@midscene/core 1.14.0 → 1.15.0

package/dist/es/ai-model/service-caller/openai/chat-completion/stream.mjs

@@ -7,18 +7,15 @@ const callChatCompletionStream = async ({ completion, openAIRequestContext, extr
     let usage;
     let responseModelName;
     requestSignal.throwIfAborted();
-    const stream = await completion.create({
+    const stream = completion.stream({
         ...requestBodyParams,
-        stream: true,
         stream_options: {
             ...requestBodyParams.stream_options,
             include_usage: true
         }
     }, {
-        stream: true,
         signal: requestSignal
     });
-    const requestId = stream._request_id ?? openAIRequestContext.responseRequestId?.requestId;
     let chunkSequence = 0;
     for await (const chunk of stream){
         requestSignal.throwIfAborted();
@@ -28,7 +25,8 @@ const callChatCompletionStream = async ({ completion, openAIRequestContext, extr
             sequence: chunkSequence,
             chunk
         });
-        const parsedChunk = extractContentAndReasoning(chunk.choices?.[0]?.delta);
+        const delta = chunk.choices?.[0]?.delta;
+        const parsedChunk = extractContentAndReasoning(delta);
         const content = parsedChunk.content || '';
         const reasoning_content = parsedChunk.reasoning_content || '';
         if (chunk.usage) usage = chunk.usage;
@@ -46,6 +44,9 @@ const callChatCompletionStream = async ({ completion, openAIRequestContext, extr
             onChunk(chunkData);
         }
     }
+    const finalCompletion = await stream.finalChatCompletion();
+    const assistantMessage = finalCompletion.choices[0].message;
+    const requestId = openAIRequestContext.responseRequestId?.requestId;
     const finalAccumulated = resolveContentWithReasoningFallback({
         content: accumulated,
         reasoningContent: accumulatedReasoning,
@@ -63,7 +64,10 @@ const callChatCompletionStream = async ({ completion, openAIRequestContext, extr
     return {
         content: accumulated,
         reasoningContent: accumulatedReasoning,
-        rawAssistantOutput: void 0,
+        rawAssistantOutput: {
+            type: 'chat-completion',
+            rawValue: assistantMessage
+        },
         rawUsage: usage,
         requestId,
         responseModelName

package/dist/es/ai-model/service-caller/openai/chat-completion/stream.mjs.map

@@ -1 +1 @@
-{"version":3,"file":"ai-model/service-caller/openai/chat-completion/stream.mjs","sources":["../../../../../../src/ai-model/service-caller/openai/chat-completion/stream.ts"],"sourcesContent":["import type { CodeGenerationChunk } from '@/types';\nimport { assert } from '@midscene/shared/utils';\nimport type OpenAI from 'openai';\nimport type { Stream } from 'openai/streaming';\nimport type { OpenAIProtocolCallResult } from '../../types';\nimport type { ChatCompletionCallOptions } from './types';\nimport { resolveContentWithReasoningFallback } from './utils';\n\nexport const callChatCompletionStream = async ({\n  completion,\n  openAIRequestContext,\n  extractContentAndReasoning,\n  useReasoningAsContentFallback,\n  requestBodyParams,\n  requestSignal,\n  onChunk,\n  recordEvent,\n}: ChatCompletionCallOptions): Promise<OpenAIProtocolCallResult> => {\n  assert(\n    typeof onChunk === 'function',\n    'onChunk is required when stream is true',\n  );\n  let accumulated = '';\n  let accumulatedReasoning = '';\n  let usage: OpenAI.CompletionUsage | undefined;\n  let responseModelName: string | undefined;\n  requestSignal.throwIfAborted();\n  const stream = (await completion.create(\n    {\n      ...requestBodyParams,\n      stream: true,\n      stream_options: {\n        ...requestBodyParams.stream_options,\n        include_usage: true,\n      },\n    },\n    {\n      stream: true,\n      signal: requestSignal,\n    },\n  )) as Stream<OpenAI.Chat.Completions.ChatCompletionChunk> & {\n    _request_id?: string | null;\n  };\n\n  const requestId =\n    stream._request_id ?? openAIRequestContext.responseRequestId?.requestId;\n\n  let chunkSequence = 0;\n  for await (const chunk of stream) {\n    requestSignal.throwIfAborted();\n    chunkSequence += 1;\n    recordEvent?.({\n      type: 'chunk',\n      sequence: chunkSequence,\n      chunk,\n    });\n    const parsedChunk = extractContentAndReasoning(chunk.choices?.[0]?.delta);\n    const content = parsedChunk.content || '';\n    const reasoning_content = parsedChunk.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    if (chunk.model) {\n      responseModelName = chunk.model;\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      onChunk(chunkData);\n    }\n  }\n\n  const finalAccumulated = resolveContentWithReasoningFallback({\n    content: accumulated,\n    reasoningContent: accumulatedReasoning,\n    useReasoningAsContentFallback,\n  });\n  accumulated = finalAccumulated || '';\n\n  // Send final chunk\n  const finalChunk: CodeGenerationChunk = {\n    content: '',\n    accumulated,\n    reasoning_content: '',\n    isComplete: true,\n    usage,\n  };\n  onChunk(finalChunk);\n  return {\n    content: accumulated,\n    reasoningContent: accumulatedReasoning,\n    rawAssistantOutput: undefined,\n    rawUsage: usage,\n    requestId,\n    responseModelName,\n  };\n};\n"],"names":["callChatCompletionStream","completion","openAIRequestContext","extractContentAndReasoning","useReasoningAsContentFallback","requestBodyParams","requestSignal","onChunk","recordEvent","assert","accumulated","accumulatedReasoning","usage","responseModelName","stream","requestId","chunkSequence","chunk","parsedChunk","content","reasoning_content","chunkData","undefined","finalAccumulated","resolveContentWithReasoningFallback","finalChunk"],"mappings":";;AAQO,MAAMA,2BAA2B,OAAO,EAC7CC,UAAU,EACVC,oBAAoB,EACpBC,0BAA0B,EAC1BC,6BAA6B,EAC7BC,iBAAiB,EACjBC,aAAa,EACbC,OAAO,EACPC,WAAW,EACe;IAC1BC,OACE,AAAmB,cAAnB,OAAOF,SACP;IAEF,IAAIG,cAAc;IAClB,IAAIC,uBAAuB;IAC3B,IAAIC;IACJ,IAAIC;IACJP,cAAc,cAAc;IAC5B,MAAMQ,SAAU,MAAMb,WAAW,MAAM,CACrC;QACE,GAAGI,iBAAiB;QACpB,QAAQ;QACR,gBAAgB;YACd,GAAGA,kBAAkB,cAAc;YACnC,eAAe;QACjB;IACF,GACA;QACE,QAAQ;QACR,QAAQC;IACV;IAKF,MAAMS,YACJD,OAAO,WAAW,IAAIZ,qBAAqB,iBAAiB,EAAE;IAEhE,IAAIc,gBAAgB;IACpB,WAAW,MAAMC,SAASH,OAAQ;QAChCR,cAAc,cAAc;QAC5BU,iBAAiB;QACjBR,cAAc;YACZ,MAAM;YACN,UAAUQ;YACVC;QACF;QACA,MAAMC,cAAcf,2BAA2Bc,MAAM,OAAO,EAAE,CAAC,EAAE,EAAE;QACnE,MAAME,UAAUD,YAAY,OAAO,IAAI;QACvC,MAAME,oBAAoBF,YAAY,iBAAiB,IAAI;QAG3D,IAAID,MAAM,KAAK,EACbL,QAAQK,MAAM,KAAK;QAErB,IAAIA,MAAM,KAAK,EACbJ,oBAAoBI,MAAM,KAAK;QAGjC,IAAIE,WAAWC,mBAAmB;YAChCV,eAAeS;YACfR,wBAAwBS;YACxB,MAAMC,YAAiC;gBACrCF;gBACAC;gBACAV;gBACA,YAAY;gBACZ,OAAOY;YACT;YACAf,QAAQc;QACV;IACF;IAEA,MAAME,mBAAmBC,oCAAoC;QAC3D,SAASd;QACT,kBAAkBC;QAClBP;IACF;IACAM,cAAca,oBAAoB;IAGlC,MAAME,aAAkC;QACtC,SAAS;QACTf;QACA,mBAAmB;QACnB,YAAY;QACZE;IACF;IACAL,QAAQkB;IACR,OAAO;QACL,SAASf;QACT,kBAAkBC;QAClB,oBAAoBW;QACpB,UAAUV;QACVG;QACAF;IACF;AACF"}
+{"version":3,"file":"ai-model/service-caller/openai/chat-completion/stream.mjs","sources":["../../../../../../src/ai-model/service-caller/openai/chat-completion/stream.ts"],"sourcesContent":["import type { CodeGenerationChunk } from '@/types';\nimport { assert } from '@midscene/shared/utils';\nimport type OpenAI from 'openai';\nimport type { OpenAIProtocolCallResult } from '../../types';\nimport type { ChatCompletionCallOptions } from './types';\nimport { resolveContentWithReasoningFallback } from './utils';\n\nexport const callChatCompletionStream = async ({\n  completion,\n  openAIRequestContext,\n  extractContentAndReasoning,\n  useReasoningAsContentFallback,\n  requestBodyParams,\n  requestSignal,\n  onChunk,\n  recordEvent,\n}: ChatCompletionCallOptions): Promise<OpenAIProtocolCallResult> => {\n  assert(\n    typeof onChunk === 'function',\n    'onChunk is required when stream is true',\n  );\n  let accumulated = '';\n  let accumulatedReasoning = '';\n  let usage: OpenAI.CompletionUsage | undefined;\n  let responseModelName: string | undefined;\n  requestSignal.throwIfAborted();\n  const stream = completion.stream(\n    {\n      ...requestBodyParams,\n      stream_options: {\n        ...requestBodyParams.stream_options,\n        include_usage: true,\n      },\n    },\n    {\n      signal: requestSignal,\n    },\n  );\n\n  let chunkSequence = 0;\n  for await (const chunk of stream) {\n    requestSignal.throwIfAborted();\n    chunkSequence += 1;\n    recordEvent?.({\n      type: 'chunk',\n      sequence: chunkSequence,\n      chunk,\n    });\n    const delta = chunk.choices?.[0]?.delta;\n    const parsedChunk = extractContentAndReasoning(delta);\n    const content = parsedChunk.content || '';\n    const reasoning_content = parsedChunk.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    if (chunk.model) {\n      responseModelName = chunk.model;\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      onChunk(chunkData);\n    }\n  }\n\n  // TODO: The SDK overwrites non-standard fields such as reasoning_content,\n  // so rawAssistantOutput may lose earlier thinking chunks (e.g. Doubao).\n  // Add provider-specific handling when full reasoning replay is needed.\n  const finalCompletion = await stream.finalChatCompletion();\n  const assistantMessage = finalCompletion.choices[0].message;\n  const requestId = openAIRequestContext.responseRequestId?.requestId;\n\n  const finalAccumulated = resolveContentWithReasoningFallback({\n    content: accumulated,\n    reasoningContent: accumulatedReasoning,\n    useReasoningAsContentFallback,\n  });\n  accumulated = finalAccumulated || '';\n\n  // Send final chunk\n  const finalChunk: CodeGenerationChunk = {\n    content: '',\n    accumulated,\n    reasoning_content: '',\n    isComplete: true,\n    usage,\n  };\n  onChunk(finalChunk);\n  return {\n    content: accumulated,\n    reasoningContent: accumulatedReasoning,\n    rawAssistantOutput: {\n      type: 'chat-completion',\n      rawValue: assistantMessage,\n    },\n    rawUsage: usage,\n    requestId,\n    responseModelName,\n  };\n};\n"],"names":["callChatCompletionStream","completion","openAIRequestContext","extractContentAndReasoning","useReasoningAsContentFallback","requestBodyParams","requestSignal","onChunk","recordEvent","assert","accumulated","accumulatedReasoning","usage","responseModelName","stream","chunkSequence","chunk","delta","parsedChunk","content","reasoning_content","chunkData","undefined","finalCompletion","assistantMessage","requestId","finalAccumulated","resolveContentWithReasoningFallback","finalChunk"],"mappings":";;AAOO,MAAMA,2BAA2B,OAAO,EAC7CC,UAAU,EACVC,oBAAoB,EACpBC,0BAA0B,EAC1BC,6BAA6B,EAC7BC,iBAAiB,EACjBC,aAAa,EACbC,OAAO,EACPC,WAAW,EACe;IAC1BC,OACE,AAAmB,cAAnB,OAAOF,SACP;IAEF,IAAIG,cAAc;IAClB,IAAIC,uBAAuB;IAC3B,IAAIC;IACJ,IAAIC;IACJP,cAAc,cAAc;IAC5B,MAAMQ,SAASb,WAAW,MAAM,CAC9B;QACE,GAAGI,iBAAiB;QACpB,gBAAgB;YACd,GAAGA,kBAAkB,cAAc;YACnC,eAAe;QACjB;IACF,GACA;QACE,QAAQC;IACV;IAGF,IAAIS,gBAAgB;IACpB,WAAW,MAAMC,SAASF,OAAQ;QAChCR,cAAc,cAAc;QAC5BS,iBAAiB;QACjBP,cAAc;YACZ,MAAM;YACN,UAAUO;YACVC;QACF;QACA,MAAMC,QAAQD,MAAM,OAAO,EAAE,CAAC,EAAE,EAAE;QAClC,MAAME,cAAcf,2BAA2Bc;QAC/C,MAAME,UAAUD,YAAY,OAAO,IAAI;QACvC,MAAME,oBAAoBF,YAAY,iBAAiB,IAAI;QAG3D,IAAIF,MAAM,KAAK,EACbJ,QAAQI,MAAM,KAAK;QAErB,IAAIA,MAAM,KAAK,EACbH,oBAAoBG,MAAM,KAAK;QAGjC,IAAIG,WAAWC,mBAAmB;YAChCV,eAAeS;YACfR,wBAAwBS;YACxB,MAAMC,YAAiC;gBACrCF;gBACAC;gBACAV;gBACA,YAAY;gBACZ,OAAOY;YACT;YACAf,QAAQc;QACV;IACF;IAKA,MAAME,kBAAkB,MAAMT,OAAO,mBAAmB;IACxD,MAAMU,mBAAmBD,gBAAgB,OAAO,CAAC,EAAE,CAAC,OAAO;IAC3D,MAAME,YAAYvB,qBAAqB,iBAAiB,EAAE;IAE1D,MAAMwB,mBAAmBC,oCAAoC;QAC3D,SAASjB;QACT,kBAAkBC;QAClBP;IACF;IACAM,cAAcgB,oBAAoB;IAGlC,MAAME,aAAkC;QACtC,SAAS;QACTlB;QACA,mBAAmB;QACnB,YAAY;QACZE;IACF;IACAL,QAAQqB;IACR,OAAO;QACL,SAASlB;QACT,kBAAkBC;QAClB,oBAAoB;YAClB,MAAM;YACN,UAAUa;QACZ;QACA,UAAUZ;QACVa;QACAZ;IACF;AACF"}

package/dist/es/ai-model/service-caller/openai/responses/stream.mjs

@@ -12,6 +12,7 @@ async function callResponsesStream({ responses, requestBodyParams, requestSignal
     }).withResponse();
     let accumulated = '';
     let finalResponse;
+    const completedItems = new Map();
     let sequence = 0;
     for await (const event of stream){
         requestSignal.throwIfAborted();
@@ -22,6 +23,7 @@ async function callResponsesStream({ responses, requestBodyParams, requestSignal
         });
         if ('error' === event.type) throw new Error(event.message);
         if ('response.failed' === event.type || 'response.incomplete' === event.type) parseResponse(event.response);
+        if ('response.output_item.done' === event.type) completedItems.set(event.output_index, event.item);
         if ('response.completed' === event.type) finalResponse = event.response;
         const content = 'response.output_text.delta' === event.type ? event.delta : '';
         const reasoningContent = 'response.reasoning_summary_text.delta' === event.type || 'response.reasoning_text.delta' === event.type ? event.delta : '';
@@ -37,7 +39,12 @@ async function callResponsesStream({ responses, requestBodyParams, requestSignal
     }
     requestSignal.throwIfAborted();
     assert(finalResponse, 'Responses stream ended without a completed response');
-    const result = parseResponse(finalResponse);
+    const result = parseResponse(finalResponse.output?.length ? finalResponse : {
+        ...finalResponse,
+        output: [
+            ...completedItems.entries()
+        ].sort(([left], [right])=>left - right).map(([, item])=>item)
+    });
     if (!hasUsableText(result.content)) throw new AIResponseParseError('empty content from AI model', result.content, void 0, result.rawAssistantOutput);
     onChunk({
         content: '',

package/dist/es/ai-model/service-caller/openai/responses/stream.mjs.map

@@ -1 +1 @@
-{"version":3,"file":"ai-model/service-caller/openai/responses/stream.mjs","sources":["../../../../../../src/ai-model/service-caller/openai/responses/stream.ts"],"sourcesContent":["import { assert } from '@midscene/shared/utils';\nimport type { Response } from 'openai/resources/responses/responses';\nimport type { OpenAIProtocolCallResult } from '../../types';\nimport { AIResponseParseError, hasUsableText } from '../../utils';\nimport type { ResponsesCallOptions } from './types';\nimport { parseResponse } from './utils';\n\nexport async function callResponsesStream({\n  responses,\n  requestBodyParams,\n  requestSignal,\n  openAIRequestContext,\n  onChunk,\n  recordEvent,\n}: ResponsesCallOptions): Promise<OpenAIProtocolCallResult> {\n  assert(\n    typeof onChunk === 'function',\n    'onChunk is required when stream is true',\n  );\n  requestSignal.throwIfAborted();\n  const { data: stream, request_id: requestId } = await responses\n    .create({ ...requestBodyParams, stream: true }, { signal: requestSignal })\n    .withResponse();\n  let accumulated = '';\n  let finalResponse: Response | undefined;\n  let sequence = 0;\n  for await (const event of stream) {\n    requestSignal.throwIfAborted();\n    recordEvent?.({ type: 'chunk', sequence: ++sequence, chunk: event });\n    if (event.type === 'error') {\n      throw new Error(event.message);\n    }\n    if (\n      event.type === 'response.failed' ||\n      event.type === 'response.incomplete'\n    ) {\n      parseResponse(event.response);\n    }\n    if (event.type === 'response.completed') {\n      finalResponse = event.response;\n    }\n    const content =\n      event.type === 'response.output_text.delta' ? event.delta : '';\n    const reasoningContent =\n      event.type === 'response.reasoning_summary_text.delta' ||\n      event.type === 'response.reasoning_text.delta'\n        ? event.delta\n        : '';\n    if (content || reasoningContent) {\n      accumulated += content;\n      onChunk({\n        content,\n        reasoning_content: reasoningContent,\n        accumulated,\n        isComplete: false,\n      });\n    }\n  }\n  requestSignal.throwIfAborted();\n  assert(finalResponse, 'Responses stream ended without a completed response');\n  const result = parseResponse(finalResponse);\n  if (!hasUsableText(result.content)) {\n    throw new AIResponseParseError(\n      'empty content from AI model',\n      result.content,\n      undefined,\n      result.rawAssistantOutput,\n    );\n  }\n  onChunk({\n    content: '',\n    reasoning_content: '',\n    accumulated: result.content,\n    isComplete: true,\n    usage: result.rawUsage,\n  });\n  return {\n    ...result,\n    requestId: requestId ?? openAIRequestContext.responseRequestId?.requestId,\n  };\n}\n"],"names":["callResponsesStream","responses","requestBodyParams","requestSignal","openAIRequestContext","onChunk","recordEvent","assert","stream","requestId","accumulated","finalResponse","sequence","event","Error","parseResponse","content","reasoningContent","result","hasUsableText","AIResponseParseError","undefined"],"mappings":";;;AAOO,eAAeA,oBAAoB,EACxCC,SAAS,EACTC,iBAAiB,EACjBC,aAAa,EACbC,oBAAoB,EACpBC,OAAO,EACPC,WAAW,EACU;IACrBC,OACE,AAAmB,cAAnB,OAAOF,SACP;IAEFF,cAAc,cAAc;IAC5B,MAAM,EAAE,MAAMK,MAAM,EAAE,YAAYC,SAAS,EAAE,GAAG,MAAMR,UACnD,MAAM,CAAC;QAAE,GAAGC,iBAAiB;QAAE,QAAQ;IAAK,GAAG;QAAE,QAAQC;IAAc,GACvE,YAAY;IACf,IAAIO,cAAc;IAClB,IAAIC;IACJ,IAAIC,WAAW;IACf,WAAW,MAAMC,SAASL,OAAQ;QAChCL,cAAc,cAAc;QAC5BG,cAAc;YAAE,MAAM;YAAS,UAAU,EAAEM;YAAU,OAAOC;QAAM;QAClE,IAAIA,AAAe,YAAfA,MAAM,IAAI,EACZ,MAAM,IAAIC,MAAMD,MAAM,OAAO;QAE/B,IACEA,AAAe,sBAAfA,MAAM,IAAI,IACVA,AAAe,0BAAfA,MAAM,IAAI,EAEVE,cAAcF,MAAM,QAAQ;QAE9B,IAAIA,AAAe,yBAAfA,MAAM,IAAI,EACZF,gBAAgBE,MAAM,QAAQ;QAEhC,MAAMG,UACJH,AAAe,iCAAfA,MAAM,IAAI,GAAoCA,MAAM,KAAK,GAAG;QAC9D,MAAMI,mBACJJ,AAAe,4CAAfA,MAAM,IAAI,IACVA,AAAe,oCAAfA,MAAM,IAAI,GACNA,MAAM,KAAK,GACX;QACN,IAAIG,WAAWC,kBAAkB;YAC/BP,eAAeM;YACfX,QAAQ;gBACNW;gBACA,mBAAmBC;gBACnBP;gBACA,YAAY;YACd;QACF;IACF;IACAP,cAAc,cAAc;IAC5BI,OAAOI,eAAe;IACtB,MAAMO,SAASH,cAAcJ;IAC7B,IAAI,CAACQ,cAAcD,OAAO,OAAO,GAC/B,MAAM,IAAIE,qBACR,+BACAF,OAAO,OAAO,EACdG,QACAH,OAAO,kBAAkB;IAG7Bb,QAAQ;QACN,SAAS;QACT,mBAAmB;QACnB,aAAaa,OAAO,OAAO;QAC3B,YAAY;QACZ,OAAOA,OAAO,QAAQ;IACxB;IACA,OAAO;QACL,GAAGA,MAAM;QACT,WAAWT,aAAaL,qBAAqB,iBAAiB,EAAE;IAClE;AACF"}
+{"version":3,"file":"ai-model/service-caller/openai/responses/stream.mjs","sources":["../../../../../../src/ai-model/service-caller/openai/responses/stream.ts"],"sourcesContent":["import { assert } from '@midscene/shared/utils';\nimport type { Response } from 'openai/resources/responses/responses';\nimport type { OpenAIProtocolCallResult } from '../../types';\nimport { AIResponseParseError, hasUsableText } from '../../utils';\nimport type { ResponsesCallOptions } from './types';\nimport { parseResponse } from './utils';\n\nexport async function callResponsesStream({\n  responses,\n  requestBodyParams,\n  requestSignal,\n  openAIRequestContext,\n  onChunk,\n  recordEvent,\n}: ResponsesCallOptions): Promise<OpenAIProtocolCallResult> {\n  assert(\n    typeof onChunk === 'function',\n    'onChunk is required when stream is true',\n  );\n  requestSignal.throwIfAborted();\n  const { data: stream, request_id: requestId } = await responses\n    .create({ ...requestBodyParams, stream: true }, { signal: requestSignal })\n    .withResponse();\n  let accumulated = '';\n  let finalResponse: Response | undefined;\n  const completedItems = new Map<number, Response['output'][number]>();\n  let sequence = 0;\n  for await (const event of stream) {\n    requestSignal.throwIfAborted();\n    recordEvent?.({ type: 'chunk', sequence: ++sequence, chunk: event });\n    if (event.type === 'error') {\n      throw new Error(event.message);\n    }\n    if (\n      event.type === 'response.failed' ||\n      event.type === 'response.incomplete'\n    ) {\n      parseResponse(event.response);\n    }\n    if (event.type === 'response.output_item.done') {\n      completedItems.set(event.output_index, event.item);\n    }\n    if (event.type === 'response.completed') {\n      finalResponse = event.response;\n    }\n    const content =\n      event.type === 'response.output_text.delta' ? event.delta : '';\n    const reasoningContent =\n      event.type === 'response.reasoning_summary_text.delta' ||\n      event.type === 'response.reasoning_text.delta'\n        ? event.delta\n        : '';\n    if (content || reasoningContent) {\n      accumulated += content;\n      onChunk({\n        content,\n        reasoning_content: reasoningContent,\n        accumulated,\n        isComplete: false,\n      });\n    }\n  }\n  requestSignal.throwIfAborted();\n  assert(finalResponse, 'Responses stream ended without a completed response');\n  // Some ChatGPT plan responses omit output from the completion event even\n  // after delivering complete items. Preserve those items, including reasoning\n  // and replay metadata, without treating an interrupted stream as complete.\n  const result = parseResponse(\n    finalResponse.output?.length\n      ? finalResponse\n      : {\n          ...finalResponse,\n          output: [...completedItems.entries()]\n            .sort(([left], [right]) => left - right)\n            .map(([, item]) => item),\n        },\n  );\n  if (!hasUsableText(result.content)) {\n    throw new AIResponseParseError(\n      'empty content from AI model',\n      result.content,\n      undefined,\n      result.rawAssistantOutput,\n    );\n  }\n  onChunk({\n    content: '',\n    reasoning_content: '',\n    accumulated: result.content,\n    isComplete: true,\n    usage: result.rawUsage,\n  });\n  return {\n    ...result,\n    requestId: requestId ?? openAIRequestContext.responseRequestId?.requestId,\n  };\n}\n"],"names":["callResponsesStream","responses","requestBodyParams","requestSignal","openAIRequestContext","onChunk","recordEvent","assert","stream","requestId","accumulated","finalResponse","completedItems","Map","sequence","event","Error","parseResponse","content","reasoningContent","result","left","right","item","hasUsableText","AIResponseParseError","undefined"],"mappings":";;;AAOO,eAAeA,oBAAoB,EACxCC,SAAS,EACTC,iBAAiB,EACjBC,aAAa,EACbC,oBAAoB,EACpBC,OAAO,EACPC,WAAW,EACU;IACrBC,OACE,AAAmB,cAAnB,OAAOF,SACP;IAEFF,cAAc,cAAc;IAC5B,MAAM,EAAE,MAAMK,MAAM,EAAE,YAAYC,SAAS,EAAE,GAAG,MAAMR,UACnD,MAAM,CAAC;QAAE,GAAGC,iBAAiB;QAAE,QAAQ;IAAK,GAAG;QAAE,QAAQC;IAAc,GACvE,YAAY;IACf,IAAIO,cAAc;IAClB,IAAIC;IACJ,MAAMC,iBAAiB,IAAIC;IAC3B,IAAIC,WAAW;IACf,WAAW,MAAMC,SAASP,OAAQ;QAChCL,cAAc,cAAc;QAC5BG,cAAc;YAAE,MAAM;YAAS,UAAU,EAAEQ;YAAU,OAAOC;QAAM;QAClE,IAAIA,AAAe,YAAfA,MAAM,IAAI,EACZ,MAAM,IAAIC,MAAMD,MAAM,OAAO;QAE/B,IACEA,AAAe,sBAAfA,MAAM,IAAI,IACVA,AAAe,0BAAfA,MAAM,IAAI,EAEVE,cAAcF,MAAM,QAAQ;QAE9B,IAAIA,AAAe,gCAAfA,MAAM,IAAI,EACZH,eAAe,GAAG,CAACG,MAAM,YAAY,EAAEA,MAAM,IAAI;QAEnD,IAAIA,AAAe,yBAAfA,MAAM,IAAI,EACZJ,gBAAgBI,MAAM,QAAQ;QAEhC,MAAMG,UACJH,AAAe,iCAAfA,MAAM,IAAI,GAAoCA,MAAM,KAAK,GAAG;QAC9D,MAAMI,mBACJJ,AAAe,4CAAfA,MAAM,IAAI,IACVA,AAAe,oCAAfA,MAAM,IAAI,GACNA,MAAM,KAAK,GACX;QACN,IAAIG,WAAWC,kBAAkB;YAC/BT,eAAeQ;YACfb,QAAQ;gBACNa;gBACA,mBAAmBC;gBACnBT;gBACA,YAAY;YACd;QACF;IACF;IACAP,cAAc,cAAc;IAC5BI,OAAOI,eAAe;IAItB,MAAMS,SAASH,cACbN,cAAc,MAAM,EAAE,SAClBA,gBACA;QACE,GAAGA,aAAa;QAChB,QAAQ;eAAIC,eAAe,OAAO;SAAG,CAClC,IAAI,CAAC,CAAC,CAACS,KAAK,EAAE,CAACC,MAAM,GAAKD,OAAOC,OACjC,GAAG,CAAC,CAAC,GAAGC,KAAK,GAAKA;IACvB;IAEN,IAAI,CAACC,cAAcJ,OAAO,OAAO,GAC/B,MAAM,IAAIK,qBACR,+BACAL,OAAO,OAAO,EACdM,QACAN,OAAO,kBAAkB;IAG7Bf,QAAQ;QACN,SAAS;QACT,mBAAmB;QACnB,aAAae,OAAO,OAAO;QAC3B,YAAY;QACZ,OAAOA,OAAO,QAAQ;IACxB;IACA,OAAO;QACL,GAAGA,MAAM;QACT,WAAWX,aAAaL,qBAAqB,iBAAiB,EAAE;IAClE;AACF"}

