bboxSize / 2, 0)),\n    Math.round(Math.max(point.y - bboxSize / 2, 0)),\n    Math.round(Math.min(point.x + bboxSize / 2, width)),\n    Math.round(Math.min(point.y + bboxSize / 2, height)),\n  ];\n};\n\nexport async function uiTarsPlanning(\n  userInstruction: string,\n  options: {\n    conversationHistory: ConversationHistory;\n    context: UIContext;\n    modelConfig: IModelConfig;\n    actionContext?: string;\n  },\n): Promise<PlanningAIResponse> {\n  const { conversationHistory, context, modelConfig, actionContext } = options;\n  const { uiTarsModelVersion } = modelConfig;\n\n  let instruction = userInstruction;\n  if (actionContext) {\n    instruction = `<high_priority_knowledge>${actionContext}</high_priority_knowledge>\\n<user_instruction>${userInstruction}</user_instruction>`;\n  }\n\n  const systemPrompt = getUiTarsPlanningPrompt() + instruction;\n\n  const imagePayload = await resizeImageForUiTars(\n    context.screenshotBase64,\n    context.size,\n    uiTarsModelVersion,\n  );\n\n  conversationHistory.append({\n    role: 'user',\n    content: [\n      {\n        type: 'image_url',\n        image_url: {\n          url: imagePayload,\n        },\n      },\n    ],\n  });\n\n  const res = await callAIWithStringResponse(\n    [\n      {\n        role: 'user',\n        content: systemPrompt,\n      },\n      ...conversationHistory.snapshot(),\n    ],\n    modelConfig,\n  );\n  const convertedText = convertBboxToCoordinates(res.content);\n\n  const { size } = context;\n  const { parsed } = actionParser({\n    prediction: convertedText,\n    factor: [1000, 1000],\n    screenContext: {\n      width: size.width,\n      height: size.height,\n    },\n    modelVer: uiTarsModelVersion,\n  });\n\n  debug(\n    'ui-tars modelVer',\n    uiTarsModelVersion,\n    ', parsed',\n    JSON.stringify(parsed),\n  );\n\n  const transformActions: PlanningAction[] = [];\n  const unhandledActions: Array<{ type: string; thought: string }> = [];\n  let shouldContinue = true;\n  parsed.forEach((action) => {\n    const actionType = (action.action_type || '').toLowerCase();\n    if (actionType === 'click') {\n      assert(action.action_inputs.start_box, 'start_box is required');\n      const point = getPoint(action.action_inputs.start_box, size);\n      transformActions.push({\n        type: 'Tap',\n        param: {\n          locate: {\n            prompt: action.thought || '',\n            bbox: pointToBbox(\n              { x: point[0], y: point[1] },\n              size.width,\n              size.height,\n            ),\n          },\n        },\n      });\n    } else if (actionType === 'left_double') {\n      assert(action.action_inputs.start_box, 'start_box is required');\n      const point = getPoint(action.action_inputs.start_box, size);\n      transformActions.push({\n        type: 'DoubleClick',\n        param: {\n          locate: {\n            prompt: action.thought || '',\n            bbox: pointToBbox(\n              { x: point[0], y: point[1] },\n              size.width,\n              