@midscene/core 1.9.8 → 1.10.0

package/dist/es/ai-model/models/auto-glm/locate.mjs

@@ -1,113 +1,11 @@
-import { generateElementByRect } from "@midscene/shared/extractor";
-import { getDebug } from "@midscene/shared/logger";
-import { assert } from "@midscene/shared/utils";
-import { extraTextFromUserPrompt, promptsToChatParam } from "../../inspect.mjs";
-import { findElementPrompt } from "../../prompt/llm-locator.mjs";
-import { callAIWithStringResponse } from "../../service-caller/index.mjs";
-import { finalizePixelBbox } from "../../shared/model-locate-result/bbox.mjs";
-import { mapLocateResultToPixelBboxByCoordinates } from "../../shared/model-locate-result/pixel-bbox-mapper.mjs";
-import { pixelBboxToRect } from "../../workflows/inspect/locate-result-rect.mjs";
-import { mapSearchAreaPixelBboxToOriginalPixelBbox } from "../../workflows/inspect/search-area-mapping.mjs";
-import { parseAutoGLMLocateResponse } from "./parser.mjs";
-const debugInspect = getDebug('ai:inspect');
-async function autoGlmLocate(elementDescription, options, getSystemPrompt) {
-    const { context, modelRuntime } = options;
-    const screenshotBase64 = context.screenshot.base64;
-    assert(elementDescription, "cannot find the target element description");
-    const elementDescriptionText = extraTextFromUserPrompt(elementDescription);
-    const userInstructionPrompt = findElementPrompt(elementDescriptionText);
-    const locateImage = options.searchConfig?.image ?? {
-        imageBase64: screenshotBase64,
-        width: context.shotSize.width,
-        height: context.shotSize.height
-    };
-    const imagePayload = locateImage.imageBase64;
-    const imageWidth = locateImage.width;
-    const imageHeight = locateImage.height;
-    const msgs = [
-        {
-            role: 'system',
-            content: getSystemPrompt()
-        },
-        {
-            role: 'user',
-            content: [
-                {
-                    type: 'image_url',
-                    image_url: {
-                        url: imagePayload,
-                        detail: 'high'
-                    }
-                },
-                {
-                    type: 'text',
-                    text: `Tap: ${userInstructionPrompt}`
-                }
-            ]
-        }
-    ];
-    if ('string' != typeof elementDescription) {
-        const addOns = await promptsToChatParam({
-            images: elementDescription.images,
-            convertHttpImage2Base64: elementDescription.convertHttpImage2Base64
-        });
-        msgs.push(...addOns);
-    }
-    const { content: rawResponseContent, usage, rawChoiceMessage } = await callAIWithStringResponse(msgs, modelRuntime, {
-        abortSignal: options.abortSignal
-    });
-    debugInspect('auto-glm rawResponse:', rawResponseContent);
-    const parsed = parseAutoGLMLocateResponse(rawResponseContent);
-    debugInspect('auto-glm thinking:', parsed.think);
-    debugInspect('auto-glm coordinates:', parsed.coordinates);
-    let resRect;
-    let matchedElement;
-    let errors = [];
-    if (parsed.error || !parsed.coordinates) {
-        errors = [
-            parsed.error || 'Failed to parse auto-glm response'
-        ];
-        debugInspect('auto-glm parse error:', errors[0]);
-    } else {
-        const { x, y } = parsed.coordinates;
-        debugInspect('auto-glm coordinates [0-999]:', {
-            x,
-            y
-        });
-        const ctx = {
-            preparedSize: {
-                width: imageWidth,
-                height: imageHeight
-            }
-        };
-        const targetPixelBbox = finalizePixelBbox(mapLocateResultToPixelBboxByCoordinates({
-            type: 'point',
-            coordinates: [
-                x,
-                y
-            ]
-        }, ctx, {
-            shape: 'point',
-            order: 'xy',
-            normalizedBy: 1000
-        }), parsed.coordinates, ctx);
-        resRect = pixelBboxToRect(mapSearchAreaPixelBboxToOriginalPixelBbox(targetPixelBbox, options.searchConfig?.mapping));
-        debugInspect('auto-glm resRect:', resRect);
-        const element = generateElementByRect(resRect, elementDescriptionText);
-        if (element) matchedElement = element;
-    }
+import { getTapLocatedPixelBbox } from "../../shared/planning-action.mjs";
+import { getAutoGLMChineseLocatePrompt, getAutoGLMMultilingualLocatePrompt } from "./prompt.mjs";
+function createAutoGlmPlanningTapLocator(isMultilingual) {
     return {
-        rect: resRect,
-        parseResult: {
-            element: matchedElement,
-            errors
-        },
-        rawResponse: rawResponseContent,
-        rawChoiceMessage,
-        usage,
-        reasoning_content: parsed.think
+        buildSystemPrompt: ()=>isMultilingual ? getAutoGLMMultilingualLocatePrompt() : getAutoGLMChineseLocatePrompt(),
+        getLocatedPixelBbox: getTapLocatedPixelBbox
     };
 }
-export { autoGlmLocate };
+export { createAutoGlmPlanningTapLocator };
 
 //# sourceMappingURL=locate.mjs.map

package/dist/es/ai-model/models/auto-glm/locate.mjs.map

@@ -1 +1 @@
-{"version":3,"file":"ai-model/models/auto-glm/locate.mjs","sources":["../../../../../src/ai-model/models/auto-glm/locate.ts"],"sourcesContent":["import type { Rect } from '@/types';\nimport { generateElementByRect } from '@midscene/shared/extractor';\nimport { getDebug } from '@midscene/shared/logger';\nimport type { LocateResultElement } from '@midscene/shared/types';\nimport { assert } from '@midscene/shared/utils';\nimport type { TUserPrompt } from '../../../common';\nimport {\n  type InspectAIArgs,\n  extraTextFromUserPrompt,\n  promptsToChatParam,\n} from '../../inspect';\nimport { findElementPrompt } from '../../prompt/llm-locator';\nimport { callAIWithStringResponse } from '../../service-caller/index';\nimport { finalizePixelBbox } from '../../shared/model-locate-result/bbox';\nimport { mapLocateResultToPixelBboxByCoordinates } from '../../shared/model-locate-result/pixel-bbox-mapper';\nimport { pixelBboxToRect } from '../../workflows/inspect/locate-result-rect';\nimport { mapSearchAreaPixelBboxToOriginalPixelBbox } from '../../workflows/inspect/search-area-mapping';\nimport type {\n  LocateOptions,\n  LocateResult,\n} from '../../workflows/inspect/types';\nimport { parseAutoGLMLocateResponse } from './parser';\n\nconst debugInspect = getDebug('ai:inspect');\n\nexport async function autoGlmLocate(\n  elementDescription: TUserPrompt,\n  options: LocateOptions,\n  getSystemPrompt: () => string,\n): Promise<LocateResult> {\n  const { context, modelRuntime } = options;\n  const screenshotBase64 = context.screenshot.base64;\n\n  assert(elementDescription, 'cannot find the target element description');\n  const elementDescriptionText = extraTextFromUserPrompt(elementDescription);\n  const userInstructionPrompt = findElementPrompt(elementDescriptionText);\n\n  const locateImage = options.searchConfig?.image ?? {\n    imageBase64: screenshotBase64,\n    width: context.shotSize.width,\n    height: context.shotSize.height,\n  };\n  const imagePayload = locateImage.imageBase64;\n  const imageWidth = locateImage.width;\n  const imageHeight = locateImage.height;\n\n  const msgs: InspectAIArgs = [\n    { role: 'system', content: getSystemPrompt() },\n    {\n      role: 'user',\n      content: [\n        {\n          type: 'image_url',\n          image_url: {\n            url: imagePayload,\n            detail: 'high',\n          },\n        },\n        {\n          type: 'text',\n          text: `Tap: ${userInstructionPrompt}`,\n        },\n      ],\n    },\n  ];\n\n  if (typeof elementDescription !== 'string') {\n    const addOns = await promptsToChatParam({\n      images: elementDescription.images,\n      convertHttpImage2Base64: elementDescription.convertHttpImage2Base64,\n    });\n    msgs.push(...addOns);\n  }\n\n  const {\n    content: rawResponseContent,\n    usage,\n    rawChoiceMessage,\n  } = await callAIWithStringResponse(msgs, modelRuntime, {\n    abortSignal: options.abortSignal,\n  });\n\n  debugInspect('auto-glm rawResponse:', rawResponseContent);\n\n  const parsed = parseAutoGLMLocateResponse(rawResponseContent);\n\n  debugInspect('auto-glm thinking:', parsed.think);\n  debugInspect('auto-glm coordinates:', parsed.coordinates);\n\n  let resRect: Rect | undefined;\n  let matchedElement: LocateResultElement | undefined;\n  let errors: string[] = [];\n\n  if (parsed.error || !parsed.coordinates) {\n    errors = [parsed.error || 'Failed to parse auto-glm response'];\n    debugInspect('auto-glm parse error:', errors[0]);\n  } else {\n    const { x, y } = parsed.coordinates;\n\n    debugInspect('auto-glm coordinates [0-999]:', { x, y });\n\n    const ctx = { preparedSize: { width: imageWidth, height: imageHeight } };\n    const targetPixelBbox = finalizePixelBbox(\n      mapLocateResultToPixelBboxByCoordinates(\n        { type: 'point', coordinates: [x, y] },\n        ctx,\n        { shape: 'point', order: 'xy', normalizedBy: 1000 },\n      ),\n      parsed.coordinates,\n      ctx,\n    );\n    resRect = pixelBboxToRect(\n      mapSearchAreaPixelBboxToOriginalPixelBbox(\n        targetPixelBbox,\n        options.searchConfig?.mapping,\n      ),\n    );\n\n    debugInspect('auto-glm resRect:', resRect);\n\n    const element: LocateResultElement = generateElementByRect(\n      resRect,\n      elementDescriptionText as string,\n    );\n\n    if (element) {\n      matchedElement = element;\n    }\n  }\n\n  return {\n    rect: resRect,\n    parseResult: {\n      element: matchedElement,\n      errors,\n    },\n    rawResponse: rawResponseContent,\n    rawChoiceMessage,\n    usage,\n    reasoning_content: parsed.think,\n  };\n}\n"],"names":["debugInspect","getDebug","autoGlmLocate","elementDescription","options","getSystemPrompt","context","modelRuntime","screenshotBase64","assert","elementDescriptionText","extraTextFromUserPrompt","userInstructionPrompt","findElementPrompt","locateImage","imagePayload","imageWidth","imageHeight","msgs","addOns","promptsToChatParam","rawResponseContent","usage","rawChoiceMessage","callAIWithStringResponse","parsed","parseAutoGLMLocateResponse","resRect","matchedElement","errors","x","y","ctx","targetPixelBbox","finalizePixelBbox","mapLocateResultToPixelBboxByCoordinates","pixelBboxToRect","mapSearchAreaPixelBboxToOriginalPixelBbox","element","generateElementByRect"],"mappings":";;;;;;;;;;;AAuBA,MAAMA,eAAeC,SAAS;AAEvB,eAAeC,cACpBC,kBAA+B,EAC/BC,OAAsB,EACtBC,eAA6B;IAE7B,MAAM,EAAEC,OAAO,EAAEC,YAAY,EAAE,GAAGH;IAClC,MAAMI,mBAAmBF,QAAQ,UAAU,CAAC,MAAM;IAElDG,OAAON,oBAAoB;IAC3B,MAAMO,yBAAyBC,wBAAwBR;IACvD,MAAMS,wBAAwBC,kBAAkBH;IAEhD,MAAMI,cAAcV,QAAQ,YAAY,EAAE,SAAS;QACjD,aAAaI;QACb,OAAOF,QAAQ,QAAQ,CAAC,KAAK;QAC7B,QAAQA,QAAQ,QAAQ,CAAC,MAAM;IACjC;IACA,MAAMS,eAAeD,YAAY,WAAW;IAC5C,MAAME,aAAaF,YAAY,KAAK;IACpC,MAAMG,cAAcH,YAAY,MAAM;IAEtC,MAAMI,OAAsB;QAC1B;YAAE,MAAM;YAAU,SAASb;QAAkB;QAC7C;YACE,MAAM;YACN,SAAS;gBACP;oBACE,MAAM;oBACN,WAAW;wBACT,KAAKU;wBACL,QAAQ;oBACV;gBACF;gBACA;oBACE,MAAM;oBACN,MAAM,CAAC,KAAK,EAAEH,uBAAuB;gBACvC;aACD;QACH;KACD;IAED,IAAI,AAA8B,YAA9B,OAAOT,oBAAiC;QAC1C,MAAMgB,SAAS,MAAMC,mBAAmB;YACtC,QAAQjB,mBAAmB,MAAM;YACjC,yBAAyBA,mBAAmB,uBAAuB;QACrE;QACAe,KAAK,IAAI,IAAIC;IACf;IAEA,MAAM,EACJ,SAASE,kBAAkB,EAC3BC,KAAK,EACLC,gBAAgB,EACjB,GAAG,MAAMC,yBAAyBN,MAAMX,cAAc;QACrD,aAAaH,QAAQ,WAAW;IAClC;IAEAJ,aAAa,yBAAyBqB;IAEtC,MAAMI,SAASC,2BAA2BL;IAE1CrB,aAAa,sBAAsByB,OAAO,KAAK;IAC/CzB,aAAa,yBAAyByB,OAAO,WAAW;IAExD,IAAIE;IACJ,IAAIC;IACJ,IAAIC,SAAmB,EAAE;IAEzB,IAAIJ,OAAO,KAAK,IAAI,CAACA,OAAO,WAAW,EAAE;QACvCI,SAAS;YAACJ,OAAO,KAAK,IAAI;SAAoC;QAC9DzB,aAAa,yBAAyB6B,MAAM,CAAC,EAAE;IACjD,OAAO;QACL,MAAM,EAAEC,CAAC,EAAEC,CAAC,EAAE,GAAGN,OAAO,WAAW;QAEnCzB,aAAa,iCAAiC;YAAE8B;YAAGC;QAAE;QAErD,MAAMC,MAAM;YAAE,cAAc;gBAAE,OAAOhB;gBAAY,QAAQC;YAAY;QAAE;QACvE,MAAMgB,kBAAkBC,kBACtBC,wCACE;YAAE,MAAM;YAAS,aAAa;gBAACL;gBAAGC;aAAE;QAAC,GACrCC,KACA;YAAE,OAAO;YAAS,OAAO;YAAM,cAAc;QAAK,IAEpDP,OAAO,WAAW,EAClBO;QAEFL,UAAUS,gBACRC,0CACEJ,iBACA7B,QAAQ,YAAY,EAAE;QAI1BJ,aAAa,qBAAqB2B;QAElC,MAAMW,UAA+BC,sBACnCZ,SACAjB;QAGF,IAAI4B,SACFV,iBAAiBU;IAErB;IAEA,OAAO;QACL,MAAMX;QACN,aAAa;YACX,SAASC;YACTC;QACF;QACA,aAAaR;QACbE;QACAD;QACA,mBAAmBG,OAAO,KAAK;IACjC;AACF"}
+{"version":3,"file":"ai-model/models/auto-glm/locate.mjs","sources":["../../../../../src/ai-model/models/auto-glm/locate.ts"],"sourcesContent":["import type { PlanningTapLocatorDefinition } from '../../model-adapter/types';\nimport { getTapLocatedPixelBbox } from '../../shared/planning-action';\nimport {\n  getAutoGLMChineseLocatePrompt,\n  getAutoGLMMultilingualLocatePrompt,\n} from './prompt';\n\nexport function createAutoGlmPlanningTapLocator(\n  isMultilingual: boolean,\n): PlanningTapLocatorDefinition {\n  return {\n    buildSystemPrompt: () =>\n      isMultilingual\n        ? getAutoGLMMultilingualLocatePrompt()\n        : getAutoGLMChineseLocatePrompt(),\n    getLocatedPixelBbox: getTapLocatedPixelBbox,\n  };\n}\n"],"names":["createAutoGlmPlanningTapLocator","isMultilingual","getAutoGLMMultilingualLocatePrompt","getAutoGLMChineseLocatePrompt","getTapLocatedPixelBbox"],"mappings":";;AAOO,SAASA,gCACdC,cAAuB;IAEvB,OAAO;QACL,mBAAmB,IACjBA,iBACIC,uCACAC;QACN,qBAAqBC;IACvB;AACF"}