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

@@ -53,12 +53,13 @@ function appendAIRequestFailureSummary(error, attemptErrors, maxAttempts) {
     error.message = `${error.message}\nPrevious AI call attempt errors:\n${details}`;
     return error;
 }
-const buildUsageInfo = ({ apiType, usageData, requestId, timeCost, totalTimeCost, retryCount, modelName, modelDescription, responseModelName, slot, internalCallId })=>{
+const buildUsageInfo = ({ apiProtocol, stream, usageData, requestId, timeCost, totalTimeCost, retryCount, modelName, modelDescription, responseModelName, slot, internalCallId })=>{
     if (!usageData) return;
     const cachedInputTokens = usageData?.prompt_tokens_details?.cached_tokens;
     return {
         ...usageData,
-        api_type: apiType,
+        api_protocol: apiProtocol,
+        stream,
         prompt_tokens: usageData.prompt_tokens ?? 0,
         completion_tokens: usageData.completion_tokens ?? 0,
         total_tokens: usageData.total_tokens ?? 0,

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

@@ -1 +1 @@
-{"version":3,"file":"ai-model/service-caller/utils.mjs","sources":["../../../../src/ai-model/service-caller/utils.ts"],"sourcesContent":["import type { RawAssistantOutput } from '@/types';\nimport type { AIUsageInfo } from '@/types';\nimport type OpenAI from 'openai';\nimport type { ModelRuntime } from '../models';\n\n// Error class that preserves usage and rawResponse when AI call parsing fails\nexport class AIResponseParseError extends Error {\n  usage?: AIUsageInfo;\n  /**\n   * Adapter-extracted content used by Midscene for parsing. This is not the\n   * full provider response or choices[0].message.\n   */\n  rawResponse: string;\n  rawAssistantOutput?: RawAssistantOutput;\n  reasoningContent?: string;\n\n  constructor(\n    message: string,\n    rawResponse: string,\n    usage?: AIUsageInfo,\n    rawAssistantOutput?: RawAssistantOutput,\n    reasoningContent?: string,\n  ) {\n    super(message);\n    this.name = 'AIResponseParseError';\n    this.rawResponse = rawResponse;\n    this.usage = usage;\n    this.rawAssistantOutput = rawAssistantOutput;\n    this.reasoningContent = reasoningContent;\n  }\n}\n\n/**\n * Internal field name stamped onto every AIUsageInfo shaped by callAI().\n * Used for cross-path dedup when the provider does not return a request_id.\n */\nexport const INTERNAL_CALL_ID_FIELD = '_midscene_call_id';\n\nlet internalCallIdCounter = 0;\nexport function nextInternalCallId(): string {\n  internalCallIdCounter += 1;\n  return `call_${internalCallIdCounter}`;\n}\n\nexport function stringifyForDebug(value: unknown): string {\n  try {\n    return JSON.stringify(value);\n  } catch (_error) {\n    return String(value);\n  }\n}\n\nfunction getErrorMessage(error: unknown): string {\n  return error instanceof Error ? error.message : String(error);\n}\n\nexport function toError(error: unknown): Error {\n  return error instanceof Error ? error : new Error(String(error));\n}\n\nexport function normalizeRetryCount(retryCount: unknown): number {\n  if (typeof retryCount !== 'number' || !Number.isFinite(retryCount)) {\n    return 1;\n  }\n\n  return Math.max(0, Math.floor(retryCount));\n}\n\nexport function appendAIRequestFailureSummary<T extends Error>(\n  error: T,\n  attemptErrors: Array<{ attempt: number; error: unknown }>,\n  maxAttempts: number,\n): T {\n  const failedAttempts = attemptErrors.length;\n  const retries = Math.max(0, failedAttempts - 1);\n  const retryLabel = retries === 1 ? 'retry' : 'retries';\n  const originalMessage = error.message;\n  const previousAttemptErrors = attemptErrors.slice(0, -1);\n\n  error.message = `AI model request failed after ${retries} ${retryLabel} (${failedAttempts}/${maxAttempts} attempts). Last error: ${originalMessage}`;\n\n  if (previousAttemptErrors.length === 0) {\n    return error;\n  }\n\n  const details = previousAttemptErrors\n    .map(\n      ({ attempt, error }) => `Attempt ${attempt}: ${getErrorMessage(error)}`,\n    )\n    .join('\\n');\n\n  error.message = `${error.message}\\nPrevious AI call attempt errors:\\n${details}`;\n  return error;\n}\n\nexport const buildUsageInfo = ({\n  apiType,\n  usageData,\n  requestId,\n  timeCost,\n  totalTimeCost,\n  retryCount,\n  modelName,\n  modelDescription,\n  responseModelName,\n  slot,\n  internalCallId,\n}: {\n  apiType: NonNullable<AIUsageInfo['api_type']>;\n  usageData?: OpenAI.CompletionUsage;\n  requestId?: string | null;\n  timeCost?: number;\n  totalTimeCost: number;\n  retryCount: number;\n  modelName: string;\n  modelDescription: string;\n  responseModelName?: string;\n  slot: ModelRuntime['config']['slot'];\n  internalCallId: string;\n}): AIUsageInfo | undefined => {\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    ...usageData,\n    api_type: apiType,\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    total_time_cost: totalTimeCost,\n    retry_count: retryCount,\n    model_name: modelName,\n    model_description: modelDescription,\n    response_model_name: responseModelName,\n    slot,\n    // Left undefined at the raw call layer. The agent's onUsage callback\n    // fills it from modelConfig.slot for metrics collection, and task\n    // layers use withUsageIntent() to stamp a more specific semantic\n    // intent (e.g. 'planning', 'insight') when attaching usage to tasks.\n    intent: undefined,\n    request_id: requestId ?? undefined,\n    // Internal stable ID for cross-path dedup when request_id is absent.\n    [INTERNAL_CALL_ID_FIELD]: internalCallId,\n  } satisfies AIUsageInfo;\n};\n\nexport const hasUsableText = (\n  value: string | null | undefined,\n): value is string => typeof value === 'string' && value.trim().length > 0;\n"],"names":["AIResponseParseError","Error","message","rawResponse","usage","rawAssistantOutput","reasoningContent","INTERNAL_CALL_ID_FIELD","internalCallIdCounter","nextInternalCallId","stringifyForDebug","value","JSON","_error","String","getErrorMessage","error","toError","normalizeRetryCount","retryCount","Number","Math","appendAIRequestFailureSummary","attemptErrors","maxAttempts","failedAttempts","retries","retryLabel","originalMessage","previousAttemptErrors","details","attempt","buildUsageInfo","apiType","usageData","requestId","timeCost","totalTimeCost","modelName","modelDescription","responseModelName","slot","internalCallId","cachedInputTokens","undefined","hasUsableText"],"mappings":";;;;;;;;;;AAMO,MAAMA,6BAA6BC;IAUxC,YACEC,OAAe,EACfC,WAAmB,EACnBC,KAAmB,EACnBC,kBAAuC,EACvCC,gBAAyB,CACzB;QACA,KAAK,CAACJ,UAhBR,yCAKA,+CACA,sDACA;QAUE,IAAI,CAAC,IAAI,GAAG;QACZ,IAAI,CAAC,WAAW,GAAGC;QACnB,IAAI,CAAC,KAAK,GAAGC;QACb,IAAI,CAAC,kBAAkB,GAAGC;QAC1B,IAAI,CAAC,gBAAgB,GAAGC;IAC1B;AACF;AAMO,MAAMC,yBAAyB;AAEtC,IAAIC,wBAAwB;AACrB,SAASC;IACdD,yBAAyB;IACzB,OAAO,CAAC,KAAK,EAAEA,uBAAuB;AACxC;AAEO,SAASE,kBAAkBC,KAAc;IAC9C,IAAI;QACF,OAAOC,KAAK,SAAS,CAACD;IACxB,EAAE,OAAOE,QAAQ;QACf,OAAOC,OAAOH;IAChB;AACF;AAEA,SAASI,gBAAgBC,KAAc;IACrC,OAAOA,iBAAiBf,QAAQe,MAAM,OAAO,GAAGF,OAAOE;AACzD;AAEO,SAASC,QAAQD,KAAc;IACpC,OAAOA,iBAAiBf,QAAQe,QAAQ,IAAIf,MAAMa,OAAOE;AAC3D;AAEO,SAASE,oBAAoBC,UAAmB;IACrD,IAAI,AAAsB,YAAtB,OAAOA,cAA2B,CAACC,OAAO,QAAQ,CAACD,aACrD,OAAO;IAGT,OAAOE,KAAK,GAAG,CAAC,GAAGA,KAAK,KAAK,CAACF;AAChC;AAEO,SAASG,8BACdN,KAAQ,EACRO,aAAyD,EACzDC,WAAmB;IAEnB,MAAMC,iBAAiBF,cAAc,MAAM;IAC3C,MAAMG,UAAUL,KAAK,GAAG,CAAC,GAAGI,iBAAiB;IAC7C,MAAME,aAAaD,AAAY,MAAZA,UAAgB,UAAU;IAC7C,MAAME,kBAAkBZ,MAAM,OAAO;IACrC,MAAMa,wBAAwBN,cAAc,KAAK,CAAC,GAAG;IAErDP,MAAM,OAAO,GAAG,CAAC,8BAA8B,EAAEU,QAAQ,CAAC,EAAEC,WAAW,EAAE,EAAEF,eAAe,CAAC,EAAED,YAAY,wBAAwB,EAAEI,iBAAiB;IAEpJ,IAAIC,AAAiC,MAAjCA,sBAAsB,MAAM,EAC9B,OAAOb;IAGT,MAAMc,UAAUD,sBACb,GAAG,CACF,CAAC,EAAEE,OAAO,EAAEf,KAAK,EAAE,GAAK,CAAC,QAAQ,EAAEe,QAAQ,EAAE,EAAEhB,gBAAgBC,QAAQ,EAExE,IAAI,CAAC;IAERA,MAAM,OAAO,GAAG,GAAGA,MAAM,OAAO,CAAC,oCAAoC,EAAEc,SAAS;IAChF,OAAOd;AACT;AAEO,MAAMgB,iBAAiB,CAAC,EAC7BC,OAAO,EACPC,SAAS,EACTC,SAAS,EACTC,QAAQ,EACRC,aAAa,EACblB,UAAU,EACVmB,SAAS,EACTC,gBAAgB,EAChBC,iBAAiB,EACjBC,IAAI,EACJC,cAAc,EAaf;IACC,IAAI,CAACR,WAAW;IAEhB,MAAMS,oBACJT,WACC,uBAAuB;IAE1B,OAAO;QACL,GAAGA,SAAS;QACZ,UAAUD;QACV,eAAeC,UAAU,aAAa,IAAI;QAC1C,mBAAmBA,UAAU,iBAAiB,IAAI;QAClD,cAAcA,UAAU,YAAY,IAAI;QACxC,cAAcS,qBAAqB;QACnC,WAAWP,YAAY;QACvB,iBAAiBC;QACjB,aAAalB;QACb,YAAYmB;QACZ,mBAAmBC;QACnB,qBAAqBC;QACrBC;QAKA,QAAQG;QACR,YAAYT,aAAaS;QAEzB,CAACrC,uBAAuB,EAAEmC;IAC5B;AACF;AAEO,MAAMG,gBAAgB,CAC3BlC,QACoB,AAAiB,YAAjB,OAAOA,SAAsBA,MAAM,IAAI,GAAG,MAAM,GAAG"}
+{"version":3,"file":"ai-model/service-caller/utils.mjs","sources":["../../../../src/ai-model/service-caller/utils.ts"],"sourcesContent":["import type { RawAssistantOutput } from '@/types';\nimport type { AIUsageInfo } from '@/types';\nimport type OpenAI from 'openai';\nimport type { ModelRuntime } from '../models';\n\n// Error class that preserves usage and rawResponse when AI call parsing fails\nexport class AIResponseParseError extends Error {\n  usage?: AIUsageInfo;\n  /**\n   * Adapter-extracted content used by Midscene for parsing. This is not the\n   * full provider response or choices[0].message.\n   */\n  rawResponse: string;\n  rawAssistantOutput?: RawAssistantOutput;\n  reasoningContent?: string;\n\n  constructor(\n    message: string,\n    rawResponse: string,\n    usage?: AIUsageInfo,\n    rawAssistantOutput?: RawAssistantOutput,\n    reasoningContent?: string,\n  ) {\n    super(message);\n    this.name = 'AIResponseParseError';\n    this.rawResponse = rawResponse;\n    this.usage = usage;\n    this.rawAssistantOutput = rawAssistantOutput;\n    this.reasoningContent = reasoningContent;\n  }\n}\n\n/**\n * Internal field name stamped onto every AIUsageInfo shaped by callAI().\n * Used for cross-path dedup when the provider does not return a request_id.\n */\nexport const INTERNAL_CALL_ID_FIELD = '_midscene_call_id';\n\nlet internalCallIdCounter = 0;\nexport function nextInternalCallId(): string {\n  internalCallIdCounter += 1;\n  return `call_${internalCallIdCounter}`;\n}\n\nexport function stringifyForDebug(value: unknown): string {\n  try {\n    return JSON.stringify(value);\n  } catch (_error) {\n    return String(value);\n  }\n}\n\nfunction getErrorMessage(error: unknown): string {\n  return error instanceof Error ? error.message : String(error);\n}\n\nexport function toError(error: unknown): Error {\n  return error instanceof Error ? error : new Error(String(error));\n}\n\nexport function normalizeRetryCount(retryCount: unknown): number {\n  if (typeof retryCount !== 'number' || !Number.isFinite(retryCount)) {\n    return 1;\n  }\n\n  return Math.max(0, Math.floor(retryCount));\n}\n\nexport function appendAIRequestFailureSummary<T extends Error>(\n  error: T,\n  attemptErrors: Array<{ attempt: number; error: unknown }>,\n  maxAttempts: number,\n): T {\n  const failedAttempts = attemptErrors.length;\n  const retries = Math.max(0, failedAttempts - 1);\n  const retryLabel = retries === 1 ? 'retry' : 'retries';\n  const originalMessage = error.message;\n  const previousAttemptErrors = attemptErrors.slice(0, -1);\n\n  error.message = `AI model request failed after ${retries} ${retryLabel} (${failedAttempts}/${maxAttempts} attempts). Last error: ${originalMessage}`;\n\n  if (previousAttemptErrors.length === 0) {\n    return error;\n  }\n\n  const details = previousAttemptErrors\n    .map(\n      ({ attempt, error }) => `Attempt ${attempt}: ${getErrorMessage(error)}`,\n    )\n    .join('\\n');\n\n  error.message = `${error.message}\\nPrevious AI call attempt errors:\\n${details}`;\n  return error;\n}\n\nexport const buildUsageInfo = ({\n  apiProtocol,\n  stream,\n  usageData,\n  requestId,\n  timeCost,\n  totalTimeCost,\n  retryCount,\n  modelName,\n  modelDescription,\n  responseModelName,\n  slot,\n  internalCallId,\n}: {\n  apiProtocol: NonNullable<AIUsageInfo['api_protocol']>;\n  stream: boolean;\n  usageData?: OpenAI.CompletionUsage;\n  requestId?: string | null;\n  timeCost?: number;\n  totalTimeCost: number;\n  retryCount: number;\n  modelName: string;\n  modelDescription: string;\n  responseModelName?: string;\n  slot: ModelRuntime['config']['slot'];\n  internalCallId: string;\n}): AIUsageInfo | undefined => {\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    ...usageData,\n    api_protocol: apiProtocol,\n    stream,\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    total_time_cost: totalTimeCost,\n    retry_count: retryCount,\n    model_name: modelName,\n    model_description: modelDescription,\n    response_model_name: responseModelName,\n    slot,\n    // Left undefined at the raw call layer. The agent's onUsage callback\n    // fills it from modelConfig.slot for metrics collection, and task\n    // layers use withUsageIntent() to stamp a more specific semantic\n    // intent (e.g. 'planning', 'insight') when attaching usage to tasks.\n    intent: undefined,\n    request_id: requestId ?? undefined,\n    // Internal stable ID for cross-path dedup when request_id is absent.\n    [INTERNAL_CALL_ID_FIELD]: internalCallId,\n  } satisfies AIUsageInfo;\n};\n\nexport const hasUsableText = (\n  value: string | null | undefined,\n): value is string => typeof value === 'string' && value.trim().length > 0;\n"],"names":["AIResponseParseError","Error","message","rawResponse","usage","rawAssistantOutput","reasoningContent","INTERNAL_CALL_ID_FIELD","internalCallIdCounter","nextInternalCallId","stringifyForDebug","value","JSON","_error","String","getErrorMessage","error","toError","normalizeRetryCount","retryCount","Number","Math","appendAIRequestFailureSummary","attemptErrors","maxAttempts","failedAttempts","retries","retryLabel","originalMessage","previousAttemptErrors","details","attempt","buildUsageInfo","apiProtocol","stream","usageData","requestId","timeCost","totalTimeCost","modelName","modelDescription","responseModelName","slot","internalCallId","cachedInputTokens","undefined","hasUsableText"],"mappings":";;;;;;;;;;AAMO,MAAMA,6BAA6BC;IAUxC,YACEC,OAAe,EACfC,WAAmB,EACnBC,KAAmB,EACnBC,kBAAuC,EACvCC,gBAAyB,CACzB;QACA,KAAK,CAACJ,UAhBR,yCAKA,+CACA,sDACA;QAUE,IAAI,CAAC,IAAI,GAAG;QACZ,IAAI,CAAC,WAAW,GAAGC;QACnB,IAAI,CAAC,KAAK,GAAGC;QACb,IAAI,CAAC,kBAAkB,GAAGC;QAC1B,IAAI,CAAC,gBAAgB,GAAGC;IAC1B;AACF;AAMO,MAAMC,yBAAyB;AAEtC,IAAIC,wBAAwB;AACrB,SAASC;IACdD,yBAAyB;IACzB,OAAO,CAAC,KAAK,EAAEA,uBAAuB;AACxC;AAEO,SAASE,kBAAkBC,KAAc;IAC9C,IAAI;QACF,OAAOC,KAAK,SAAS,CAACD;IACxB,EAAE,OAAOE,QAAQ;QACf,OAAOC,OAAOH;IAChB;AACF;AAEA,SAASI,gBAAgBC,KAAc;IACrC,OAAOA,iBAAiBf,QAAQe,MAAM,OAAO,GAAGF,OAAOE;AACzD;AAEO,SAASC,QAAQD,KAAc;IACpC,OAAOA,iBAAiBf,QAAQe,QAAQ,IAAIf,MAAMa,OAAOE;AAC3D;AAEO,SAASE,oBAAoBC,UAAmB;IACrD,IAAI,AAAsB,YAAtB,OAAOA,cAA2B,CAACC,OAAO,QAAQ,CAACD,aACrD,OAAO;IAGT,OAAOE,KAAK,GAAG,CAAC,GAAGA,KAAK,KAAK,CAACF;AAChC;AAEO,SAASG,8BACdN,KAAQ,EACRO,aAAyD,EACzDC,WAAmB;IAEnB,MAAMC,iBAAiBF,cAAc,MAAM;IAC3C,MAAMG,UAAUL,KAAK,GAAG,CAAC,GAAGI,iBAAiB;IAC7C,MAAME,aAAaD,AAAY,MAAZA,UAAgB,UAAU;IAC7C,MAAME,kBAAkBZ,MAAM,OAAO;IACrC,MAAMa,wBAAwBN,cAAc,KAAK,CAAC,GAAG;IAErDP,MAAM,OAAO,GAAG,CAAC,8BAA8B,EAAEU,QAAQ,CAAC,EAAEC,WAAW,EAAE,EAAEF,eAAe,CAAC,EAAED,YAAY,wBAAwB,EAAEI,iBAAiB;IAEpJ,IAAIC,AAAiC,MAAjCA,sBAAsB,MAAM,EAC9B,OAAOb;IAGT,MAAMc,UAAUD,sBACb,GAAG,CACF,CAAC,EAAEE,OAAO,EAAEf,KAAK,EAAE,GAAK,CAAC,QAAQ,EAAEe,QAAQ,EAAE,EAAEhB,gBAAgBC,QAAQ,EAExE,IAAI,CAAC;IAERA,MAAM,OAAO,GAAG,GAAGA,MAAM,OAAO,CAAC,oCAAoC,EAAEc,SAAS;IAChF,OAAOd;AACT;AAEO,MAAMgB,iBAAiB,CAAC,EAC7BC,WAAW,EACXC,MAAM,EACNC,SAAS,EACTC,SAAS,EACTC,QAAQ,EACRC,aAAa,EACbnB,UAAU,EACVoB,SAAS,EACTC,gBAAgB,EAChBC,iBAAiB,EACjBC,IAAI,EACJC,cAAc,EAcf;IACC,IAAI,CAACR,WAAW;IAEhB,MAAMS,oBACJT,WACC,uBAAuB;IAE1B,OAAO;QACL,GAAGA,SAAS;QACZ,cAAcF;QACdC;QACA,eAAeC,UAAU,aAAa,IAAI;QAC1C,mBAAmBA,UAAU,iBAAiB,IAAI;QAClD,cAAcA,UAAU,YAAY,IAAI;QACxC,cAAcS,qBAAqB;QACnC,WAAWP,YAAY;QACvB,iBAAiBC;QACjB,aAAanB;QACb,YAAYoB;QACZ,mBAAmBC;QACnB,qBAAqBC;QACrBC;QAKA,QAAQG;QACR,YAAYT,aAAaS;QAEzB,CAACtC,uBAAuB,EAAEoC;IAC5B;AACF;AAEO,MAAMG,gBAAgB,CAC3BnC,QACoB,AAAiB,YAAjB,OAAOA,SAAsBA,MAAM,IAAI,GAAG,MAAM,GAAG"}

package/dist/es/device/index.mjs

@@ -23,6 +23,7 @@ function resolveSwipeInputPrimitive(input) {
     };
 }
 class AbstractInterface {
+    async defaultActionWait(_actionName, _param) {}
     constructor(){
         _define_property(this, "mjpegStreamUrl", void 0);
         _define_property(this, "inputPrimitives", void 0);
@@ -167,7 +168,7 @@ const defineActionInput = (keyboard)=>defineAction({
                 prompt: 'the email input field'
             }
         },
-        call: async (param)=>{
+        call: async (param, context)=>{
             if ('append' === param.mode) param.mode = 'typeOnly';
             if ('clear' === param.mode) return void await keyboard.clearInput(param.locate);
             if (!param || !param.value) return;
@@ -176,7 +177,10 @@ const defineActionInput = (keyboard)=>defineAction({
                 replace: 'typeOnly' !== param.mode,
                 autoDismissKeyboard: param.autoDismissKeyboard,
                 keyboardTypeDelay: param.keyboardTypeDelay,
-                inputStrategy: param.inputStrategy
+                inputStrategy: param.inputStrategy,
+                ...context?.skipDefaultWait ? {
+                    skipDefaultWait: true
+                } : {}
             });
         }
     });
@@ -227,8 +231,11 @@ const defineActionScroll = (scroll)=>defineAction({
                 prompt: 'the center of the product list area'
             }
         },
-        call: async (param)=>{
-            await scroll(param);
+        call: async (param, context)=>{
+            if (context?.skipDefaultWait) await scroll(param, {
+                skipDefaultWait: true
+            });
+            else await scroll(param);
         }
     });
 const actionDragAndDropParamSchema = z.object({