size.height,\n            ),\n          },\n        },\n        thought: action.thought || '',\n      });\n    } else if (actionType === 'right_single') {\n      assert(action.action_inputs.start_box, 'start_box is required');\n      const point = getPoint(action.action_inputs.start_box, size);\n      transformActions.push({\n        type: 'RightClick',\n        param: {\n          locate: {\n            prompt: action.thought || '',\n            bbox: pointToBbox(\n              { x: point[0], y: point[1] },\n              size.width,\n              size.height,\n            ),\n          },\n        },\n        thought: action.thought || '',\n      });\n    } else if (actionType === 'drag') {\n      assert(action.action_inputs.start_box, 'start_box is required');\n      assert(action.action_inputs.end_box, 'end_box is required');\n      const startPoint = getPoint(action.action_inputs.start_box, size);\n      const endPoint = getPoint(action.action_inputs.end_box, size);\n      transformActions.push({\n        type: 'DragAndDrop',\n        param: {\n          from: {\n            prompt: action.thought || '',\n            bbox: pointToBbox(\n              { x: startPoint[0], y: startPoint[1] },\n              size.width,\n              size.height,\n            ),\n          },\n          to: {\n            prompt: action.thought || '',\n            bbox: pointToBbox(\n              { x: endPoint[0], y: endPoint[1] },\n              size.width,\n              size.height,\n            ),\n          },\n        },\n        thought: action.thought || '',\n      });\n    } else if (actionType === 'type') {\n      transformActions.push({\n        type: 'Input',\n        param: {\n          value: action.action_inputs.content,\n        },\n        thought: action.thought || '',\n      });\n    } else if (actionType === 'scroll') {\n      transformActions.push({\n        type: 'Scroll',\n        param: {\n          direction: action.action_inputs.direction,\n        },\n        thought: action.thought || '',\n      });\n    } else if (actionType === 'finished') {\n      shouldContinue = false;\n      transformActions.push({\n        type: 'Finished',\n        param: {},\n        thought: action.thought || '',\n      });\n    } else if (actionType === 'hotkey') {\n      if (!action.action_inputs.key) {\n        console.warn(\n          'No key found in action: hotkey. Will not perform action.',\n        );\n      } else {\n        const keys = transformHotkeyInput(action.action_inputs.key);\n\n        transformActions.push({\n          type: 'KeyboardPress',\n          param: {\n            keyName: keys,\n          },\n          thought: action.thought || '',\n        });\n      }\n    } else if (actionType === 'wait') {\n      transformActions.push({\n        type: 'Sleep',\n        param: {\n          timeMs: 1000,\n        },\n        thought: action.thought || '',\n      });\n    } else if (actionType) {\n      // Track unhandled action types\n      