package/dist/es/ai-model/models/auto-glm/parser.mjs

@@ -7,7 +7,7 @@ const extractValueAfter = (src, key)=>{
     if (rest.endsWith('")')) rest = rest.slice(0, -2);
     return rest;
 };
-function parseAction(response) {
+function parseAutoGLMPlanningAction(response) {
     debug('Parsing action:', response);
     let trimmedResponse = '';
     try {
@@ -170,70 +170,48 @@ function parseAction(response) {
     }
 }
 function parseAutoGLMResponse(content) {
+    let parsedResponse;
     if (content.includes('finish(message=')) {
         const parts = content.split('finish(message=');
         const think = parts[0].trim();
         const actionContent = `finish(message=${parts[1]}`;
-        return {
+        parsedResponse = {
             think,
             content: actionContent
         };
-    }
-    if (content.includes('do(action=')) {
+    } else if (content.includes('do(action=')) {
         const parts = content.split('do(action=');
         const think = parts[0].trim();
         const actionContent = `do(action=${parts[1]}`;
-        return {
+        parsedResponse = {
             think,
             content: actionContent
         };
-    }
-    if (content.includes('<answer>')) {
+    } else if (content.includes('<answer>')) {
         const parts = content.split('<answer>');
         const think = parts[0].replace(/<think>/g, '').replace(/<\/think>/g, '').trim();
         const actionContent = parts[1].replace(/<\/answer>/g, '').trim();
-        return {
+        parsedResponse = {
             think,
             content: actionContent
         };
-    }
-    return {
+    } else parsedResponse = {
         think: '',
         content
     };
+    debug('autoGLM rawResponse:', content);
+    debug('thinking in response:', parsedResponse.think);
+    debug('action in response:', parsedResponse.content);
+    return parsedResponse;
 }
-function parseAutoGLMLocateResponse(rawResponse) {
-    const { think, content: actionContent } = parseAutoGLMResponse(rawResponse);
-    if (!actionContent.startsWith('do(action="Tap"')) return {
-        think,
-        coordinates: null,
-        error: `Unexpected action type in auto-glm locate response: ${actionContent}`
+function parseAutoGLMPlanningResponse(content) {
+    const response = parseAutoGLMResponse(content);
+    const action = parseAutoGLMPlanningAction(response);
+    return {
+        response,
+        action
     };
-    try {
-        const elementMatch = actionContent.match(/element=\[(\d+),(\d+)\]/);
-        if (!elementMatch) return {
-            think,
-            coordinates: null,
-            error: `Failed to extract element coordinates from auto-glm response: ${actionContent}`
-        };
-        const x = Number(elementMatch[1]);
-        const y = Number(elementMatch[2]);
-        return {
-            think,
-            coordinates: {
-                x,
-                y
-            }
-        };
-    } catch (e) {
-        const errorMessage = e instanceof Error ? e.message : String(e);
-        return {
-            think,
-            coordinates: null,
-            error: `Failed to parse coordinates "${actionContent}" with errorMessage: ${errorMessage}`
-        };
-    }
 }
-export { extractValueAfter, parseAction, parseAutoGLMLocateResponse, parseAutoGLMResponse };
+export { extractValueAfter, parseAutoGLMPlanningAction, parseAutoGLMPlanningResponse, parseAutoGLMResponse };
 