package/dist/es/device/index.mjs.map

@@ -1 +1 @@
-{"version":3,"file":"device/index.mjs","sources":["../../../src/device/index.ts"],"sourcesContent":["import type { ModelRuntime } from '@/ai-model/models';\nimport { getMidsceneLocationSchema } from '@/common';\nimport type {\n  ActionScrollParam,\n  DeviceAction,\n  ExecutorContext,\n  LocateResultElement,\n} from '@/types';\nimport type { ElementNode } from '@midscene/shared/extractor';\nimport { getDebug } from '@midscene/shared/logger';\nimport { _keyDefinitions } from '@midscene/shared/us-keyboard-layout';\nimport { z } from 'zod';\nimport type {\n  ElementCacheFeature,\n  Rect,\n  Size,\n  UIContext,\n  UITreeSnapshot,\n} from '../types';\nimport { type InputStrategy, inputStrategies } from './input-strategy';\n\nexport interface FileChooserHandler {\n  accept(files: string[]): Promise<void>;\n}\n\nexport interface FileChooserRegistration {\n  dispose: () => void;\n  getError: () => Error | undefined | Promise<Error | undefined>;\n}\n\nexport interface MjpegStreamFrame {\n  /** Raw base64-encoded image bytes WITHOUT a `data:image/...;base64,` prefix. */\n  data: string;\n  contentType?: string;\n}\n\nexport interface MjpegStreamHandle {\n  stop(): void | Promise<void>;\n}\n\nexport interface MjpegStreamOptions {\n  signal?: AbortSignal;\n  onFrame(frame: MjpegStreamFrame): void;\n  onError?(error: unknown): void;\n}\n\n/**\n * A cheap, not-yet-decoded handle to one screen frame from a\n * {@link DeviceFrameSource}. `ref` is platform-specific (a raw H.264 keyframe\n * buffer on Android, an already-encoded JPEG data URL on iOS/web) and must not\n * be interpreted by callers — pass it back to `decode()` to materialize.\n */\nexport interface DeviceFrameRef {\n  ref: unknown;\n  capturedAt: number;\n}\n\n/**\n * A continuous screen-frame source opened via\n * {@link AbstractInterface.openFrameSource}. Designed for deferred decoding:\n * grabbing `latest()` is near-zero cost, so observers can sample at a steady\n * cadence and pay any decode cost only once, for the frames they keep.\n */\nexport interface DeviceFrameSource {\n  /** Latest frame handle, near-zero cost. Null until the first frame arrives. */\n  latest(): DeviceFrameRef | null;\n  /**\n   * Materialize frame handles into `data:image/...;base64,` URLs, preserving\n   * order. Possibly expensive (e.g. one ffmpeg run per unique frame on\n   * Android) — call once with the sampled handles, never per tick.\n   */\n  decode(refs: DeviceFrameRef[]): Promise<string[]>;\n  /** Release the source (stop streams/subscriptions it started). */\n  stop(): Promise<void> | void;\n}\n\n/** A point in device-pixel coordinates on the screen. */\nexport interface PointerPoint {\n  x: number;\n  y: number;\n}\n\n/** Low-level swipe gesture implemented by touch or pointer-backed devices. */\nexport type SwipeInputPrimitive = (\n  start: PointerPoint,\n  end: PointerPoint,\n  opts?: { duration?: number; repeat?: number },\n) => Promise<void>;\n\n/** The physical input source used to execute and describe a swipe. */\nexport type SwipeInputMode = 'touch' | 'mouse';\n\n/** A resolved swipe capability and the input semantics exposed to the model. */\nexport interface ResolvedSwipeInputPrimitive {\n  swipe: SwipeInputPrimitive;\n  inputMode: SwipeInputMode;\n}\n\nexport interface PointerInputPrimitives {\n  tap(p: PointerPoint, opts?: { duration?: number }): Promise<void>;\n  doubleClick?(p: PointerPoint): Promise<void>;\n  rightClick?(p: PointerPoint): Promise<void>;\n  hover?(p: PointerPoint): Promise<void>;\n  longPress?(p: PointerPoint, opts?: { duration?: number }): Promise<void>;\n  dragAndDrop?(from: PointerPoint, to: PointerPoint): Promise<void>;\n  swipe?: SwipeInputPrimitive;\n}\n\nexport interface TouchInputPrimitives {\n  swipe: SwipeInputPrimitive;\n  pinch?(\n    center: PointerPoint,\n    opts: { startDistance: number; endDistance: number; duration: number },\n  ): Promise<void>;\n}\n\nexport interface KeyboardInputPrimitives {\n  keyboardPress(keyName: string, opts?: { target?: unknown }): Promise<void>;\n  cursorMove?(direction: 'left' | 'right', times?: number): Promise<void>;\n  typeText(\n    value: string,\n    opts?: {\n      autoDismissKeyboard?: boolean;\n      keyboardDismissStrategy?: 'esc-first' | 'back-first';\n      keyboardTypeDelay?: number;\n      inputStrategy?: InputStrategy;\n      target?: unknown;\n      replace?: boolean;\n      focusOnly?: boolean;\n    },\n  ): Promise<void>;\n  clearInput(target?: unknown): Promise<void>;\n}\n\nexport interface ScrollInputPrimitives {\n  scroll(param: ActionScrollParam): Promise<void>;\n}\n\nexport interface SystemInputPrimitives {\n  backButton?(): Promise<void>;\n  homeButton?(): Promise<void>;\n  recentAppsButton?(): Promise<void>;\n}\n\nexport interface InputPrimitives {\n  pointer?: PointerInputPrimitives;\n  keyboard?: KeyboardInputPrimitives;\n  touch?: TouchInputPrimitives;\n  scroll?: ScrollInputPrimitives;\n  system?: SystemInputPrimitives;\n}\n\n/**\n * Resolve the canonical swipe capability for an input collection.\n * Touch remains preferred for devices that expose both capabilities; desktop\n * devices can fall back to a pointer-backed primary-button gesture.\n */\nexport function resolveSwipeInputPrimitive(\n  input: Pick<InputPrimitives, 'pointer' | 'touch'>,\n): ResolvedSwipeInputPrimitive | undefined {\n  if (input.touch?.swipe) {\n    return { swipe: input.touch.swipe, inputMode: 'touch' };\n  }\n  if (input.pointer?.swipe) {\n    return { swipe: input.pointer.swipe, inputMode: 'mouse' };\n  }\n  return undefined;\n}\n\nexport interface MobileInputPrimitives extends InputPrimitives {\n  pointer: PointerInputPrimitives & {\n    doubleClick(p: PointerPoint): Promise<void>;\n    longPress(p: PointerPoint, opts?: { duration?: number }): Promise<void>;\n    dragAndDrop(from: PointerPoint, to: PointerPoint): Promise<void>;\n  };\n  keyboard: KeyboardInputPrimitives;\n  touch: TouchInputPrimitives;\n}\n\nexport interface BrowserInputPrimitives extends InputPrimitives {\n  pointer: PointerInputPrimitives & {\n    doubleClick(p: PointerPoint): Promise<void>;\n    rightClick(p: PointerPoint): Promise<void>;\n    hover(p: PointerPoint): Promise<void>;\n    dragAndDrop(from: PointerPoint, to: PointerPoint): Promise<void>;\n    longPress(p: PointerPoint, opts?: { duration?: number }): Promise<void>;\n  };\n  keyboard: KeyboardInputPrimitives;\n  scroll: ScrollInputPrimitives;\n  touch: TouchInputPrimitives;\n}\n\nexport interface ComputerInputPrimitives extends InputPrimitives {\n  pointer: PointerInputPrimitives & {\n    doubleClick(p: PointerPoint): Promise<void>;\n    rightClick(p: PointerPoint): Promise<void>;\n    hover(p: PointerPoint): Promise<void>;\n    dragAndDrop(from: PointerPoint, to: PointerPoint): Promise<void>;\n  };\n  keyboard: KeyboardInputPrimitives;\n  scroll: ScrollInputPrimitives;\n}\n\nexport abstract class AbstractInterface {\n  abstract interfaceType: string;\n\n  abstract screenshotBase64(): Promise<string>;\n  abstract size(): Promise<Size>;\n  abstract actionSpace(): DeviceAction[];\n\n  abstract cacheFeatureForPoint?(\n    center: [number, number],\n    options?: {\n      targetDescription?: string;\n      modelRuntime?: ModelRuntime;\n    },\n  ): Promise<ElementCacheFeature>;\n  abstract rectMatchesCacheFeature?(\n    feature: ElementCacheFeature,\n  ): Promise<Rect>;\n\n  abstract getUITree?(): Promise<UITreeSnapshot>;\n\n  abstract destroy?(): Promise<void>;\n\n  abstract describe?(): string;\n  abstract beforeInvokeAction?(actionName: string, param: any): Promise<void>;\n  abstract afterInvokeAction?(actionName: string, param: any): Promise<void>;\n\n  // for web only\n  registerFileChooserListener?(\n    handler: (chooser: FileChooserHandler) => Promise<void>,\n  ): Promise<FileChooserRegistration>;\n\n  // @deprecated do NOT extend this method\n  abstract getElementsNodeTree?: () => Promise<ElementNode>;\n\n  // @deprecated do NOT extend this method\n  abstract url?: () => string | Promise<string>;\n\n  // @deprecated do NOT extend this method\n  abstract evaluateJavaScript?<T = any>(script: string): Promise<T>;\n\n  /**\n   * Get the current device-local time as a formatted string.\n   * Prefer this for user-visible time because timestamps alone do not preserve\n   * the target device's timezone when formatted on the host machine.\n   */\n  getDeviceLocalTimeString?(format?: string): Promise<string>;\n\n  /** URL of native MJPEG stream for real-time screen preview (e.g. WDA MJPEG server) */\n  mjpegStreamUrl?: string;\n\n  /**\n   * Optional continuous frame source for UI observation (`startObserving`).\n   * Devices that maintain a continuous frame stream — scrcpy on Android, WDA\n   * MJPEG on iOS, CDP screencast on web — implement this so an observer can\n   * sample the screen far faster than repeated `screenshotBase64()` calls,\n   * catching short-lived UI (toasts, carousels, transitions).\n   *\n   * The contract enables DEFERRED decoding: `latest()` returns a cheap opaque\n   * handle (e.g. a raw H.264 keyframe on Android) with no per-frame decode\n   * cost, and `decode()` materializes only the handles that were actually\n   * sampled — once, at the end of the observation window.\n   */\n  openFrameSource?(): Promise<DeviceFrameSource | undefined>;\n\n  /**\n   * Optional in-process MJPEG frame producer. Implementations can push raw\n   * base64 frames here when there is no standalone native MJPEG URL, e.g.\n   * Chromium CDP Page.startScreencast for web previews.\n   */\n  startMjpegStream?(\n    options: MjpegStreamOptions,\n  ): MjpegStreamHandle | undefined | Promise<MjpegStreamHandle | undefined>;\n\n  /**\n   * Optional hook used after a UI action to push a fresh frame on the active\n   * MJPEG stream. Set `force` after navigation to replace a transient loading\n   * frame even when the screencast has already emitted one. Implementations\n   * should be a no-op when no stream is active.\n   */\n  flushPendingVisualUpdate?(force?: boolean): Promise<void>;\n\n  /**\n   * Optional non-blocking variant of `flushPendingVisualUpdate`. Keyboard-\n   * heavy preview interactions can schedule a coalesced refresh here without\n   * stalling the input hot path; `force` preserves a requested navigation\n   * refresh while work is already queued.\n   */\n  schedulePendingVisualUpdate?(force?: boolean): void;\n\n  /**\n   * Optional navigation state probe for browser-like interfaces, used to drive\n   * loading indicators in playground UIs. Returning `undefined` means the\n   * interface does not expose this concept.\n   */\n  navigationState?(): Promise<{ isLoading: boolean }>;\n\n  /**\n   * Low-level device input surface. Platform implementations expose transport\n   * primitives here; higher-level AI actions and manual pointer dispatch should\n   * adapt to this instead of duplicating platform gesture logic.\n   */\n  inputPrimitives?: InputPrimitives;\n}\n\n// Generic function to define actions with proper type inference\n// TRuntime allows specifying a different type for the runtime parameter (after location resolution)\n// TReturn allows specifying the return type of the action\nexport const defineAction = <\n  TSchema extends z.ZodType | undefined = undefined,\n  TRuntime = TSchema extends z.ZodType ? z.infer<TSchema> : undefined,\n  TReturn = any,\n>(\n  config: {\n    name: string;\n    description: string;\n    interfaceAlias?: string;\n    paramSchema?: TSchema;\n    call: (\n      param: TRuntime,\n      context?: ExecutorContext,\n    ) => Promise<TReturn> | TReturn;\n  } & Partial<\n    Omit<\n      DeviceAction<TRuntime, TReturn>,\n      'name' | 'description' | 'interfaceAlias' | 'paramSchema' | 'call'\n    >\n  >,\n): DeviceAction<TRuntime, TReturn> => {\n  return config as any; // Type assertion needed because schema validation type differs from runtime type\n};\n\nfunction pointFromLocate(\n  locate: LocateResultElement | undefined,\n  missingMessage: string,\n): PointerPoint {\n  if (!locate) {\n    throw new Error(missingMessage);\n  }\n  return { x: locate.center[0], y: locate.center[1] };\n}\n\nfunction defineLocatedPointAction<\n  TSchema extends z.ZodType,\n  TParam extends { locate: LocateResultElement },\n>(config: {\n  name: string;\n  description: string;\n  interfaceAlias?: string;\n  paramSchema: TSchema;\n  sample: DeviceAction<TParam>['sample'];\n  missingLocateMessage: string;\n  call: (point: PointerPoint, param: TParam) => Promise<void>;\n}): DeviceAction<TParam> {\n  return defineAction<TSchema, TParam>({\n    name: config.name,\n    description: config.description,\n    interfaceAlias: config.interfaceAlias,\n    paramSchema: config.paramSchema,\n    sample: config.sample,\n    call: async (param) => {\n      await config.call(\n        pointFromLocate(param.locate, config.missingLocateMessage),\n        param,\n      );\n    },\n  });\n}\n\n// Tap\nexport const actionTapParamSchema = z.object({\n  locate: getMidsceneLocationSchema().describe('The element to be tapped'),\n});\nexport type ActionTapParam = {\n  locate: LocateResultElement;\n};\n\nexport const defineActionTap = (\n  tap: PointerInputPrimitives['tap'],\n): DeviceAction<ActionTapParam> => {\n  return defineLocatedPointAction<typeof actionTapParamSchema, ActionTapParam>({\n    name: 'Tap',\n    description: 'Tap the element',\n    interfaceAlias: 'aiTap',\n    paramSchema: actionTapParamSchema,\n    sample: {\n      locate: { prompt: 'the \"Submit\" button' },\n    },\n    missingLocateMessage: 'Element not found, cannot tap',\n    call: async (point) => {\n      await tap(point);\n    },\n  });\n};\n\nexport const registerFileChooserAcceptParamSchema = z.object({\n  files: z\n    .union([z.string(), z.array(z.string())])\n    .describe(\n      \"File path(s) within the current aiAct call's fileChooserAllowedDir to use whenever a later action triggers a file chooser. fileChooserAllowedDir must be provided for this action. This setting replaces any previously registered file path(s).\",\n    ),\n});\nexport type RegisterFileChooserAcceptParam = {\n  files: string | string[];\n};\n\nexport const defineActionRegisterFileChooserAccept = (\n  register: (files: string | string[]) => Promise<void>,\n): DeviceAction<RegisterFileChooserAcceptParam> => {\n  return defineAction<\n    typeof registerFileChooserAcceptParamSchema,\n    RegisterFileChooserAcceptParam\n  >({\n    name: 'RegisterFileChooserAccept',\n    description:\n      'Configure files for file chooser dialogs triggered by later actions in this aiAct',\n    interfaceAlias: 'registerFileChooserAccept',\n    paramSchema: registerFileChooserAcceptParamSchema,\n    sample: {\n      files: ['fixtures/document.pdf'],\n    },\n    call: async (param) => {\n      await register(param.files);\n    },\n  });\n};\n\n// RightClick\nexport const actionRightClickParamSchema = z.object({\n  locate: getMidsceneLocationSchema().describe(\n    'The element to be right clicked',\n  ),\n});\nexport type ActionRightClickParam = {\n  locate: LocateResultElement;\n};\n\nexport const defineActionRightClick = (\n  rightClick: NonNullable<PointerInputPrimitives['rightClick']>,\n): DeviceAction<ActionRightClickParam> => {\n  return defineLocatedPointAction<\n    typeof actionRightClickParamSchema,\n    ActionRightClickParam\n  >({\n    name: 'RightClick',\n    description: 'Right click the element',\n    interfaceAlias: 'aiRightClick',\n    paramSchema: actionRightClickParamSchema,\n    sample: {\n      locate: { prompt: 'the file icon on the desktop' },\n    },\n    missingLocateMessage: 'Element not found, cannot right click',\n    call: async (point) => {\n      await rightClick(point);\n    },\n  });\n};\n\n// DoubleClick\nexport const actionDoubleClickParamSchema = z.object({\n  locate: getMidsceneLocationSchema().describe(\n    'The element to be double clicked',\n  ),\n});\nexport type ActionDoubleClickParam = {\n  locate: LocateResultElement;\n};\n\nexport const defineActionDoubleClick = (\n  doubleClick: NonNullable<PointerInputPrimitives['doubleClick']>,\n): DeviceAction<ActionDoubleClickParam> => {\n  return defineLocatedPointAction<\n    typeof actionDoubleClickParamSchema,\n    ActionDoubleClickParam\n  >({\n    name: 'DoubleClick',\n    description: 'Double click the element',\n    interfaceAlias: 'aiDoubleClick',\n    paramSchema: actionDoubleClickParamSchema,\n    sample: {\n      locate: { prompt: 'the folder icon' },\n    },\n    missingLocateMessage: 'Element not found, cannot double click',\n    call: async (point) => {\n      await doubleClick(point);\n    },\n  });\n};\n\n// Hover\nexport const actionHoverParamSchema = z.object({\n  locate: getMidsceneLocationSchema().describe('The element to be hovered'),\n});\nexport type ActionHoverParam = {\n  locate: LocateResultElement;\n};\n\nexport const defineActionHover = (\n  hover: NonNullable<PointerInputPrimitives['hover']>,\n): DeviceAction<ActionHoverParam> => {\n  return defineLocatedPointAction<\n    typeof actionHoverParamSchema,\n    ActionHoverParam\n  >({\n    name: 'Hover',\n    description: 'Move the mouse to the element',\n    interfaceAlias: 'aiHover',\n    paramSchema: actionHoverParamSchema,\n    sample: {\n      locate: { prompt: 'the navigation menu item \"Products\"' },\n    },\n    missingLocateMessage: 'Element not found, cannot hover',\n    call: async (point) => {\n      await hover(point);\n    },\n  });\n};\n\n// Input\nconst inputLocateDescription =\n  'the position of the placeholder or text content in the target input field. If there is no content, locate the center of the input field.';\nexport const actionInputParamSchema = z.object({\n  value: z\n    .union([z.string(), z.number()])\n    .transform((val) => String(val))\n    .describe(\n      'The text to input. Provide the final content for replace mode, only the inserted characters for typeOnly mode, or an empty string when using clear mode to remove existing text.',\n    ),\n  locate: getMidsceneLocationSchema()\n    .describe(inputLocateDescription)\n    .optional(),\n  mode: z\n    .enum(['replace', 'clear', 'typeOnly'])\n    .default('replace')\n    .describe(\n      'Input mode: \"replace\" (default) - clear the field and input the value; \"typeOnly\" - type the value directly without clearing the field first, and should be set explicitly for incremental edits after moving the cursor; \"clear\" - clear the field without inputting new text.',\n    ),\n  autoDismissKeyboard: z\n    .boolean()\n    .optional()\n    .describe(\n      'If true, the keyboard will be dismissed after the input is completed. Do not set it unless the user asks you to do so.',\n    ),\n  keyboardTypeDelay: z\n    .number()\n    .finite()\n    .nonnegative()\n    .optional()\n    .describe(\n      'Delay in milliseconds between keystrokes when typing. Passed through from device/user configuration. Do not set it unless the user asks you to do so.',\n    ),\n  inputStrategy: z\n    .enum(inputStrategies)\n    .optional()\n    .describe(\n      'Text input strategy: \"legacy\" (default) preserves the current platform behavior; \"sequential\" enters one Unicode code point at a time; \"bulk\" sends the complete text through one platform input operation when supported and requires keyboardTypeDelay to be omitted or set to 0. Do not set it unless the user asks you to do so.',\n    ),\n});\nexport type ActionInputParam = {\n  value: string;\n  locate?: LocateResultElement;\n  mode?: 'replace' | 'clear' | 'typeOnly' | 'append';\n  autoDismissKeyboard?: boolean;\n  keyboardTypeDelay?: number;\n  inputStrategy?: InputStrategy;\n};\n\nexport const defineActionInput = (\n  keyboard: KeyboardInputPrimitives,\n): DeviceAction<ActionInputParam> => {\n  return defineAction<typeof actionInputParamSchema, ActionInputParam>({\n    name: 'Input',\n    description: 'Input the value into the element',\n    interfaceAlias: 'aiInput',\n    paramSchema: actionInputParamSchema,\n    sample: {\n      value: 'test@example.com',\n      locate: { prompt: 'the email input field' },\n    },\n    call: async (param) => {\n      // backward compat: convert deprecated 'append' to 'typeOnly'\n      if ((param.mode as string) === 'append') {\n        param.mode = 'typeOnly';\n      }\n\n      if (param.mode === 'clear') {\n        await keyboard.clearInput(param.locate);\n        return;\n      }\n\n      if (!param || !param.value) {\n        return;\n      }\n\n      await keyboard.typeText(param.value, {\n        target: param.locate,\n        replace: param.mode !