unhandledActions.push({\n        type: actionType,\n        thought: action.thought || '',\n      });\n      debug('Unhandled action type:', actionType, 'thought:', action.thought);\n    }\n  });\n\n  if (transformActions.length === 0) {\n    const errorDetails: string[] = [];\n\n    // Check if parsing failed\n    if (parsed.length === 0) {\n      errorDetails.push('Action parser returned no actions');\n\n      // Check if response has Thought but no Action\n      if (\n        res.content.includes('Thought:') &&\n        !res.content.includes('Action:')\n      ) {\n        errorDetails.push(\n          'Response contains \"Thought:\" but missing \"Action:\" line',\n        );\n      } else {\n        errorDetails.push('Response may be malformed or empty');\n      }\n    }\n\n    // Check if we have unhandled action types\n    if (unhandledActions.length > 0) {\n      const types = unhandledActions.map((a) => a.type).join(', ');\n      errorDetails.push(`Unhandled action types: ${types}`);\n    }\n\n    const errorMessage = [\n      'No actions found in UI-TARS response.',\n      ...errorDetails,\n      `\\nRaw response: ${res.content}`,\n    ].join('\\n');\n\n    throw new Error(errorMessage, {\n      cause: {\n        prediction: res.content,\n        parsed,\n        unhandledActions,\n        convertedText,\n      },\n    });\n  }\n\n  debug('transformActions', JSON.stringify(transformActions, null, 2));\n  const log = getSummary(res.content);\n\n  conversationHistory.append({\n    role: 'assistant',\n    content: log,\n  });\n\n  return {\n    actions: transformActions,\n    log,\n    usage: res.usage,\n    rawResponse: JSON.stringify(res.content, undefined, 2),\n    more_actions_needed_by_instruction: shouldContinue,\n  };\n}\n\n/**\n * Converts bounding box notation to coordinate points\n * @param text - The text containing bbox tags to be converted\n * @returns The text with bbox tags replaced by coordinate points\n */\nfunction convertBboxToCoordinates(text: string): string {\n  // Match the four numbers after <bbox>\n  const pattern = /<bbox>(\\d+)\\s+(\\d+)\\s+(\\d+)\\s+(\\d+)<\\/bbox>/g;\n\n  function replaceMatch(\n    match: string,\n    x1: string,\n    y1: string,\n    x2: string,\n    y2: string,\n  ): string {\n    // Convert strings to numbers and calculate center point\n    const x1Num = Number.parseInt(x1, 10);\n    const y1Num = Number.parseInt(y1, 10);\n    const x2Num = Number.parseInt(x2, 10);\n    const y2Num = Number.parseInt(y2, 10);\n\n    // Use Math.floor to truncate and calculate center point\n    const x = Math.floor((x1Num + x2Num) / 2);\n    const y = Math.floor((y1Num + y2Num) / 2);\n\n    // Return formatted coordinate string\n    return `(${x},${y})`;\n  }\n\n  // Remove [EOS] and replace <bbox> coordinates\n  const cleanedText = text.replace(/\\[EOS\\]/g, '');\n  return cleanedText.replace(pattern, replaceMatch).trim();\n}\n\nfunction getPoint(startBox: string, size: { width: number; height: number }) {\n  const [x, y] = JSON.parse(startBox);\n  