 //# sourceMappingURL=parser.mjs.map

package/dist/es/ai-model/models/auto-glm/parser.mjs.map

@@ -1 +1 @@
-{"version":3,"file":"ai-model/models/auto-glm/parser.mjs","sources":["../../../../../src/ai-model/models/auto-glm/parser.ts"],"sourcesContent":["import { getDebug } from '@midscene/shared/logger';\nimport type { ParsedAction } from './actions';\n\nconst debug = getDebug('auto-glm-parser');\n\n// Do not rely on regex extraction here; regex can fail on malformed input.\n// Bad Case: finish(message=\"Finished! Now There is a contact whose name is \\\"Tom\\\" in the list.\")\nexport const extractValueAfter = (src: string, key: string): string => {\n  const idx = src.indexOf(key);\n  if (idx === -1) {\n    throw new Error(`Missing key ${key} in action payload ${src}`);\n  }\n  let rest = src.slice(idx + key.length).trim();\n  if (rest.endsWith('\")')) {\n    rest = rest.slice(0, -2);\n  }\n  return rest;\n};\n\nexport function parseAction(response: {\n  think: string;\n  content: string;\n}): ParsedAction {\n  debug('Parsing action:', response);\n  let trimmedResponse = '';\n  try {\n    trimmedResponse = response.content.trim();\n\n    if (\n      trimmedResponse.startsWith('do(action=\"Type\"') ||\n      trimmedResponse.startsWith('do(action=\"Type_Name\"')\n    ) {\n      const text = extractValueAfter(trimmedResponse, 'text=\"');\n      return {\n        _metadata: 'do',\n        action: 'Type',\n        text,\n        think: response.think,\n      } as ParsedAction;\n    }\n\n    if (trimmedResponse.startsWith('finish(message=')) {\n      let message = extractValueAfter(trimmedResponse, 'finish(message=\"');\n      if (message.endsWith(')')) message = message.slice(0, -1);\n      return {\n        _metadata: 'finish',\n        message,\n        think: response.think,\n      } as ParsedAction;\n    }\n\n    if (trimmedResponse.startsWith('do(')) {\n      const actionMatch = trimmedResponse.match(/do\\(action=\"([^\"]+)\"/);\n      if (!actionMatch)\n        throw new Error(\n          `Failed to extract action type from do() call; raw=\"${trimmedResponse}\"`,\n        );\n      const actionType = actionMatch[1];\n\n      const baseAction = { _metadata: 'do' as const, think: response.think };\n      switch (actionType) {\n        case 'Tap': {\n          const elementMatch = trimmedResponse.match(/element=\\[(\\d+),(\\d+)\\]/);\n          if (!elementMatch)\n            throw new Error(\n              `Failed to extract element coordinates for Tap; raw=\"${trimmedResponse}\"`,\n            );\n          return {\n            ...baseAction,\n            action: 'Tap',\n            element: [Number(elementMatch[1]), Number(elementMatch[2])],\n          } as ParsedAction;\n        }\n        case 'Double Tap': {\n          const elementMatch = trimmedResponse.match(/element=\\[(\\d+),(\\d+)\\]/);\n          if (!elementMatch)\n            throw new Error(\n              `Failed to extract element coordinates for Double Tap; raw=\"${trimmedResponse}\"`,\n            );\n          return {\n            ...baseAction,\n            action: 'Double Tap',\n            element: [Number(elementMatch[1]), Number(elementMatch[2])],\n          } as ParsedAction;\n        }\n        case 'Swipe': {\n          const startMatch = trimmedResponse.match(/start=\\[(\\d+),(\\d+)\\]/);\n          const endMatch = trimmedResponse.match(/end=\\[(\\d+),(\\d+)\\]/);\n          if (!startMatch || !endMatch)\n            throw new Error(\n              `Failed to extract start/end coordinates for Swipe; raw=\"${trimmedResponse}\"`,\n            );\n          return {\n            ...baseAction,\n            action: 'Swipe',\n            start: [Number(startMatch[1]), Number(startMatch[2])],\n            end: [Number(endMatch[1]), Number(endMatch[2])],\n          } as ParsedAction;\n        }\n        case 'Long Press': {\n          const elementMatch = trimmedResponse.match(/element=\\[(\\d+),(\\d+)\\]/);\n          if (!elementMatch)\n            throw new Error(\n              `Failed to extract element coordinates for Long Press; raw=\"${trimmedResponse}\"`,\n            );\n          return {\n            ...baseAction,\n            action: 'Long Press',\n            element: [Number(elementMatch[1]), Number(elementMatch[2])],\n          } as ParsedAction;\n        }\n        case 'Launch': {\n          const app = extractValueAfter(trimmedResponse, 'app=\"');\n          return { ...baseAction, action: 'Launch', app } as ParsedAction;\n        }\n        case 'Back': {\n          return { ...baseAction, action: 'Back' } as ParsedAction;\n        }\n        case 'Home': {\n          return { ...baseAction, action: 'Home' } as ParsedAction;\n        }\n        case 'Wait': {\n          const durationMatch = trimmedResponse.match(\n            /duration=(?:[\"\\[])?(\\d+)/,\n          );\n          if (!durationMatch) {\n            throw new Error(\n              `Failed to extract duration for Wait; raw=\"${trimmedResponse}\"`,\n            );\n          }\n          const seconds = Number.parseInt(durationMatch[1], 10);\n          const durationMs = seconds * 1000;\n          return {\n            ...baseAction,\n            action: 'Wait',\n            durationMs,\n          } as ParsedAction;\n        }\n        case 'Interact': {\n          return { ...baseAction, action: 'Interact' } as ParsedAction;\n        }\n        case 'Call_API': {\n          const instruction = extractValueAfter(\n            trimmedResponse,\n            'instruction=\"',\n          );\n          return {\n            ...baseAction,\n            action: 'Call_API',\n            instruction,\n          } as ParsedAction;\n        }\n        case 'Take_over': {\n          const message = extractValueAfter(trimmedResponse, 'message=\"');\n          return {\n            ...baseAction,\n            action: 'Take_over',\n            message,\n          } as ParsedAction;\n        }\n        case 'Note': {\n          const message = extractValueAfter(trimmedResponse, 'message=\"');\n          return {\n            ...baseAction,\n            action: 'Note',\n            message,\n          } as ParsedAction;\n        }\n        default:\n          throw new Error(\n            `Unknown action type: ${actionType}; raw=\"${trimmedResponse}\"`,\n          );\n      }\n    }\n    throw new Error(`Failed to parse action: ${trimmedResponse}`);\n  } catch (error) {\n    const errorMessage = error instanceof Error ? error.message : String(error);\n    throw new Error(\n      `Failed to parse action: ${errorMessage}; raw=\"${trimmedResponse}\"`,\n    );\n  }\n}\n\nexport function parseAutoGLMResponse(content: string): {\n  think: string;\n  content: string;\n} {\n  if (content.includes('finish(message=')) {\n    const parts = content.split('finish(message=');\n    const think = parts[0].trim();\n    const actionContent = `finish(message=${parts[1]}`;\n    return { think, content: actionContent };\n  }\n  if (content.includes('do(action=')) {\n    const parts = content.split('do(action=');\n    const think = parts[0].trim();\n    const actionContent = `do(action=${parts[1]}`;\n    return { think, content: actionContent };\n  }\n  if (content.includes('<answer>')) {\n    const parts = content.split('<answer>');\n    const think = parts[0]\n      .replace(/<think>/g, '')\n      .replace(/<\\/think>/g, '')\n      .trim();\n    const actionContent = parts[1].replace(/<\\/answer>/g, '').trim();\n    return { think, content: actionContent };\n  }\n  return { think: '', content };\n}\n\nexport function parseAutoGLMLocateResponse(rawResponse: string): {\n  think: string;\n  coordinates: { x: number; y: number } | null;\n  error?: string;\n} {\n  const { think, content: actionContent } = parseAutoGLMResponse(rawResponse);\n  if (!actionContent.startsWith('do(action=\"Tap\"')) {\n    return {\n      think,\n      coordinates: null,\n      error: `Unexpected action type in auto-glm locate response: ${actionContent}`,\n    };\n  }\n  try {\n    const elementMatch = actionContent.match(/element=\\[(\\d+),(\\d+)\\]/);\n    if (!elementMatch) {\n      return {\n        think,\n        coordinates: null,\n        error: `Failed to extract element coordinates from auto-glm response: ${actionContent}`,\n      };\n    }\n    const x = Number(elementMatch[1]);\n    const y = Number(elementMatch[2]);\n    return { think, coordinates: { x, y } };\n  } catch (e) {\n    const errorMessage = e instanceof Error ? e.message : String(e);\n    return {\n      think,\n      coordinates: null,\n      error: `Failed to parse coordinates \"${actionContent}\" with errorMessage: ${errorMessage}`,\n    };\n  }\n}\n"],"names":["debug","getDebug","extractValueAfter","src","key","idx","Error","rest","parseAction","response","trimmedResponse","text","message","actionMatch","actionType","baseAction","elementMatch","Number","startMatch","endMatch","app","durationMatch","seconds","durationMs","instruction","error","errorMessage","String","parseAutoGLMResponse","content","parts","think","actionContent","parseAutoGLMLocateResponse","rawResponse","x","y","e"],"mappings":";AAGA,MAAMA,QAAQC,SAAS;AAIhB,MAAMC,oBAAoB,CAACC,KAAaC;IAC7C,MAAMC,MAAMF,IAAI,OAAO,CAACC;IACxB,IAAIC,AAAQ,OAARA,KACF,MAAM,IAAIC,MAAM,CAAC,YAAY,EAAEF,IAAI,mBAAmB,EAAED,KAAK;IAE/D,IAAII,OAAOJ,IAAI,KAAK,CAACE,MAAMD,IAAI,MAAM,EAAE,IAAI;IAC3C,IAAIG,KAAK,QAAQ,CAAC,OAChBA,OAAOA,KAAK,KAAK,CAAC,GAAG;IAEvB,OAAOA;AACT;AAEO,SAASC,YAAYC,QAG3B;IACCT,MAAM,mBAAmBS;IACzB,IAAIC,kBAAkB;IACtB,IAAI;QACFA,kBAAkBD,SAAS,OAAO,CAAC,IAAI;QAEvC,IACEC,gBAAgB,UAAU,CAAC,uBAC3BA,gBAAgB,UAAU,CAAC,0BAC3B;YACA,MAAMC,OAAOT,kBAAkBQ,iBAAiB;YAChD,OAAO;gBACL,WAAW;gBACX,QAAQ;gBACRC;gBACA,OAAOF,SAAS,KAAK;YACvB;QACF;QAEA,IAAIC,gBAAgB,UAAU,CAAC,oBAAoB;YACjD,IAAIE,UAAUV,kBAAkBQ,iBAAiB;YACjD,IAAIE,QAAQ,QAAQ,CAAC,MAAMA,UAAUA,QAAQ,KAAK,CAAC,GAAG;YACtD,OAAO;gBACL,WAAW;gBACXA;gBACA,OAAOH,SAAS,KAAK;YACvB;QACF;QAEA,IAAIC,gBAAgB,UAAU,CAAC,QAAQ;YACrC,MAAMG,cAAcH,gBAAgB,KAAK,CAAC;YAC1C,IAAI,CAACG,aACH,MAAM,IAAIP,MACR,CAAC,mDAAmD,EAAEI,gBAAgB,CAAC,CAAC;YAE5E,MAAMI,aAAaD,WAAW,CAAC,EAAE;YAEjC,MAAME,aAAa;gBAAE,WAAW;gBAAe,OAAON,SAAS,KAAK;YAAC;YACrE,OAAQK;gBACN,KAAK;oBAAO;wBACV,MAAME,eAAeN,gBAAgB,KAAK,CAAC;wBAC3C,IAAI,CAACM,cACH,MAAM,IAAIV,MACR,CAAC,oDAAoD,EAAEI,gBAAgB,CAAC,CAAC;wBAE7E,OAAO;4BACL,GAAGK,UAAU;4BACb,QAAQ;4BACR,SAAS;gCAACE,OAAOD,YAAY,CAAC,EAAE;gCAAGC,OAAOD,YAAY,CAAC,EAAE;6BAAE;wBAC7D;oBACF;gBACA,KAAK;oBAAc;wBACjB,MAAMA,eAAeN,gBAAgB,KAAK,CAAC;wBAC3C,IAAI,CAACM,cACH,MAAM,IAAIV,MACR,CAAC,2DAA2D,EAAEI,gBAAgB,CAAC,CAAC;wBAEpF,OAAO;4BACL,GAAGK,UAAU;4BACb,QAAQ;4BACR,SAAS;gCAACE,OAAOD,YAAY,CAAC,EAAE;gCAAGC,OAAOD,YAAY,CAAC,EAAE;6BAAE;wBAC7D;oBACF;gBACA,KAAK;oBAAS;wBACZ,MAAME,aAAaR,gBAAgB,KAAK,CAAC;wBACzC,MAAMS,WAAWT,gBAAgB,KAAK,CAAC;wBACvC,IAAI,CAACQ,cAAc,CAACC,UAClB,MAAM,IAAIb,MACR,CAAC,wDAAwD,EAAEI,gBAAgB,CAAC,CAAC;wBAEjF,OAAO;4BACL,GAAGK,UAAU;4BACb,QAAQ;4BACR,OAAO;gCAACE,OAAOC,UAAU,CAAC,EAAE;gCAAGD,OAAOC,UAAU,CAAC,EAAE;6BAAE;4BACrD,KAAK;gCAACD,OAAOE,QAAQ,CAAC,EAAE;gCAAGF,OAAOE,QAAQ,CAAC,EAAE;6BAAE;wBACjD;oBACF;gBACA,KAAK;oBAAc;wBACjB,MAAMH,eAAeN,gBAAgB,KAAK,CAAC;wBAC3C,IAAI,CAACM,cACH,MAAM,IAAIV,MACR,CAAC,2DAA2D,EAAEI,gBAAgB,CAAC,CAAC;wBAEpF,OAAO;4BACL,GAAGK,UAAU;4BACb,QAAQ;4BACR,SAAS;gCAACE,OAAOD,YAAY,CAAC,EAAE;gCAAGC,OAAOD,YAAY,CAAC,EAAE;6BAAE;wBAC7D;oBACF;gBACA,KAAK;oBAAU;wBACb,MAAMI,MAAMlB,kBAAkBQ,iBAAiB;wBAC/C,OAAO;4BAAE,GAAGK,UAAU;4BAAE,QAAQ;4BAAUK;wBAAI;oBAChD;gBACA,KAAK;oBACH,OAAO;wBAAE,GAAGL,UAAU;wBAAE,QAAQ;oBAAO;gBAEzC,KAAK;oBACH,OAAO;wBAAE,GAAGA,UAAU;wBAAE,QAAQ;oBAAO;gBAEzC,KAAK;oBAAQ;wBACX,MAAMM,gBAAgBX,gBAAgB,KAAK,CACzC;wBAEF,IAAI,CAACW,eACH,MAAM,IAAIf,MACR,CAAC,0CAA0C,EAAEI,gBAAgB,CAAC,CAAC;wBAGnE,MAAMY,UAAUL,OAAO,QAAQ,CAACI,aAAa,CAAC,EAAE,EAAE;wBAClD,MAAME,aAAaD,AAAU,OAAVA;wBACnB,OAAO;4BACL,GAAGP,UAAU;4BACb,QAAQ;4BACRQ;wBACF;oBACF;gBACA,KAAK;oBACH,OAAO;wBAAE,GAAGR,UAAU;wBAAE,QAAQ;oBAAW;gBAE7C,KAAK;oBAAY;wBACf,MAAMS,cAActB,kBAClBQ,iBACA;wBAEF,OAAO;4BACL,GAAGK,UAAU;4BACb,QAAQ;4BACRS;wBACF;oBACF;gBACA,KAAK;oBAAa;wBAChB,MAAMZ,UAAUV,kBAAkBQ,iBAAiB;wBACnD,OAAO;4BACL,GAAGK,UAAU;4BACb,QAAQ;4BACRH;wBACF;oBACF;gBACA,KAAK;oBAAQ;wBACX,MAAMA,UAAUV,kBAAkBQ,iBAAiB;wBACnD,OAAO;4BACL,GAAGK,UAAU;4BACb,QAAQ;4BACRH;wBACF;oBACF;gBACA;oBACE,MAAM,IAAIN,MACR,CAAC,qBAAqB,EAAEQ,WAAW,OAAO,EAAEJ,gBAAgB,CAAC,CAAC;YAEpE;QACF;QACA,MAAM,IAAIJ,MAAM,CAAC,wBAAwB,EAAEI,iBAAiB;IAC9D,EAAE,OAAOe,OAAO;QACd,MAAMC,eAAeD,iBAAiBnB,QAAQmB,MAAM,OAAO,GAAGE,OAAOF;QACrE,MAAM,IAAInB,MACR,CAAC,wBAAwB,EAAEoB,aAAa,OAAO,EAAEhB,gBAAgB,CAAC,CAAC;IAEvE;AACF;AAEO,SAASkB,qBAAqBC,OAAe;IAIlD,IAAIA,QAAQ,QAAQ,CAAC,oBAAoB;QACvC,MAAMC,QAAQD,QAAQ,KAAK,CAAC;QAC5B,MAAME,QAAQD,KAAK,CAAC,EAAE,CAAC,IAAI;QAC3B,MAAME,gBAAgB,CAAC,eAAe,EAAEF,KAAK,CAAC,EAAE,EAAE;QAClD,OAAO;YAAEC;YAAO,SAASC;QAAc;IACzC;IACA,IAAIH,QAAQ,QAAQ,CAAC,eAAe;QAClC,MAAMC,QAAQD,QAAQ,KAAK,CAAC;QAC5B,MAAME,QAAQD,KAAK,CAAC,EAAE,CAAC,IAAI;QAC3B,MAAME,gBAAgB,CAAC,UAAU,EAAEF,KAAK,CAAC,EAAE,EAAE;QAC7C,OAAO;YAAEC;YAAO,SAASC;QAAc;IACzC;IACA,IAAIH,QAAQ,QAAQ,CAAC,aAAa;QAChC,MAAMC,QAAQD,QAAQ,KAAK,CAAC;QAC5B,MAAME,QAAQD,KAAK,CAAC,EAAE,CACnB,OAAO,CAAC,YAAY,IACpB,OAAO,CAAC,cAAc,IACtB,IAAI;QACP,MAAME,gBAAgBF,KAAK,CAAC,EAAE,CAAC,OAAO,CAAC,eAAe,IAAI,IAAI;QAC9D,OAAO;YAAEC;YAAO,SAASC;QAAc;IACzC;IACA,OAAO;QAAE,OAAO;QAAIH;IAAQ;AAC9B;AAEO,SAASI,2BAA2BC,WAAmB;IAK5D,MAAM,EAAEH,KAAK,EAAE,SAASC,aAAa,EAAE,GAAGJ,qBAAqBM;IAC/D,IAAI,CAACF,cAAc,UAAU,CAAC,oBAC5B,OAAO;QACLD;QACA,aAAa;QACb,OAAO,CAAC,oDAAoD,EAAEC,eAAe;IAC/E;IAEF,IAAI;QACF,MAAMhB,eAAegB,cAAc,KAAK,CAAC;QACzC,IAAI,CAAChB,cACH,OAAO;YACLe;YACA,aAAa;YACb,OAAO,CAAC,8DAA8D,EAAEC,eAAe;QACzF;QAEF,MAAMG,IAAIlB,OAAOD,YAAY,CAAC,EAAE;QAChC,MAAMoB,IAAInB,OAAOD,YAAY,CAAC,EAAE;QAChC,OAAO;YAAEe;YAAO,aAAa;gBAAEI;gBAAGC;YAAE;QAAE;IACxC,EAAE,OAAOC,GAAG;QACV,MAAMX,eAAeW,aAAa/B,QAAQ+B,EAAE,OAAO,GAAGV,OAAOU;QAC7D,OAAO;YACLN;YACA,aAAa;YACb,OAAO,CAAC,6BAA6B,EAAEC,cAAc,qBAAqB,EAAEN,cAAc;QAC5F;IACF;AACF"}