== 'typeOnly',\n        autoDismissKeyboard: param.autoDismissKeyboard,\n        keyboardTypeDelay: param.keyboardTypeDelay,\n        inputStrategy: param.inputStrategy,\n      });\n    },\n  });\n};\n\n// KeyboardPress\nexport const actionKeyboardPressParamSchema = z.object({\n  locate: getMidsceneLocationSchema()\n    .describe(\n      'The optional element to click before pressing the key. Omit this when the key should operate on the currently focused element, especially when copying or cutting an existing text selection.',\n    )\n    .optional(),\n  keyName: z\n    .string()\n    .describe(\n      \"The key to be pressed. Use '+' for key combinations, e.g., 'Control+A', 'Shift+Enter'\",\n    ),\n});\nexport type ActionKeyboardPressParam = {\n  locate?: LocateResultElement;\n  keyName: string;\n};\n\nexport const defineActionKeyboardPress = (\n  keyboardPress: KeyboardInputPrimitives['keyboardPress'],\n): DeviceAction<ActionKeyboardPressParam> => {\n  return defineAction<\n    typeof actionKeyboardPressParamSchema,\n    ActionKeyboardPressParam\n  >({\n    name: 'KeyboardPress',\n    description:\n      'Press a key or key combination, like \"Enter\", \"Tab\", \"Escape\", or \"Control+A\", \"Shift+Enter\". Do not use this to type text. Omit locate to operate on the current focus without clicking again, especially for Copy or Cut after text has been selected.',\n    interfaceAlias: 'aiKeyboardPress',\n    paramSchema: actionKeyboardPressParamSchema,\n    sample: {\n      keyName: 'Enter',\n    },\n    call: async (param) => {\n      await keyboardPress(param.keyName, {\n        target: param.locate,\n      });\n    },\n  });\n};\n\n// Scroll\nexport const actionScrollParamSchema = z.object({\n  scrollType: z\n    .enum([\n      'singleAction',\n      'scrollToBottom',\n      'scrollToTop',\n      'scrollToRight',\n      'scrollToLeft',\n    ])\n    .default('singleAction')\n    .describe(\n      'The scroll behavior: \"singleAction\" for a single scroll action, \"scrollToBottom\" for scrolling all the way to the bottom by rapidly scrolling 5-10 times (skipping intermediate content until reaching the bottom), \"scrollToTop\" for scrolling all the way to the top by rapidly scrolling 5-10 times (skipping intermediate content until reaching the top), \"scrollToRight\" for scrolling all the way to the right by rapidly scrolling multiple times, \"scrollToLeft\" for scrolling all the way to the left by rapidly scrolling multiple times',\n    ),\n  direction: z\n    .enum(['down', 'up', 'right', 'left'])\n    .default('down')\n    .describe(\n      'The direction toward the off-screen content to reveal. \"down\" reveals content below the current viewport, \"up\" reveals content above, \"right\" reveals content to the right, and \"left\" reveals content to the left. This does not describe the movement direction of the content currently visible on the screen. Only effective when scrollType is \"singleAction\".',\n    ),\n  distance: z\n    .number()\n    .positive()\n    .nullable()\n    .optional()\n    .describe(\n      'Positive requested scroll amount in screen-coordinate units. On touch platforms, it controls the length of the swipe gesture used to scroll; on web and desktop platforms, it controls or approximates the wheel scroll delta. Only effective when scrollType is \"singleAction\".',\n    ),\n  locate: getMidsceneLocationSchema()\n    .optional()\n    .describe(\n      'Describe the target element to be scrolled on, like \"the table\" or \"the list\" or \"the content area\" or \"the scrollable area\". Do NOT provide a general intent like \"scroll to find some element\"',\n    ),\n});\n\nexport const defineActionScroll = (\n  scroll: ScrollInputPrimitives['scroll'],\n): DeviceAction<ActionScrollParam> => {\n  return defineAction<typeof actionScrollParamSchema, ActionScrollParam>({\n    name: 'Scroll',\n    description:\n      'Scroll a page or scrollable region to reveal off-screen content. Use Scroll when the goal is to browse content outside the current viewport. For direct gesture interactions, such as adjusting a slider or wheel picker, switching between paged cards or images, following an on-screen swipe gesture to continue or dismiss, or swiping an item to delete it, use Swipe instead if available in the current Action Space. Supports scrollToBottom/scrollToTop for boundary navigation. Default: direction `down`, scrollType `singleAction`, distance `null`.',\n    interfaceAlias: 'aiScroll',\n    paramSchema: actionScrollParamSchema,\n    sample: {\n      direction: 'down',\n      scrollType: 'singleAction',\n      locate: { prompt: 'the center of the product list area' },\n    },\n    call: async (param) => {\n      await scroll(param);\n    },\n  });\n};\n\n// DragAndDrop\nexport const actionDragAndDropParamSchema = z.object({\n  from: getMidsceneLocationSchema().describe('The position to be dragged'),\n  to: getMidsceneLocationSchema().describe('The position to be dropped'),\n});\nexport type ActionDragAndDropParam = {\n  from: LocateResultElement;\n  to: LocateResultElement;\n};\n\nexport const defineActionDragAndDrop = (\n  dragAndDrop: NonNullable<PointerInputPrimitives['dragAndDrop']>,\n): DeviceAction<ActionDragAndDropParam> => {\n  return defineAction<\n    typeof actionDragAndDropParamSchema,\n    ActionDragAndDropParam\n  >({\n    name: 'DragAndDrop',\n    description:\n      'Pick up a specific UI element and move it to a new position (e.g., reorder a card, move a file into a folder, sort list items). The element itself moves with your finger/mouse.',\n    interfaceAlias: 'aiDragAndDrop',\n    paramSchema: actionDragAndDropParamSchema,\n    sample: {\n      from: { prompt: 'the \"report.pdf\" file icon' },\n      to: { prompt: 'the upload drop zone' },\n    },\n    call: async (param) => {\n      const from = param.from;\n      const to = param.to;\n      if (!from) {\n        throw new Error('missing \"from\" param for drag and drop');\n      }\n      if (!to) {\n        throw new Error('missing \"to\" param for drag and drop');\n      }\n      await dragAndDrop(\n        { x: from.center[0], y: from.center[1] },\n        { x: to.center[0], y: to.center[1] },\n      );\n    },\n  });\n};\n\nexport const ActionLongPressParamSchema = z.object({\n  locate: getMidsceneLocationSchema().describe(\n    'The element to be long pressed',\n  ),\n  duration: z\n    .number()\n    .optional()\n    .describe('Long press duration in milliseconds'),\n});\n\nexport type ActionLongPressParam = {\n  locate: LocateResultElement;\n  duration?: number;\n};\nexport const defineActionLongPress = (\n  longPress: NonNullable<PointerInputPrimitives['longPress']>,\n): DeviceAction<ActionLongPressParam> => {\n  return defineLocatedPointAction<\n    typeof ActionLongPressParamSchema,\n    ActionLongPressParam\n  >({\n    name: 'LongPress',\n    description: 'Long press the element',\n    interfaceAlias: 'aiLongPress',\n    paramSchema: ActionLongPressParamSchema,\n    sample: {\n      locate: { prompt: 'the message bubble' },\n    },\n    missingLocateMessage: 'LongPress requires an element to be located',\n    call: async (point, param) => {\n      await longPress(point, { duration: param.duration });\n    },\n  });\n};\n\nfunction createActionSwipeParamSchema(inputMode: SwipeInputMode) {\n  const movementSource = inputMode === 'touch' ? 'finger' : 'pointer';\n\n  return z.object({\n    start: getMidsceneLocationSchema()\n      .optional()\n      .describe(\n        `Optional starting point of the ${movementSource} movement. Available in both relative and endpoint forms. If omitted, the center of the page is used.`,\n      ),\n    direction: z\n      .enum(['up', 'down', 'left', 'right'])\n      .optional()\n      .describe(\n        `${movementSource === 'finger' ? 'Finger' : 'Pointer'} movement direction. Required together with a positive distance for a relative swipe. Omit when using end.`,\n      ),\n    distance: z\n      .number()\n      .positive()\n      .optional()\n      .describe(\n        `Positive length of the ${movementSource} movement in pixels. Required together with direction for a relative swipe. Omit when using end.`,\n      ),\n    end: getMidsceneLocationSchema()\n      .optional()\n      .describe(\n        `Endpoint of the ${movementSource} movement. Use for an endpoint swipe, optionally with start. Do not provide direction or distance when using end.`,\n      ),\n    duration: z\n      .number()\n      .default(300)\n      .describe('Duration of the swipe gesture in milliseconds'),\n    repeat: z\n      .number()\n      .optional()\n      .describe(\n        'The number of times to repeat the swipe gesture. 1 for default, 0 for infinite (e.g. endless swipe until the end of the page)',\n      ),\n  });\n}\n\n// Preserve the historical touch-oriented public schema. Desktop actions use\n// an equivalent schema with mouse-specific descriptions at definition time.\nexport const ActionSwipeParamSchema = createActionSwipeParamSchema('touch');\n\nexport type ActionSwipeParam = {\n  start?: LocateResultElement;\n  direction?: 'up' | 'down' | 'left' | 'right';\n  distance?: number;\n  end?: LocateResultElement;\n  duration?: number;\n  repeat?: number;\n};\n\nexport function normalizeSwipeParam(\n  param: ActionSwipeParam,\n  screenSize: { width: number; height: number },\n): {\n  startPoint: { x: number; y: number };\n  endPoint: { x: number; y: number };\n  duration: number;\n  repeatCount: number;\n} {\n  const { width, height } = screenSize;\n  const { start, end } = param;\n\n  const startPoint = start\n    ? { x: start.center[0], y: start.center[1] }\n    : { x: width / 2, y: height / 2 };\n\n  const endPoint = (() => {\n    if (end) {\n      if (param.direction !== undefined || param.distance !== undefined) {\n        throw new Error(\n          'Invalid Swipe parameters: \"end\" cannot be combined with \"direction\" or \"distance\". Use \"end\", optionally with \"start\", for an endpoint swipe.',\n        );\n      }\n\n      return { x: end.center[0], y: end.center[1] };\n    }\n\n    if (param.direction === undefined || param.distance === undefined) {\n      throw new Error(\n        'Invalid Swipe parameters: a relative swipe requires both \"direction\" and a positive \"distance\".',\n      );\n    }\n\n    if (param.distance <= 0) {\n      throw new Error(\n        'Invalid Swipe parameters: \"distance\" must be a positive number.',\n      );\n    }\n\n    return {\n      x:\n        startPoint.x +\n        (param.direction === 'right'\n          ? param.distance\n          : param.direction === 'left'\n            ? -param.distance\n            : 0),\n      y:\n        startPoint.y +\n        (param.direction === 'down'\n          ? param.distance\n          : param.direction === 'up'\n            ? -param.distance\n            : 0),\n    };\n  })();\n\n  endPoint.x = Math.max(0, Math.min(endPoint.x, width));\n  endPoint.y = Math.max(0, Math.min(endPoint.y, height));\n\n  const duration = param.duration ?? 300;\n\n  let repeatCount = typeof param.repeat === 'number' ? param.repeat : 1;\n  if (repeatCount === 0) {\n    repeatCount = 10;\n  }\n\n  return { startPoint, endPoint, duration, repeatCount };\n}\n\n/** @deprecated Use {@link normalizeSwipeParam} instead. */\nexport const normalizeMobileSwipeParam = normalizeSwipeParam;\n\nexport const defineActionSwipe = (config: {\n  swipe: SwipeInputPrimitive;\n  size(): Promise<Size>;\n  inputMode?: SwipeInputMode;\n}): DeviceAction<ActionSwipeParam> => {\n  const inputMode = config.inputMode ?? 'touch';\n  return defineAction<typeof ActionSwipeParamSchema, ActionSwipeParam>({\n    name: 'Swipe',\n    description:\n      inputMode === 'touch'\n        ? 'Perform a touch gesture that directly manipulates the UI (e.g., adjust a continuous control such as a slider or wheel picker, switch between paged cards or images, follow an on-screen swipe gesture to continue or dismiss, or swipe an item to delete it). For browsing off-screen content in a page or scrollable region, use Scroll instead. Choose exactly one movement form: (1) relative swipe — provide \"direction\" and a positive \"distance\"; or (2) endpoint swipe — provide \"end\". \"start\" is optional for both forms and defaults to the center of the page. Do not combine \"end\" with \"direction\" or \"distance\".'\n        : 'Perform a primary-mouse-button gesture that continuously presses and moves the pointer across the desktop UI. Use it to adjust a continuous control such as a slider or wheel picker, switch between paged cards or images, follow an on-screen swipe gesture to continue or dismiss, or swipe an item to delete it. For browsing off-screen content in a page or scrollable region, use Scroll instead. Choose exactly one movement form: (1) relative swipe — provide \"direction\" and a positive \"distance\"; or (2) endpoint swipe — provide \"end\". \"start\" is optional for both forms and defaults to the center of the page. Do not combine \"end\" with \"direction\" or \"distance\".',\n    interfaceAlias: 'aiSwipe',\n    paramSchema:\n      inputMode === 'touch'\n        ? ActionSwipeParamSchema\n        : createActionSwipeParamSchema(inputMode),\n    sample: {\n      start: { prompt: 'center of the notification' },\n      end: { prompt: 'upper edge of the screen' },\n    },\n    call: async (param) => {\n      const { startPoint, endPoint, duration, repeatCount } =\n        normalizeSwipeParam(param, await config.size());\n      for (let i = 0; i < repeatCount; i++) {\n        await config.swipe(startPoint, endPoint, { duration });\n      }\n    },\n  });\n};\n\n// ClearInput\nexport const actionClearInputParamSchema = z.object({\n  locate: getMidsceneLocationSchema()\n    .describe('The input field to be cleared')\n    .optional(),\n});\nexport type ActionClearInputParam = {\n  locate?: LocateResultElement;\n};\n\nexport const defineActionClearInput = (\n  clearInput: KeyboardInputPrimitives['clearInput'],\n): DeviceAction<ActionClearInputParam> => {\n  return defineAction<\n    typeof actionClearInputParamSchema,\n    ActionClearInputParam\n  >({\n    name: 'ClearInput',\n    description: inputLocateDescription,\n    interfaceAlias: 'aiClearInput',\n    paramSchema: actionClearInputParamSchema,\n    sample: {\n      locate: { prompt: 'the search input field' },\n    },\n    call: async (param) => {\n      await clearInput(param.locate);\n    },\n  });\n};\n\n// CursorMove\nexport const actionCursorMoveParamSchema = z.object({\n  direction: z\n    .enum(['left', 'right'])\n    .describe('The direction to move the cursor'),\n  times: z\n    .number()\n    .int()\n    .min(1)\n    .default(1)\n    .describe(\n      'The number of times to move the cursor in the specified direction',\n    ),\n});\nexport type ActionCursorMoveParam = {\n  direction: 'left' | 'right';\n  times?: number;\n};\n\nexport const defineActionCursorMove = (config: {\n  keyboard: Pick<KeyboardInputPrimitives, 'keyboardPress' | 'cursorMove'>;\n  sleep?(timeMs: number): Promise<void>;\n}): DeviceAction<ActionCursorMoveParam> => {\n  return defineAction<\n    typeof actionCursorMoveParamSchema,\n    ActionCursorMoveParam\n  >({\n    name: 'CursorMove',\n    description:\n      'Move the text cursor (caret) left or right within an input field or text area. Use this to reposition the cursor without selecting text.',\n    paramSchema: actionCursorMoveParamSchema,\n    sample: {\n      direction: 'left',\n      times: 3,\n    },\n    call: async (param) => {\n      const times = param.times ?? 1;\n      if (config.keyboard.cursorMove) {\n        await config.keyboard.cursorMove(param.direction, times);\n        return;\n      }\n\n      const wait =\n        config.sleep ??\n        ((timeMs: number) =>\n          new Promise<void>((resolve) => setTimeout(resolve, timeMs)));\n      const arrowKey = param.direction === 'left' ? 'ArrowLeft' : 'ArrowRight';\n      for (let i = 0; i < times; i++) {\n        await config.keyboard.keyboardPress(arrowKey);\n        await wait(100);\n      }\n    },\n  });\n};\n\n// Pinch\nexport const ActionPinchParamSchema = z.object({\n  locate: getMidsceneLocationSchema()\n    .optional()\n    .describe(\n      'The element to pinch on. If not specified, the center of the screen will be used',\n    ),\n  direction: z\n    .enum(['in', 'out'])\n    .describe(\n      'Pinch direction. \"in\" = pinch fingers together (zoom out / shrink), \"out\" = spread fingers apart (zoom in / enlarge).',\n    ),\n  distance: z\n    .number()\n    .positive()\n    .optional()\n    .describe(\n      'How far each finger moves in pixels. Defaults to a quarter of the shorter screen dimension.',\n    ),\n  duration: z\n    .number()\n    .default(500)\n    .optional()\n    .describe('Duration of the pinch gesture in milliseconds'),\n});\n\nexport type ActionPinchParam = {\n  locate?: LocateResultElement;\n  direction: 'in' | 'out';\n  distance?: number;\n  duration?: number;\n};\n\nexport const defineActionPinch = (config: {\n  pinch: TouchInputPrimitives['pinch'];\n  size(): Promise<Size>;\n}): DeviceAction<ActionPinchParam> | undefined => {\n  if (!config.pinch) {\n    return undefined;\n  }\n\n  return defineAction<typeof ActionPinchParamSchema, ActionPinchParam>({\n    name: 'Pinch',\n    description:\n      'Perform a two-finger pinch gesture. Use direction \"in\" to pinch fingers together (zoom out), or \"out\" to spread fingers apart (zoom in). Optionally specify distance for how far each finger moves.',\n    interfaceAlias: 'aiPinch',\n    paramSchema: ActionPinchParamSchema,\n    sample: {\n      locate: { prompt: 'the map area' },\n      direction: 'out',\n      distance: 200,\n    },\n    call: async (param) => {\n      const { centerX, centerY, startDistance, endDistance, duration } =\n        normalizePinchParam(param, await config.size());\n      await config.pinch?.(\n        { x: centerX, y: centerY },\n        { startDistance, endDistance, duration },\n      );\n    },\n  });\n};\n\nexport function normalizePinchParam(\n  param: ActionPinchParam,\n  screenSize: { width: number; height: number },\n): {\n  centerX: number;\n  centerY: number;\n  startDistance: number;\n  endDistance: number;\n  duration: number;\n} {\n  const { width, height } = screenSize;\n  const element = param.locate;\n  const centerX = element\n    ? Math.round(element.center[0])\n    : Math.round(width / 2);\n  const centerY = element\n    ? Math.round(element.center[1])\n    : Math.round(height / 2);\n  const duration = param.duration ?? 500;\n\n  const baseDistance = Math.round(Math.min(width, height) / 4);\n  const fingerDistance = param.distance ?? baseDistance;\n\n  const startDistance = baseDistance;\n  const endDistance =\n    param.direction === 'out'\n      ? baseDistance + fingerDistance\n      : Math.max(10, baseDistance - fingerDistance);\n\n  return { centerX, centerY, startDistance, endDistance, duration };\n}\n\nexport interface MobileInputActionContext {\n  input: MobileInputPrimitives;\n  size(): Promise<Size>;\n  sleep?(timeMs: number): Promise<void>;\n  getDefaultAutoDismissKeyboard?(): boolean | undefined;\n  systemActions?: SystemInputActionOptions;\n}\n\nexport interface SystemInputActionConfig {\n  name: string;\n  description: string;\n  interfaceAlias?: string;\n  delayBeforeRunner?: number;\n  delayAfterRunner?: number;\n}\n\nexport interface SystemInputActionOptions {\n  backButton?: SystemInputActionConfig;\n  homeButton?: SystemInputActionConfig;\n  recentAppsButton?: SystemInputActionConfig;\n}\n\nexport interface InputPrimitiveActionOptions {\n  size?: () => Promise<Size>;\n  sleep?: (timeMs: number) => Promise<void>;\n  includeSwipe?: boolean;\n  includePinch?: boolean;\n  systemActions?: SystemInputActionOptions;\n}\n\nfunction defineSystemInputAction(\n  config: SystemInputActionConfig,\n  call: () => Promise<void>,\n): DeviceAction<undefined, void> {\n  return defineAction<undefined, undefined, void>({\n    name: config.name,\n    description: config.description,\n    interfaceAlias: config.interfaceAlias,\n    delayBeforeRunner: config.delayBeforeRunner,\n    delayAfterRunner: config.delayAfterRunner,\n    call,\n  });\n}\n\nexport function defineActionsFromInputPrimitives(\n  input: InputPrimitives,\n  options: InputPrimitiveActionOptions = {},\n): DeviceAction<any>[] {\n  const actions: Array<DeviceAction<any> | undefined> = [];\n  const { pointer, keyboard, scroll, touch, system } = input;\n\n  if (pointer) {\n    actions.push(defineActionTap(pointer.tap));\n    if (pointer.doubleClick) {\n      actions.push(defineActionDoubleClick(pointer.doubleClick));\n    }\n    if (pointer.rightClick) {\n      actions.push(defineActionRightClick(pointer.rightClick));\n    }\n    if (pointer.hover) {\n      actions.push(defineActionHover(pointer.hover));\n    }\n    if (pointer.dragAndDrop) {\n      actions.push(defineActionDragAndDrop(pointer.dragAndDrop));\n    }\n    if (pointer.longPress) {\n      actions.push(defineActionLongPress(pointer.longPress));\n    }\n  }\n\n  if (keyboard) {\n    actions.push(\n      defineActionInput(keyboard),\n      defineActionClearInput(keyboard.clearInput),\n      defineActionKeyboardPress(keyboard.keyboardPress),\n      defineActionCursorMove({ keyboard, sleep: options.sleep }),\n    );\n  }\n\n  if (scroll) {\n    actions.push(defineActionScroll(scroll.scroll));\n  }\n\n  const swipeConfig = resolveSwipeInputPrimitive({ pointer, touch });\n  if (swipeConfig && options.size && options.includeSwipe !== false) {\n    actions.push(defineActionSwipe({ ...swipeConfig, size: options.size }));\n  }\n\n  if (touch?.pinch && options.size && options.includePinch !== false) {\n    actions.push(defineActionPinch({ pinch: touch.pinch, size: options.size }));\n  }\n\n  if (system && options.systemActions) {\n    const { systemActions } = options;\n    if (system.backButton && systemActions.backButton) {\n      actions.push(\n        defineSystemInputAction(systemActions.backButton, system.backButton),\n      );\n    }\n    if (system.homeButton && systemActions.homeButton) {\n      actions.push(\n        defineSystemInputAction(systemActions.homeButton, system.homeButton),\n      );\n    }\n    if (system.recentAppsButton && systemActions.recentAppsButton) {\n      actions.push(\n        defineSystemInputAction(\n          systemActions.recentAppsButton,\n          system.recentAppsButton,\n        ),\n      );\n    }\n  }\n\n  return actions.filter((action): action is DeviceAction<any> =>\n    Boolean(action),\n  );\n}\n\nexport function createDefaultMobileActions(\n  context: MobileInputActionContext,\n): DeviceAction<any>[] {\n  return defineActionsFromInputPrimitives(context.input, {\n    size: context.size,\n    sleep: context.sleep,\n    systemActions: context.systemActions,\n  });\n}\n\n// Sleep\nexport const ActionSleepParamSchema = z.object({\n  timeMs: z\n    .number()\n    .default(1000)\n    .optional()\n    .describe('Sleep duration in milliseconds, defaults to 1000ms (1 second)'),\n});\n\nexport type ActionSleepParam = {\n  timeMs?: number;\n};\n\nexport const defineActionSleep = (\n  abortSignal?: AbortSignal,\n): DeviceAction<ActionSleepParam> => {\n  return defineAction<typeof ActionSleepParamSchema, ActionSleepParam>({\n    name: 'Sleep',\n    description:\n      'Wait for a specified duration before continuing. Defaults to 1 second (1000ms) if not specified.',\n    paramSchema: ActionSleepParamSchema,\n    sample: {\n      timeMs: 2000,\n    },\n    call: async (param) => {\n      const duration = param?.timeMs ?? 