return [x * size.width, y * size.height];\n}\n\ninterface BaseAction {\n  action_type: ActionType;\n  action_inputs: Record<string, any>;\n  reflection: string | null;\n  thought: string | null;\n}\n\ninterface ClickAction extends BaseAction {\n  action_type: 'click';\n  action_inputs: {\n    start_box: string; // JSON string of [x, y] coordinates\n  };\n}\n\ninterface DragAction extends BaseAction {\n  action_type: 'drag';\n  action_inputs: {\n    start_box: string; // JSON string of [x, y] coordinates\n    end_box: string; // JSON string of [x, y] coordinates\n  };\n}\n\ninterface WaitAction extends BaseAction {\n  action_type: 'wait';\n  action_inputs: {\n    time: string; // JSON string of [x, y] coordinates\n  };\n}\n\ninterface LeftDoubleAction extends BaseAction {\n  action_type: 'left_double';\n  action_inputs: {\n    start_box: string; // JSON string of [x, y] coordinates\n  };\n}\n\ninterface RightSingleAction extends BaseAction {\n  action_type: 'right_single';\n  action_inputs: {\n    start_box: string; // JSON string of [x, y] coordinates\n  };\n}\n\ninterface TypeAction extends BaseAction {\n  action_type: 'type';\n  action_inputs: {\n    content: string;\n  };\n}\n\ninterface HotkeyAction extends BaseAction {\n  action_type: 'hotkey';\n  action_inputs: {\n    key: string;\n  };\n}\n\ninterface ScrollAction extends BaseAction {\n  action_type: 'scroll';\n  action_inputs: {\n    direction: 'up' | 'down';\n  };\n}\n\ninterface FinishedAction extends BaseAction {\n  action_type: 'finished';\n  action_inputs: Record<string, never>;\n}\n\nexport type Action =\n  | ClickAction\n  | LeftDoubleAction\n  | RightSingleAction\n  | DragAction\n  | TypeAction\n  | HotkeyAction\n  | ScrollAction\n  | FinishedAction\n  | WaitAction;\n\nexport async function resizeImageForUiTars(\n  imageBase64: string,\n  size: Size,\n  uiTarsVersion: UITarsModelVersion | undefined,\n) {\n  if (uiTarsVersion === UITarsModelVersion.V1_5) {\n    debug('ui-tars-v1.5, will check image size', size);\n    const currentPixels = size.width * size.height;\n    const maxPixels = 16384 * 28 * 28; //\n    if (currentPixels > maxPixels) {\n      const resizeFactor = Math.sqrt(maxPixels / currentPixels);\n      const newWidth = Math.floor(size.width * resizeFactor);\n      const newHeight = Math.floor(size.height * resizeFactor);\n      debug(\n        'resize image for ui-tars, new width: %s, new height: %s',\n        newWidth,\n        newHeight,\n      );\n      const resizedImage = await resizeImgBase64(imageBase64, {\n        width: newWidth,\n        height: newHeight,\n      });\n      return resizedImage;\n    }\n  }\n  return