+{"version":3,"file":"ai-model/models/auto-glm/parser.mjs","sources":["../../../../../src/ai-model/models/auto-glm/parser.ts"],"sourcesContent":["import { getDebug } from '@midscene/shared/logger';\nimport type { AutoGLMParsedAction } from './actions';\n\nconst debug = getDebug('auto-glm-parser');\n\n// Do not rely on regex extraction here; regex can fail on malformed input.\n// Bad Case: finish(message=\"Finished! Now There is a contact whose name is \\\"Tom\\\" in the list.\")\nexport const extractValueAfter = (src: string, key: string): string => {\n  const idx = src.indexOf(key);\n  if (idx === -1) {\n    throw new Error(`Missing key ${key} in action payload ${src}`);\n  }\n  let rest = src.slice(idx + key.length).trim();\n  if (rest.endsWith('\")')) {\n    rest = rest.slice(0, -2);\n  }\n  return rest;\n};\n\nexport function parseAutoGLMPlanningAction(response: {\n  think: string;\n  content: string;\n}): AutoGLMParsedAction {\n  debug('Parsing action:', response);\n  let trimmedResponse = '';\n  try {\n    trimmedResponse = response.content.trim();\n\n    if (\n      trimmedResponse.startsWith('do(action=\"Type\"') ||\n      trimmedResponse.startsWith('do(action=\"Type_Name\"')\n    ) {\n      const text = extractValueAfter(trimmedResponse, 'text=\"');\n      return {\n        _metadata: 'do',\n        action: 'Type',\n        text,\n        think: response.think,\n      } as AutoGLMParsedAction;\n    }\n\n    if (trimmedResponse.startsWith('finish(message=')) {\n      let message = extractValueAfter(trimmedResponse, 'finish(message=\"');\n      if (message.endsWith(')')) message = message.slice(0, -1);\n      return {\n        _metadata: 'finish',\n        message,\n        think: response.think,\n      } as AutoGLMParsedAction;\n    }\n\n    if (trimmedResponse.startsWith('do(')) {\n      const actionMatch = trimmedResponse.match(/do\\(action=\"([^\"]+)\"/);\n      if (!actionMatch)\n        throw new Error(\n          `Failed to extract action type from do() call; raw=\"${trimmedResponse}\"`,\n        );\n      const actionType = actionMatch[1];\n\n      const baseAction = { _metadata: 'do' as const, think: response.think };\n      switch (actionType) {\n        case 'Tap': {\n          const elementMatch = trimmedResponse.match(/element=\\[(\\d+),(\\d+)\\]/);\n          if (!elementMatch)\n            throw new Error(\n              `Failed to extract element coordinates for Tap; raw=\"${trimmedResponse}\"`,\n            );\n          return {\n            ...baseAction,\n            action: 'Tap',\n            element: [Number(elementMatch[1]), Number(elementMatch[2])],\n          } as AutoGLMParsedAction;\n        }\n        case 'Double Tap': {\n          const elementMatch = trimmedResponse.match(/element=\\[(\\d+),(\\d+)\\]/);\n          if (!elementMatch)\n            throw new Error(\n              `Failed to extract element coordinates for Double Tap; raw=\"${trimmedResponse}\"`,\n            );\n          return {\n            ...baseAction,\n            action: 'Double Tap',\n            element: [Number(elementMatch[1]), Number(elementMatch[2])],\n          } as AutoGLMParsedAction;\n        }\n        case 'Swipe': {\n          const startMatch = trimmedResponse.match(/start=\\[(\\d+),(\\d+)\\]/);\n          const endMatch = trimmedResponse.match(/end=\\[(\\d+),(\\d+)\\]/);\n          if (!startMatch || !endMatch)\n            throw new Error(\n              `Failed to extract start/end coordinates for Swipe; raw=\"${trimmedResponse}\"`,\n            );\n          return {\n            ...baseAction,\n            action: 'Swipe',\n            start: [Number(startMatch[1]), Number(startMatch[2])],\n            end: [Number(endMatch[1]), Number(endMatch[2])],\n          } as AutoGLMParsedAction;\n        }\n        case 'Long Press': {\n          const elementMatch = trimmedResponse.match(/element=\\[(\\d+),(\\d+)\\]/);\n          if (!elementMatch)\n            throw new Error(\n              `Failed to extract element coordinates for Long Press; raw=\"${trimmedResponse}\"`,\n            );\n          return {\n            ...baseAction,\n            action: 'Long Press',\n            element: [Number(elementMatch[1]), Number(elementMatch[2])],\n          } as AutoGLMParsedAction;\n        }\n        case 'Launch': {\n          const app = extractValueAfter(trimmedResponse, 'app=\"');\n          return {\n            ...baseAction,\n            action: 'Launch',\n            app,\n          } as AutoGLMParsedAction;\n        }\n        case 'Back': {\n          return { ...baseAction, action: 'Back' } as AutoGLMParsedAction;\n        }\n        case 'Home': {\n          return { ...baseAction, action: 'Home' } as AutoGLMParsedAction;\n        }\n        case 'Wait': {\n          const durationMatch = trimmedResponse.match(\n            /duration=(?:[\"\\[])?(\\d+)/,\n          );\n          if (!durationMatch) {\n            throw new Error(\n              `Failed to extract duration for Wait; raw=\"${trimmedResponse}\"`,\n            );\n          }\n          const seconds = Number.parseInt(durationMatch[1], 10);\n          const durationMs = seconds * 1000;\n          return {\n            ...baseAction,\n            action: 'Wait',\n            durationMs,\n          } as AutoGLMParsedAction;\n        }\n        case 'Interact': {\n          return { ...baseAction, action: 'Interact' } as AutoGLMParsedAction;\n        }\n        case 'Call_API': {\n          const instruction = extractValueAfter(\n            trimmedResponse,\n            'instruction=\"',\n          );\n          return {\n            ...baseAction,\n            action: 'Call_API',\n            instruction,\n          } as AutoGLMParsedAction;\n        }\n        case 'Take_over': {\n          const message = extractValueAfter(trimmedResponse, 'message=\"');\n          return {\n            ...baseAction,\n            action: 'Take_over',\n            message,\n          } as AutoGLMParsedAction;\n        }\n        case 'Note': {\n          const message = extractValueAfter(trimmedResponse, 'message=\"');\n          return {\n            ...baseAction,\n            action: 'Note',\n            message,\n          } as AutoGLMParsedAction;\n        }\n        default:\n          throw new Error(\n            `Unknown action type: ${actionType}; raw=\"${trimmedResponse}\"`,\n          );\n      }\n    }\n    throw new Error(`Failed to parse action: ${trimmedResponse}`);\n  } catch (error) {\n    const errorMessage = error instanceof Error ? error.message : String(error);\n    throw new Error(\n      `Failed to parse action: ${errorMessage}; raw=\"${trimmedResponse}\"`,\n    );\n  }\n}\n\nexport function parseAutoGLMResponse(content: string): {\n  think: string;\n  content: string;\n} {\n  let parsedResponse: { think: string; content: string };\n\n  if (content.includes('finish(message=')) {\n    const parts = content.split('finish(message=');\n    const think = parts[0].trim();\n    const actionContent = `finish(message=${parts[1]}`;\n    parsedResponse = { think, content: actionContent };\n  } else if (content.includes('do(action=')) {\n    const parts = content.split('do(action=');\n    const think = parts[0].trim();\n    const actionContent = `do(action=${parts[1]}`;\n    parsedResponse = { think, content: actionContent };\n  } else if (content.includes('<answer>')) {\n    const parts = content.split('<answer>');\n    const think = parts[0]\n      .replace(/<think>/g, '')\n      .replace(/<\\/think>/g, '')\n      .trim();\n    const actionContent = parts[1].replace(/<\\/answer>/g, '').trim();\n    parsedResponse = { think, content: actionContent };\n  } else {\n    parsedResponse = { think: '', content };\n  }\n\n  debug('autoGLM rawResponse:', content);\n  debug('thinking in response:', parsedResponse.think);\n  debug('action in response:', parsedResponse.content);\n  return parsedResponse;\n}\n\nexport function parseAutoGLMPlanningResponse(content: string): {\n  response: ReturnType<typeof parseAutoGLMResponse>;\n  action: AutoGLMParsedAction;\n} {\n  const response = parseAutoGLMResponse(content);\n  const action = parseAutoGLMPlanningAction(response);\n  return { response, action };\n}\n"],"names":["debug","getDebug","extractValueAfter","src","key","idx","Error","rest","parseAutoGLMPlanningAction","response","trimmedResponse","text","message","actionMatch","actionType","baseAction","elementMatch","Number","startMatch","endMatch","app","durationMatch","seconds","durationMs","instruction","error","errorMessage","String","parseAutoGLMResponse","content","parsedResponse","parts","think","actionContent","parseAutoGLMPlanningResponse","action"],"mappings":";AAGA,MAAMA,QAAQC,SAAS;AAIhB,MAAMC,oBAAoB,CAACC,KAAaC;IAC7C,MAAMC,MAAMF,IAAI,OAAO,CAACC;IACxB,IAAIC,AAAQ,OAARA,KACF,MAAM,IAAIC,MAAM,CAAC,YAAY,EAAEF,IAAI,mBAAmB,EAAED,KAAK;IAE/D,IAAII,OAAOJ,IAAI,KAAK,CAACE,MAAMD,IAAI,MAAM,EAAE,IAAI;IAC3C,IAAIG,KAAK,QAAQ,CAAC,OAChBA,OAAOA,KAAK,KAAK,CAAC,GAAG;IAEvB,OAAOA;AACT;AAEO,SAASC,2BAA2BC,QAG1C;IACCT,MAAM,mBAAmBS;IACzB,IAAIC,kBAAkB;IACtB,IAAI;QACFA,kBAAkBD,SAAS,OAAO,CAAC,IAAI;QAEvC,IACEC,gBAAgB,UAAU,CAAC,uBAC3BA,gBAAgB,UAAU,CAAC,0BAC3B;YACA,MAAMC,OAAOT,kBAAkBQ,iBAAiB;YAChD,OAAO;gBACL,WAAW;gBACX,QAAQ;gBACRC;gBACA,OAAOF,SAAS,KAAK;YACvB;QACF;QAEA,IAAIC,gBAAgB,UAAU,CAAC,oBAAoB;YACjD,IAAIE,UAAUV,kBAAkBQ,iBAAiB;YACjD,IAAIE,QAAQ,QAAQ,CAAC,MAAMA,UAAUA,QAAQ,KAAK,CAAC,GAAG;YACtD,OAAO;gBACL,WAAW;gBACXA;gBACA,OAAOH,SAAS,KAAK;YACvB;QACF;QAEA,IAAIC,gBAAgB,UAAU,CAAC,QAAQ;YACrC,MAAMG,cAAcH,gBAAgB,KAAK,CAAC;YAC1C,IAAI,CAACG,aACH,MAAM,IAAIP,MACR,CAAC,mDAAmD,EAAEI,gBAAgB,CAAC,CAAC;YAE5E,MAAMI,aAAaD,WAAW,CAAC,EAAE;YAEjC,MAAME,aAAa;gBAAE,WAAW;gBAAe,OAAON,SAAS,KAAK;YAAC;YACrE,OAAQK;gBACN,KAAK;oBAAO;wBACV,MAAME,eAAeN,gBAAgB,KAAK,CAAC;wBAC3C,IAAI,CAACM,cACH,MAAM,IAAIV,MACR,CAAC,oDAAoD,EAAEI,gBAAgB,CAAC,CAAC;wBAE7E,OAAO;4BACL,GAAGK,UAAU;4BACb,QAAQ;4BACR,SAAS;gCAACE,OAAOD,YAAY,CAAC,EAAE;gCAAGC,OAAOD,YAAY,CAAC,EAAE;6BAAE;wBAC7D;oBACF;gBACA,KAAK;oBAAc;wBACjB,MAAMA,eAAeN,gBAAgB,KAAK,CAAC;wBAC3C,IAAI,CAACM,cACH,MAAM,IAAIV,MACR,CAAC,2DAA2D,EAAEI,gBAAgB,CAAC,CAAC;wBAEpF,OAAO;4BACL,GAAGK,UAAU;4BACb,QAAQ;4BACR,SAAS;gCAACE,OAAOD,YAAY,CAAC,EAAE;gCAAGC,OAAOD,YAAY,CAAC,EAAE;6BAAE;wBAC7D;oBACF;gBACA,KAAK;oBAAS;wBACZ,MAAME,aAAaR,gBAAgB,KAAK,CAAC;wBACzC,MAAMS,WAAWT,gBAAgB,KAAK,CAAC;wBACvC,IAAI,CAACQ,cAAc,CAACC,UAClB,MAAM,IAAIb,MACR,CAAC,wDAAwD,EAAEI,gBAAgB,CAAC,CAAC;wBAEjF,OAAO;4BACL,GAAGK,UAAU;4BACb,QAAQ;4BACR,OAAO;gCAACE,OAAOC,UAAU,CAAC,EAAE;gCAAGD,OAAOC,UAAU,CAAC,EAAE;6BAAE;4BACrD,KAAK;gCAACD,OAAOE,QAAQ,CAAC,EAAE;gCAAGF,OAAOE,QAAQ,CAAC,EAAE;6BAAE;wBACjD;oBACF;gBACA,KAAK;oBAAc;wBACjB,MAAMH,eAAeN,gBAAgB,KAAK,CAAC;wBAC3C,IAAI,CAACM,cACH,MAAM,IAAIV,MACR,CAAC,2DAA2D,EAAEI,gBAAgB,CAAC,CAAC;wBAEpF,OAAO;4BACL,GAAGK,UAAU;4BACb,QAAQ;4BACR,SAAS;gCAACE,OAAOD,YAAY,CAAC,EAAE;gCAAGC,OAAOD,YAAY,CAAC,EAAE;6BAAE;wBAC7D;oBACF;gBACA,KAAK;oBAAU;wBACb,MAAMI,MAAMlB,kBAAkBQ,iBAAiB;wBAC/C,OAAO;4BACL,GAAGK,UAAU;4BACb,QAAQ;4BACRK;wBACF;oBACF;gBACA,KAAK;oBACH,OAAO;wBAAE,GAAGL,UAAU;wBAAE,QAAQ;oBAAO;gBAEzC,KAAK;oBACH,OAAO;wBAAE,GAAGA,UAAU;wBAAE,QAAQ;oBAAO;gBAEzC,KAAK;oBAAQ;wBACX,MAAMM,gBAAgBX,gBAAgB,KAAK,CACzC;wBAEF,IAAI,CAACW,eACH,MAAM,IAAIf,MACR,CAAC,0CAA0C,EAAEI,gBAAgB,CAAC,CAAC;wBAGnE,MAAMY,UAAUL,OAAO,QAAQ,CAACI,aAAa,CAAC,EAAE,EAAE;wBAClD,MAAME,aAAaD,AAAU,OAAVA;wBACnB,OAAO;4BACL,GAAGP,UAAU;4BACb,QAAQ;4BACRQ;wBACF;oBACF;gBACA,KAAK;oBACH,OAAO;wBAAE,GAAGR,UAAU;wBAAE,QAAQ;oBAAW;gBAE7C,KAAK;oBAAY;wBACf,MAAMS,cAActB,kBAClBQ,iBACA;wBAEF,OAAO;4BACL,GAAGK,UAAU;4BACb,QAAQ;4BACRS;wBACF;oBACF;gBACA,KAAK;oBAAa;wBAChB,MAAMZ,UAAUV,kBAAkBQ,iBAAiB;wBACnD,OAAO;4BACL,GAAGK,UAAU;4BACb,QAAQ;4BACRH;wBACF;oBACF;gBACA,KAAK;oBAAQ;wBACX,MAAMA,UAAUV,kBAAkBQ,iBAAiB;wBACnD,OAAO;4BACL,GAAGK,UAAU;4BACb,QAAQ;4BACRH;wBACF;oBACF;gBACA;oBACE,MAAM,IAAIN,MACR,CAAC,qBAAqB,EAAEQ,WAAW,OAAO,EAAEJ,gBAAgB,CAAC,CAAC;YAEpE;QACF;QACA,MAAM,IAAIJ,MAAM,CAAC,wBAAwB,EAAEI,iBAAiB;IAC9D,EAAE,OAAOe,OAAO;QACd,MAAMC,eAAeD,iBAAiBnB,QAAQmB,MAAM,OAAO,GAAGE,OAAOF;QACrE,MAAM,IAAInB,MACR,CAAC,wBAAwB,EAAEoB,aAAa,OAAO,EAAEhB,gBAAgB,CAAC,CAAC;IAEvE;AACF;AAEO,SAASkB,qBAAqBC,OAAe;IAIlD,IAAIC;IAEJ,IAAID,QAAQ,QAAQ,CAAC,oBAAoB;QACvC,MAAME,QAAQF,QAAQ,KAAK,CAAC;QAC5B,MAAMG,QAAQD,KAAK,CAAC,EAAE,CAAC,IAAI;QAC3B,MAAME,gBAAgB,CAAC,eAAe,EAAEF,KAAK,CAAC,EAAE,EAAE;QAClDD,iBAAiB;YAAEE;YAAO,SAASC;QAAc;IACnD,OAAO,IAAIJ,QAAQ,QAAQ,CAAC,eAAe;QACzC,MAAME,QAAQF,QAAQ,KAAK,CAAC;QAC5B,MAAMG,QAAQD,KAAK,CAAC,EAAE,CAAC,IAAI;QAC3B,MAAME,gBAAgB,CAAC,UAAU,EAAEF,KAAK,CAAC,EAAE,EAAE;QAC7CD,iBAAiB;YAAEE;YAAO,SAASC;QAAc;IACnD,OAAO,IAAIJ,QAAQ,QAAQ,CAAC,aAAa;QACvC,MAAME,QAAQF,QAAQ,KAAK,CAAC;QAC5B,MAAMG,QAAQD,KAAK,CAAC,EAAE,CACnB,OAAO,CAAC,YAAY,IACpB,OAAO,CAAC,cAAc,IACtB,IAAI;QACP,MAAME,gBAAgBF,KAAK,CAAC,EAAE,CAAC,OAAO,CAAC,eAAe,IAAI,IAAI;QAC9DD,iBAAiB;YAAEE;YAAO,SAASC;QAAc;IACnD,OACEH,iBAAiB;QAAE,OAAO;QAAID;IAAQ;IAGxC7B,MAAM,wBAAwB6B;IAC9B7B,MAAM,yBAAyB8B,eAAe,KAAK;IACnD9B,MAAM,uBAAuB8B,eAAe,OAAO;IACnD,OAAOA;AACT;AAEO,SAASI,6BAA6BL,OAAe;IAI1D,MAAMpB,WAAWmB,qBAAqBC;IACtC,MAAMM,SAAS3B,2BAA2BC;IAC1C,OAAO;QAAEA;QAAU0B;IAAO;AAC5B"}