1000;\n      getDebug('device:common-action')(`Sleeping for ${duration}ms`);\n      abortSignal?.throwIfAborted();\n      await new Promise<void>((resolve, reject) => {\n        const abort = () => {\n          clearTimeout(timer);\n          reject(abortSignal?.reason);\n        };\n        const timer = setTimeout(() => {\n          abortSignal?.removeEventListener('abort', abort);\n          resolve();\n        }, duration);\n        abortSignal?.addEventListener('abort', abort, { once: true });\n      });\n    },\n  });\n};\n\nexport type { DeviceAction } from '../types';\nexport type { ActionScrollParam, LocateResultElement, Size } from '../types';\nexport { getMidsceneLocationSchema } from '../common';\nexport { z };\nexport {\n  inputStrategies,\n  type InputStrategy,\n  type ResolvedTextInputOptions,\n  type SequentialTextInputHandlers,\n  type TextInputOptions,\n  resolveInputStrategy,\n  resolveTextInputOptions,\n  sendTextSequentially,\n  shouldInputSequentially,\n} from './input-strategy';\nexport type {\n  AndroidDeviceOpt,\n  AndroidDeviceInputOpt,\n  IOSDeviceOpt,\n  IOSDeviceInputOpt,\n  HarmonyDeviceOpt,\n  HarmonyDeviceInputOpt,\n} from './device-options';\n"],"names":["resolveSwipeInputPrimitive","input","AbstractInterface","defineAction","config","pointFromLocate","locate","missingMessage","Error","defineLocatedPointAction","param","actionTapParamSchema","z","getMidsceneLocationSchema","defineActionTap","tap","point","registerFileChooserAcceptParamSchema","defineActionRegisterFileChooserAccept","register","actionRightClickParamSchema","defineActionRightClick","rightClick","actionDoubleClickParamSchema","defineActionDoubleClick","doubleClick","actionHoverParamSchema","defineActionHover","hover","inputLocateDescription","actionInputParamSchema","val","String","inputStrategies","defineActionInput","keyboard","actionKeyboardPressParamSchema","defineActionKeyboardPress","keyboardPress","actionScrollParamSchema","defineActionScroll","scroll","actionDragAndDropParamSchema","defineActionDragAndDrop","dragAndDrop","from","to","ActionLongPressParamSchema","defineActionLongPress","longPress","createActionSwipeParamSchema","inputMode","movementSource","ActionSwipeParamSchema","normalizeSwipeParam","screenSize","width","height","start","end","startPoint","endPoint","undefined","Math","duration","repeatCount","normalizeMobileSwipeParam","defineActionSwipe","i","actionClearInputParamSchema","defineActionClearInput","clearInput","actionCursorMoveParamSchema","defineActionCursorMove","times","wait","timeMs","Promise","resolve","setTimeout","arrowKey","ActionPinchParamSchema","defineActionPinch","centerX","centerY","startDistance","endDistance","normalizePinchParam","element","baseDistance","fingerDistance","defineSystemInputAction","call","defineActionsFromInputPrimitives","options","actions","pointer","touch","system","swipeConfig","systemActions","action","Boolean","createDefaultMobileActions","context","ActionSleepParamSchema","defineActionSleep","abortSignal","getDebug","reject","abort","clearTimeout","timer"],"mappings":";;;;;;;;;;;;;;AA6JO,SAASA,2BACdC,KAAiD;IAEjD,IAAIA,MAAM,KAAK,EAAE,OACf,OAAO;QAAE,OAAOA,MAAM,KAAK,CAAC,KAAK;QAAE,WAAW;IAAQ;IAExD,IAAIA,MAAM,OAAO,EAAE,OACjB,OAAO;QAAE,OAAOA,MAAM,OAAO,CAAC,KAAK;QAAE,WAAW;IAAQ;AAG5D;AAoCO,MAAeC;;QAgDpB;QAqDA;;AACF;AAKO,MAAMC,eAAe,CAK1BC,SAgBOA;AAGT,SAASC,gBACPC,MAAuC,EACvCC,cAAsB;IAEtB,IAAI,CAACD,QACH,MAAM,IAAIE,MAAMD;IAElB,OAAO;QAAE,GAAGD,OAAO,MAAM,CAAC,EAAE;QAAE,GAAGA,OAAO,MAAM,CAAC,EAAE;IAAC;AACpD;AAEA,SAASG,yBAGPL,MAQD;IACC,OAAOD,aAA8B;QACnC,MAAMC,OAAO,IAAI;QACjB,aAAaA,OAAO,WAAW;QAC/B,gBAAgBA,OAAO,cAAc;QACrC,aAAaA,OAAO,WAAW;QAC/B,QAAQA,OAAO,MAAM;QACrB,MAAM,OAAOM;YACX,MAAMN,OAAO,IAAI,CACfC,gBAAgBK,MAAM,MAAM,EAAEN,OAAO,oBAAoB,GACzDM;QAEJ;IACF;AACF;AAGO,MAAMC,uBAAuBC,EAAE,MAAM,CAAC;IAC3C,QAAQC,4BAA4B,QAAQ,CAAC;AAC/C;AAKO,MAAMC,kBAAkB,CAC7BC,MAEON,yBAAsE;QAC3E,MAAM;QACN,aAAa;QACb,gBAAgB;QAChB,aAAaE;QACb,QAAQ;YACN,QAAQ;gBAAE,QAAQ;YAAsB;QAC1C;QACA,sBAAsB;QACtB,MAAM,OAAOK;YACX,MAAMD,IAAIC;QACZ;IACF;AAGK,MAAMC,uCAAuCL,EAAE,MAAM,CAAC;IAC3D,OAAOA,EAAAA,KACC,CAAC;QAACA,EAAE,MAAM;QAAIA,EAAE,KAAK,CAACA,EAAE,MAAM;KAAI,EACvC,QAAQ,CACP;AAEN;AAKO,MAAMM,wCAAwC,CACnDC,WAEOhB,aAGL;QACA,MAAM;QACN,aACE;QACF,gBAAgB;QAChB,aAAac;QACb,QAAQ;YACN,OAAO;gBAAC;aAAwB;QAClC;QACA,MAAM,OAAOP;YACX,MAAMS,SAAST,MAAM,KAAK;QAC5B;IACF;AAIK,MAAMU,8BAA8BR,EAAE,MAAM,CAAC;IAClD,QAAQC,4BAA4B,QAAQ,CAC1C;AAEJ;AAKO,MAAMQ,yBAAyB,CACpCC,aAEOb,yBAGL;QACA,MAAM;QACN,aAAa;QACb,gBAAgB;QAChB,aAAaW;QACb,QAAQ;YACN,QAAQ;gBAAE,QAAQ;YAA+B;QACnD;QACA,sBAAsB;QACtB,MAAM,OAAOJ;YACX,MAAMM,WAAWN;QACnB;IACF;AAIK,MAAMO,+BAA+BX,EAAE,MAAM,CAAC;IACnD,QAAQC,4BAA4B,QAAQ,CAC1C;AAEJ;AAKO,MAAMW,0BAA0B,CACrCC,cAEOhB,yBAGL;QACA,MAAM;QACN,aAAa;QACb,gBAAgB;QAChB,aAAac;QACb,QAAQ;YACN,QAAQ;gBAAE,QAAQ;YAAkB;QACtC;QACA,sBAAsB;QACtB,MAAM,OAAOP;YACX,MAAMS,YAAYT;QACpB;IACF;AAIK,MAAMU,yBAAyBd,EAAE,MAAM,CAAC;IAC7C,QAAQC,4BAA4B,QAAQ,CAAC;AAC/C;AAKO,MAAMc,oBAAoB,CAC/BC,QAEOnB,yBAGL;QACA,MAAM;QACN,aAAa;QACb,gBAAgB;QAChB,aAAaiB;QACb,QAAQ;YACN,QAAQ;gBAAE,QAAQ;YAAsC;QAC1D;QACA,sBAAsB;QACtB,MAAM,OAAOV;YACX,MAAMY,MAAMZ;QACd;IACF;AAIF,MAAMa,yBACJ;AACK,MAAMC,yBAAyBlB,EAAE,MAAM,CAAC;IAC7C,OAAOA,EAAAA,KACC,CAAC;QAACA,EAAE,MAAM;QAAIA,EAAE,MAAM;KAAG,EAC9B,SAAS,CAAC,CAACmB,MAAQC,OAAOD,MAC1B,QAAQ,CACP;IAEJ,QAAQlB,4BACL,QAAQ,CAACgB,wBACT,QAAQ;IACX,MAAMjB,CAAC,CAADA,OACC,CAAC;QAAC;QAAW;QAAS;KAAW,EACrC,OAAO,CAAC,WACR,QAAQ,CACP;IAEJ,qBAAqBA,EAAAA,OACX,GACP,QAAQ,GACR,QAAQ,CACP;IAEJ,mBAAmBA,EAAAA,MACV,GACN,MAAM,GACN,WAAW,GACX,QAAQ,GACR,QAAQ,CACP;IAEJ,eAAeA,CAAC,CAADA,OACR,CAACqB,iBACL,QAAQ,GACR,QAAQ,CACP;AAEN;AAUO,MAAMC,oBAAoB,CAC/BC,WAEOhC,aAA8D;QACnE,MAAM;QACN,aAAa;QACb,gBAAgB;QAChB,aAAa2B;QACb,QAAQ;YACN,OAAO;YACP,QAAQ;gBAAE,QAAQ;YAAwB;QAC5C;QACA,MAAM,OAAOpB;YAEX,IAAKA,AAA0B,aAA1BA,MAAM,IAAI,EACbA,MAAM,IAAI,GAAG;YAGf,IAAIA,AAAe,YAAfA,MAAM,IAAI,EAAc,YAC1B,MAAMyB,SAAS,UAAU,CAACzB,MAAM,MAAM;YAIxC,IAAI,CAACA,SAAS,CAACA,MAAM,KAAK,EACxB;YAGF,MAAMyB,SAAS,QAAQ,CAACzB,MAAM,KAAK,EAAE;gBACnC,QAAQA,MAAM,MAAM;gBACpB,SAASA,AAAe,eAAfA,MAAM,IAAI;gBACnB,qBAAqBA,MAAM,mBAAmB;gBAC9C,mBAAmBA,MAAM,iBAAiB;gBAC1C,eAAeA,MAAM,aAAa;YACpC;QACF;IACF;AAIK,MAAM0B,iCAAiCxB,EAAE,MAAM,CAAC;IACrD,QAAQC,4BACL,QAAQ,CACP,iMAED,QAAQ;IACX,SAASD,EAAAA,MACA,GACN,QAAQ,CACP;AAEN;AAMO,MAAMyB,4BAA4B,CACvCC,gBAEOnC,aAGL;QACA,MAAM;QACN,aACE;QACF,gBAAgB;QAChB,aAAaiC;QACb,QAAQ;YACN,SAAS;QACX;QACA,MAAM,OAAO1B;YACX,MAAM4B,cAAc5B,MAAM,OAAO,EAAE;gBACjC,QAAQA,MAAM,MAAM;YACtB;QACF;IACF;AAIK,MAAM6B,0BAA0B3B,EAAE,MAAM,CAAC;IAC9C,YAAYA,CAAC,CAADA,OACL,CAAC;QACJ;QACA;QACA;QACA;QACA;KACD,EACA,OAAO,CAAC,gBACR,QAAQ,CACP;IAEJ,WAAWA,CAAC,CAADA,OACJ,CAAC;QAAC;QAAQ;QAAM;QAAS;KAAO,EACpC,OAAO,CAAC,QACR,QAAQ,CACP;IAEJ,UAAUA,EAAAA,MACD,GACN,QAAQ,GACR,QAAQ,GACR,QAAQ,GACR,QAAQ,CACP;IAEJ,QAAQC,4BACL,QAAQ,GACR,QAAQ,CACP;AAEN;AAEO,MAAM2B,qBAAqB,CAChCC,SAEOtC,aAAgE;QACrE,MAAM;QACN,aACE;QACF,gBAAgB;QAChB,aAAaoC;QACb,QAAQ;YACN,WAAW;YACX,YAAY;YACZ,QAAQ;gBAAE,QAAQ;YAAsC;QAC1D;QACA,MAAM,OAAO7B;YACX,MAAM+B,OAAO/B;QACf;IACF;AAIK,MAAMgC,+BAA+B9B,EAAE,MAAM,CAAC;IACnD,MAAMC,4BAA4B,QAAQ,CAAC;IAC3C,IAAIA,4BAA4B,QAAQ,CAAC;AAC3C;AAMO,MAAM8B,0BAA0B,CACrCC,cAEOzC,aAGL;QACA,MAAM;QACN,aACE;QACF,gBAAgB;QAChB,aAAauC;QACb,QAAQ;YACN,MAAM;gBAAE,QAAQ;YAA6B;YAC7C,IAAI;gBAAE,QAAQ;YAAuB;QACvC;QACA,MAAM,OAAOhC;YACX,MAAMmC,OAAOnC,MAAM,IAAI;YACvB,MAAMoC,KAAKpC,MAAM,EAAE;YACnB,IAAI,CAACmC,MACH,MAAM,IAAIrC,MAAM;YAElB,IAAI,CAACsC,IACH,MAAM,IAAItC,MAAM;YAElB,MAAMoC,YACJ;gBAAE,GAAGC,KAAK,MAAM,CAAC,EAAE;gBAAE,GAAGA,KAAK,MAAM,CAAC,EAAE;YAAC,GACvC;gBAAE,GAAGC,GAAG,MAAM,CAAC,EAAE;gBAAE,GAAGA,GAAG,MAAM,CAAC,EAAE;YAAC;QAEvC;IACF;AAGK,MAAMC,6BAA6BnC,EAAE,MAAM,CAAC;IACjD,QAAQC,4BAA4B,QAAQ,CAC1C;IAEF,UAAUD,EAAAA,MACD,GACN,QAAQ,GACR,QAAQ,CAAC;AACd;AAMO,MAAMoC,wBAAwB,CACnCC,YAEOxC,yBAGL;QACA,MAAM;QACN,aAAa;QACb,gBAAgB;QAChB,aAAasC;QACb,QAAQ;YACN,QAAQ;gBAAE,QAAQ;YAAqB;QACzC;QACA,sBAAsB;QACtB,MAAM,OAAO/B,OAAON;YAClB,MAAMuC,UAAUjC,OAAO;gBAAE,UAAUN,MAAM,QAAQ;YAAC;QACpD;IACF;AAGF,SAASwC,6BAA6BC,SAAyB;IAC7D,MAAMC,iBAAiBD,AAAc,YAAdA,YAAwB,WAAW;IAE1D,OAAOvC,EAAE,MAAM,CAAC;QACd,OAAOC,4BACJ,QAAQ,GACR,QAAQ,CACP,CAAC,+BAA+B,EAAEuC,eAAe,qGAAqG,CAAC;QAE3J,WAAWxC,CAAC,CAADA,OACJ,CAAC;YAAC;YAAM;YAAQ;YAAQ;SAAQ,EACpC,QAAQ,GACR,QAAQ,CACP,GAAGwC,AAAmB,aAAnBA,iBAA8B,WAAW,UAAU,0GAA0G,CAAC;QAErK,UAAUxC,EAAAA,MACD,GACN,QAAQ,GACR,QAAQ,GACR,QAAQ,CACP,CAAC,uBAAuB,EAAEwC,eAAe,gGAAgG,CAAC;QAE9I,KAAKvC,4BACF,QAAQ,GACR,QAAQ,CACP,CAAC,gBAAgB,EAAEuC,eAAe,iHAAiH,CAAC;QAExJ,UAAUxC,EAAAA,MACD,GACN,OAAO,CAAC,KACR,QAAQ,CAAC;QACZ,QAAQA,EAAAA,MACC,GACN,QAAQ,GACR,QAAQ,CACP;IAEN;AACF;AAIO,MAAMyC,yBAAyBH,6BAA6B;AAW5D,SAASI,oBACd5C,KAAuB,EACvB6C,UAA6C;IAO7C,MAAM,EAAEC,KAAK,EAAEC,MAAM,EAAE,GAAGF;IAC1B,MAAM,EAAEG,KAAK,EAAEC,GAAG,EAAE,GAAGjD;IAEvB,MAAMkD,aAAaF,QACf;QAAE,GAAGA,MAAM,MAAM,CAAC,EAAE;QAAE,GAAGA,MAAM,MAAM,CAAC,EAAE;IAAC,IACzC;QAAE,GAAGF,QAAQ;QAAG,GAAGC,SAAS;IAAE;IAElC,MAAMI,WAAY,AAAC;QACjB,IAAIF,KAAK;YACP,IAAIjD,AAAoBoD,WAApBpD,MAAM,SAAS,IAAkBA,AAAmBoD,WAAnBpD,MAAM,QAAQ,EACjD,MAAM,IAAIF,MACR;YAIJ,OAAO;gBAAE,GAAGmD,IAAI,MAAM,CAAC,EAAE;gBAAE,GAAGA,IAAI,MAAM,CAAC,EAAE;YAAC;QAC9C;QAEA,IAAIjD,AAAoBoD,WAApBpD,MAAM,SAAS,IAAkBA,AAAmBoD,WAAnBpD,MAAM,QAAQ,EACjD,MAAM,IAAIF,MACR;QAIJ,IAAIE,MAAM,QAAQ,IAAI,GACpB,MAAM,IAAIF,MACR;QAIJ,OAAO;YACL,GACEoD,WAAW,CAAC,GACXlD,CAAAA,AAAoB,YAApBA,MAAM,SAAS,GACZA,MAAM,QAAQ,GACdA,AAAoB,WAApBA,MAAM,SAAS,GACb,CAACA,MAAM,QAAQ,GACf;YACR,GACEkD,WAAW,CAAC,GACXlD,CAAAA,AAAoB,WAApBA,MAAM,SAAS,GACZA,MAAM,QAAQ,GACdA,AAAoB,SAApBA,MAAM,SAAS,GACb,CAACA,MAAM,QAAQ,GACf;QACV;IACF;IAEAmD,SAAS,CAAC,GAAGE,KAAK,GAAG,CAAC,GAAGA,KAAK,GAAG,CAACF,SAAS,CAAC,EAAEL;IAC9CK,SAAS,CAAC,GAAGE,KAAK,GAAG,CAAC,GAAGA,KAAK,GAAG,CAACF,SAAS,CAAC,EAAEJ;IAE9C,MAAMO,WAAWtD,MAAM,QAAQ,IAAI;IAEnC,IAAIuD,cAAc,AAAwB,YAAxB,OAAOvD,MAAM,MAAM,GAAgBA,MAAM,MAAM,GAAG;IACpE,IAAIuD,AAAgB,MAAhBA,aACFA,cAAc;IAGhB,OAAO;QAAEL;QAAYC;QAAUG;QAAUC;IAAY;AACvD;AAGO,MAAMC,4BAA4BZ;AAElC,MAAMa,oBAAoB,CAAC/D;IAKhC,MAAM+C,YAAY/C,OAAO,SAAS,IAAI;IACtC,OAAOD,aAA8D;QACnE,MAAM;QACN,aACEgD,AAAc,YAAdA,YACI,mmBACA;QACN,gBAAgB;QAChB,aACEA,AAAc,YAAdA,YACIE,yBACAH,6BAA6BC;QACnC,QAAQ;YACN,OAAO;gBAAE,QAAQ;YAA6B;YAC9C,KAAK;gBAAE,QAAQ;YAA2B;QAC5C;QACA,MAAM,OAAOzC;YACX,MAAM,EAAEkD,UAAU,EAAEC,QAAQ,EAAEG,QAAQ,EAAEC,WAAW,EAAE,GACnDX,oBAAoB5C,OAAO,MAAMN,OAAO,IAAI;YAC9C,IAAK,IAAIgE,IAAI,GAAGA,IAAIH,aAAaG,IAC/B,MAAMhE,OAAO,KAAK,CAACwD,YAAYC,UAAU;gBAAEG;YAAS;QAExD;IACF;AACF;AAGO,MAAMK,8BAA8BzD,EAAE,MAAM,CAAC;IAClD,QAAQC,4BACL,QAAQ,CAAC,iCACT,QAAQ;AACb;AAKO,MAAMyD,yBAAyB,CACpCC,aAEOpE,aAGL;QACA,MAAM;QACN,aAAa0B;QACb,gBAAgB;QAChB,aAAawC;QACb,QAAQ;YACN,QAAQ;gBAAE,QAAQ;YAAyB;QAC7C;QACA,MAAM,OAAO3D;YACX,MAAM6D,WAAW7D,MAAM,MAAM;QAC/B;IACF;AAIK,MAAM8D,8BAA8B5D,EAAE,MAAM,CAAC;IAClD,WAAWA,CAAC,CAADA,OACJ,CAAC;QAAC;QAAQ;KAAQ,EACtB,QAAQ,CAAC;IACZ,OAAOA,EAAAA,MACE,GACN,GAAG,GACH,GAAG,CAAC,GACJ,OAAO,CAAC,GACR,QAAQ,CACP;AAEN;AAMO,MAAM6D,yBAAyB,CAACrE,SAI9BD,aAGL;QACA,MAAM;QACN,aACE;QACF,aAAaqE;QACb,QAAQ;YACN,WAAW;YACX,OAAO;QACT;QACA,MAAM,OAAO9D;YACX,MAAMgE,QAAQhE,MAAM,KAAK,IAAI;YAC7B,IAAIN,OAAO,QAAQ,CAAC,UAAU,EAAE,YAC9B,MAAMA,OAAO,QAAQ,CAAC,UAAU,CAACM,MAAM,SAAS,EAAEgE;YAIpD,MAAMC,OACJvE,OAAO,KAAK,IACV,EAAAwE,SACA,IAAIC,QAAc,CAACC,UAAYC,WAAWD,SAASF,QAAO;YAC9D,MAAMI,WAAWtE,AAAoB,WAApBA,MAAM,SAAS,GAAc,cAAc;YAC5D,IAAK,IAAI0D,IAAI,GAAGA,IAAIM,OAAON,IAAK;gBAC9B,MAAMhE,OAAO,QAAQ,CAAC,aAAa,CAAC4E;gBACpC,MAAML,KAAK;YACb;QACF;IACF;AAIK,MAAMM,yBAAyBrE,EAAE,MAAM,CAAC;IAC7C,QAAQC,4BACL,QAAQ,GACR,QAAQ,CACP;IAEJ,WAAWD,CAAC,CAADA,OACJ,CAAC;QAAC;QAAM;KAAM,EAClB,QAAQ,CACP;IAEJ,UAAUA,EAAAA,MACD,GACN,QAAQ,GACR,QAAQ,GACR,QAAQ,CACP;IAEJ,UAAUA,EAAAA,MACD,GACN,OAAO,CAAC,KACR,QAAQ,GACR,QAAQ,CAAC;AACd;AASO,MAAMsE,oBAAoB,CAAC9E;IAIhC,IAAI,CAACA,OAAO,KAAK,EACf;IAGF,OAAOD,aAA8D;QACnE,MAAM;QACN,aACE;QACF,gBAAgB;QAChB,aAAa8E;QACb,QAAQ;YACN,QAAQ;gBAAE,QAAQ;YAAe;YACjC,WAAW;YACX,UAAU;QACZ;QACA,MAAM,OAAOvE;YACX,MAAM,EAAEyE,OAAO,EAAEC,OAAO,EAAEC,aAAa,EAAEC,WAAW,EAAEtB,QAAQ,EAAE,GAC9DuB,oBAAoB7E,OAAO,MAAMN,OAAO,IAAI;YAC9C,MAAMA,OAAO,KAAK,GAChB;gBAAE,GAAG+E;gBAAS,GAAGC;YAAQ,GACzB;gBAAEC;gBAAeC;gBAAatB;YAAS;QAE3C;IACF;AACF;AAEO,SAASuB,oBACd7E,KAAuB,EACvB6C,UAA6C;IAQ7C,MAAM,EAAEC,KAAK,EAAEC,MAAM,EAAE,GAAGF;IAC1B,MAAMiC,UAAU9E,MAAM,MAAM;IAC5B,MAAMyE,UAAUK,UACZzB,KAAK,KAAK,CAACyB,QAAQ,MAAM,CAAC,EAAE,IAC5BzB,KAAK,KAAK,CAACP,QAAQ;IACvB,MAAM4B,UAAUI,UACZzB,KAAK,KAAK,CAACyB,QAAQ,MAAM,CAAC,EAAE,IAC5BzB,KAAK,KAAK,CAACN,SAAS;IACxB,MAAMO,WAAWtD,MAAM,QAAQ,IAAI;IAEnC,MAAM+E,eAAe1B,KAAK,KAAK,CAACA,KAAK,GAAG,CAACP,OAAOC,UAAU;IAC1D,MAAMiC,iBAAiBhF,MAAM,QAAQ,IAAI+E;IAEzC,MAAMJ,gBAAgBI;IACtB,MAAMH,cACJ5E,AAAoB,UAApBA,MAAM,SAAS,GACX+E,eAAeC,iBACf3B,KAAK,GAAG,CAAC,IAAI0B,eAAeC;IAElC,OAAO;QAAEP;QAASC;QAASC;QAAeC;QAAatB;IAAS;AAClE;AAgCA,SAAS2B,wBACPvF,MAA+B,EAC/BwF,IAAyB;IAEzB,OAAOzF,aAAyC;QAC9C,MAAMC,OAAO,IAAI;QACjB,aAAaA,OAAO,WAAW;QAC/B,gBAAgBA,OAAO,cAAc;QACrC,mBAAmBA,OAAO,iBAAiB;QAC3C,kBAAkBA,OAAO,gBAAgB;QACzCwF;IACF;AACF;AAEO,SAASC,iCACd5F,KAAsB,EACtB6F,UAAuC,CAAC,CAAC;IAEzC,MAAMC,UAAgD,EAAE;IACxD,MAAM,EAAEC,OAAO,EAAE7D,QAAQ,EAAEM,MAAM,EAAEwD,KAAK,EAAEC,MAAM,EAAE,GAAGjG;IAErD,IAAI+F,SAAS;QACXD,QAAQ,IAAI,CAACjF,gBAAgBkF,QAAQ,GAAG;QACxC,IAAIA,QAAQ,WAAW,EACrBD,QAAQ,IAAI,CAACvE,wBAAwBwE,QAAQ,WAAW;QAE1D,IAAIA,QAAQ,UAAU,EACpBD,QAAQ,IAAI,CAAC1E,uBAAuB2E,QAAQ,UAAU;QAExD,IAAIA,QAAQ,KAAK,EACfD,QAAQ,IAAI,CAACpE,kBAAkBqE,QAAQ,KAAK;QAE9C,IAAIA,QAAQ,WAAW,EACrBD,QAAQ,IAAI,CAACpD,wBAAwBqD,QAAQ,WAAW;QAE1D,IAAIA,QAAQ,SAAS,EACnBD,QAAQ,IAAI,CAAC/C,sBAAsBgD,QAAQ,SAAS;IAExD;IAEA,IAAI7D,UACF4D,QAAQ,IAAI,CACV7D,kBAAkBC,WAClBmC,uBAAuBnC,SAAS,UAAU,GAC1CE,0BAA0BF,SAAS,aAAa,GAChDsC,uBAAuB;QAAEtC;QAAU,OAAO2D,QAAQ,KAAK;IAAC;IAI5D,IAAIrD,QACFsD,QAAQ,IAAI,CAACvD,mBAAmBC,OAAO,MAAM;IAG/C,MAAM0D,cAAcnG,2BAA2B;QAAEgG;QAASC;IAAM;IAChE,IAAIE,eAAeL,QAAQ,IAAI,IAAIA,AAAyB,UAAzBA,QAAQ,YAAY,EACrDC,QAAQ,IAAI,CAAC5B,kBAAkB;QAAE,GAAGgC,WAAW;QAAE,MAAML,QAAQ,IAAI;IAAC;IAGtE,IAAIG,OAAO,SAASH,QAAQ,IAAI,IAAIA,AAAyB,UAAzBA,QAAQ,YAAY,EACtDC,QAAQ,IAAI,CAACb,kBAAkB;QAAE,OAAOe,MAAM,KAAK;QAAE,MAAMH,QAAQ,IAAI;IAAC;IAG1E,IAAII,UAAUJ,QAAQ,aAAa,EAAE;QACnC,MAAM,EAAEM,aAAa,EAAE,GAAGN;QAC1B,IAAII,OAAO,UAAU,IAAIE,cAAc,UAAU,EAC/CL,QAAQ,IAAI,CACVJ,wBAAwBS,cAAc,UAAU,EAAEF,OAAO,UAAU;QAGvE,IAAIA,OAAO,UAAU,IAAIE,cAAc,UAAU,EAC/CL,QAAQ,IAAI,CACVJ,wBAAwBS,cAAc,UAAU,EAAEF,OAAO,UAAU;QAGvE,IAAIA,OAAO,gBAAgB,IAAIE,cAAc,gBAAgB,EAC3DL,QAAQ,IAAI,CACVJ,wBACES,cAAc,gBAAgB,EAC9BF,OAAO,gBAAgB;IAI/B;IAEA,OAAOH,QAAQ,MAAM,CAAC,CAACM,SACrBC,QAAQD;AAEZ;AAEO,SAASE,2BACdC,OAAiC;IAEjC,OAAOX,iCAAiCW,QAAQ,KAAK,EAAE;QACrD,MAAMA,QAAQ,IAAI;QAClB,OAAOA,QAAQ,KAAK;QACpB,eAAeA,QAAQ,aAAa;IACtC;AACF;AAGO,MAAMC,yBAAyB7F,EAAE,MAAM,CAAC;IAC7C,QAAQA,EAAAA,MACC,GACN,OAAO,CAAC,MACR,QAAQ,GACR,QAAQ,CAAC;AACd;AAMO,MAAM8F,oBAAoB,CAC/BC,cAEOxG,aAA8D;QACnE,MAAM;QACN,aACE;QACF,aAAasG;QACb,QAAQ;YACN,QAAQ;QACV;QACA,MAAM,OAAO/F;YACX,MAAMsD,WAAWtD,OAAO,UAAU;YAClCkG,SAAS,wBAAwB,CAAC,aAAa,EAAE5C,SAAS,EAAE,CAAC;YAC7D2C,aAAa;YACb,MAAM,IAAI9B,QAAc,CAACC,SAAS+B;gBAChC,MAAMC,QAAQ;oBACZC,aAAaC;oBACbH,OAAOF,aAAa;gBACtB;gBACA,MAAMK,QAAQjC,WAAW;oBACvB4B,aAAa,oBAAoB,SAASG;oBAC1ChC;gBACF,GAAGd;gBACH2C,aAAa,iBAAiB,SAASG,OAAO;oBAAE,MAAM;gBAAK;YAC7D;QACF;IACF"}
+{"version":3,"file":"device/index.mjs","sources":["../../../src/device/index.ts"],"sourcesContent":["import type { ModelRuntime } from '@/ai-model/models';\nimport { getMidsceneLocationSchema } from '@/common';\nimport type {\n  ActionScrollParam,\n  DeviceAction,\n  ExecutorContext,\n  LocateResultElement,\n} from '@/types';\nimport type { ElementNode } from '@midscene/shared/extractor';\nimport { getDebug } from '@midscene/shared/logger';\nimport { _keyDefinitions } from '@midscene/shared/us-keyboard-layout';\nimport { z } from 'zod';\nimport type {\n  ElementCacheFeature,\n  Rect,\n  Size,\n  UIContext,\n  UITreeSnapshot,\n} from '../types';\nimport { type InputStrategy, inputStrategies } from './input-strategy';\n\nexport interface FileChooserHandler {\n  accept(files: string[]): Promise<void>;\n}\n\nexport interface FileChooserRegistration {\n  dispose: () => void;\n  getError: () => Error | undefined | Promise<Error | undefined>;\n}\n\nexport interface MjpegStreamFrame {\n  /** Raw base64-encoded image bytes WITHOUT a `data:image/...;base64,` prefix. */\n  data: string;\n  contentType?: string;\n}\n\nexport interface MjpegStreamHandle {\n  stop(): void | Promise<void>;\n}\n\nexport interface MjpegStreamOptions {\n  signal?: AbortSignal;\n  onFrame(frame: MjpegStreamFrame): void;\n  onError?(error: unknown): void;\n}\n\n/**\n * A cheap, not-yet-decoded handle to one screen frame from a\n * {@link DeviceFrameSource}. `ref` is platform-specific (a raw H.264 keyframe\n * buffer on Android, an already-encoded JPEG data URL on iOS/web) and must not\n * be interpreted by callers — pass it back to `decode()` to materialize.\n */\nexport interface DeviceFrameRef {\n  ref: unknown;\n  capturedAt: number;\n}\n\n/**\n * A continuous screen-frame source opened via\n * {@link AbstractInterface.openFrameSource}. Designed for deferred decoding:\n * grabbing `latest()` is near-zero cost, so observers can sample at a steady\n * cadence and pay any decode cost only once, for the frames they keep.\n */\nexport interface DeviceFrameSource {\n  /** Latest frame handle, near-zero cost. Null until the first frame arrives. */\n  latest(): DeviceFrameRef | null;\n  /**\n   * Materialize frame handles into `data:image/...;base64,` URLs, preserving\n   * order. Possibly expensive (e.g. one ffmpeg run per unique frame on\n   * Android) — call once with the sampled handles, never per tick.\n   */\n  decode(refs: DeviceFrameRef[]): Promise<string[]>;\n  /** Release the source (stop streams/subscriptions it started). */\n  stop(): Promise<void> | void;\n}\n\n/** A point in device-pixel coordinates on the screen. */\nexport interface PointerPoint {\n  x: number;\n  y: number;\n}\n\n/** Low-level swipe gesture implemented by touch or pointer-backed devices. */\nexport type SwipeInputPrimitive = (\n  start: PointerPoint,\n  end: PointerPoint,\n  opts?: { duration?: number; repeat?: number },\n) => Promise<void>;\n\n/** The physical input source used to execute and describe a swipe. */\nexport type SwipeInputMode = 'touch' | 'mouse';\n\n/** A resolved swipe capability and the input semantics exposed to the model. */\nexport interface ResolvedSwipeInputPrimitive {\n  swipe: SwipeInputPrimitive;\n  inputMode: SwipeInputMode;\n}\n\nexport interface PointerInputPrimitives {\n  tap(p: PointerPoint, opts?: { duration?: number }): Promise<void>;\n  doubleClick?(p: PointerPoint): Promise<void>;\n  rightClick?(p: PointerPoint): Promise<void>;\n  hover?(p: PointerPoint): Promise<void>;\n  longPress?(p: PointerPoint, opts?: { duration?: number }): Promise<void>;\n  dragAndDrop?(from: PointerPoint, to: PointerPoint): Promise<void>;\n  swipe?: SwipeInputPrimitive;\n}\n\nexport interface TouchInputPrimitives {\n  swipe: SwipeInputPrimitive;\n  pinch?(\n    center: PointerPoint,\n    opts: { startDistance: number; endDistance: number; duration: number },\n  ): Promise<void>;\n}\n\nexport interface KeyboardInputPrimitives {\n  keyboardPress(keyName: string, opts?: { target?: unknown }): Promise<void>;\n  cursorMove?(direction: 'left' | 'right', times?: number): Promise<void>;\n  typeText(\n    value: string,\n    opts?: {\n      autoDismissKeyboard?: boolean;\n      keyboardDismissStrategy?: 'esc-first' | 'back-first';\n      keyboardTypeDelay?: number;\n      inputStrategy?: InputStrategy;\n      target?: unknown;\n      replace?: boolean;\n      focusOnly?: boolean;\n      /** @internal Custom readiness replaces optional post-input settling. */\n      skipDefaultWait?: boolean;\n    },\n  ): Promise<void>;\n  clearInput(target?: unknown): Promise<void>;\n}\n\nexport interface ScrollInputPrimitives {\n  scroll(\n    param: ActionScrollParam,\n    opts?: { skipDefaultWait?: boolean },\n  ): Promise<void>;\n}\n\nexport interface SystemInputPrimitives {\n  backButton?(): Promise<void>;\n  homeButton?(): Promise<void>;\n  recentAppsButton?(): Promise<void>;\n}\n\nexport interface InputPrimitives {\n  pointer?: PointerInputPrimitives;\n  keyboard?: KeyboardInputPrimitives;\n  touch?: TouchInputPrimitives;\n  scroll?: ScrollInputPrimitives;\n  system?: SystemInputPrimitives;\n}\n\n/**\n * Resolve the canonical swipe capability for an input collection.\n * Touch remains preferred for devices that expose both capabilities; desktop\n * devices can fall back to a pointer-backed primary-button gesture.\n */\nexport function resolveSwipeInputPrimitive(\n  input: Pick<InputPrimitives, 'pointer' | 'touch'>,\n): ResolvedSwipeInputPrimitive | undefined {\n  if (input.touch?.swipe) {\n    return { swipe: input.touch.swipe, inputMode: 'touch' };\n  }\n  if (input.pointer?.swipe) {\n    return { swipe: input.pointer.swipe, inputMode: 'mouse' };\n  }\n  return undefined;\n}\n\nexport interface MobileInputPrimitives extends InputPrimitives {\n  pointer: PointerInputPrimitives & {\n    doubleClick(p: PointerPoint): Promise<void>;\n    longPress(p: PointerPoint, opts?: { duration?: number }): Promise<void>;\n    dragAndDrop(from: PointerPoint, to: PointerPoint): Promise<void>;\n  };\n  keyboard: KeyboardInputPrimitives;\n  touch: TouchInputPrimitives;\n}\n\nexport interface BrowserInputPrimitives extends InputPrimitives {\n  pointer: PointerInputPrimitives & {\n    doubleClick(p: PointerPoint): Promise<void>;\n    rightClick(p: PointerPoint): Promise<void>;\n    hover(p: PointerPoint): Promise<void>;\n    dragAndDrop(from: PointerPoint, to: PointerPoint): Promise<void>;\n    longPress(p: PointerPoint, opts?: { duration?: number }): Promise<void>;\n  };\n  keyboard: KeyboardInputPrimitives;\n  scroll: ScrollInputPrimitives;\n  touch: TouchInputPrimitives;\n}\n\nexport interface ComputerInputPrimitives extends InputPrimitives {\n  pointer: PointerInputPrimitives & {\n    doubleClick(p: PointerPoint): Promise<void>;\n    rightClick(p: PointerPoint): Promise<void>;\n    hover(p: PointerPoint): Promise<void>;\n    dragAndDrop(from: PointerPoint, to: PointerPoint): Promise<void>;\n  };\n  keyboard: KeyboardInputPrimitives;\n  scroll: ScrollInputPrimitives;\n}\n\nexport abstract class AbstractInterface {\n  abstract interfaceType: string;\n\n  abstract screenshotBase64(): Promise<string>;\n  abstract size(): Promise<Size>;\n  abstract actionSpace(): DeviceAction[];\n\n  abstract cacheFeatureForPoint?(\n    center: [number, number],\n    options?: {\n      targetDescription?: string;\n      modelRuntime?: ModelRuntime;\n    },\n  ): Promise<ElementCacheFeature>;\n  abstract rectMatchesCacheFeature?(\n    feature: ElementCacheFeature,\n  ): Promise<Rect>;\n\n  abstract getUITree?(): Promise<UITreeSnapshot>;\n\n  abstract destroy?(): Promise<void>;\n\n  abstract describe?(): string;\n  abstract beforeInvokeAction?(actionName: string, param: any): Promise<void>;\n  abstract afterInvokeAction?(actionName: string, param: any): Promise<void>;\n\n  /** Platform readiness heuristics, skipped when a custom action waiter takes over. */\n  async defaultActionWait?(_actionName: string, _param: any): Promise<void> {}\n\n  // for web only\n  registerFileChooserListener?(\n    handler: (chooser: FileChooserHandler) => Promise<void>,\n  ): Promise<FileChooserRegistration>;\n\n  // @deprecated do NOT extend this method\n  abstract getElementsNodeTree?: () => Promise<ElementNode>;\n\n  // @deprecated do NOT extend this method\n  abstract url?: () => string | Promise<string>;\n\n  // @deprecated do NOT extend this method\n  abstract evaluateJavaScript?<T = any>(script: string): Promise<T>;\n\n  /**\n   * Get the current device-local time as a formatted string.\n   * Prefer this for user-visible time because timestamps alone do not preserve\n   * the target device's timezone when formatted on the host machine.\n   */\n  getDeviceLocalTimeString?(format?: string): Promise<string>;\n\n  /** URL of native MJPEG stream for real-time screen preview (e.g. WDA MJPEG server) */\n  mjpegStreamUrl?: string;\n\n  /**\n   * Optional continuous frame source for UI observation (`startObserving`).\n   * Devices that maintain a continuous frame stream — scrcpy on Android, WDA\n   * MJPEG on iOS, CDP screencast on web — implement this so an observer can\n   * sample the screen far faster than repeated `screenshotBase64()` calls,\n   * catching short-lived UI (toasts, carousels, transitions).\n   *\n   * The contract enables DEFERRED decoding: `latest()` returns a cheap opaque\n   * handle (e.g. a raw H.264 keyframe on Android) with no per-frame decode\n   * cost, and `decode()` materializes only the handles that were actually\n   * sampled — once, at the end of the observation window.\n   */\n  openFrameSource?(): Promise<DeviceFrameSource | undefined>;\n\n  /**\n   * Optional in-process MJPEG frame producer. Implementations can push raw\n   * base64 frames here when there is no standalone native MJPEG URL, e.g.\n   * Chromium CDP Page.startScreencast for web previews.\n   */\n  startMjpegStream?(\n    options: MjpegStreamOptions,\n  ): MjpegStreamHandle | undefined | Promise<MjpegStreamHandle | undefined>;\n\n  /**\n   * Optional hook used after a UI action to push a fresh frame on the active\n   * MJPEG stream. Set `force` after navigation to replace a transient loading\n   * frame even when the screencast has already emitted one. Implementations\n   * should be a no-op when no stream is active.\n   */\n  flushPendingVisualUpdate?(force?: boolean): Promise<void>;\n\n  /**\n   * Optional non-blocking variant of `flushPendingVisualUpdate`. Keyboard-\n   * heavy preview interactions can schedule a coalesced refresh here without\n   * stalling the input hot path; `force` preserves a requested navigation\n   * refresh while work is already queued.\n   */\n  schedulePendingVisualUpdate?(force?: boolean): void;\n\n  /**\n   * Optional navigation state probe for browser-like interfaces, used to drive\n   * loading indicators in playground UIs. Returning `undefined` means the\n   * interface does not expose this concept.\n   */\n  navigationState?(): Promise<{ isLoading: boolean }>;\n\n  /**\n   * Low-level device input surface. Platform implementations expose transport\n   * primitives here; higher-level AI actions and manual pointer dispatch should\n   * adapt to this instead of duplicating platform gesture logic.