imageBase64;\n}\n"],"names":["debug","getDebug","bboxSize","pointToBbox","point","width","height","Math","uiTarsPlanning","userInstruction","options","conversationHistory","context","modelConfig","actionContext","uiTarsModelVersion","instruction","systemPrompt","getUiTarsPlanningPrompt","imagePayload","resizeImageForUiTars","res","callAIWithStringResponse","convertedText","convertBboxToCoordinates","size","parsed","actionParser","JSON","transformActions","unhandledActions","shouldContinue","action","actionType","assert","getPoint","startPoint","endPoint","keys","transformHotkeyInput","console","errorDetails","types","a","errorMessage","Error","log","getSummary","undefined","text","pattern","replaceMatch","match","x1","y1","x2","y2","x1Num","Number","y1Num","x2Num","y2Num","x","y","cleanedText","startBox","imageBase64","uiTarsVersion","UITarsModelVersion","currentPixels","maxPixels","resizeFactor","newWidth","newHeight","resizedImage","resizeImgBase64"],"mappings":";;;;;;;;AA0BA,MAAMA,QAAQC,SAAS;AACvB,MAAMC,WAAW;AACjB,MAAMC,cAAc,CAClBC,OACAC,OACAC,SAEO;QACLC,KAAK,KAAK,CAACA,KAAK,GAAG,CAACH,MAAM,CAAC,GAAGF,WAAW,GAAG;QAC5CK,KAAK,KAAK,CAACA,KAAK,GAAG,CAACH,MAAM,CAAC,GAAGF,WAAW,GAAG;QAC5CK,KAAK,KAAK,CAACA,KAAK,GAAG,CAACH,MAAM,CAAC,GAAGF,WAAW,GAAGG;QAC5CE,KAAK,KAAK,CAACA,KAAK,GAAG,CAACH,MAAM,CAAC,GAAGF,WAAW,GAAGI;KAC7C;AAGI,eAAeE,eACpBC,eAAuB,EACvBC,OAKC;IAED,MAAM,EAAEC,mBAAmB,EAAEC,OAAO,EAAEC,WAAW,EAAEC,aAAa,EAAE,GAAGJ;IACrE,MAAM,EAAEK,kBAAkB,EAAE,GAAGF;IAE/B,IAAIG,cAAcP;IAClB,IAAIK,eACFE,cAAc,CAAC,yBAAyB,EAAEF,cAAc,8CAA8C,EAAEL,gBAAgB,mBAAmB,CAAC;IAG9I,MAAMQ,eAAeC,4BAA4BF;IAEjD,MAAMG,eAAe,MAAMC,qBACzBR,QAAQ,gBAAgB,EACxBA,QAAQ,IAAI,EACZG;IAGFJ,oBAAoB,MAAM,CAAC;QACzB,MAAM;QACN,SAAS;YACP;gBACE,MAAM;gBACN,WAAW;oBACT,KAAKQ;gBACP;YACF;SACD;IACH;IAEA,MAAME,MAAM,MAAMC,yBAChB;QACE;YACE,MAAM;YACN,SAASL;QACX;WACGN,oBAAoB,QAAQ;KAChC,EACDE;IAEF,MAAMU,gBAAgBC,yBAAyBH,IAAI,OAAO;IAE1D,MAAM,EAAEI,IAAI,EAAE,GAAGb;IACjB,MAAM,EAAEc,MAAM,EAAE,GAAGC,aAAa;QAC9B,YAAYJ;QACZ,QAAQ;YAAC;YAAM;SAAK;QACpB,eAAe;YACb,OAAOE,KAAK,KAAK;YACjB,QAAQA,KAAK,MAAM;QACrB;QACA,UAAUV;IACZ;IAEAf,MACE,oBACAe,oBACA,YACAa,KAAK,SAAS,CAACF;IAGjB,MAAMG,mBAAqC,EAAE;IAC7C,MAAMC,mBAA6D,EAAE;IACrE,IAAIC,iBAAiB;IACrBL,OAAO,OAAO,CAAC,CAACM;QACd,MAAMC,aAAcD,AAAAA,CAAAA,OAAO,WAAW,IAAI,EAAC,EAAG,WAAW;QACzD,IAAIC,AAAe,YAAfA,YAAwB;YAC1BC,OAAOF,OAAO,aAAa,CAAC,SAAS,EAAE;YACvC,MAAM5B,QAAQ+B,SAASH,OAAO,aAAa,CAAC,SAAS,EAAEP;YACvDI,iBAAiB,IAAI,CAAC;gBACpB,MAAM;gBACN,OAAO;oBACL,QAAQ;wBACN,QAAQG,OAAO,OAAO,IAAI;wBAC1B,MAAM7B,YACJ;4BAAE,GAAGC,KAAK,CAAC,EAAE;4BAAE,GAAGA,KAAK,CAAC,EAAE;wBAAC,GAC3BqB,KAAK,KAAK,EACVA,KAAK,MAAM;oBAEf;gBACF;YACF;QACF,OAAO,IAAIQ,AAAe,kBAAfA,YAA8B;YACvCC,OAAOF,OAAO,aAAa,CAAC,SAAS,EAAE;YACvC,MAAM5B,QAAQ+B,SAASH,OAAO,aAAa,CAAC,SAAS,EAAEP;YACvDI,iBAAiB,IAAI,CAAC;gBACpB,MAAM;gBACN,OAAO;oBACL,QAAQ;wBACN,QAAQG,OAAO,OAAO,IAAI;wBAC1B,MAAM7B,YACJ;4BAAE,GAAGC,KAAK,CAAC,EAAE;4BAAE,GAAGA,KAAK,CAAC,EAAE;wBAAC,GAC3BqB,KAAK,KAAK,EACVA,KAAK,MAAM;oBAEf;gBACF;gBACA,SAASO,OAAO,OAAO,IAAI;YAC7B;QACF,OAAO,IAAIC,AAAe,mBAAfA,YAA+B;YACxCC,OAAOF,OAAO,aAAa,CAAC,SAAS,EAAE;YACvC,MAAM5B,QAAQ+B,SAASH,OAAO,aAAa,CAAC,SAAS,EAAEP;YACvDI,iBAAiB,IAAI,CAAC;gBACpB,MAAM;gBACN,OAAO;oBACL,QAAQ;wBACN,QAAQG,OAAO,OAAO,IAAI;wBAC1B,MAAM7B,YACJ;4BAAE,GAAGC,KAAK,CAAC,EAAE;4BAAE,GAAGA,KAAK,CAAC,EAAE;wBAAC,GAC3BqB,KAAK,KAAK,EACVA,KAAK,MAAM;oBAEf;gBACF;gBACA,SAASO,OAAO,OAAO,IAAI;YAC7B;QACF,OAAO,IAAIC,AAAe,WAAfA,YAAuB;YAChCC,OAAOF,OAAO,