package/dist/es/ai-model/models/auto-glm/planning.mjs

@@ -1,76 +1,33 @@
-import { userPromptToString } from "../../../common.mjs";
-import { getDebug } from "@midscene/shared/logger";
-import { AIResponseParseError, callAIWithStringResponse } from "../../service-caller/index.mjs";
+import { assert } from "@midscene/shared/utils";
+import { createCoordinateDistanceToPixels } from "../../shared/model-locate-result/index.mjs";
 import { transformAutoGLMAction } from "./actions.mjs";
-import { parseAction, parseAutoGLMResponse } from "./parser.mjs";
-const debug = getDebug('auto-glm-planning');
-async function autoGlmPlanning(userInstruction, options, getSystemPrompt) {
-    const { conversationHistory, context, actionContext } = options;
-    const systemPrompt = getSystemPrompt() + (actionContext ? `<high_priority_knowledge>${actionContext}</high_priority_knowledge>` : '');
-    const imagePayloadBase64 = context.screenshot.base64;
-    const userInstructionText = userPromptToString(userInstruction);
-    const referenceImageMessages = options.referenceImageMessages ?? [];
-    const userInstructionMessage = {
-        role: 'user',
-        content: [
-            {
-                type: 'text',
-                text: userInstructionText
-            }
-        ]
-    };
-    conversationHistory.append({
-        role: 'user',
-        content: [
-            {
-                type: 'image_url',
-                image_url: {
-                    url: imagePayloadBase64
-                }
-            }
-        ]
-    });
-    const msgs = [
-        {
-            role: 'system',
-            content: systemPrompt
-        },
-        userInstructionMessage,
-        ...referenceImageMessages,
-        ...conversationHistory.snapshot(1)
-    ];
-    const { content: rawResponse, usage, rawChoiceMessage } = await callAIWithStringResponse(msgs, options.modelRuntime, {
-        abortSignal: options.abortSignal
-    });
-    debug('autoGLMPlanning rawResponse:', rawResponse);
-    let parsedResponse;
-    let transformedActions;
-    try {
-        parsedResponse = parseAutoGLMResponse(rawResponse);
-        debug('thinking in response:', parsedResponse.think);
-        debug('action in response:', parsedResponse.content);
-        const parsedAction = parseAction(parsedResponse);
-        debug('Parsed action object:', parsedAction);
-        transformedActions = transformAutoGLMAction(parsedAction, context.shotSize, options.actionSpace);
-        debug('Transformed actions:', transformedActions);
-    } catch (parseError) {
-        const errorMessage = parseError instanceof Error ? parseError.message : String(parseError);
-        throw new AIResponseParseError(`Parse error: ${errorMessage}`, JSON.stringify(rawResponse, void 0, 2), usage, rawChoiceMessage);
-    }
-    conversationHistory.append({
-        role: 'assistant',
-        content: `<think>${parsedResponse.think}</think><answer>${parsedResponse.content}</answer>`
-    });
-    const shouldContinuePlanning = !parsedResponse.content.startsWith('finish(');
+import { parseAutoGLMPlanningResponse } from "./parser.mjs";
+import { getAutoGLMChinesePlanPrompt, getAutoGLMMultilingualPlanPrompt } from "./prompt.mjs";
+function createAutoGlmPlanner(isMultilingual) {
     return {
-        actions: transformedActions,
-        log: rawResponse,
-        usage,
-        shouldContinuePlanning,
-        rawResponse: JSON.stringify(rawResponse, void 0, 2),
-        rawChoiceMessage
+        messages: {
+            systemPromptPlacement: 'system-message',
+            buildSystemPrompt: ()=>isMultilingual ? getAutoGLMMultilingualPlanPrompt() : getAutoGLMChinesePlanPrompt(),
+            historyImageLimit: 1,
+            buildAssistantContent: (parsedResponse)=>`<think>${parsedResponse.response.think}</think><answer>${parsedResponse.response.content}</answer>`
+        },
+        coordinates: {
+            shape: 'point',
+            order: 'xy',
+            normalizedBy: 1000
+        },
+        parseResponse: (rawResponse)=>parseAutoGLMPlanningResponse(rawResponse),
+        transformActions: (parsedResponse, { options, coordinateSystem })=>{
+            assert(coordinateSystem, 'Auto-GLM planning requires coordinate system');
+            return transformAutoGLMAction(parsedResponse.action, {
+                actionSpace: options.actionSpace,
+                coordinateDistanceToPixels: createCoordinateDistanceToPixels(options.context.shotSize, coordinateSystem)
+            });
+        },
+        shouldContinuePlanning: (parsedResponse)=>'finish' !== parsedResponse.action._metadata,
+        buildResponseLog: (_parsedResponse, rawResponse)=>rawResponse
     };
 }
-export { autoGlmPlanning };
+export { createAutoGlmPlanner };
 