\n   */\n  inputPrimitives?: InputPrimitives;\n}\n\n// Generic function to define actions with proper type inference\n// TRuntime allows specifying a different type for the runtime parameter (after location resolution)\n// TReturn allows specifying the return type of the action\nexport const defineAction = <\n  TSchema extends z.ZodType | undefined = undefined,\n  TRuntime = TSchema extends z.ZodType ? z.infer<TSchema> : undefined,\n  TReturn = any,\n>(\n  config: {\n    name: string;\n    description: string;\n    interfaceAlias?: string;\n    paramSchema?: TSchema;\n    call: (\n      param: TRuntime,\n      context?: ExecutorContext,\n    ) => Promise<TReturn> | TReturn;\n  } & Partial<\n    Omit<\n      DeviceAction<TRuntime, TReturn>,\n      'name' | 'description' | 'interfaceAlias' | 'paramSchema' | 'call'\n    >\n  >,\n): DeviceAction<TRuntime, TReturn> => {\n  return config as any; // Type assertion needed because schema validation type differs from runtime type\n};\n\nfunction pointFromLocate(\n  locate: LocateResultElement | undefined,\n  missingMessage: string,\n): PointerPoint {\n  if (!locate) {\n    throw new Error(missingMessage);\n  }\n  return { x: locate.center[0], y: locate.center[1] };\n}\n\nfunction defineLocatedPointAction<\n  TSchema extends z.ZodType,\n  TParam extends { locate: LocateResultElement },\n>(config: {\n  name: string;\n  description: string;\n  interfaceAlias?: string;\n  paramSchema: TSchema;\n  sample: DeviceAction<TParam>['sample'];\n  missingLocateMessage: string;\n  call: (point: PointerPoint, param: TParam) => Promise<void>;\n}): DeviceAction<TParam> {\n  return defineAction<TSchema, TParam>({\n    name: config.name,\n    description: config.description,\n    interfaceAlias: config.interfaceAlias,\n    paramSchema: config.paramSchema,\n    sample: config.sample,\n    call: async (param) => {\n      await config.call(\n        pointFromLocate(param.locate, config.missingLocateMessage),\n        param,\n      );\n    },\n  });\n}\n\n// Tap\nexport const actionTapParamSchema = z.object({\n  locate: getMidsceneLocationSchema().describe('The element to be tapped'),\n});\nexport type ActionTapParam = {\n  locate: LocateResultElement;\n};\n\nexport const defineActionTap = (\n  tap: PointerInputPrimitives['tap'],\n): DeviceAction<ActionTapParam> => {\n  return defineLocatedPointAction<typeof actionTapParamSchema, ActionTapParam>({\n    name: 'Tap',\n    description: 'Tap the element',\n    interfaceAlias: 'aiTap',\n    paramSchema: actionTapParamSchema,\n    sample: {\n      locate: { prompt: 'the \"Submit\" button' },\n    },\n    missingLocateMessage: 'Element not found, cannot tap',\n    call: async (point) => {\n      await tap(point);\n    },\n  });\n};\n\nexport const registerFileChooserAcceptParamSchema = z.object({\n  files: z\n    .union([z.string(), z.array(z.string())])\n    .describe(\n      \"File path(s) within the current aiAct call's fileChooserAllowedDir to use whenever a later action triggers a file chooser. fileChooserAllowedDir must be provided for this action. This setting replaces any previously registered file path(s).\",\n    ),\n});\nexport type RegisterFileChooserAcceptParam = {\n  files: string | string[];\n};\n\nexport const defineActionRegisterFileChooserAccept = (\n  register: (files: string | string[]) => Promise<void>,\n): DeviceAction<RegisterFileChooserAcceptParam> => {\n  return defineAction<\n    typeof registerFileChooserAcceptParamSchema,\n    RegisterFileChooserAcceptParam\n  >({\n    name: 'RegisterFileChooserAccept',\n    description:\n      'Configure files for file chooser dialogs triggered by later actions in this aiAct',\n    interfaceAlias: 'registerFileChooserAccept',\n    paramSchema: registerFileChooserAcceptParamSchema,\n    sample: {\n      files: ['fixtures/document.pdf'],\n    },\n    call: async (param) => {\n      await register(param.files);\n    },\n  });\n};\n\n// RightClick\nexport const actionRightClickParamSchema = z.object({\n  locate: getMidsceneLocationSchema().describe(\n    'The element to be right clicked',\n  ),\n});\nexport type ActionRightClickParam = {\n  locate: LocateResultElement;\n};\n\nexport const defineActionRightClick = (\n  rightClick: NonNullable<PointerInputPrimitives['rightClick']>,\n): DeviceAction<ActionRightClickParam> => {\n  return defineLocatedPointAction<\n    typeof actionRightClickParamSchema,\n    ActionRightClickParam\n  >({\n    name: 'RightClick',\n    description: 'Right click the element',\n    interfaceAlias: 'aiRightClick',\n    paramSchema: actionRightClickParamSchema,\n    sample: {\n      locate: { prompt: 'the file icon on the desktop' },\n    },\n    missingLocateMessage: 'Element not found, cannot right click',\n    call: async (point) => {\n      await rightClick(point);\n    },\n  });\n};\n\n// DoubleClick\nexport const actionDoubleClickParamSchema = z.object({\n  locate: getMidsceneLocationSchema().describe(\n    'The element to be double clicked',\n  ),\n});\nexport type ActionDoubleClickParam = {\n  locate: LocateResultElement;\n};\n\nexport const defineActionDoubleClick = (\n  doubleClick: NonNullable<PointerInputPrimitives['doubleClick']>,\n): DeviceAction<ActionDoubleClickParam> => {\n  return defineLocatedPointAction<\n    typeof actionDoubleClickParamSchema,\n    ActionDoubleClickParam\n  >({\n    name: 'DoubleClick',\n    description: 'Double click the element',\n    interfaceAlias: 'aiDoubleClick',\n    paramSchema: actionDoubleClickParamSchema,\n    sample: {\n      locate: { prompt: 'the folder icon' },\n    },\n    missingLocateMessage: 'Element not found, cannot double click',\n    call: async (point) => {\n      await doubleClick(point);\n    },\n  });\n};\n\n// Hover\nexport const actionHoverParamSchema = z.object({\n  locate: getMidsceneLocationSchema().describe('The element to be hovered'),\n});\nexport type ActionHoverParam = {\n  locate: LocateResultElement;\n};\n\nexport const defineActionHover = (\n  hover: NonNullable<PointerInputPrimitives['hover']>,\n): DeviceAction<ActionHoverParam> => {\n  return defineLocatedPointAction<\n    typeof actionHoverParamSchema,\n    ActionHoverParam\n  >({\n    name: 'Hover',\n    description: 'Move the mouse to the element',\n    interfaceAlias: 'aiHover',\n    paramSchema: actionHoverParamSchema,\n    sample: {\n      locate: { prompt: 'the navigation menu item \"Products\"' },\n    },\n    missingLocateMessage: 'Element not found, cannot hover',\n    call: async (point) => {\n      await hover(point);\n    },\n  });\n};\n\n// Input\nconst inputLocateDescription =\n  'the position of the placeholder or text content in the target input field. If there is no content, locate the center of the input field.';\nexport const actionInputParamSchema = z.object({\n  value: z\n    .union([z.string(), z.number()])\n    .transform((val) => String(val))\n    .describe(\n      'The text to input. Provide the final content for replace mode, only the inserted characters for typeOnly mode, or an empty string when using clear mode to remove existing text.',\n    ),\n  locate: getMidsceneLocationSchema()\n    .describe(inputLocateDescription)\n    .optional(),\n  mode: z\n    .enum(['replace', 'clear', 'typeOnly'])\n    .default('replace')\n    .describe(\n      'Input mode: \"replace\" (default) - clear the field and input the value; \"typeOnly\" - type the value directly without clearing the field first, and should be set explicitly for incremental edits after moving the cursor; \"clear\" - clear the field without inputting new text.',\n    ),\n  autoDismissKeyboard: z\n    .boolean()\n    .optional()\n    .describe(\n      'If true, the keyboard will be dismissed after the input is completed. Do not set it unless the user asks you to do so.',\n    ),\n  keyboardTypeDelay: z\n    .number()\n    .finite()\n    .nonnegative()\n    .optional()\n    .describe(\n      'Delay in milliseconds between keystrokes when typing. Passed through from device/user configuration. Do not set it unless the user asks you to do so.',\n    ),\n  inputStrategy: z\n    .enum(inputStrategies)\n    .optional()\n    .describe(\n      'Text input strategy: \"legacy\" (default) preserves the current platform behavior; \"sequential\" enters one Unicode code point at a time; \"bulk\" sends the complete text through one platform input operation when supported and requires keyboardTypeDelay to be omitted or set to 0. Do not set it unless the user asks you to do so.',\n    ),\n});\nexport type ActionInputParam = {\n  value: string;\n  locate?: LocateResultElement;\n  mode?: 'replace' | 'clear' | 'typeOnly' | 'append';\n  autoDismissKeyboard?: boolean;\n  keyboardTypeDelay?: number;\n  inputStrategy?: InputStrategy;\n};\n\nexport const defineActionInput = (\n  keyboard: KeyboardInputPrimitives,\n): DeviceAction<ActionInputParam> => {\n  return defineAction<typeof actionInputParamSchema, ActionInputParam>({\n    name: 'Input',\n    description: 'Input the value into the element',\n    interfaceAlias: 'aiInput',\n    paramSchema: actionInputParamSchema,\n    sample: {\n      value: 'test@example.com',\n      locate: { prompt: 'the email input field' },\n    },\n    call: async (param, context) => {\n      // backward compat: convert deprecated 'append' to 'typeOnly'\n      if ((param.mode as string) === 'append') {\n        param.mode = 'typeOnly';\n      }\n\n      if (param.mode === 'clear') {\n        await keyboard.clearInput(param.locate);\n        return;\n      }\n\n      if (!param || !param.value) {\n        return;\n      }\n\n      await keyboard.typeText(param.value, {\n        target: param.locate,\n        replace: param.mode !== 'typeOnly',\n        autoDismissKeyboard: param.autoDismissKeyboard,\n        keyboardTypeDelay: param.keyboardTypeDelay,\n        inputStrategy: param.inputStrategy,\n        ...(context?.skipDefaultWait ? { skipDefaultWait: true } : {}),\n      });\n    },\n  });\n};\n\n// KeyboardPress\nexport const actionKeyboardPressParamSchema = z.object({\n  locate: getMidsceneLocationSchema()\n    .describe(\n      'The optional element to click before pressing the key. Omit this when the key should operate on the currently focused element, especially when copying or cutting an existing text selection.',\n    )\n    .optional(),\n  keyName: z\n    .string()\n    .describe(\n      \"The key to be pressed. Use '+' for key combinations, e.g., 'Control+A', 'Shift+Enter'\",\n    ),\n});\nexport type ActionKeyboardPressParam = {\n  locate?: LocateResultElement;\n  keyName: string;\n};\n\nexport const defineActionKeyboardPress = (\n  keyboardPress: KeyboardInputPrimitives['keyboardPress'],\n): DeviceAction<ActionKeyboardPressParam> => {\n  return defineAction<\n    typeof actionKeyboardPressParamSchema,\n    ActionKeyboardPressParam\n  >({\n    name: 'KeyboardPress',\n    description:\n      'Press a key or key combination, like \"Enter\", \"Tab\", \"Escape\", or \"Control+A\", \"Shift+Enter\". Do not use this to type text. Omit locate to operate on the current focus without clicking again, especially for Copy or Cut after text has been selected.',\n    interfaceAlias: 'aiKeyboardPress',\n    paramSchema: actionKeyboardPressParamSchema,\n    sample: {\n      keyName: 'Enter',\n    },\n    call: async (param) => {\n      await keyboardPress(param.keyName, {\n        target: param.locate,\n      });\n    },\n  });\n};\n\n// Scroll\nexport const actionScrollParamSchema = z.object({\n  scrollType: z\n    .enum([\n      'singleAction',\n      'scrollToBottom',\n      'scrollToTop',\n      'scrollToRight',\n      'scrollToLeft',\n    ])\n    .default('singleAction')\n    .describe(\n      'The scroll behavior: \"singleAction\" for a single scroll action, \"scrollToBottom\" for scrolling all the way to the bottom by rapidly scrolling 5-10 times (skipping intermediate content until reaching the bottom), \"scrollToTop\" for scrolling all the way to the top by rapidly scrolling 5-10 times (skipping intermediate content until reaching the top), \"scrollToRight\" for scrolling all the way to the right by rapidly scrolling multiple times, \"scrollToLeft\" for scrolling all the way to the left by rapidly scrolling multiple times',\n    ),\n  direction: z\n    .enum(['down', 'up', 'right', 'left'])\n    .default('down')\n    .describe(\n      'The direction toward the off-screen content to reveal. \"down\" reveals content below the current viewport, \"up\" reveals content above, \"right\" reveals content to the right, and \"left\" reveals content to the left. This does not describe the movement direction of the content currently visible on the screen. Only effective when scrollType is \"singleAction\".',\n    ),\n  distance: z\n    .number()\n    .positive()\n    .nullable()\n    .optional()\n    .describe(\n      'Positive requested scroll amount in screen-coordinate units. On touch platforms, it controls the length of the swipe gesture used to scroll; on web and desktop platforms, it controls or approximates the wheel scroll delta. Only effective when scrollType is \"singleAction\".',\n    ),\n  locate: getMidsceneLocationSchema()\n    .optional()\n    .describe(\n      'Describe the target element to be scrolled on, like \"the table\" or \"the list\" or \"the content area\" or \"the scrollable area\". Do NOT provide a general intent like \"scroll to find some element\"',\n    ),\n});\n\nexport const defineActionScroll = (\n  scroll: ScrollInputPrimitives['scroll'],\n): DeviceAction<ActionScrollParam> => {\n  return defineAction<typeof actionScrollParamSchema, ActionScrollParam>({\n    name: 'Scroll',\n    description:\n      'Scroll a page or scrollable region to reveal off-screen content. Use Scroll when the goal is to browse content outside the current viewport. For direct gesture interactions, such as adjusting a slider or wheel picker, switching between paged cards or images, following an on-screen swipe gesture to continue or dismiss, or swiping an item to delete it, use Swipe instead if available in the current Action Space. Supports scrollToBottom/scrollToTop for boundary navigation. Default: direction `down`, scrollType `singleAction`, distance `null`.',\n    interfaceAlias: 'aiScroll',\n    paramSchema: actionScrollParamSchema,\n    sample: {\n      direction: 'down',\n      scrollType: 'singleAction',\n      locate: { prompt: 'the center of the product list area' },\n    },\n    call: async (param, context) => {\n      if (context?.skipDefaultWait)\n        await scroll(param, { skipDefaultWait: true });\n      else await scroll(param);\n    },\n  });\n};\n\n// DragAndDrop\nexport const actionDragAndDropParamSchema = z.object({\n  from: getMidsceneLocationSchema().describe('The position to be dragged'),\n  to: getMidsceneLocationSchema().describe('The position to be dropped'),\n});\nexport type ActionDragAndDropParam = {\n  from: LocateResultElement;\n  to: LocateResultElement;\n};\n\nexport const defineActionDragAndDrop = (\n  dragAndDrop: NonNullable<PointerInputPrimitives['dragAndDrop']>,\n): DeviceAction<ActionDragAndDropParam> => {\n  return defineAction<\n    typeof actionDragAndDropParamSchema,\n    ActionDragAndDropParam\n  >({\n    name: 'DragAndDrop',\n    description:\n      'Pick up a specific UI element and move it to a new position (e.g., reorder a card, move a file into a folder, sort list items). The element itself moves with your finger/mouse.',\n    interfaceAlias: 'aiDragAndDrop',\n    paramSchema: actionDragAndDropParamSchema,\n    sample: {\n      from: { prompt: 'the \"report.pdf\" file icon' },\n      to: { prompt: 'the upload drop zone' },\n    },\n    call: async (param) => {\n      const from = param.from;\n      const to = param.to;\n      if (!from) {\n        throw new Error('missing \"from\" param for drag and drop');\n      }\n      if (!to) {\n        throw new Error('missing \"to\" param for drag and drop');\n      }\n      await dragAndDrop(\n        { x: from.center[0], y: from.center[1] },\n        { x: to.center[0], y: to.center[1] },\n      );\n    },\n  });\n};\n\nexport const ActionLongPressParamSchema = z.object({\n  locate: getMidsceneLocationSchema().describe(\n    'The element to be long pressed',\n  ),\n  duration: z\n    .number()\n    .optional()\n    .describe('Long press duration in milliseconds'),\n});\n\nexport type ActionLongPressParam = {\n  locate: LocateResultElement;\n  duration?: number;\n};\nexport const defineActionLongPress = (\n  longPress: NonNullable<PointerInputPrimitives['longPress']>,\n): DeviceAction<ActionLongPressParam> => {\n  return defineLocatedPointAction<\n    typeof ActionLongPressParamSchema,\n    ActionLongPressParam\n  >({\n    name: 'LongPress',\n    description: 'Long press the element',\n    interfaceAlias: 'aiLongPress',\n    paramSchema: ActionLongPressParamSchema,\n    sample: {\n      locate: { prompt: 'the message bubble' },\n    },\n    missingLocateMessage: 'LongPress requires an element to be located',\n    call: async (point, param) => {\n      await longPress(point, { duration: param.duration });\n    },\n  });\n};\n\nfunction createActionSwipeParamSchema(inputMode: SwipeInputMode) {\n  const movementSource = inputMode === 'touch' ? 'finger' : 'pointer';\n\n  return z.object({\n    start: getMidsceneLocationSchema()\n      .optional()\n      .describe(\n        `Optional starting point of the ${movementSource} movement. Available in both relative and endpoint forms. If omitted, the center of the page is used.`,\n      ),\n    direction: z\n      .enum(['up', 'down', 'left', 'right'])\n      .optional()\n      .describe(\n        `${movementSource === 'finger' ? 'Finger' : 'Pointer'} movement direction. Required together with a positive distance for a relative swipe. Omit when using end.`,\n      ),\n    distance: z\n      .number()\n      .positive()\n      .optional()\n      .describe(\n        `Positive length of the ${movementSource} movement in pixels. Required together with direction for a relative swipe. Omit when using end.`,\n      ),\n    end: getMidsceneLocationSchema()\n      .optional()\n      .describe(\n        `Endpoint of the ${movementSource} movement. Use for an endpoint swipe, optionally with start. Do not provide direction or distance when using end.`,\n      ),\n    duration: z\n      .number()\n      .default(300)\n      .describe('Duration of the swipe gesture in milliseconds'),\n    repeat: z\n      .number()\n      .optional()\n      .describe(\n        'The number of times to repeat the swipe gesture. 1 for default, 0 for infinite (e.g. endless swipe until the end of the page)',\n      ),\n  });\n}\n\n// Preserve the historical touch-oriented public schema. Desktop actions use\n// an equivalent schema with mouse-specific descriptions at definition time.\nexport const ActionSwipeParamSchema = createActionSwipeParamSchema('touch');\n\nexport type ActionSwipeParam = {\n  start?: LocateResultElement;\n  direction?: 'up' | 'down' | 'left' | 'right';\n  distance?: number;\n  end?: LocateResultElement;\n  duration?: number;\n  repeat?: number;\n};\n\nexport function normalizeSwipeParam(\n  param: ActionSwipeParam,\n  screenSize: { width: number; height: number },\n): {\n  startPoint: { x: number; y: number };\n  endPoint: { x: number; y: number };\n  duration: number;\n  repeatCount: number;\n} {\n  const { width, height } = screenSize;\n  const { start, end } = param;\n\n  const startPoint = start\n    ? { x: start.center[0], y: start.center[1] }\n    : { x: width / 2, y: height / 2 };\n\n  const endPoint = (() => {\n    if (end) {\n      if (param.direction !== undefined || param.distance !== undefined) {\n        throw new Error(\n          'Invalid Swipe parameters: \"end\" cannot be combined with \"direction\" or \"distance\". Use \"end\", optionally with \"start\", for an endpoint swipe.',\n        );\n      }\n\n      return { x: end.center[0], y: end.center[1] };\n    }\n\n    if (param.direction === undefined || param.distance === undefined) {\n      throw new Error(\n        'Invalid Swipe parameters: a relative swipe requires both \"direction\" and a positive \"distance\".',\n      );\n    }\n\n    if (param.distance <= 0) {\n      throw new Error(\n        'Invalid Swipe parameters: \"distance\" must be a positive number.',\n      );\n    }\n\n    return {\n      x:\n        startPoint.x +\n        (param.direction === 'right'\n          ? param.distance\n          : param.direction === 'left'\n            ? -param.distance\n            : 0),\n      y:\n        startPoint.y +\n        (param.direction === 'down'\n          ? param.distance\n          : param.direction === 'up'\n            ? -param.distance\n            : 0),\n    };\n  })();\n\n  endPoint.x = Math.max(0, Math.min(endPoint.x, width));\n  endPoint.y = Math.max(0, Math.min(endPoint.y, height));\n\n  const duration = param.duration ?? 300;\n\n  let repeatCount = typeof param.repeat === 'number' ? param.repeat : 1;\n  if (repeatCount === 0) {\n    repeatCount = 10;\n  }\n\n  return { startPoint, endPoint, duration, repeatCount };\n}\n\n/** @deprecated Use {@link normalizeSwipeParam} instead. */\nexport const normalizeMobileSwipeParam = normalizeSwipeParam;\n\nexport const defineActionSwipe = (config: {\n  swipe: SwipeInputPrimitive;\n  size(): Promise<Size>;\n  inputMode?: SwipeInputMode;\n}): DeviceAction<ActionSwipeParam> => {\n  const inputMode = config.inputMode ?? 'touch';\n  return defineAction<typeof ActionSwipeParamSchema, ActionSwipeParam>({\n    name: 'Swipe',\n    description:\n      inputMode === 'touch'\n        ? 'Perform a touch gesture that directly manipulates the UI (e.g., adjust a continuous control such as a slider or wheel picker, switch between paged cards or images, follow an on-screen swipe gesture to continue or dismiss, or swipe an item to delete it). For browsing off-screen content in a page or scrollable region, use Scroll instead. Choose exactly one movement form: (1) relative swipe — provide \"direction\" and a positive \"distance\"; or (2) endpoint swipe — provide \"end\". \"start\" is optional for both forms and defaults to the center of the page. Do not combine \"end\" with \"direction\" or \"distance\".'\n        : 'Perform a primary-mouse-button gesture that continuously presses and moves the pointer across the desktop UI. Use it to adjust a continuous control such as a slider or wheel picker, switch between paged cards or images, follow an on-screen swipe gesture to continue or dismiss, or swipe an item to delete it. For browsing off-screen content in a page or scrollable region, use Scroll instead. Choose exactly one movement form: (1) relative swipe — provide \"direction\" and a positive \"distance\"; or (2) endpoint swipe — provide \"end\". \"start\" is optional for both forms and defaults to the center of the page. Do not combine \"end\" with \"direction\" or \"distance\".',\n    interfaceAlias: 'aiSwipe',\n    paramSchema:\n      inputMode === 'touch'\n        ? ActionSwipeParamSchema\n        : createActionSwipeParamSchema(inputMode),\n    sample: {\n      start: { prompt: 'center of the notification' },\n      end: { prompt: 'upper edge of the screen' },\n    },\n    call: async (param) => {\n      const { startPoint, endPoint, duration, repeatCount } =\n        normalizeSwipeParam(param, await config.size());\n      for (let i = 0; i < repeatCount; i++) {\n        await config.swipe(startPoint, endPoint, { duration });\n      }\n    },\n  });\n};\n\n// ClearInput\nexport const actionClearInputParamSchema = z.object({\n  locate: getMidsceneLocationSchema()\n    .describe('The input field to be cleared')\n    .optional(),\n});\nexport type ActionClearInputParam = {\n  locate?: LocateResultElement;\n};\n\nexport const defineActionClearInput = (\n  clearInput: KeyboardInputPrimitives['clearInput'],\n): DeviceAction<ActionClearInputParam> => {\n  return defineAction<\n    typeof actionClearInputParamSchema,\n    ActionClearInputParam\n  >({\n    name: 'ClearInput',\n    description: inputLocateDescription,\n    interfaceAlias: 'aiClearInput',\n    paramSchema: actionClearInputParamSchema,\n    sample: {\n      locate: { prompt: 'the search input field' },\n    },\n    call: async (param) => {\n      await clearInput(param.locate);\n    },\n  });\n};\n\n// CursorMove\nexport const actionCursorMoveParamSchema = z.object({\n  direction: z\n    .enum(['left', 'right'])\n    .describe('The direction to move the cursor'),\n  times: z\n    .number()\n    .int()\n    .min(1)\n    .default(1)\n    .describe(\n      'The number of times to move the cursor in the specified direction',\n    ),\n});\nexport type ActionCursorMoveParam = {\n  direction: 'left' | 'right';\n  times?: number;\n};\n\nexport const defineActionCursorMove = (config: {\n  keyboard: Pick<KeyboardInputPrimitives, 'keyboardPress' | 'cursorMove'>;\n  sleep?