aAAa,CAAC,SAAS,EAAE;YACvCE,OAAOF,OAAO,aAAa,CAAC,OAAO,EAAE;YACrC,MAAMI,aAAaD,SAASH,OAAO,aAAa,CAAC,SAAS,EAAEP;YAC5D,MAAMY,WAAWF,SAASH,OAAO,aAAa,CAAC,OAAO,EAAEP;YACxDI,iBAAiB,IAAI,CAAC;gBACpB,MAAM;gBACN,OAAO;oBACL,MAAM;wBACJ,QAAQG,OAAO,OAAO,IAAI;wBAC1B,MAAM7B,YACJ;4BAAE,GAAGiC,UAAU,CAAC,EAAE;4BAAE,GAAGA,UAAU,CAAC,EAAE;wBAAC,GACrCX,KAAK,KAAK,EACVA,KAAK,MAAM;oBAEf;oBACA,IAAI;wBACF,QAAQO,OAAO,OAAO,IAAI;wBAC1B,MAAM7B,YACJ;4BAAE,GAAGkC,QAAQ,CAAC,EAAE;4BAAE,GAAGA,QAAQ,CAAC,EAAE;wBAAC,GACjCZ,KAAK,KAAK,EACVA,KAAK,MAAM;oBAEf;gBACF;gBACA,SAASO,OAAO,OAAO,IAAI;YAC7B;QACF,OAAO,IAAIC,AAAe,WAAfA,YACTJ,iBAAiB,IAAI,CAAC;YACpB,MAAM;YACN,OAAO;gBACL,OAAOG,OAAO,aAAa,CAAC,OAAO;YACrC;YACA,SAASA,OAAO,OAAO,IAAI;QAC7B;aACK,IAAIC,AAAe,aAAfA,YACTJ,iBAAiB,IAAI,CAAC;YACpB,MAAM;YACN,OAAO;gBACL,WAAWG,OAAO,aAAa,CAAC,SAAS;YAC3C;YACA,SAASA,OAAO,OAAO,IAAI;QAC7B;aACK,IAAIC,AAAe,eAAfA,YAA2B;YACpCF,iBAAiB;YACjBF,iBAAiB,IAAI,CAAC;gBACpB,MAAM;gBACN,OAAO,CAAC;gBACR,SAASG,OAAO,OAAO,IAAI;YAC7B;QACF,OAAO,IAAIC,AAAe,aAAfA,YACT,IAAKD,OAAO,aAAa,CAAC,GAAG,EAItB;YACL,MAAMM,OAAOC,qBAAqBP,OAAO,aAAa,CAAC,GAAG;YAE1DH,iBAAiB,IAAI,CAAC;gBACpB,MAAM;gBACN,OAAO;oBACL,SAASS;gBACX;gBACA,SAASN,OAAO,OAAO,IAAI;YAC7B;QACF,OAbEQ,QAAQ,IAAI,CACV;aAaC,IAAIP,AAAe,WAAfA,YACTJ,iBAAiB,IAAI,CAAC;YACpB,MAAM;YACN,OAAO;gBACL,QAAQ;YACV;YACA,SAASG,OAAO,OAAO,IAAI;QAC7B;aACK,IAAIC,YAAY;YAErBH,iBAAiB,IAAI,CAAC;gBACpB,MAAMG;gBACN,SAASD,OAAO,OAAO,IAAI;YAC7B;YACAhC,MAAM,0BAA0BiC,YAAY,YAAYD,OAAO,OAAO;QACxE;IACF;IAEA,IAAIH,AAA4B,MAA5BA,iBAAiB,MAAM,EAAQ;QACjC,MAAMY,eAAyB,EAAE;QAGjC,IAAIf,AAAkB,MAAlBA,OAAO,MAAM,EAAQ;YACvBe,aAAa,IAAI,CAAC;YAGlB,IACEpB,IAAI,OAAO,CAAC,QAAQ,CAAC,eACrB,CAACA,IAAI,OAAO,CAAC,QAAQ,CAAC,YAEtBoB,aAAa,IAAI,CACf;iBAGFA,aAAa,IAAI,CAAC;QAEtB;QAGA,IAAIX,iBAAiB,MAAM,GAAG,GAAG;YAC/B,MAAMY,QAAQZ,iBAAiB,GAAG,CAAC,CAACa,IAAMA,EAAE,IAAI,EAAE,IAAI,CAAC;YACvDF,aAAa,IAAI,CAAC,CAAC,wBAAwB,EAAEC,OAAO;QACtD;QAEA,MAAME,eAAe;YACnB;eACGH;YACH,CAAC,gBAAgB,EAAEpB,IAAI,OAAO,EAAE;SACjC,CAAC,IAAI,CAAC;QAEP,MAAM,IAAIwB,MAAMD,cAAc;YAC5B,OAAO;gBACL,YAAYvB,IAAI,OAAO;gBACvBK;gBACAI;gBACAP;YACF;QACF;IACF;IAEAvB,MAAM,oBAAoB4B,KAAK,SAAS,CAACC,kBAAkB,MAAM;IACjE,MAAMiB,MAAMC,WAAW1B,IAAI,OAAO;IAElCV,oBAAoB,MAAM,CAAC;QACzB,MAAM;QACN,SAASmC;IACX;IAEA,OAAO;QACL,SAASjB;QACTiB;QACA,OAAOzB,IAAI,KAAK;QAChB,aAAaO,KAAK,SAAS,CAACP,IAAI,OAAO,EAAE2B,QAAW;QACpD,oCAAoCjB;IACtC;AACF;AAOA,SAASP,yBAAyByB,IAAY;IAE5C,MAAMC,UAAU;IAEhB,SAASC,aACPC,KAAa,EACbC,EAAU,EACVC,EAAU,EACVC,EAAU,EACVC,EAAU;QAGV,MAAMC,QAAQC,OAAO,QAAQ,CAACL,IAAI;QAClC,MAAMM,QAAQD,OAAO,QAAQ,CAACJ,IAAI;QAClC,MAAMM,QAAQF,OAAO,QAAQ,CAACH,IAAI;QAClC,MAAMM,QAAQH,OAAO,QAAQ,CAACF,IAAI;QAGlC,MAAMM,IAAIvD,KAAK,KAAK,CAAEkD,AAAAA,CAAAA,QAAQG,KAAI,IAAK;QACvC,MAAMG,IAAIxD,KAAK,KAAK,CAAEoD,AAAAA,CAAAA,QAAQE,KAAI,IAAK;QAGvC,OAAO,CAAC,CAAC,EAAEC,EAAE,CAAC,EAAEC,EAAE,CAAC,CAAC;IACtB;IAGA,MAAMC,cAAcf,KAAK,OAAO,CAAC,YAAY;IAC7C,OAAOe,YAAY,OAAO,CAACd,SAASC,cAAc,IAAI;AACxD;AAEA,SAAShB,SAAS8B,QAAgB,EAAExC,IAAuC;IACzE,MAAM,CAACqC,GAAGC,EAAE,GAAGnC,KAAK,KAAK,CAACqC;IAC1B,OAAO;QAACH,IAAIrC,KAAK,KAAK;QAAEsC,IAAItC,KAAK,MAAM;KAAC;AAC1C;AAkFO,eAAeL,qBACpB8C,WAAmB,EACnBzC,IAAU,EACV0C,aAA6C;IAE7C,IAAIA,kBAAkBC,mBAAmB,IAAI,EAAE;QAC7CpE,MAAM,uCAAuCyB;QAC7C,MAAM4C,gBAAgB5C,KAAK,KAAK,GAAGA,KAAK,MAAM;QAC9C,MAAM6C,YAAY;QAClB,IAAID,gBAAgBC,WAAW;YAC7B,MAAMC,eAAehE,KAAK,IAAI,CAAC+D,YAAYD;YAC3C,MAAMG,WAAWjE,KAAK,KAAK,CAACkB,KAAK,KAAK,GAAG8C;YACzC,MAAME,YAAYlE,KAAK,KAAK,CAACkB,KAAK,MAAM,GAAG8C;YAC3CvE,MACE,2DACAwE,UACAC;YAEF,MAAMC,eAAe,MAAMC,gBAAgBT,aAAa;gBACtD,OAAOM;gBACP,QAAQC;YACV;YACA,OAAOC;QACT;IACF;IACA,OAAOR;AACT"}

package/dist/es/common.mjs

@@ -4,15 +4,6 @@ import { treeToList } from "@midscene/shared/extractor";
 import { compositeElementInfoImg } from "@midscene/shared/img";
 import { getDebug } from "@midscene/shared/logger";
 import { z } from "zod";
-var common_AIActionType = /*#__PURE__*/ function(AIActionType) {
-    AIActionType[AIActionType["ASSERT"] = 0] = "ASSERT";
-    AIActionType[AIActionType["INSPECT_ELEMENT"] = 1] = "INSPECT_ELEMENT";
-    AIActionType[AIActionType["EXTRACT_DATA"] = 2] = "EXTRACT_DATA";
-    AIActionType[AIActionType["PLAN"] = 3] = "PLAN";
-    AIActionType[AIActionType["DESCRIBE_ELEMENT"] = 4] = "DESCRIBE_ELEMENT";
-    AIActionType[AIActionType["TEXT"] = 5] = "TEXT";
-    return AIActionType;
-}({});
 const defaultBboxSize = 20;
 const debugInspectUtils = getDebug('ai:common');
 function fillBboxParam(locate, width, height, rightLimit, bottomLimit, vlMode) {
@@ -154,7 +145,7 @@ function mergeRects(rects) {
 }
 function expandSearchArea(rect, screenSize, vlMode) {
     let minEdgeSize = 500;
-    if ('qwen3-vl' === vlMode) minEdgeSize = 1000;
+    if ('qwen3-vl' === vlMode) minEdgeSize = 1200;
     const defaultPadding = 160;
     const paddingSizeHorizontal = rect.width < minEdgeSize ? Math.ceil((minEdgeSize - rect.width) / 2) : defaultPadding;
     const paddingSizeVertical = rect.height < minEdgeSize ? Math.ceil((minEdgeSize - rect.height) / 2) : defaultPadding;
@@ -339,6 +330,6 @@ const parseActionParam = (rawParam, zodSchema)=>{
     for(const fieldName in locateFieldValues)validated[fieldName] = locateFieldValues[fieldName];
     return validated;
 };
-export { common_AIActionType as AIActionType, PointSchema, RectSchema, SizeSchema, TMultimodalPromptSchema, TUserPromptSchema, adaptBbox, adaptBboxToRect, adaptDoubaoBbox, adaptGeminiBbox, adaptQwen2_5Bbox, buildYamlFlowFromPlans, dumpActionParam, dumpMidsceneLocatorField, expandSearchArea, fillBboxParam, findAllMidsceneLocatorField, getMidsceneLocationSchema, ifMidsceneLocatorField, loadActionParam, markupImageForLLM, mergeRects, normalized01000, parseActionParam };
+export { PointSchema, RectSchema, SizeSchema, TMultimodalPromptSchema, TUserPromptSchema, adaptBbox, adaptBboxToRect, adaptDoubaoBbox, adaptGeminiBbox, adaptQwen2_5Bbox, buildYamlFlowFromPlans, dumpActionParam, dumpMidsceneLocatorField, expandSearchArea, fillBboxParam, findAllMidsceneLocatorField, getMidsceneLocationSchema, ifMidsceneLocatorField, loadActionParam, markupImageForLLM, mergeRects, normalized01000, parseActionParam };
 
 //# sourceMappingURL=common.mjs.map