 //# sourceMappingURL=planning.mjs.map

package/dist/es/ai-model/models/auto-glm/planning.mjs.map

@@ -1 +1 @@
-{"version":3,"file":"ai-model/models/auto-glm/planning.mjs","sources":["../../../../../src/ai-model/models/auto-glm/planning.ts"],"sourcesContent":["import { type TUserPrompt, userPromptToString } from '@/common';\nimport type { PlanningAIResponse } from '@/types';\nimport { getDebug } from '@midscene/shared/logger';\nimport type { ChatCompletionMessageParam } from 'openai/resources/index';\nimport {\n  AIResponseParseError,\n  callAIWithStringResponse,\n} from '../../service-caller/index';\nimport type { PlanOptions } from '../../workflows/planning/types';\nimport { transformAutoGLMAction } from './actions';\nimport { parseAction, parseAutoGLMResponse } from './parser';\n\nconst debug = getDebug('auto-glm-planning');\n\nexport async function autoGlmPlanning(\n  userInstruction: TUserPrompt,\n  options: PlanOptions,\n  getSystemPrompt: () => string,\n): Promise<PlanningAIResponse> {\n  const { conversationHistory, context, actionContext } = options;\n\n  const systemPrompt =\n    getSystemPrompt() +\n    (actionContext\n      ? `<high_priority_knowledge>${actionContext}</high_priority_knowledge>`\n      : '');\n\n  const imagePayloadBase64 = context.screenshot.base64;\n  const userInstructionText = userPromptToString(userInstruction);\n  const referenceImageMessages = options.referenceImageMessages ?? [];\n\n  const userInstructionMessage: ChatCompletionMessageParam = {\n    role: 'user',\n    content: [{ type: 'text', text: userInstructionText }],\n  };\n  conversationHistory.append({\n    role: 'user',\n    content: [{ type: 'image_url', image_url: { url: imagePayloadBase64 } }],\n  });\n\n  const msgs: ChatCompletionMessageParam[] = [\n    { role: 'system', content: systemPrompt },\n    userInstructionMessage,\n    ...referenceImageMessages,\n    ...conversationHistory.snapshot(1),\n  ];\n\n  const {\n    content: rawResponse,\n    usage,\n    rawChoiceMessage,\n  } = await callAIWithStringResponse(msgs, options.modelRuntime, {\n    abortSignal: options.abortSignal,\n  });\n\n  debug('autoGLMPlanning rawResponse:', rawResponse);\n\n  let parsedResponse: ReturnType<typeof parseAutoGLMResponse>;\n  let transformedActions: ReturnType<typeof transformAutoGLMAction>;\n\n  try {\n    parsedResponse = parseAutoGLMResponse(rawResponse);\n    debug('thinking in response:', parsedResponse.think);\n    debug('action in response:', parsedResponse.content);\n\n    const parsedAction = parseAction(parsedResponse);\n    debug('Parsed action object:', parsedAction);\n    transformedActions = transformAutoGLMAction(\n      parsedAction,\n      context.shotSize,\n      options.actionSpace,\n    );\n    debug('Transformed actions:', transformedActions);\n  } catch (parseError) {\n    // Throw AIResponseParseError with usage and rawResponse preserved\n    const errorMessage =\n      parseError instanceof Error ? parseError.message : String(parseError);\n    throw new AIResponseParseError(\n      `Parse error: ${errorMessage}`,\n      JSON.stringify(rawResponse, undefined, 2),\n      usage,\n      rawChoiceMessage,\n    );\n  }\n\n  conversationHistory.append({\n    role: 'assistant',\n    content: `<think>${parsedResponse.think}</think><answer>${parsedResponse.content}</answer>`,\n  });\n\n  const shouldContinuePlanning = !parsedResponse.content.startsWith('finish(');\n\n  return {\n    actions: transformedActions,\n    log: rawResponse,\n    usage,\n    shouldContinuePlanning,\n    rawResponse: JSON.stringify(rawResponse, undefined, 2),\n    rawChoiceMessage,\n  };\n}\n"],"names":["debug","getDebug","autoGlmPlanning","userInstruction","options","getSystemPrompt","conversationHistory","context","actionContext","systemPrompt","imagePayloadBase64","userInstructionText","userPromptToString","referenceImageMessages","userInstructionMessage","msgs","rawResponse","usage","rawChoiceMessage","callAIWithStringResponse","parsedResponse","transformedActions","parseAutoGLMResponse","parsedAction","parseAction","transformAutoGLMAction","parseError","errorMessage","Error","String","AIResponseParseError","JSON","undefined","shouldContinuePlanning"],"mappings":";;;;;AAYA,MAAMA,QAAQC,SAAS;AAEhB,eAAeC,gBACpBC,eAA4B,EAC5BC,OAAoB,EACpBC,eAA6B;IAE7B,MAAM,EAAEC,mBAAmB,EAAEC,OAAO,EAAEC,aAAa,EAAE,GAAGJ;IAExD,MAAMK,eACJJ,oBACCG,CAAAA,gBACG,CAAC,yBAAyB,EAAEA,cAAc,0BAA0B,CAAC,GACrE,EAAC;IAEP,MAAME,qBAAqBH,QAAQ,UAAU,CAAC,MAAM;IACpD,MAAMI,sBAAsBC,mBAAmBT;IAC/C,MAAMU,yBAAyBT,QAAQ,sBAAsB,IAAI,EAAE;IAEnE,MAAMU,yBAAqD;QACzD,MAAM;QACN,SAAS;YAAC;gBAAE,MAAM;gBAAQ,MAAMH;YAAoB;SAAE;IACxD;IACAL,oBAAoB,MAAM,CAAC;QACzB,MAAM;QACN,SAAS;YAAC;gBAAE,MAAM;gBAAa,WAAW;oBAAE,KAAKI;gBAAmB;YAAE;SAAE;IAC1E;IAEA,MAAMK,OAAqC;QACzC;YAAE,MAAM;YAAU,SAASN;QAAa;QACxCK;WACGD;WACAP,oBAAoB,QAAQ,CAAC;KACjC;IAED,MAAM,EACJ,SAASU,WAAW,EACpBC,KAAK,EACLC,gBAAgB,EACjB,GAAG,MAAMC,yBAAyBJ,MAAMX,QAAQ,YAAY,EAAE;QAC7D,aAAaA,QAAQ,WAAW;IAClC;IAEAJ,MAAM,gCAAgCgB;IAEtC,IAAII;IACJ,IAAIC;IAEJ,IAAI;QACFD,iBAAiBE,qBAAqBN;QACtChB,MAAM,yBAAyBoB,eAAe,KAAK;QACnDpB,MAAM,uBAAuBoB,eAAe,OAAO;QAEnD,MAAMG,eAAeC,YAAYJ;QACjCpB,MAAM,yBAAyBuB;QAC/BF,qBAAqBI,uBACnBF,cACAhB,QAAQ,QAAQ,EAChBH,QAAQ,WAAW;QAErBJ,MAAM,wBAAwBqB;IAChC,EAAE,OAAOK,YAAY;QAEnB,MAAMC,eACJD,sBAAsBE,QAAQF,WAAW,OAAO,GAAGG,OAAOH;QAC5D,MAAM,IAAII,qBACR,CAAC,aAAa,EAAEH,cAAc,EAC9BI,KAAK,SAAS,CAACf,aAAagB,QAAW,IACvCf,OACAC;IAEJ;IAEAZ,oBAAoB,MAAM,CAAC;QACzB,MAAM;QACN,SAAS,CAAC,OAAO,EAAEc,eAAe,KAAK,CAAC,gBAAgB,EAAEA,eAAe,OAAO,CAAC,SAAS,CAAC;IAC7F;IAEA,MAAMa,yBAAyB,CAACb,eAAe,OAAO,CAAC,UAAU,CAAC;IAElE,OAAO;QACL,SAASC;QACT,KAAKL;QACLC;QACAgB;QACA,aAAaF,KAAK,SAAS,CAACf,aAAagB,QAAW;QACpDd;IACF;AACF"}
+{"version":3,"file":"ai-model/models/auto-glm/planning.mjs","sources":["../../../../../src/ai-model/models/auto-glm/planning.ts"],"sourcesContent":["import { assert } from '@midscene/shared/utils';\nimport type { CustomPlanningDefinition } from '../../model-adapter/custom-planning-types';\nimport { createCoordinateDistanceToPixels } from '../../shared/model-locate-result';\nimport { transformAutoGLMAction } from './actions';\nimport { parseAutoGLMPlanningResponse } from './parser';\nimport {\n  getAutoGLMChinesePlanPrompt,\n  getAutoGLMMultilingualPlanPrompt,\n} from './prompt';\n\ntype AutoGLMParsedResponse = ReturnType<typeof parseAutoGLMPlanningResponse>;\n\nexport function createAutoGlmPlanner(\n  isMultilingual: boolean,\n): CustomPlanningDefinition<AutoGLMParsedResponse> {\n  return {\n    messages: {\n      systemPromptPlacement: 'system-message',\n      buildSystemPrompt: () =>\n        isMultilingual\n          ? getAutoGLMMultilingualPlanPrompt()\n          : getAutoGLMChinesePlanPrompt(),\n      historyImageLimit: 1,\n      buildAssistantContent: (parsedResponse) =>\n        `<think>${parsedResponse.response.think}</think><answer>${parsedResponse.response.content}</answer>`,\n    },\n    coordinates: {\n      shape: 'point',\n      order: 'xy',\n      normalizedBy: 1000,\n    },\n    parseResponse: (rawResponse) => {\n      return parseAutoGLMPlanningResponse(rawResponse);\n    },\n    transformActions: (parsedResponse, { options, coordinateSystem }) => {\n      assert(coordinateSystem, 'Auto-GLM planning requires coordinate system');\n      return transformAutoGLMAction(parsedResponse.action, {\n        actionSpace: options.actionSpace,\n        coordinateDistanceToPixels: createCoordinateDistanceToPixels(\n          options.context.shotSize,\n          coordinateSystem,\n        ),\n      });\n    },\n    shouldContinuePlanning: (parsedResponse) =>\n      parsedResponse.action._metadata !== 'finish',\n    buildResponseLog: (_parsedResponse, rawResponse) => rawResponse,\n  };\n}\n"],"names":["createAutoGlmPlanner","isMultilingual","getAutoGLMMultilingualPlanPrompt","getAutoGLMChinesePlanPrompt","parsedResponse","rawResponse","parseAutoGLMPlanningResponse","options","coordinateSystem","assert","transformAutoGLMAction","createCoordinateDistanceToPixels","_parsedResponse"],"mappings":";;;;;AAYO,SAASA,qBACdC,cAAuB;IAEvB,OAAO;QACL,UAAU;YACR,uBAAuB;YACvB,mBAAmB,IACjBA,iBACIC,qCACAC;YACN,mBAAmB;YACnB,uBAAuB,CAACC,iBACtB,CAAC,OAAO,EAAEA,eAAe,QAAQ,CAAC,KAAK,CAAC,gBAAgB,EAAEA,eAAe,QAAQ,CAAC,OAAO,CAAC,SAAS,CAAC;QACxG;QACA,aAAa;YACX,OAAO;YACP,OAAO;YACP,cAAc;QAChB;QACA,eAAe,CAACC,cACPC,6BAA6BD;QAEtC,kBAAkB,CAACD,gBAAgB,EAAEG,OAAO,EAAEC,gBAAgB,EAAE;YAC9DC,OAAOD,kBAAkB;YACzB,OAAOE,uBAAuBN,eAAe,MAAM,EAAE;gBACnD,aAAaG,QAAQ,WAAW;gBAChC,4BAA4BI,iCAC1BJ,QAAQ,OAAO,CAAC,QAAQ,EACxBC;YAEJ;QACF;QACA,wBAAwB,CAACJ,iBACvBA,AAAoC,aAApCA,eAAe,MAAM,CAAC,SAAS;QACjC,kBAAkB,CAACQ,iBAAiBP,cAAgBA;IACtD;AACF"}

package/dist/es/ai-model/models/default.mjs.map

@@ -1 +1 @@
-{"version":3,"file":"ai-model/models/default.mjs","sources":["../../../../src/ai-model/models/default.ts"],"sourcesContent":["import type { ModelAdapterDefinition } from './types';\n\nexport const defaultOpenAICompatibleAdapterConfig: ModelAdapterDefinition = {\n  chatCompletion: {\n    unsupportedUserConfig: [\n      'reasoningEnabled',\n      'reasoningEffort',\n      'reasoningBudget',\n    ],\n  },\n};\n"],"names":["defaultOpenAICompatibleAdapterConfig"],"mappings":"AAEO,MAAMA,uCAA+D;IAC1E,gBAAgB;QACd,uBAAuB;YACrB;YACA;YACA;SACD;IACH;AACF"}
+{"version":3,"file":"ai-model/models/default.mjs","sources":["../../../../src/ai-model/models/default.ts"],"sourcesContent":["import type { ModelAdapterDefinition } from '../model-adapter/types';\n\nexport const defaultOpenAICompatibleAdapterConfig: ModelAdapterDefinition = {\n  chatCompletion: {\n    unsupportedUserConfig: [\n      'reasoningEnabled',\n      'reasoningEffort',\n      'reasoningBudget',\n    ],\n  },\n};\n"],"names":["defaultOpenAICompatibleAdapterConfig"],"mappings":"AAEO,MAAMA,uCAA+D;IAC1E,gBAAgB;QACd,uBAAuB;YACrB;YACA;YACA;SACD;IACH;AACF"}