(timeMs: number): Promise<void>;\n}): DeviceAction<ActionCursorMoveParam> => {\n  return defineAction<\n    typeof actionCursorMoveParamSchema,\n    ActionCursorMoveParam\n  >({\n    name: 'CursorMove',\n    description:\n      'Move the text cursor (caret) left or right within an input field or text area. Use this to reposition the cursor without selecting text.',\n    paramSchema: actionCursorMoveParamSchema,\n    sample: {\n      direction: 'left',\n      times: 3,\n    },\n    call: async (param) => {\n      const times = param.times ?? 1;\n      if (config.keyboard.cursorMove) {\n        await config.keyboard.cursorMove(param.direction, times);\n        return;\n      }\n\n      const wait =\n        config.sleep ??\n        ((timeMs: number) =>\n          new Promise<void>((resolve) => setTimeout(resolve, timeMs)));\n      const arrowKey = param.direction === 'left' ? 'ArrowLeft' : 'ArrowRight';\n      for (let i = 0; i < times; i++) {\n        await config.keyboard.keyboardPress(arrowKey);\n        await wait(100);\n      }\n    },\n  });\n};\n\n// Pinch\nexport const ActionPinchParamSchema = z.object({\n  locate: getMidsceneLocationSchema()\n    .optional()\n    .describe(\n      'The element to pinch on. If not specified, the center of the screen will be used',\n    ),\n  direction: z\n    .enum(['in', 'out'])\n    .describe(\n      'Pinch direction. \"in\" = pinch fingers together (zoom out / shrink), \"out\" = spread fingers apart (zoom in / enlarge).',\n    ),\n  distance: z\n    .number()\n    .positive()\n    .optional()\n    .describe(\n      'How far each finger moves in pixels. Defaults to a quarter of the shorter screen dimension.',\n    ),\n  duration: z\n    .number()\n    .default(500)\n    .optional()\n    .describe('Duration of the pinch gesture in milliseconds'),\n});\n\nexport type ActionPinchParam = {\n  locate?: LocateResultElement;\n  direction: 'in' | 'out';\n  distance?: number;\n  duration?: number;\n};\n\nexport const defineActionPinch = (config: {\n  pinch: TouchInputPrimitives['pinch'];\n  size(): Promise<Size>;\n}): DeviceAction<ActionPinchParam> | undefined => {\n  if (!config.pinch) {\n    return undefined;\n  }\n\n  return defineAction<typeof ActionPinchParamSchema, ActionPinchParam>({\n    name: 'Pinch',\n    description:\n      'Perform a two-finger pinch gesture. Use direction \"in\" to pinch fingers together (zoom out), or \"out\" to spread fingers apart (zoom in). Optionally specify distance for how far each finger moves.',\n    interfaceAlias: 'aiPinch',\n    paramSchema: ActionPinchParamSchema,\n    sample: {\n      locate: { prompt: 'the map area' },\n      direction: 'out',\n      distance: 200,\n    },\n    call: async (param) => {\n      const { centerX, centerY, startDistance, endDistance, duration } =\n        normalizePinchParam(param, await config.size());\n      await config.pinch?.(\n        { x: centerX, y: centerY },\n        { startDistance, endDistance, duration },\n      );\n    },\n  });\n};\n\nexport function normalizePinchParam(\n  param: ActionPinchParam,\n  screenSize: { width: number; height: number },\n): {\n  centerX: number;\n  centerY: number;\n  startDistance: number;\n  endDistance: number;\n  duration: number;\n} {\n  const { width, height } = screenSize;\n  const element = param.locate;\n  const centerX = element\n    ? Math.round(element.center[0])\n    : Math.round(width / 2);\n  const centerY = element\n    ? Math.round(element.center[1])\n    : Math.round(height / 2);\n  const duration = param.duration ?? 500;\n\n  const baseDistance = Math.round(Math.min(width, height) / 4);\n  const fingerDistance = param.distance ?? baseDistance;\n\n  const startDistance = baseDistance;\n  const endDistance =\n    param.direction === 'out'\n      ? baseDistance + fingerDistance\n      : Math.max(10, baseDistance - fingerDistance);\n\n  return { centerX, centerY, startDistance, endDistance, duration };\n}\n\nexport interface MobileInputActionContext {\n  input: MobileInputPrimitives;\n  size(): Promise<Size>;\n  sleep?(timeMs: number): Promise<void>;\n  getDefaultAutoDismissKeyboard?(): boolean | undefined;\n  systemActions?: SystemInputActionOptions;\n}\n\nexport interface SystemInputActionConfig {\n  name: string;\n  description: string;\n  interfaceAlias?: string;\n  delayBeforeRunner?: number;\n  delayAfterRunner?: number;\n}\n\nexport interface SystemInputActionOptions {\n  backButton?: SystemInputActionConfig;\n  homeButton?: SystemInputActionConfig;\n  recentAppsButton?: SystemInputActionConfig;\n}\n\nexport interface InputPrimitiveActionOptions {\n  size?: () => Promise<Size>;\n  sleep?: (timeMs: number) => Promise<void>;\n  includeSwipe?: boolean;\n  includePinch?: boolean;\n  systemActions?: SystemInputActionOptions;\n}\n\nfunction defineSystemInputAction(\n  config: SystemInputActionConfig,\n  call: () => Promise<void>,\n): DeviceAction<undefined, void> {\n  return defineAction<undefined, undefined, void>({\n    name: config.name,\n    description: config.description,\n    interfaceAlias: config.interfaceAlias,\n    delayBeforeRunner: config.delayBeforeRunner,\n    delayAfterRunner: config.delayAfterRunner,\n    call,\n  });\n}\n\nexport function defineActionsFromInputPrimitives(\n  input: InputPrimitives,\n  options: InputPrimitiveActionOptions = {},\n): DeviceAction<any>[] {\n  const actions: Array<DeviceAction<any> | undefined> = [];\n  const { pointer, keyboard, scroll, touch, system } = input;\n\n  if (pointer) {\n    actions.push(defineActionTap(pointer.tap));\n    if (pointer.doubleClick) {\n      actions.push(defineActionDoubleClick(pointer.doubleClick));\n    }\n    if (pointer.rightClick) {\n      actions.push(defineActionRightClick(pointer.rightClick));\n    }\n    if (pointer.hover) {\n      actions.push(defineActionHover(pointer.hover));\n    }\n    if (pointer.dragAndDrop) {\n      actions.push(defineActionDragAndDrop(pointer.dragAndDrop));\n    }\n    if (pointer.longPress) {\n      actions.push(defineActionLongPress(pointer.longPress));\n    }\n  }\n\n  if (keyboard) {\n    actions.push(\n      defineActionInput(keyboard),\n      defineActionClearInput(keyboard.clearInput),\n      defineActionKeyboardPress(keyboard.keyboardPress),\n      defineActionCursorMove({ keyboard, sleep: options.sleep }),\n    );\n  }\n\n  if (scroll) {\n    actions.push(defineActionScroll(scroll.scroll));\n  }\n\n  const swipeConfig = resolveSwipeInputPrimitive({ pointer, touch });\n  if (swipeConfig && options.size && options.includeSwipe !== false) {\n    actions.push(defineActionSwipe({ ...swipeConfig, size: options.size }));\n  }\n\n  if (touch?.pinch && options.size && options.includePinch !== false) {\n    actions.push(defineActionPinch({ pinch: touch.pinch, size: options.size }));\n  }\n\n  if (system && options.systemActions) {\n    const { systemActions } = options;\n    if (system.backButton && systemActions.backButton) {\n      actions.push(\n        defineSystemInputAction(systemActions.backButton, system.backButton),\n      );\n    }\n    if (system.homeButton && systemActions.homeButton) {\n      actions.push(\n        defineSystemInputAction(systemActions.homeButton, system.homeButton),\n      );\n    }\n    if (system.recentAppsButton && systemActions.recentAppsButton) {\n      actions.push(\n        defineSystemInputAction(\n          systemActions.recentAppsButton,\n          system.recentAppsButton,\n        ),\n      );\n    }\n  }\n\n  return actions.filter((action): action is DeviceAction<any> =>\n    Boolean(action),\n  );\n}\n\nexport function createDefaultMobileActions(\n  context: MobileInputActionContext,\n): DeviceAction<any>[] {\n  return defineActionsFromInputPrimitives(context.input, {\n    size: context.size,\n    sleep: context.sleep,\n    systemActions: context.systemActions,\n  });\n}\n\n// Sleep\nexport const ActionSleepParamSchema = z.object({\n  timeMs: z\n    .number()\n    .default(1000)\n    .optional()\n    .describe('Sleep duration in milliseconds, defaults to 1000ms (1 second)'),\n});\n\nexport type ActionSleepParam = {\n  timeMs?: number;\n};\n\nexport const defineActionSleep = (\n  abortSignal?: AbortSignal,\n): DeviceAction<ActionSleepParam> => {\n  return defineAction<typeof ActionSleepParamSchema, ActionSleepParam>({\n    name: 'Sleep',\n    description:\n      'Wait for a specified duration before continuing. Defaults to 1 second (1000ms) if not specified.',\n    paramSchema: ActionSleepParamSchema,\n    sample: {\n      timeMs: 2000,\n    },\n    call: async (param) => {\n      const duration = param?.timeMs ?? 1000;\n      getDebug('device:common-action')(`Sleeping for ${duration}ms`);\n      abortSignal?.throwIfAborted();\n      await new Promise<void>((resolve, reject) => {\n        const abort = () => {\n          clearTimeout(timer);\n          reject(abortSignal?.reason);\n        };\n        const timer = setTimeout(() => {\n          abortSignal?.removeEventListener('abort', abort);\n          resolve();\n        }, duration);\n        abortSignal?.addEventListener('abort', abort, { once: true });\n      });\n    },\n  });\n};\n\nexport type { DeviceAction } from '../types';\nexport type { ActionScrollParam, LocateResultElement, Size } from '../types';\nexport { getMidsceneLocationSchema } from '../common';\nexport { z };\nexport {\n  inputStrategies,\n  type InputStrategy,\n  type ResolvedTextInputOptions,\n  type SequentialTextInputHandlers,\n  type TextInputOptions,\n  resolveInputStrategy,\n  resolveTextInputOptions,\n  sendTextSequentially,\n  shouldInputSequentially,\n} from './input-strategy';\nexport type {\n  AndroidDeviceOpt,\n  AndroidDeviceInputOpt,\n  IOSDeviceOpt,\n  IOSDeviceInputOpt,\n  HarmonyDeviceOpt,\n  HarmonyDeviceInputOpt,\n} from './device-options';\n"],"names":["resolveSwipeInputPrimitive","input","AbstractInterface","_actionName","_param","defineAction","config","pointFromLocate","locate","missingMessage","Error","defineLocatedPointAction","param","actionTapParamSchema","z","getMidsceneLocationSchema","defineActionTap","tap","point","registerFileChooserAcceptParamSchema","defineActionRegisterFileChooserAccept","register","actionRightClickParamSchema","defineActionRightClick","rightClick","actionDoubleClickParamSchema","defineActionDoubleClick","doubleClick","actionHoverParamSchema","defineActionHover","hover","inputLocateDescription","actionInputParamSchema","val","String","inputStrategies","defineActionInput","keyboard","context","actionKeyboardPressParamSchema","defineActionKeyboardPress","keyboardPress","actionScrollParamSchema","defineActionScroll","scroll","actionDragAndDropParamSchema","defineActionDragAndDrop","dragAndDrop","from","to","ActionLongPressParamSchema","defineActionLongPress","longPress","createActionSwipeParamSchema","inputMode","movementSource","ActionSwipeParamSchema","normalizeSwipeParam","screenSize","width","height","start","end","startPoint","endPoint","undefined","Math","duration","repeatCount","normalizeMobileSwipeParam","defineActionSwipe","i","actionClearInputParamSchema","defineActionClearInput","clearInput","actionCursorMoveParamSchema","defineActionCursorMove","times","wait","timeMs","Promise","resolve","setTimeout","arrowKey","ActionPinchParamSchema","defineActionPinch","centerX","centerY","startDistance","endDistance","normalizePinchParam","element","baseDistance","fingerDistance","defineSystemInputAction","call","defineActionsFromInputPrimitives","options","actions","pointer","touch","system","swipeConfig","systemActions","action","Boolean","createDefaultMobileActions","ActionSleepParamSchema","defineActionSleep","abortSignal","getDebug","reject","abort","clearTimeout","timer"],"mappings":";;;;;;;;;;;;;;AAkKO,SAASA,2BACdC,KAAiD;IAEjD,IAAIA,MAAM,KAAK,EAAE,OACf,OAAO;QAAE,OAAOA,MAAM,KAAK,CAAC,KAAK;QAAE,WAAW;IAAQ;IAExD,IAAIA,MAAM,OAAO,EAAE,OACjB,OAAO;QAAE,OAAOA,MAAM,OAAO,CAAC,KAAK;QAAE,WAAW;IAAQ;AAG5D;AAoCO,MAAeC;IA2BpB,MAAM,kBAAmBC,WAAmB,EAAEC,MAAW,EAAiB,CAAC;;QAwB3E;QAqDA;;AACF;AAKO,MAAMC,eAAe,CAK1BC,SAgBOA;AAGT,SAASC,gBACPC,MAAuC,EACvCC,cAAsB;IAEtB,IAAI,CAACD,QACH,MAAM,IAAIE,MAAMD;IAElB,OAAO;QAAE,GAAGD,OAAO,MAAM,CAAC,EAAE;QAAE,GAAGA,OAAO,MAAM,CAAC,EAAE;IAAC;AACpD;AAEA,SAASG,yBAGPL,MAQD;IACC,OAAOD,aAA8B;QACnC,MAAMC,OAAO,IAAI;QACjB,aAAaA,OAAO,WAAW;QAC/B,gBAAgBA,OAAO,cAAc;QACrC,aAAaA,OAAO,WAAW;QAC/B,QAAQA,OAAO,MAAM;QACrB,MAAM,OAAOM;YACX,MAAMN,OAAO,IAAI,CACfC,gBAAgBK,MAAM,MAAM,EAAEN,OAAO,oBAAoB,GACzDM;QAEJ;IACF;AACF;AAGO,MAAMC,uBAAuBC,EAAE,MAAM,CAAC;IAC3C,QAAQC,4BAA4B,QAAQ,CAAC;AAC/C;AAKO,MAAMC,kBAAkB,CAC7BC,MAEON,yBAAsE;QAC3E,MAAM;QACN,aAAa;QACb,gBAAgB;QAChB,aAAaE;QACb,QAAQ;YACN,QAAQ;gBAAE,QAAQ;YAAsB;QAC1C;QACA,sBAAsB;QACtB,MAAM,OAAOK;YACX,MAAMD,IAAIC;QACZ;IACF;AAGK,MAAMC,uCAAuCL,EAAE,MAAM,CAAC;IAC3D,OAAOA,EAAAA,KACC,CAAC;QAACA,EAAE,MAAM;QAAIA,EAAE,KAAK,CAACA,EAAE,MAAM;KAAI,EACvC,QAAQ,CACP;AAEN;AAKO,MAAMM,wCAAwC,CACnDC,WAEOhB,aAGL;QACA,MAAM;QACN,aACE;QACF,gBAAgB;QAChB,aAAac;QACb,QAAQ;YACN,OAAO;gBAAC;aAAwB;QAClC;QACA,MAAM,OAAOP;YACX,MAAMS,SAAST,MAAM,KAAK;QAC5B;IACF;AAIK,MAAMU,8BAA8BR,EAAE,MAAM,CAAC;IAClD,QAAQC,4BAA4B,QAAQ,CAC1C;AAEJ;AAKO,MAAMQ,yBAAyB,CACpCC,aAEOb,yBAGL;QACA,MAAM;QACN,aAAa;QACb,gBAAgB;QAChB,aAAaW;QACb,QAAQ;YACN,QAAQ;gBAAE,QAAQ;YAA+B;QACnD;QACA,sBAAsB;QACtB,MAAM,OAAOJ;YACX,MAAMM,WAAWN;QACnB;IACF;AAIK,MAAMO,+BAA+BX,EAAE,MAAM,CAAC;IACnD,QAAQC,4BAA4B,QAAQ,CAC1C;AAEJ;AAKO,MAAMW,0BAA0B,CACrCC,cAEOhB,yBAGL;QACA,MAAM;QACN,aAAa;QACb,gBAAgB;QAChB,aAAac;QACb,QAAQ;YACN,QAAQ;gBAAE,QAAQ;YAAkB;QACtC;QACA,sBAAsB;QACtB,MAAM,OAAOP;YACX,MAAMS,YAAYT;QACpB;IACF;AAIK,MAAMU,yBAAyBd,EAAE,MAAM,CAAC;IAC7C,QAAQC,4BAA4B,QAAQ,CAAC;AAC/C;AAKO,MAAMc,oBAAoB,CAC/BC,QAEOnB,yBAGL;QACA,MAAM;QACN,aAAa;QACb,gBAAgB;QAChB,aAAaiB;QACb,QAAQ;YACN,QAAQ;gBAAE,QAAQ;YAAsC;QAC1D;QACA,sBAAsB;QACtB,MAAM,OAAOV;YACX,MAAMY,MAAMZ;QACd;IACF;AAIF,MAAMa,yBACJ;AACK,MAAMC,yBAAyBlB,EAAE,MAAM,CAAC;IAC7C,OAAOA,EAAAA,KACC,CAAC;QAACA,EAAE,MAAM;QAAIA,EAAE,MAAM;KAAG,EAC9B,SAAS,CAAC,CAACmB,MAAQC,OAAOD,MAC1B,QAAQ,CACP;IAEJ,QAAQlB,4BACL,QAAQ,CAACgB,wBACT,QAAQ;IACX,MAAMjB,CAAC,CAADA,OACC,CAAC;QAAC;QAAW;QAAS;KAAW,EACrC,OAAO,CAAC,WACR,QAAQ,CACP;IAEJ,qBAAqBA,EAAAA,OACX,GACP,QAAQ,GACR,QAAQ,CACP;IAEJ,mBAAmBA,EAAAA,MACV,GACN,MAAM,GACN,WAAW,GACX,QAAQ,GACR,QAAQ,CACP;IAEJ,eAAeA,CAAC,CAADA,OACR,CAACqB,iBACL,QAAQ,GACR,QAAQ,CACP;AAEN;AAUO,MAAMC,oBAAoB,CAC/BC,WAEOhC,aAA8D;QACnE,MAAM;QACN,aAAa;QACb,gBAAgB;QAChB,aAAa2B;QACb,QAAQ;YACN,OAAO;YACP,QAAQ;gBAAE,QAAQ;YAAwB;QAC5C;QACA,MAAM,OAAOpB,OAAO0B;YAElB,IAAK1B,AAA0B,aAA1BA,MAAM,IAAI,EACbA,MAAM,IAAI,GAAG;YAGf,IAAIA,AAAe,YAAfA,MAAM,IAAI,EAAc,YAC1B,MAAMyB,SAAS,UAAU,CAACzB,MAAM,MAAM;YAIxC,IAAI,CAACA,SAAS,CAACA,MAAM,KAAK,EACxB;YAGF,MAAMyB,SAAS,QAAQ,CAACzB,MAAM,KAAK,EAAE;gBACnC,QAAQA,MAAM,MAAM;gBACpB,SAASA,AAAe,eAAfA,MAAM,IAAI;gBACnB,qBAAqBA,MAAM,mBAAmB;gBAC9C,mBAAmBA,MAAM,iBAAiB;gBAC1C,eAAeA,MAAM,aAAa;gBAClC,GAAI0B,SAAS,kBAAkB;oBAAE,iBAAiB;gBAAK,IAAI,CAAC,CAAC;YAC/D;QACF;IACF;AAIK,MAAMC,iCAAiCzB,EAAE,MAAM,CAAC;IACrD,QAAQC,4BACL,QAAQ,CACP,iMAED,QAAQ;IACX,SAASD,EAAAA,MACA,GACN,QAAQ,CACP;AAEN;AAMO,MAAM0B,4BAA4B,CACvCC,gBAEOpC,aAGL;QACA,MAAM;QACN,aACE;QACF,gBAAgB;QAChB,aAAakC;QACb,QAAQ;YACN,SAAS;QACX;QACA,MAAM,OAAO3B;YACX,MAAM6B,cAAc7B,MAAM,OAAO,EAAE;gBACjC,QAAQA,MAAM,MAAM;YACtB;QACF;IACF;AAIK,MAAM8B,0BAA0B5B,EAAE,MAAM,CAAC;IAC9C,YAAYA,CAAC,CAADA,OACL,CAAC;QACJ;QACA;QACA;QACA;QACA;KACD,EACA,OAAO,CAAC,gBACR,QAAQ,CACP;IAEJ,WAAWA,CAAC,CAADA,OACJ,CAAC;QAAC;QAAQ;QAAM;QAAS;KAAO,EACpC,OAAO,CAAC,QACR,QAAQ,CACP;IAEJ,UAAUA,EAAAA,MACD,GACN,QAAQ,GACR,QAAQ,GACR,QAAQ,GACR,QAAQ,CACP;IAEJ,QAAQC,4BACL,QAAQ,GACR,QAAQ,CACP;AAEN;AAEO,MAAM4B,qBAAqB,CAChCC,SAEOvC,aAAgE;QACrE,MAAM;QACN,aACE;QACF,gBAAgB;QAChB,aAAaqC;QACb,QAAQ;YACN,WAAW;YACX,YAAY;YACZ,QAAQ;gBAAE,QAAQ;YAAsC;QAC1D;QACA,MAAM,OAAO9B,OAAO0B;YAClB,IAAIA,SAAS,iBACX,MAAMM,OAAOhC,OAAO;gBAAE,iBAAiB;YAAK;iBACzC,MAAMgC,OAAOhC;QACpB;IACF;AAIK,MAAMiC,+BAA+B/B,EAAE,MAAM,CAAC;IACnD,MAAMC,4BAA4B,QAAQ,CAAC;IAC3C,IAAIA,4BAA4B,QAAQ,CAAC;AAC3C;AAMO,MAAM+B,0BAA0B,CACrCC,cAEO1C,aAGL;QACA,MAAM;QACN,aACE;QACF,gBAAgB;QAChB,aAAawC;QACb,QAAQ;YACN,MAAM;gBAAE,QAAQ;YAA6B;YAC7C,IAAI;gBAAE,QAAQ;YAAuB;QACvC;QACA,MAAM,OAAOjC;YACX,MAAMoC,OAAOpC,MAAM,IAAI;YACvB,MAAMqC,KAAKrC,MAAM,EAAE;YACnB,IAAI,CAACoC,MACH,MAAM,IAAItC,MAAM;YAElB,IAAI,CAACuC,IACH,MAAM,IAAIvC,MAAM;YAElB,MAAMqC,YACJ;gBAAE,GAAGC,KAAK,MAAM,CAAC,EAAE;gBAAE,GAAGA,KAAK,MAAM,CAAC,EAAE;YAAC,GACvC;gBAAE,GAAGC,GAAG,MAAM,CAAC,EAAE;gBAAE,GAAGA,GAAG,MAAM,CAAC,EAAE;YAAC;QAEvC;IACF;AAGK,MAAMC,6BAA6BpC,EAAE,MAAM,CAAC;IACjD,QAAQC,4BAA4B,QAAQ,CAC1C;IAEF,UAAUD,EAAAA,MACD,GACN,QAAQ,GACR,QAAQ,CAAC;AACd;AAMO,MAAMqC,wBAAwB,CACnCC,YAEOzC,yBAGL;QACA,MAAM;QACN,aAAa;QACb,gBAAgB;QAChB,aAAauC;QACb,QAAQ;YACN,QAAQ;gBAAE,QAAQ;YAAqB;QACzC;QACA,sBAAsB;QACtB,MAAM,OAAOhC,OAAON;YAClB,MAAMwC,UAAUlC,OAAO;gBAAE,UAAUN,MAAM,QAAQ;YAAC;QACpD;IACF;AAGF,SAASyC,6BAA6BC,SAAyB;IAC7D,MAAMC,iBAAiBD,AAAc,YAAdA,YAAwB,WAAW;IAE1D,OAAOxC,EAAE,MAAM,CAAC;QACd,OAAOC,4BACJ,QAAQ,GACR,QAAQ,CACP,CAAC,+BAA+B,EAAEwC,eAAe,qGAAqG,CAAC;QAE3J,WAAWzC,CAAC,CAADA,OACJ,CAAC;YAAC;YAAM;YAAQ;YAAQ;SAAQ,EACpC,QAAQ,GACR,QAAQ,CACP,GAAGyC,AAAmB,aAAnBA,iBAA8B,WAAW,UAAU,0GAA0G,CAAC;QAErK,UAAUzC,EAAAA,MACD,GACN,QAAQ,GACR,QAAQ,GACR,QAAQ,CACP,CAAC,uBAAuB,EAAEyC,eAAe,gGAAgG,CAAC;QAE9I,KAAKxC,4BACF,QAAQ,GACR,QAAQ,CACP,CAAC,gBAAgB,EAAEwC,eAAe,iHAAiH,CAAC;QAExJ,UAAUzC,EAAAA,MACD,GACN,OAAO,CAAC,KACR,QAAQ,CAAC;QACZ,QAAQA,EAAAA,MACC,GACN,QAAQ,GACR,QAAQ,CACP;IAEN;AACF;AAIO,MAAM0C,yBAAyBH,6BAA6B;AAW5D,SAASI,oBACd7C,KAAuB,EACvB8C,UAA6C;IAO7C,MAAM,EAAEC,KAAK,EAAEC,MAAM,EAAE,GAAGF;IAC1B,MAAM,EAAEG,KAAK,EAAEC,GAAG,EAAE,GAAGlD;IAEvB,MAAMmD,aAAaF,QACf;QAAE,GAAGA,MAAM,MAAM,CAAC,EAAE;QAAE,GAAGA,MAAM,MAAM,CAAC,EAAE;IAAC,IACzC;QAAE,GAAGF,QAAQ;QAAG,GAAGC,SAAS;IAAE;IAElC,MAAMI,WAAY,AAAC;QACjB,IAAIF,KAAK;YACP,IAAIlD,AAAoBqD,WAApBrD,MAAM,SAAS,IAAkBA,AAAmBqD,WAAnBrD,MAAM,QAAQ,EACjD,MAAM,IAAIF,MACR;YAIJ,OAAO;gBAAE,GAAGoD,IAAI,MAAM,CAAC,EAAE;gBAAE,GAAGA,IAAI,MAAM,CAAC,EAAE;YAAC;QAC9C;QAEA,IAAIlD,AAAoBqD,WAApBrD,MAAM,SAAS,IAAkBA,AAAmBqD,WAAnBrD,MAAM,QAAQ,EACjD,MAAM,IAAIF,MACR;QAIJ,IAAIE,MAAM,QAAQ,IAAI,GACpB,MAAM,IAAIF,MACR;QAIJ,OAAO;YACL,GACEqD,WAAW,CAAC,GACXnD,CAAAA,AAAoB,YAApBA,MAAM,SAAS,GACZA,MAAM,QAAQ,GACdA,AAAoB,WAApBA,MAAM,SAAS,GACb,CAACA,MAAM,QAAQ,GACf;YACR,GACEmD,WAAW,CAAC,GACXnD,CAAAA,AAAoB,WAApBA,MAAM,SAAS,GACZA,MAAM,QAAQ,GACdA,AAAoB,SAApBA,MAAM,SAAS,GACb,CAACA,MAAM,QAAQ,GACf;QACV;IACF;IAEAoD,SAAS,CAAC,GAAGE,KAAK,GAAG,CAAC,GAAGA,KAAK,GAAG,CAACF,SAAS,CAAC,EAAEL;IAC9CK,SAAS,CAAC,GAAGE,KAAK,GAAG,CAAC,GAAGA,KAAK,GAAG,CAACF,SAAS,CAAC,EAAEJ;IAE9C,MAAMO,WAAWvD,MAAM,QAAQ,IAAI;IAEnC,IAAIwD,cAAc,AAAwB,YAAxB,OAAOxD,MAAM,MAAM,GAAgBA,MAAM,MAAM,GAAG;IACpE,IAAIwD,AAAgB,MAAhBA,aACFA,cAAc;IAGhB,OAAO;QAAEL;QAAYC;QAAUG;QAAUC;IAAY;AACvD;AAGO,MAAMC,4BAA4BZ;AAElC,MAAMa,oBAAoB,CAAChE;IAKhC,MAAMgD,YAAYhD,OAAO,SAAS,IAAI;IACtC,OAAOD,aAA8D;QACnE,MAAM;QACN,aACEiD,AAAc,YAAdA,YACI,mmBACA;QACN,gBAAgB;QAChB,aACEA,AAAc,YAAdA,YACIE,yBACAH,6BAA6BC;QACnC,QAAQ;YACN,OAAO;gBAAE,QAAQ;YAA6B;YAC9C,KAAK;gBAAE,QAAQ;YAA2B;QAC5C;QACA,MAAM,OAAO1C;YACX,MAAM,EAAEmD,UAAU,EAAEC,QAAQ,EAAEG,QAAQ,EAAEC,WAAW,EAAE,GACnDX,oBAAoB7C,OAAO,MAAMN,OAAO,IAAI;YAC9C,IAAK,IAAIiE,IAAI,GAAGA,IAAIH,aAAaG,IAC/B,MAAMjE,OAAO,KAAK,CAACyD,YAAYC,UAAU;gBAAEG;YAAS;QAExD;IACF;AACF;AAGO,MAAMK,8BAA8B1D,EAAE,MAAM,CAAC;IAClD,QAAQC,4BACL,QAAQ,CAAC,iCACT,QAAQ;AACb;AAKO,MAAM0D,yBAAyB,CACpCC,aAEOrE,aAGL;QACA,MAAM;QACN,aAAa0B;QACb,gBAAgB;QAChB,aAAayC;QACb,QAAQ;YACN,QAAQ;gBAAE,QAAQ;YAAyB;QAC7C;QACA,MAAM,OAAO5D;YACX,MAAM8D,WAAW9D,MAAM,MAAM;QAC/B;IACF;AAIK,MAAM+D,8BAA8B7D,EAAE,MAAM,CAAC;IAClD,WAAWA,CAAC,CAADA,OACJ,CAAC;QAAC;QAAQ;KAAQ,EACtB,QAAQ,CAAC;IACZ,OAAOA,EAAAA,MACE,GACN,GAAG,GACH,GAAG,CAAC,GACJ,OAAO,CAAC,GACR,QAAQ,CACP;AAEN;AAMO,MAAM8D,yBAAyB,CAACtE,SAI9BD,aAGL;QACA,MAAM;QACN,aACE;QACF,aAAasE;QACb,QAAQ;YACN,WAAW;YACX,OAAO;QACT;QACA,MAAM,OAAO/D;YACX,MAAMiE,QAAQjE,MAAM,KAAK,IAAI;YAC7B,IAAIN,OAAO,QAAQ,CAAC,UAAU,EAAE,YAC9B,MAAMA,OAAO,QAAQ,CAAC,UAAU,CAACM,MAAM,SAAS,EAAEiE;YAIpD,MAAMC,OACJxE,OAAO,KAAK,IACV,EAAAyE,SACA,IAAIC,QAAc,CAACC,UAAYC,WAAWD,SAASF,QAAO;YAC9D,MAAMI,WAAWvE,AAAoB,WAApBA,MAAM,SAAS,GAAc,cAAc;YAC5D,IAAK,IAAI2D,IAAI,GAAGA,IAAIM,OAAON,IAAK;gBAC9B,MAAMjE,OAAO,QAAQ,CAAC,aAAa,CAAC6E;gBACpC,MAAML,KAAK;YACb;QACF;IACF;AAIK,MAAMM,yBAAyBtE,EAAE,MAAM,CAAC;IAC7C,QAAQC,4BACL,QAAQ,GACR,QAAQ,CACP;IAEJ,WAAWD,CAAC,CAADA,OACJ,CAAC;QAAC;QAAM;KAAM,EAClB,QAAQ,CACP;IAEJ,UAAUA,EAAAA,MACD,GACN,QAAQ,GACR,QAAQ,GACR,QAAQ,CACP;IAEJ,UAAUA,EAAAA,MACD,GACN,OAAO,CAAC,KACR,QAAQ,GACR,QAAQ,CAAC;AACd;AASO,MAAMuE,oBAAoB,CAAC/E;IAIhC,IAAI,CAACA,OAAO,KAAK,EACf;IAGF,OAAOD,aAA8D;QACnE,MAAM;QACN,aACE;QACF,gBAAgB;QAChB,aAAa+E;QACb,QAAQ;YACN,QAAQ;gBAAE,QAAQ;YAAe;YACjC,WAAW;YACX,UAAU;QACZ;QACA,MAAM,OAAOxE;YACX,MAAM,EAAE0E,OAAO,EAAEC,OAAO,EAAEC,aAAa,EAAEC,WAAW,EAAEtB,QAAQ,EAAE,GAC9DuB,oBAAoB9E,OAAO,MAAMN,OAAO,IAAI;YAC9C,MAAMA,OAAO,KAAK,GAChB;gBAAE,GAAGgF;gBAAS,GAAGC;YAAQ,GACzB;gBAAEC;gBAAeC;gBAAatB;YAAS;QAE3C;IACF;AACF;AAEO,SAASuB,oBACd9E,KAAuB,EACvB8C,UAA6C;IAQ7C,MAAM,EAAEC,KAAK,EAAEC,MAAM,EAAE,GAAGF;IAC1B,MAAMiC,UAAU/E,MAAM,MAAM;IAC5B,MAAM0E,UAAUK,UACZzB,KAAK,KAAK,CAACyB,QAAQ,MAAM,CAAC,EAAE,IAC5BzB,KAAK,KAAK,CAACP,QAAQ;IACvB,MAAM4B,UAAUI,UACZzB,KAAK,KAAK,CAACyB,QAAQ,MAAM,CAAC,EAAE,IAC5BzB,KAAK,KAAK,CAACN,SAAS;IACxB,MAAMO,WAAWvD,MAAM,QAAQ,IAAI;IAEnC,MAAMgF,eAAe1B,KAAK,KAAK,CAACA,KAAK,GAAG,CAACP,OAAOC,UAAU;IAC1D,MAAMiC,iBAAiBjF,MAAM,QAAQ,IAAIgF;IAEzC,MAAMJ,gBAAgBI;IACtB,MAAMH,cACJ7E,AAAoB,UAApBA,MAAM,SAAS,GACXgF,eAAeC,iBACf3B,KAAK,GAAG,CAAC,IAAI0B,eAAeC;IAElC,OAAO;QAAEP;QAASC;QAASC;QAAeC;QAAatB;IAAS;AAClE;AAgCA,SAAS2B,wBACPxF,MAA+B,EAC/ByF,IAAyB;IAEzB,OAAO1F,aAAyC;QAC9C,MAAMC,OAAO,IAAI;QACjB,aAAaA,OAAO,WAAW;QAC/B,gBAAgBA,OAAO,cAAc;QACrC,mBAAmBA,OAAO,iBAAiB;QAC3C,kBAAkBA,OAAO,gBAAgB;QACzCyF;IACF;AACF;AAEO,SAASC,iCACd/F,KAAsB,EACtBgG,UAAuC,CAAC,CAAC;IAEzC,MAAMC,UAAgD,EAAE;IACxD,MAAM,EAAEC,OAAO,EAAE9D,QAAQ,EAAEO,MAAM,EAAEwD,KAAK,EAAEC,MAAM,EAAE,GAAGpG;IAErD,IAAIkG,SAAS;QACXD,QAAQ,IAAI,CAAClF,gBAAgBmF,QAAQ,GAAG;QACxC,IAAIA,QAAQ,WAAW,EACrBD,QAAQ,IAAI,CAACxE,wBAAwByE,QAAQ,WAAW;QAE1D,IAAIA,QAAQ,UAAU,EACpBD,QAAQ,IAAI,CAAC3E,uBAAuB4E,QAAQ,UAAU;QAExD,IAAIA,QAAQ,KAAK,EACfD,QAAQ,IAAI,CAACrE,kBAAkBsE,QAAQ,KAAK;QAE9C,IAAIA,QAAQ,WAAW,EACrBD,QAAQ,IAAI,CAACpD,wBAAwBqD,QAAQ,WAAW;QAE1D,IAAIA,QAAQ,SAAS,EACnBD,QAAQ,IAAI,CAAC/C,sBAAsBgD,QAAQ,SAAS;IAExD;IAEA,IAAI9D,UACF6D,QAAQ,IAAI,CACV9D,kBAAkBC,WAClBoC,uBAAuBpC,SAAS,UAAU,GAC1CG,0BAA0BH,SAAS,aAAa,GAChDuC,uBAAuB;QAAEvC;QAAU,OAAO4D,QAAQ,KAAK;IAAC;IAI5D,IAAIrD,QACFsD,QAAQ,IAAI,CAACvD,mBAAmBC,OAAO,MAAM;IAG/C,MAAM0D,cAActG,2BAA2B;QAAEmG;QAASC;IAAM;IAChE,IAAIE,eAAeL,QAAQ,IAAI,IAAIA,AAAyB,UAAzBA,QAAQ,YAAY,EACrDC,QAAQ,IAAI,CAAC5B,kBAAkB;QAAE,GAAGgC,WAAW;QAAE,MAAML,QAAQ,IAAI;IAAC;IAGtE,IAAIG,OAAO,SAASH,QAAQ,IAAI,IAAIA,AAAyB,UAAzBA,QAAQ,YAAY,EACtDC,QAAQ,IAAI,CAACb,kBAAkB;QAAE,OAAOe,MAAM,KAAK;QAAE,MAAMH,QAAQ,IAAI;IAAC;IAG1E,IAAII,UAAUJ,QAAQ,aAAa,EAAE;QACnC,MAAM,EAAEM,aAAa,EAAE,GAAGN;QAC1B,IAAII,OAAO,UAAU,IAAIE,cAAc,UAAU,EAC/CL,QAAQ,IAAI,CACVJ,wBAAwBS,cAAc,UAAU,EAAEF,OAAO,UAAU;QAGvE,IAAIA,OAAO,UAAU,IAAIE,cAAc,UAAU,EAC/CL,QAAQ,IAAI,CACVJ,wBAAwBS,cAAc,UAAU,EAAEF,OAAO,UAAU;QAGvE,IAAIA,OAAO,gBAAgB,IAAIE,cAAc,gBAAgB,EAC3DL,QAAQ,IAAI,CACVJ,wBACES,cAAc,gBAAgB,EAC9BF,OAAO,gBAAgB;IAI/B;IAEA,OAAOH,QAAQ,MAAM,CAAC,CAACM,SACrBC,QAAQD;AAEZ;AAEO,SAASE,2BACdpE,OAAiC;IAEjC,OAAO0D,iCAAiC1D,QAAQ,KAAK,EAAE;QACrD,MAAMA,QAAQ,IAAI;QAClB,OAAOA,QAAQ,KAAK;QACpB,eAAeA,QAAQ,aAAa;IACtC;AACF;AAGO,MAAMqE,yBAAyB7F,EAAE,MAAM,CAAC;IAC7C,QAAQA,EAAAA,MACC,GACN,OAAO,CAAC,MACR,QAAQ,GACR,QAAQ,CAAC;AACd;AAMO,MAAM8F,oBAAoB,CAC/BC,cAEOxG,aAA8D;QACnE,MAAM;QACN,aACE;QACF,aAAasG;QACb,QAAQ;YACN,QAAQ;QACV;QACA,MAAM,OAAO/F;YACX,MAAMuD,WAAWvD,OAAO,UAAU;YAClCkG,SAAS,wBAAwB,CAAC,aAAa,EAAE3C,SAAS,EAAE,CAAC;YAC7D0C,aAAa;YACb,MAAM,IAAI7B,QAAc,CAACC,SAAS8B;gBAChC,MAAMC,QAAQ;oBACZC,aAAaC;oBACbH,OAAOF,aAAa;gBACtB;gBACA,MAAMK,QAAQhC,WAAW;oBACvB2B,aAAa,oBAAoB,SAASG;oBAC1C/B;gBACF,GAAGd;gBACH0C,aAAa,iBAAiB,SAASG,OAAO;oBAAE,MAAM;gBAAK;YAC7D;QACF;IACF"}

package/dist/es/index.mjs.map

@@ -1 +1 @@
-{"version":3,"file":"index.mjs","sources":["../../src/index.ts"],"sourcesContent":["import { z } from 'zod';\nimport Service from './service/index';\nimport { TaskRunner } from './task-runner';\nimport { getVersion } from './utils';\n\nexport {\n  standardPlan,\n  AiLocateElement,\n  runConnectivityTest,\n  getMidsceneLocationSchema,\n  PointSchema,\n  SizeSchema,\n  RectSchema,\n  TMultimodalPromptSchema,\n  TUserPromptSchema,\n  type TMultimodalPrompt,\n  type TUserPrompt,\n  type ConnectivityTestConfig,\n  type ConnectivityTestResult,\n} from './ai-model/index';\n\nexport {\n  MIDSCENE_MODEL_NAME,\n  type CreateOpenAIClientFn,\n} from '@midscene/shared/env';\n\nexport type * from './types';\nexport {\n  ServiceError,\n  ExecutionDump,\n  ReportActionDump,\n  GroupedActionDump,\n  type IExecutionDump,\n  type IReportActionDump,\n  type IGroupedActionDump,\n  type ReportMeta,\n  type GroupMeta,\n} from './types';\n\nexport { z };\n\nexport default Service;\nexport { TaskRunner, Service, getVersion };\n\nexport type {\n  MidsceneYamlScript,\n  MidsceneYamlTask,\n  MidsceneYamlFlowItem,\n  MidsceneYamlConfigResult,\n  MidsceneYamlConfig,\n  MidsceneYamlScriptWebEnv,\n  MidsceneYamlScriptAndroidEnv,\n  MidsceneYamlScriptIOSEnv,\n  MidsceneYamlScriptHarmonyEnv,\n  MidsceneYamlTargetConfig,\n  MidsceneYamlTargetKey,\n  MidsceneYamlScriptEnv,\n  LocateOption,\n  DetailedLocateParam,\n} from './yaml';\n\nexport {\n  Agent,\n  type AgentOpt,\n  type AiActOptions,\n  type GherkinStepKeyword,\n  type MidsceneUsageMetrics,\n  type RunGherkinScenarioOptions,\n  type UsageBucket,\n  createAgent,\n  type UIObservation,\n  type UIObserver,\n  type UIObserverOption,\n} from './agent';\nexport {\n  describeElementAtPoint,\n  verifyElementDescriptionAtPoint,\n  verifyLocator,\n  type DescribeElementAtPointOptions,\n  type DescribeElementCoordinateSpace,\n  type ElementDescriberRuntime,\n  type VerifyElementDescriptionAtPointOptions,\n} from './element-describer';\n\n// Dump utilities\nexport {\n  createInlineImageResolver,\n  restoreImageReferences,\n  restoreReportImageReferences,\n  escapeContent,\n  unescapeContent,\n  parseImageScripts,\n  parseDumpScript,\n  parseDumpScriptAttributes,\n  generateImageScriptTag,\n  generateDumpScriptTag,\n  deriveTaskStatus,\n  deriveCaseStatus,\n} from './dump';\nexport type {\n  TaskStatusFields,\n  DerivedTaskStatus,\n  RestoredScreenshotReference,\n  StoredImageReferenceResolver,\n} from './dump';\nexport {\n  getTaskSearchArea,\n  getTaskServiceDump,\n} from './dump/task-service-dump';\n\n// Report generator\nexport type { IReportGenerator } from './report-generator';\nexport { ReportGenerator, nullReportGenerator } from './report-generator';\nexport {\n  collectDedupedExecutions,\n  ReportMergingTool,\n  TestRunReportAssembler,\n  dedupeExecutionsKeepLatest,\n  splitReportHtmlByExecution,\n} from './report';\nexport type * from './test-run-report';\nexport { TEST_RUN_REPORT_SCRIPT_TYPE } from './test-run-report';\nexport type * from './test-run-health';\nexport {\n  calculateTestRunHealth,\n  classifyTestRunCase,\n} from './test-run-health';\nexport {\n  createReportCliCommands,\n  reportFileToMarkdown,\n  splitReportFile,\n  mergeReportFiles,\n  type ConsumeReportFileAction,\n  type ReportFileToMarkdownOptions,\n  type ReportCliCommandDefinition,\n  type ReportCliCommandEntry,\n  type SplitReportFileOptions,\n  type MergeReportFilesOptions,\n  type MergeReportFilesResult,\n} from './report-cli';\n\n// ScreenshotItem\nexport { ScreenshotItem } from './screenshot-item';\nexport type {\n  ImageUrlRef,\n  ScreenshotRef,\n  StoredImageRef,\n} from './dump/image-reference';\nexport {\n  ReportImageStore,\n  ScreenshotStore,\n} from './dump/screenshot-store';\n\nexport {\n  executionToMarkdown,\n  reportToMarkdown,\n  type ExecutionMarkdownOptions,\n  type ExecutionMarkdownResult,\n  type ReportMarkdownOptions,\n  type ReportMarkdownResult,\n  type MarkdownAttachment,\n} from './report-markdown';\n"],"names":["Service"],"mappings":";;;;;;;;;;;;;;;;;;;AAyCA,YAAeA"}
+{"version":3,"file":"index.mjs","sources":["../../src/index.ts"],"sourcesContent":["import { z } from 'zod';\nimport Service from './service/index';\nimport { TaskRunner } from './task-runner';\nimport { getVersion } from './utils';\n\nexport {\n  standardPlan,\n  AiLocateElement,\n  runConnectivityTest,\n  getMidsceneLocationSchema,\n  PointSchema,\n  SizeSchema,\n  RectSchema,\n  TMultimodalPromptSchema,\n  TUserPromptSchema,\n  type TMultimodalPrompt,\n  type TUserPrompt,\n  type ConnectivityTestConfig,\n  type ConnectivityTestResult,\n  type ConnectivityCheckResultItem,\n} from './ai-model/index';\n\nexport {\n  MIDSCENE_MODEL_NAME,\n  type CreateOpenAIClientFn,\n} from '@midscene/shared/env';\n\nexport type * from './types';\nexport {\n  ServiceError,\n  ExecutionDump,\n  ReportActionDump,\n  GroupedActionDump,\n  type IExecutionDump,\n  type IReportActionDump,\n  type IGroupedActionDump,\n  type ReportMeta,\n  type GroupMeta,\n} from './types';\n\nexport { z };\n\nexport default Service;\nexport { TaskRunner, Service, getVersion };\n\nexport type {\n  MidsceneYamlScript,\n  MidsceneYamlTask,\n  MidsceneYamlFlowItem,\n  MidsceneYamlConfigResult,\n  MidsceneYamlConfig,\n  MidsceneYamlScriptWebEnv,\n  MidsceneYamlScriptAndroidEnv,\n  MidsceneYamlScriptIOSEnv,\n  MidsceneYamlScriptHarmonyEnv,\n  MidsceneYamlTargetConfig,\n  MidsceneYamlTargetKey,\n  MidsceneYamlScriptEnv,\n  LocateOption,\n  DetailedLocateParam,\n} from './yaml';\n\nexport {\n  Agent,\n  type AgentOpt,\n  type AiActOptions,\n  type GherkinStepKeyword,\n  type MidsceneUsageMetrics,\n  type RunGherkinScenarioOptions,\n  type UsageBucket,\n  createAgent,\n  type UIObservation,\n  type UIObserver,\n  type UIObserverOption,\n} from './agent';\nexport {\n  describeElementAtPoint,\n  verifyElementDescriptionAtPoint,\n  verifyLocator,\n  type DescribeElementAtPointOptions,\n  type DescribeElementCoordinateSpace,\n  type ElementDescriberRuntime,\n  type VerifyElementDescriptionAtPointOptions,\n} from './element-describer';\n\n// Dump utilities\nexport {\n  createInlineImageResolver,\n  restoreImageReferences,\n  restoreReportImageReferences,\n  escapeContent,\n  unescapeContent,\n  parseImageScripts,\n  parseDumpScript,\n  parseDumpScriptAttributes,\n  generateImageScriptTag,\n  generateDumpScriptTag,\n  deriveTaskStatus,\n  deriveCaseStatus,\n} from './dump';\nexport type {\n  TaskStatusFields,\n  DerivedTaskStatus,\n  RestoredScreenshotReference,\n  StoredImageReferenceResolver,\n} from './dump';\nexport {\n  getTaskSearchArea,\n  getTaskServiceDump,\n} from './dump/task-service-dump';\n\n// Report generator\nexport type { IReportGenerator } from './report-generator';\nexport { ReportGenerator, nullReportGenerator } from './report-generator';\nexport {\n  collectDedupedExecutions,\n  ReportMergingTool,\n  TestRunReportAssembler,\n  dedupeExecutionsKeepLatest,\n  splitReportHtmlByExecution,\n} from './report';\nexport type * from './test-run-report';\nexport { TEST_RUN_REPORT_SCRIPT_TYPE } from './test-run-report';\nexport type * from './test-run-health';\nexport {\n  calculateTestRunHealth,\n  classifyTestRunCase,\n} from './test-run-health';\nexport {\n  createReportCliCommands,\n  reportFileToMarkdown,\n  splitReportFile,\n  mergeReportFiles,\n  type ConsumeReportFileAction,\n  type ReportFileToMarkdownOptions,\n  type ReportCliCommandDefinition,\n  type ReportCliCommandEntry,\n  type SplitReportFileOptions,\n  type MergeReportFilesOptions,\n  type MergeReportFilesResult,\n} from './report-cli';\n\n// ScreenshotItem\nexport { ScreenshotItem } from './screenshot-item';\nexport type {\n  ImageUrlRef,\n  ScreenshotRef,\n  StoredImageRef,\n} from './dump/image-reference';\nexport {\n  ReportImageStore,\n  ScreenshotStore,\n} from './dump/screenshot-store';\n\nexport {\n  executionToMarkdown,\n  reportToMarkdown,\n  type ExecutionMarkdownOptions,\n  type ExecutionMarkdownResult,\n  type ReportMarkdownOptions,\n  type ReportMarkdownResult,\n  type MarkdownAttachment,\n} from './report-markdown';\n"],"names":["Service"],"mappings":";;;;;;;;;;;;;;;;;;;AA0CA,YAAeA"}