package/dist/es/ai-model/models/doubao.mjs.map

@@ -1 +1 @@
-{"version":3,"file":"ai-model/models/doubao.mjs","sources":["../../../../src/ai-model/models/doubao.ts"],"sourcesContent":["import type { TModelFamily } from '@midscene/shared/env';\nimport { assert } from '@midscene/shared/utils';\nimport { jsonrepair } from 'jsonrepair';\nimport {\n  extractJSONFromCodeBlock,\n  safeParseJson,\n} from '../service-caller/json';\nimport {\n  type LocateResultValue,\n  unwrapCoordinateListLikeInput,\n} from '../shared/model-locate-result';\nimport type {\n  ChatCompletionCallContext,\n  ChatCompletionParamsResult,\n  JsonParserContext,\n  JsonParserSource,\n  ModelAdapterDefinition,\n} from './types';\n\nexport function normalizeDoubaoJsonObject(\n  obj: any,\n  context: Pick<JsonParserContext, 'preserveStringValueKeys'> = {},\n): any {\n  if (obj === null || obj === undefined) {\n    return obj;\n  }\n\n  if (Array.isArray(obj)) {\n    return obj.map((item) => normalizeDoubaoJsonObject(item, context));\n  }\n\n  if (typeof obj === 'object') {\n    const normalized: any = {};\n    for (const [key, value] of Object.entries(obj)) {\n      const trimmedKey = key.trim();\n      const preserveStringValue =\n        context.preserveStringValueKeys?.includes(trimmedKey) ?? false;\n      const normalizedValue =\n        typeof value === 'string'\n          ? preserveStringValue\n            ? value\n            : value.trim()\n          : normalizeDoubaoJsonObject(value, context);\n      normalized[trimmedKey] = normalizedValue;\n    }\n    return normalized;\n  }\n\n  return typeof obj === 'string' ? obj.trim() : obj;\n}\n\nexport function shouldRepairDoubaoLocateJson(source: JsonParserSource) {\n  return (\n    source === 'locate' ||\n    source === 'section-locator' ||\n    source === 'planning-action-param'\n  );\n}\n\nexport function preprocessDoubaoLocateJson(input: string) {\n  if (input.includes('bbox')) {\n    while (/\\d+\\s+\\d+/.test(input)) {\n      input = input.replace(/(\\d+)\\s+(\\d+)/g, '$1,$2');\n    }\n  }\n  return input;\n}\n\nconst doubaoJsonParser: ModelAdapterDefinition['jsonParser'] = (\n  raw,\n  context = { source: 'generic-object' },\n) => {\n  const { source } = context;\n  try {\n    return safeParseJson(raw, context);\n  } catch (firstError) {\n    if (!shouldRepairDoubaoLocateJson(source)) {\n      throw firstError;\n    }\n\n    const jsonString = preprocessDoubaoLocateJson(\n      extractJSONFromCodeBlock(raw),\n    );\n    try {\n      return normalizeDoubaoJsonObject(\n        JSON.parse(jsonrepair(jsonString)),\n        context,\n      );\n    } catch (error) {\n      throw Error(\n        `failed to parse LLM response into JSON. Error - ${String(\n          error ?? firstError ?? 'unknown error',\n        )}. Response - \\n ${raw}`,\n      );\n    }\n  }\n};\n\nexport function parseDoubaoRawLocateValue(input: unknown): LocateResultValue {\n  const bbox = unwrapCoordinateListLikeInput(input as any);\n  if (typeof bbox === 'string') {\n    assert(\n      /^(\\d+)\\s(\\d+)\\s(\\d+)\\s(\\d+)$/.test(bbox.trim()),\n      `invalid bbox data string for doubao-vision mode: ${bbox}`,\n    );\n    const splitted = bbox.split(' ');\n    if (splitted.length === 4) {\n      return {\n        type: 'bbox',\n        coordinates: [\n          Number(splitted[0]),\n          Number(splitted[1]),\n          Number(splitted[2]),\n          Number(splitted[3]),\n        ],\n      };\n    }\n    throw new Error(`invalid bbox data string for doubao-vision mode: ${bbox}`);\n  }\n\n  let bboxList: number[] = [];\n  if (Array.isArray(bbox) && typeof bbox[0] === 'string') {\n    bbox.forEach((item) => {\n      if (typeof item === 'string' && item.includes(',')) {\n        const [x, y] = item.split(',');\n        bboxList.push(Number(x.trim()), Number(y.trim()));\n      } else if (typeof item === 'string' && item.includes(' ')) {\n        const [x, y] = item.split(' ');\n        bboxList.push(Number(x.trim()), Number(y.trim()));\n      } else {\n        bboxList.push(Number(item));\n      }\n    });\n  } else {\n    bboxList = bbox as number[];\n  }\n\n  if (bboxList.length === 4 || bboxList.length === 5) {\n    return {\n      type: 'bbox',\n      coordinates: [bboxList[0], bboxList[1], bboxList[2], bboxList[3]],\n    };\n  }\n\n  if (\n    bboxList.length === 6 ||\n    bboxList.length === 2 ||\n    bboxList.length === 3 ||\n    bboxList.length === 7\n  ) {\n    return { type: 'point', coordinates: [bboxList[0], bboxList[1]] };\n  }\n\n  if (bbox.length === 8) {\n    return {\n      type: 'bbox',\n      coordinates: [bboxList[0], bboxList[1], bboxList[4], bboxList[5]],\n    };\n  }\n\n  const msg = `invalid bbox data for doubao-vision mode: ${JSON.stringify(bbox)} `;\n  throw new Error(msg);\n}\n\nconst buildDoubaoChatCompletionParams = (\n  input: ChatCompletionCallContext,\n): ChatCompletionParamsResult => {\n  const { midsceneDefaults, userConfig } = input;\n  const { reasoningEnabled, reasoningEffort } = userConfig;\n  const commonOverrideConfig: Record<string, unknown> = {};\n\n  if (userConfig.temperature !== undefined) {\n    commonOverrideConfig.temperature = userConfig.temperature;\n  }\n\n  const modelSpecificConfig: Record<string, unknown> = {};\n\n  if (reasoningEnabled !== 'default') {\n    modelSpecificConfig.thinking = {\n      type: (reasoningEnabled ?? false) ? 'enabled' : 'disabled',\n    };\n    if (reasoningEffort) {\n      modelSpecificConfig.reasoning_effort = reasoningEffort;\n    }\n  }\n\n  return {\n    config: {\n      ...midsceneDefaults,\n      ...commonOverrideConfig,\n      ...modelSpecificConfig,\n    },\n  };\n};\n\nconst doubaoVisionAdapter: ModelAdapterDefinition = {\n  jsonParser: doubaoJsonParser,\n  chatCompletion: {\n    unsupportedUserConfig: ['reasoningBudget'],\n    buildChatCompletionParams: buildDoubaoChatCompletionParams,\n  },\n  locate: {\n    resultAdapter: {\n      coordinates: { shape: 'bbox', order: 'xy', normalizedBy: 1000 },\n      parseRawLocateValue: parseDoubaoRawLocateValue,\n    },\n  },\n};\n\nexport const doubaoAdapters = {\n  'doubao-vision': doubaoVisionAdapter,\n  'doubao-seed': doubaoVisionAdapter,\n} satisfies Pick<\n  Record<TModelFamily, ModelAdapterDefinition>,\n  'doubao-vision' | 'doubao-seed'\n>;\n"],"names":["normalizeDoubaoJsonObject","obj","context","Array","item","normalized","key","value","Object","trimmedKey","preserveStringValue","normalizedValue","shouldRepairDoubaoLocateJson","source","preprocessDoubaoLocateJson","input","doubaoJsonParser","raw","safeParseJson","firstError","jsonString","extractJSONFromCodeBlock","JSON","jsonrepair","error","Error","String","parseDoubaoRawLocateValue","bbox","unwrapCoordinateListLikeInput","assert","splitted","Number","bboxList","x","y","msg","buildDoubaoChatCompletionParams","midsceneDefaults","userConfig","reasoningEnabled","reasoningEffort","commonOverrideConfig","undefined","modelSpecificConfig","doubaoVisionAdapter","doubaoAdapters"],"mappings":";;;;AAmBO,SAASA,0BACdC,GAAQ,EACRC,UAA8D,CAAC,CAAC;IAEhE,IAAID,QAAAA,KACF,OAAOA;IAGT,IAAIE,MAAM,OAAO,CAACF,MAChB,OAAOA,IAAI,GAAG,CAAC,CAACG,OAASJ,0BAA0BI,MAAMF;IAG3D,IAAI,AAAe,YAAf,OAAOD,KAAkB;QAC3B,MAAMI,aAAkB,CAAC;QACzB,KAAK,MAAM,CAACC,KAAKC,MAAM,IAAIC,OAAO,OAAO,CAACP,KAAM;YAC9C,MAAMQ,aAAaH,IAAI,IAAI;YAC3B,MAAMI,sBACJR,QAAQ,uBAAuB,EAAE,SAASO,eAAe;YAC3D,MAAME,kBACJ,AAAiB,YAAjB,OAAOJ,QACHG,sBACEH,QACAA,MAAM,IAAI,KACZP,0BAA0BO,OAAOL;YACvCG,UAAU,CAACI,WAAW,GAAGE;QAC3B;QACA,OAAON;IACT;IAEA,OAAO,AAAe,YAAf,OAAOJ,MAAmBA,IAAI,IAAI,KAAKA;AAChD;AAEO,SAASW,6BAA6BC,MAAwB;IACnE,OACEA,AAAW,aAAXA,UACAA,AAAW,sBAAXA,UACAA,AAAW,4BAAXA;AAEJ;AAEO,SAASC,2BAA2BC,KAAa;IACtD,IAAIA,MAAM,QAAQ,CAAC,SACjB,MAAO,YAAY,IAAI,CAACA,OACtBA,QAAQA,MAAM,OAAO,CAAC,kBAAkB;IAG5C,OAAOA;AACT;AAEA,MAAMC,mBAAyD,CAC7DC,KACAf,UAAU;IAAE,QAAQ;AAAiB,CAAC;IAEtC,MAAM,EAAEW,MAAM,EAAE,GAAGX;IACnB,IAAI;QACF,OAAOgB,cAAcD,KAAKf;IAC5B,EAAE,OAAOiB,YAAY;QACnB,IAAI,CAACP,6BAA6BC,SAChC,MAAMM;QAGR,MAAMC,aAAaN,2BACjBO,yBAAyBJ;QAE3B,IAAI;YACF,OAAOjB,0BACLsB,KAAK,KAAK,CAACC,WAAWH,cACtBlB;QAEJ,EAAE,OAAOsB,OAAO;YACd,MAAMC,MACJ,CAAC,gDAAgD,EAAEC,OACjDF,SAASL,cAAc,iBACvB,gBAAgB,EAAEF,KAAK;QAE7B;IACF;AACF;AAEO,SAASU,0BAA0BZ,KAAc;IACtD,MAAMa,OAAOC,8BAA8Bd;IAC3C,IAAI,AAAgB,YAAhB,OAAOa,MAAmB;QAC5BE,OACE,+BAA+B,IAAI,CAACF,KAAK,IAAI,KAC7C,CAAC,iDAAiD,EAAEA,MAAM;QAE5D,MAAMG,WAAWH,KAAK,KAAK,CAAC;QAC5B,IAAIG,AAAoB,MAApBA,SAAS,MAAM,EACjB,OAAO;YACL,MAAM;YACN,aAAa;gBACXC,OAAOD,QAAQ,CAAC,EAAE;gBAClBC,OAAOD,QAAQ,CAAC,EAAE;gBAClBC,OAAOD,QAAQ,CAAC,EAAE;gBAClBC,OAAOD,QAAQ,CAAC,EAAE;aACnB;QACH;QAEF,MAAM,IAAIN,MAAM,CAAC,iDAAiD,EAAEG,MAAM;IAC5E;IAEA,IAAIK,WAAqB,EAAE;IAC3B,IAAI9B,MAAM,OAAO,CAACyB,SAAS,AAAmB,YAAnB,OAAOA,IAAI,CAAC,EAAE,EACvCA,KAAK,OAAO,CAAC,CAACxB;QACZ,IAAI,AAAgB,YAAhB,OAAOA,QAAqBA,KAAK,QAAQ,CAAC,MAAM;YAClD,MAAM,CAAC8B,GAAGC,EAAE,GAAG/B,KAAK,KAAK,CAAC;YAC1B6B,SAAS,IAAI,CAACD,OAAOE,EAAE,IAAI,KAAKF,OAAOG,EAAE,IAAI;QAC/C,OAAO,IAAI,AAAgB,YAAhB,OAAO/B,QAAqBA,KAAK,QAAQ,CAAC,MAAM;YACzD,MAAM,CAAC8B,GAAGC,EAAE,GAAG/B,KAAK,KAAK,CAAC;YAC1B6B,SAAS,IAAI,CAACD,OAAOE,EAAE,IAAI,KAAKF,OAAOG,EAAE,IAAI;QAC/C,OACEF,SAAS,IAAI,CAACD,OAAO5B;IAEzB;SAEA6B,WAAWL;IAGb,IAAIK,AAAoB,MAApBA,SAAS,MAAM,IAAUA,AAAoB,MAApBA,SAAS,MAAM,EAC1C,OAAO;QACL,MAAM;QACN,aAAa;YAACA,QAAQ,CAAC,EAAE;YAAEA,QAAQ,CAAC,EAAE;YAAEA,QAAQ,CAAC,EAAE;YAAEA,QAAQ,CAAC,EAAE;SAAC;IACnE;IAGF,IACEA,AAAoB,MAApBA,SAAS,MAAM,IACfA,AAAoB,MAApBA,SAAS,MAAM,IACfA,AAAoB,MAApBA,SAAS,MAAM,IACfA,AAAoB,MAApBA,SAAS,MAAM,EAEf,OAAO;QAAE,MAAM;QAAS,aAAa;YAACA,QAAQ,CAAC,EAAE;YAAEA,QAAQ,CAAC,EAAE;SAAC;IAAC;IAGlE,IAAIL,AAAgB,MAAhBA,KAAK,MAAM,EACb,OAAO;QACL,MAAM;QACN,aAAa;YAACK,QAAQ,CAAC,EAAE;YAAEA,QAAQ,CAAC,EAAE;YAAEA,QAAQ,CAAC,EAAE;YAAEA,QAAQ,CAAC,EAAE;SAAC;IACnE;IAGF,MAAMG,MAAM,CAAC,0CAA0C,EAAEd,KAAK,SAAS,CAACM,MAAM,CAAC,CAAC;IAChF,MAAM,IAAIH,MAAMW;AAClB;AAEA,MAAMC,kCAAkC,CACtCtB;IAEA,MAAM,EAAEuB,gBAAgB,EAAEC,UAAU,EAAE,GAAGxB;IACzC,MAAM,EAAEyB,gBAAgB,EAAEC,eAAe,EAAE,GAAGF;IAC9C,MAAMG,uBAAgD,CAAC;IAEvD,IAAIH,AAA2BI,WAA3BJ,WAAW,WAAW,EACxBG,qBAAqB,WAAW,GAAGH,WAAW,WAAW;IAG3D,MAAMK,sBAA+C,CAAC;IAEtD,IAAIJ,AAAqB,cAArBA,kBAAgC;QAClCI,oBAAoB,QAAQ,GAAG;YAC7B,MAAOJ,oBAAoB,QAAS,YAAY;QAClD;QACA,IAAIC,iBACFG,oBAAoB,gBAAgB,GAAGH;IAE3C;IAEA,OAAO;QACL,QAAQ;YACN,GAAGH,gBAAgB;YACnB,GAAGI,oBAAoB;YACvB,GAAGE,mBAAmB;QACxB;IACF;AACF;AAEA,MAAMC,sBAA8C;IAClD,YAAY7B;IACZ,gBAAgB;QACd,uBAAuB;YAAC;SAAkB;QAC1C,2BAA2BqB;IAC7B;IACA,QAAQ;QACN,eAAe;YACb,aAAa;gBAAE,OAAO;gBAAQ,OAAO;gBAAM,cAAc;YAAK;YAC9D,qBAAqBV;QACvB;IACF;AACF;AAEO,MAAMmB,iBAAiB;IAC5B,iBAAiBD;IACjB,eAAeA;AACjB"}
+{"version":3,"file":"ai-model/models/doubao.mjs","sources":["../../../../src/ai-model/models/doubao.ts"],"sourcesContent":["import type { TModelFamily } from '@midscene/shared/env';\nimport { assert } from '@midscene/shared/utils';\nimport { jsonrepair } from 'jsonrepair';\nimport type {\n  ChatCompletionCallContext,\n  ChatCompletionParamsResult,\n  JsonParserContext,\n  JsonParserSource,\n  ModelAdapterDefinition,\n} from '../model-adapter/types';\nimport {\n  extractJSONFromCodeBlock,\n  safeParseJson,\n} from '../service-caller/json';\nimport {\n  type LocateResultValue,\n  unwrapCoordinateListLikeInput,\n} from '../shared/model-locate-result';\n\nexport function normalizeDoubaoJsonObject(\n  obj: any,\n  context: Pick<JsonParserContext, 'preserveStringValueKeys'> = {},\n): any {\n  if (obj === null || obj === undefined) {\n    return obj;\n  }\n\n  if (Array.isArray(obj)) {\n    return obj.map((item) => normalizeDoubaoJsonObject(item, context));\n  }\n\n  if (typeof obj === 'object') {\n    const normalized: any = {};\n    for (const [key, value] of Object.entries(obj)) {\n      const trimmedKey = key.trim();\n      const preserveStringValue =\n        context.preserveStringValueKeys?.includes(trimmedKey) ?? false;\n      const normalizedValue =\n        typeof value === 'string'\n          ? preserveStringValue\n            ? value\n            : value.trim()\n          : normalizeDoubaoJsonObject(value, context);\n      normalized[trimmedKey] = normalizedValue;\n    }\n    return normalized;\n  }\n\n  return typeof obj === 'string' ? obj.trim() : obj;\n}\n\nexport function shouldRepairDoubaoLocateJson(source: JsonParserSource) {\n  return (\n    source === 'locate' ||\n    source === 'section-locator' ||\n    source === 'planning-action-param'\n  );\n}\n\nexport function preprocessDoubaoLocateJson(input: string) {\n  if (input.includes('bbox')) {\n    while (/\\d+\\s+\\d+/.test(input)) {\n      input = input.replace(/(\\d+)\\s+(\\d+)/g, '$1,$2');\n    }\n  }\n  return input;\n}\n\nconst doubaoJsonParser: ModelAdapterDefinition['jsonParser'] = (\n  raw,\n  context = { source: 'generic-object' },\n) => {\n  const { source } = context;\n  try {\n    return safeParseJson(raw, context);\n  } catch (firstError) {\n    if (!shouldRepairDoubaoLocateJson(source)) {\n      throw firstError;\n    }\n\n    const jsonString = preprocessDoubaoLocateJson(\n      extractJSONFromCodeBlock(raw),\n    );\n    try {\n      return normalizeDoubaoJsonObject(\n        JSON.parse(jsonrepair(jsonString)),\n        context,\n      );\n    } catch (error) {\n      throw Error(\n        `failed to parse LLM response into JSON. Error - ${String(\n          error ?? firstError ?? 'unknown error',\n        )}. Response - \\n ${raw}`,\n      );\n    }\n  }\n};\n\nexport function parseDoubaoRawLocateValue(input: unknown): LocateResultValue {\n  const bbox = unwrapCoordinateListLikeInput(input as any);\n  if (typeof bbox === 'string') {\n    assert(\n      /^(\\d+)\\s(\\d+)\\s(\\d+)\\s(\\d+)$/.test(bbox.trim()),\n      `invalid bbox data string for doubao-vision mode: ${bbox}`,\n    );\n    const splitted = bbox.split(' ');\n    if (splitted.length === 4) {\n      return {\n        type: 'bbox',\n        coordinates: [\n          Number(splitted[0]),\n          Number(splitted[1]),\n          Number(splitted[2]),\n          Number(splitted[3]),\n        ],\n      };\n    }\n    throw new Error(`invalid bbox data string for doubao-vision mode: ${bbox}`);\n  }\n\n  let bboxList: number[] = [];\n  if (Array.isArray(bbox) && typeof bbox[0] === 'string') {\n    bbox.forEach((item) => {\n      if (typeof item === 'string' && item.includes(',')) {\n        const [x, y] = item.split(',');\n        bboxList.push(Number(x.trim()), Number(y.trim()));\n      } else if (typeof item === 'string' && item.includes(' ')) {\n        const [x, y] = item.split(' ');\n        bboxList.push(Number(x.trim()), Number(y.trim()));\n      } else {\n        bboxList.push(Number(item));\n      }\n    });\n  } else {\n    bboxList = bbox as number[];\n  }\n\n  if (bboxList.length === 4 || bboxList.length === 5) {\n    return {\n      type: 'bbox',\n      coordinates: [bboxList[0], bboxList[1], bboxList[2], bboxList[3]],\n    };\n  }\n\n  if (\n    bboxList.length === 6 ||\n    bboxList.length === 2 ||\n    bboxList.length === 3 ||\n    bboxList.length === 7\n  ) {\n    return { type: 'point', coordinates: [bboxList[0], bboxList[1]] };\n  }\n\n  if (bbox.length === 8) {\n    return {\n      type: 'bbox',\n      coordinates: [bboxList[0], bboxList[1], bboxList[4], bboxList[5]],\n    };\n  }\n\n  const msg = `invalid bbox data for doubao-vision mode: ${JSON.stringify(bbox)} `;\n  throw new Error(msg);\n}\n\nconst buildDoubaoChatCompletionParams = (\n  input: ChatCompletionCallContext,\n): ChatCompletionParamsResult => {\n  const { midsceneDefaults, userConfig } = input;\n  const { reasoningEnabled, reasoningEffort } = userConfig;\n  const commonOverrideConfig: Record<string, unknown> = {};\n\n  if (userConfig.temperature !== undefined) {\n    commonOverrideConfig.temperature = userConfig.temperature;\n  }\n\n  const modelSpecificConfig: Record<string, unknown> = {};\n\n  if (reasoningEnabled !== 'default') {\n    modelSpecificConfig.thinking = {\n      type: (reasoningEnabled ?? false) ? 'enabled' : 'disabled',\n    };\n    if (reasoningEffort) {\n      modelSpecificConfig.reasoning_effort = reasoningEffort;\n    }\n  }\n\n  return {\n    config: {\n      ...midsceneDefaults,\n      ...commonOverrideConfig,\n      ...modelSpecificConfig,\n    },\n  };\n};\n\nconst doubaoVisionAdapter: ModelAdapterDefinition = {\n  jsonParser: doubaoJsonParser,\n  chatCompletion: {\n    unsupportedUserConfig: ['reasoningBudget'],\n    buildChatCompletionParams: buildDoubaoChatCompletionParams,\n  },\n  locate: {\n    resultAdapter: {\n      coordinates: { shape: 'bbox', order: 'xy', normalizedBy: 1000 },\n      parseRawLocateValue: parseDoubaoRawLocateValue,\n    },\n  },\n};\n\nexport const doubaoAdapters = {\n  'doubao-vision': doubaoVisionAdapter,\n  'doubao-seed': doubaoVisionAdapter,\n} satisfies Pick<\n  Record<TModelFamily, ModelAdapterDefinition>,\n  'doubao-vision' | 'doubao-seed'\n>;\n"],"names":["normalizeDoubaoJsonObject","obj","context","Array","item","normalized","key","value","Object","trimmedKey","preserveStringValue","normalizedValue","shouldRepairDoubaoLocateJson","source","preprocessDoubaoLocateJson","input","doubaoJsonParser","raw","safeParseJson","firstError","jsonString","extractJSONFromCodeBlock","JSON","jsonrepair","error","Error","String","parseDoubaoRawLocateValue","bbox","unwrapCoordinateListLikeInput","assert","splitted","Number","bboxList","x","y","msg","buildDoubaoChatCompletionParams","midsceneDefaults","userConfig","reasoningEnabled","reasoningEffort","commonOverrideConfig","undefined","modelSpecificConfig","doubaoVisionAdapter","doubaoAdapters"],"mappings":";;;;AAmBO,SAASA,0BACdC,GAAQ,EACRC,UAA8D,CAAC,CAAC;IAEhE,IAAID,QAAAA,KACF,OAAOA;IAGT,IAAIE,MAAM,OAAO,CAACF,MAChB,OAAOA,IAAI,GAAG,CAAC,CAACG,OAASJ,0BAA0BI,MAAMF;IAG3D,IAAI,AAAe,YAAf,OAAOD,KAAkB;QAC3B,MAAMI,aAAkB,CAAC;QACzB,KAAK,MAAM,CAACC,KAAKC,MAAM,IAAIC,OAAO,OAAO,CAACP,KAAM;YAC9C,MAAMQ,aAAaH,IAAI,IAAI;YAC3B,MAAMI,sBACJR,QAAQ,uBAAuB,EAAE,SAASO,eAAe;YAC3D,MAAME,kBACJ,AAAiB,YAAjB,OAAOJ,QACHG,sBACEH,QACAA,MAAM,IAAI,KACZP,0BAA0BO,OAAOL;YACvCG,UAAU,CAACI,WAAW,GAAGE;QAC3B;QACA,OAAON;IACT;IAEA,OAAO,AAAe,YAAf,OAAOJ,MAAmBA,IAAI,IAAI,KAAKA;AAChD;AAEO,SAASW,6BAA6BC,MAAwB;IACnE,OACEA,AAAW,aAAXA,UACAA,AAAW,sBAAXA,UACAA,AAAW,4BAAXA;AAEJ;AAEO,SAASC,2BAA2BC,KAAa;IACtD,IAAIA,MAAM,QAAQ,CAAC,SACjB,MAAO,YAAY,IAAI,CAACA,OACtBA,QAAQA,MAAM,OAAO,CAAC,kBAAkB;IAG5C,OAAOA;AACT;AAEA,MAAMC,mBAAyD,CAC7DC,KACAf,UAAU;IAAE,QAAQ;AAAiB,CAAC;IAEtC,MAAM,EAAEW,MAAM,EAAE,GAAGX;IACnB,IAAI;QACF,OAAOgB,cAAcD,KAAKf;IAC5B,EAAE,OAAOiB,YAAY;QACnB,IAAI,CAACP,6BAA6BC,SAChC,MAAMM;QAGR,MAAMC,aAAaN,2BACjBO,yBAAyBJ;QAE3B,IAAI;YACF,OAAOjB,0BACLsB,KAAK,KAAK,CAACC,WAAWH,cACtBlB;QAEJ,EAAE,OAAOsB,OAAO;YACd,MAAMC,MACJ,CAAC,gDAAgD,EAAEC,OACjDF,SAASL,cAAc,iBACvB,gBAAgB,EAAEF,KAAK;QAE7B;IACF;AACF;AAEO,SAASU,0BAA0BZ,KAAc;IACtD,MAAMa,OAAOC,8BAA8Bd;IAC3C,IAAI,AAAgB,YAAhB,OAAOa,MAAmB;QAC5BE,OACE,+BAA+B,IAAI,CAACF,KAAK,IAAI,KAC7C,CAAC,iDAAiD,EAAEA,MAAM;QAE5D,MAAMG,WAAWH,KAAK,KAAK,CAAC;QAC5B,IAAIG,AAAoB,MAApBA,SAAS,MAAM,EACjB,OAAO;YACL,MAAM;YACN,aAAa;gBACXC,OAAOD,QAAQ,CAAC,EAAE;gBAClBC,OAAOD,QAAQ,CAAC,EAAE;gBAClBC,OAAOD,QAAQ,CAAC,EAAE;gBAClBC,OAAOD,QAAQ,CAAC,EAAE;aACnB;QACH;QAEF,MAAM,IAAIN,MAAM,CAAC,iDAAiD,EAAEG,MAAM;IAC5E;IAEA,IAAIK,WAAqB,EAAE;IAC3B,IAAI9B,MAAM,OAAO,CAACyB,SAAS,AAAmB,YAAnB,OAAOA,IAAI,CAAC,EAAE,EACvCA,KAAK,OAAO,CAAC,CAACxB;QACZ,IAAI,AAAgB,YAAhB,OAAOA,QAAqBA,KAAK,QAAQ,CAAC,MAAM;YAClD,MAAM,CAAC8B,GAAGC,EAAE,GAAG/B,KAAK,KAAK,CAAC;YAC1B6B,SAAS,IAAI,CAACD,OAAOE,EAAE,IAAI,KAAKF,OAAOG,EAAE,IAAI;QAC/C,OAAO,IAAI,AAAgB,YAAhB,OAAO/B,QAAqBA,KAAK,QAAQ,CAAC,MAAM;YACzD,MAAM,CAAC8B,GAAGC,EAAE,GAAG/B,KAAK,KAAK,CAAC;YAC1B6B,SAAS,IAAI,CAACD,OAAOE,EAAE,IAAI,KAAKF,OAAOG,EAAE,IAAI;QAC/C,OACEF,SAAS,IAAI,CAACD,OAAO5B;IAEzB;SAEA6B,WAAWL;IAGb,IAAIK,AAAoB,MAApBA,SAAS,MAAM,IAAUA,AAAoB,MAApBA,SAAS,MAAM,EAC1C,OAAO;QACL,MAAM;QACN,aAAa;YAACA,QAAQ,CAAC,EAAE;YAAEA,QAAQ,CAAC,EAAE;YAAEA,QAAQ,CAAC,EAAE;YAAEA,QAAQ,CAAC,EAAE;SAAC;IACnE;IAGF,IACEA,AAAoB,MAApBA,SAAS,MAAM,IACfA,AAAoB,MAApBA,SAAS,MAAM,IACfA,AAAoB,MAApBA,SAAS,MAAM,IACfA,AAAoB,MAApBA,SAAS,MAAM,EAEf,OAAO;QAAE,MAAM;QAAS,aAAa;YAACA,QAAQ,CAAC,EAAE;YAAEA,QAAQ,CAAC,EAAE;SAAC;IAAC;IAGlE,IAAIL,AAAgB,MAAhBA,KAAK,MAAM,EACb,OAAO;QACL,MAAM;QACN,aAAa;YAACK,QAAQ,CAAC,EAAE;YAAEA,QAAQ,CAAC,EAAE;YAAEA,QAAQ,CAAC,EAAE;YAAEA,QAAQ,CAAC,EAAE;SAAC;IACnE;IAGF,MAAMG,MAAM,CAAC,0CAA0C,EAAEd,KAAK,SAAS,CAACM,MAAM,CAAC,CAAC;IAChF,MAAM,IAAIH,MAAMW;AAClB;AAEA,MAAMC,kCAAkC,CACtCtB;IAEA,MAAM,EAAEuB,gBAAgB,EAAEC,UAAU,EAAE,GAAGxB;IACzC,MAAM,EAAEyB,gBAAgB,EAAEC,eAAe,EAAE,GAAGF;IAC9C,MAAMG,uBAAgD,CAAC;IAEvD,IAAIH,AAA2BI,WAA3BJ,WAAW,WAAW,EACxBG,qBAAqB,WAAW,GAAGH,WAAW,WAAW;IAG3D,MAAMK,sBAA+C,CAAC;IAEtD,IAAIJ,AAAqB,cAArBA,kBAAgC;QAClCI,oBAAoB,QAAQ,GAAG;YAC7B,MAAOJ,oBAAoB,QAAS,YAAY;QAClD;QACA,IAAIC,iBACFG,oBAAoB,gBAAgB,GAAGH;IAE3C;IAEA,OAAO;QACL,QAAQ;YACN,GAAGH,gBAAgB;YACnB,GAAGI,oBAAoB;YACvB,GAAGE,mBAAmB;QACxB;IACF;AACF;AAEA,MAAMC,sBAA8C;IAClD,YAAY7B;IACZ,gBAAgB;QACd,uBAAuB;YAAC;SAAkB;QAC1C,2BAA2BqB;IAC7B;IACA,QAAQ;QACN,eAAe;YACb,aAAa;gBAAE,OAAO;gBAAQ,OAAO;gBAAM,cAAc;YAAK;YAC9D,qBAAqBV;QACvB;IACF;AACF;AAEO,MAAMmB,iBAAiB;IAC5B,iBAAiBD;IACjB,eAAeA;AACjB"}

package/dist/es/ai-model/models/gemini.mjs

@@ -1,4 +1,4 @@
-import { defaultExtractContentAndReasoning } from "./chat-content.mjs";
+import { defaultExtractContentAndReasoning } from "../model-adapter/chat-completion.mjs";
 const buildGeminiChatCompletionParams = (input)=>{
     const { midsceneDefaults, userConfig, intent } = input;
     const { reasoningEnabled, reasoningEffort } = userConfig;

package/dist/es/ai-model/models/gemini.mjs.map

@@ -1 +1 @@
-{"version":3,"file":"ai-model/models/gemini.mjs","sources":["../../../../src/ai-model/models/gemini.ts"],"sourcesContent":["import type { TModelFamily } from '@midscene/shared/env';\nimport type OpenAI from 'openai';\nimport { defaultExtractContentAndReasoning } from './chat-content';\nimport type {\n  ChatCompletionCallContext,\n  ChatCompletionContentSource,\n  ChatCompletionParamsResult,\n  ContentAndReasoning,\n  ModelAdapterDefinition,\n} from './types';\n\ntype GeminiContentPart = Record<string, unknown> & {\n  text?: string;\n  thought?: boolean;\n};\n\ntype GeminiContentExtension = {\n  content: string | null | GeminiContentPart[];\n  reasoning_content?: string;\n  extra_content?: unknown;\n};\n\ntype GeminiContentSource =\n  | ChatCompletionContentSource\n  | (Omit<OpenAI.Chat.Completions.ChatCompletionMessage, 'content'> &\n      GeminiContentExtension)\n  | (Omit<OpenAI.Chat.Completions.ChatCompletionChunk.Choice.Delta, 'content'> &\n      GeminiContentExtension);\n\nconst buildGeminiChatCompletionParams = (\n  input: ChatCompletionCallContext,\n): ChatCompletionParamsResult => {\n  const { midsceneDefaults, userConfig, intent } = input;\n  const { reasoningEnabled, reasoningEffort } = userConfig;\n  const commonOverrideConfig: Record<string, unknown> = {};\n\n  if (userConfig.temperature !== undefined) {\n    commonOverrideConfig.temperature = userConfig.temperature;\n  }\n\n  const modelSpecificConfig: {\n    extra_body?: {\n      google?: {\n        thinking_config: Record<string, unknown>;\n      };\n    };\n    reasoning_effort?: unknown;\n  } = {};\n\n  if (reasoningEnabled !== 'default') {\n    if (reasoningEffort) {\n      modelSpecificConfig.extra_body = {\n        google: {\n          thinking_config: {\n            thinking_level: reasoningEffort,\n            include_thoughts: true,\n          },\n        },\n      };\n    } else {\n      if (intent === 'insight') {\n        modelSpecificConfig.extra_body = {\n          google: {\n            thinking_config: {\n              // In real Gemini tests, insight calls need `include_thoughts` to get\n              // the model's thinking, and Gemini puts that thinking into `content`.\n              include_thoughts: true,\n            },\n          },\n        };\n      }\n\n      // Gemini 3.x cannot fully disable native thinking, so use the lowest\n      // supported effort unless the user explicitly requests another level.\n      modelSpecificConfig.reasoning_effort = 'minimal';\n    }\n  }\n\n  return {\n    config: {\n      ...midsceneDefaults,\n      ...commonOverrideConfig,\n      ...modelSpecificConfig,\n    },\n  };\n};\n\nconst extractInlineThought = (content: string): string | undefined => {\n  const match = content.match(/<thought>([\\s\\S]*?)<\\/thought>/i);\n  return match?.[1]?.trim() || undefined;\n};\n\nfunction isRecord(value: unknown): value is Record<string, unknown> {\n  return !!value && typeof value === 'object' && !Array.isArray(value);\n}\n\ntype GeminiExtractedContent = {\n  content: string;\n  geminiReasoningContent: string;\n};\n\nconst formatReasoningContent = ({\n  geminiReasoningContent,\n  providerReasoningContent,\n}: {\n  geminiReasoningContent: string;\n  providerReasoningContent: string;\n}): string => {\n  if (geminiReasoningContent && providerReasoningContent) {\n    return `thoughtParts：${geminiReasoningContent}; reasoning_content：${providerReasoningContent}`;\n  }\n\n  return geminiReasoningContent || providerReasoningContent || '';\n};\n\nconst extractGeminiThoughtParts = (\n  content: GeminiContentPart[],\n): GeminiExtractedContent => {\n  const contentParts: string[] = [];\n  const reasoningParts: string[] = [];\n\n  for (const part of content) {\n    if (!isRecord(part)) {\n      continue;\n    }\n\n    const text = typeof part.text === 'string' ? part.text : undefined;\n    if (!text) {\n      continue;\n    }\n\n    if (part.thought === true) {\n      reasoningParts.push(text);\n    } else {\n      contentParts.push(text);\n    }\n  }\n\n  return {\n    content: contentParts.join(''),\n    geminiReasoningContent: reasoningParts.join(''),\n  };\n};\n\nexport const extractGeminiContentAndReasoning = (\n  message: GeminiContentSource | undefined,\n): ContentAndReasoning => {\n  // Gemini native API exposes thought summaries in\n  // `response.candidates[0].content.parts`, where each text part can carry\n  // `thought: true`. Some OpenAI-compatible adapters may pass through a\n  // native-like content-parts shape, even though the Gemini OpenAI\n  // compatibility docs do not guarantee it.\n  // Docs: https://ai.google.dev/gemini-api/docs/thinking#thought-summaries\n\n  if (!message) {\n    return {\n      content: '',\n      reasoning_content: '',\n    };\n  }\n\n  if (Array.isArray(message.content)) {\n    const extracted = extractGeminiThoughtParts(message.content);\n\n    return {\n      content: extracted.content,\n      reasoning_content: formatReasoningContent({\n        geminiReasoningContent: extracted.geminiReasoningContent,\n        providerReasoningContent:\n          typeof message.reasoning_content === 'string'\n            ? message.reasoning_content\n            : '',\n      }),\n    };\n  }\n\n  const result = defaultExtractContentAndReasoning(\n    message as ChatCompletionContentSource,\n  );\n  // In real Gemini OpenAI-compatible responses we observed that\n  // `include_thoughts` can still return a plain string `message.content`,\n  // with the thought summary prepended as `<thought>...</thought>` before the\n  // visible answer. Keep content unchanged, but extract that leading thought\n  // text for report display.\n  const geminiReasoningContent = extractInlineThought(result.content) || '';\n\n  return {\n    content: result.content,\n    reasoning_content: formatReasoningContent({\n      geminiReasoningContent,\n      providerReasoningContent: result.reasoning_content,\n    }),\n  };\n};\n\nexport const geminiAdapters = {\n  gemini: {\n    chatCompletion: {\n      unsupportedUserConfig: ['reasoningEnabled', 'reasoningBudget'],\n      buildChatCompletionParams: buildGeminiChatCompletionParams,\n      extractContentAndReasoning: extractGeminiContentAndReasoning,\n    },\n    locate: {\n      resultAdapter: {\n        coordinates: { shape: 'bbox', order: 'yx', normalizedBy: 1000 },\n      },\n    },\n  },\n} satisfies Pick<Record<TModelFamily, ModelAdapterDefinition>, 'gemini'>;\n"],"names":["buildGeminiChatCompletionParams","input","midsceneDefaults","userConfig","intent","reasoningEnabled","reasoningEffort","commonOverrideConfig","undefined","modelSpecificConfig","extractInlineThought","content","match","isRecord","value","Array","formatReasoningContent","geminiReasoningContent","providerReasoningContent","extractGeminiThoughtParts","contentParts","reasoningParts","part","text","extractGeminiContentAndReasoning","message","extracted","result","defaultExtractContentAndReasoning","geminiAdapters"],"mappings":";AA6BA,MAAMA,kCAAkC,CACtCC;IAEA,MAAM,EAAEC,gBAAgB,EAAEC,UAAU,EAAEC,MAAM,EAAE,GAAGH;IACjD,MAAM,EAAEI,gBAAgB,EAAEC,eAAe,EAAE,GAAGH;IAC9C,MAAMI,uBAAgD,CAAC;IAEvD,IAAIJ,AAA2BK,WAA3BL,WAAW,WAAW,EACxBI,qBAAqB,WAAW,GAAGJ,WAAW,WAAW;IAG3D,MAAMM,sBAOF,CAAC;IAEL,IAAIJ,AAAqB,cAArBA,kBACF,IAAIC,iBACFG,oBAAoB,UAAU,GAAG;QAC/B,QAAQ;YACN,iBAAiB;gBACf,gBAAgBH;gBAChB,kBAAkB;YACpB;QACF;IACF;SACK;QACL,IAAIF,AAAW,cAAXA,QACFK,oBAAoB,UAAU,GAAG;YAC/B,QAAQ;gBACN,iBAAiB;oBAGf,kBAAkB;gBACpB;YACF;QACF;QAKFA,oBAAoB,gBAAgB,GAAG;IACzC;IAGF,OAAO;QACL,QAAQ;YACN,GAAGP,gBAAgB;YACnB,GAAGK,oBAAoB;YACvB,GAAGE,mBAAmB;QACxB;IACF;AACF;AAEA,MAAMC,uBAAuB,CAACC;IAC5B,MAAMC,QAAQD,QAAQ,KAAK,CAAC;IAC5B,OAAOC,OAAO,CAAC,EAAE,EAAE,UAAUJ;AAC/B;AAEA,SAASK,SAASC,KAAc;IAC9B,OAAO,CAAC,CAACA,SAAS,AAAiB,YAAjB,OAAOA,SAAsB,CAACC,MAAM,OAAO,CAACD;AAChE;AAOA,MAAME,yBAAyB,CAAC,EAC9BC,sBAAsB,EACtBC,wBAAwB,EAIzB;IACC,IAAID,0BAA0BC,0BAC5B,OAAO,CAAC,aAAa,EAAED,uBAAuB,oBAAoB,EAAEC,0BAA0B;IAGhG,OAAOD,0BAA0BC,4BAA4B;AAC/D;AAEA,MAAMC,4BAA4B,CAChCR;IAEA,MAAMS,eAAyB,EAAE;IACjC,MAAMC,iBAA2B,EAAE;IAEnC,KAAK,MAAMC,QAAQX,QAAS;QAC1B,IAAI,CAACE,SAASS,OACZ;QAGF,MAAMC,OAAO,AAAqB,YAArB,OAAOD,KAAK,IAAI,GAAgBA,KAAK,IAAI,GAAGd;QACzD,IAAKe,MAIL,IAAID,AAAiB,SAAjBA,KAAK,OAAO,EACdD,eAAe,IAAI,CAACE;aAEpBH,aAAa,IAAI,CAACG;IAEtB;IAEA,OAAO;QACL,SAASH,aAAa,IAAI,CAAC;QAC3B,wBAAwBC,eAAe,IAAI,CAAC;IAC9C;AACF;AAEO,MAAMG,mCAAmC,CAC9CC;IASA,IAAI,CAACA,SACH,OAAO;QACL,SAAS;QACT,mBAAmB;IACrB;IAGF,IAAIV,MAAM,OAAO,CAACU,QAAQ,OAAO,GAAG;QAClC,MAAMC,YAAYP,0BAA0BM,QAAQ,OAAO;QAE3D,OAAO;YACL,SAASC,UAAU,OAAO;YAC1B,mBAAmBV,uBAAuB;gBACxC,wBAAwBU,UAAU,sBAAsB;gBACxD,0BACE,AAAqC,YAArC,OAAOD,QAAQ,iBAAiB,GAC5BA,QAAQ,iBAAiB,GACzB;YACR;QACF;IACF;IAEA,MAAME,SAASC,kCACbH;IAOF,MAAMR,yBAAyBP,qBAAqBiB,OAAO,OAAO,KAAK;IAEvE,OAAO;QACL,SAASA,OAAO,OAAO;QACvB,mBAAmBX,uBAAuB;YACxCC;YACA,0BAA0BU,OAAO,iBAAiB;QACpD;IACF;AACF;AAEO,MAAME,iBAAiB;IAC5B,QAAQ;QACN,gBAAgB;YACd,uBAAuB;gBAAC;gBAAoB;aAAkB;YAC9D,2BAA2B7B;YAC3B,4BAA4BwB;QAC9B;QACA,QAAQ;YACN,eAAe;gBACb,aAAa;oBAAE,OAAO;oBAAQ,OAAO;oBAAM,cAAc;gBAAK;YAChE;QACF;IACF;AACF"}
+{"version":3,"file":"ai-model/models/gemini.mjs","sources":["../../../../src/ai-model/models/gemini.ts"],"sourcesContent":["import type { TModelFamily } from '@midscene/shared/env';\nimport type OpenAI from 'openai';\nimport { defaultExtractContentAndReasoning } from '../model-adapter/chat-completion';\nimport type {\n  ChatCompletionCallContext,\n  ChatCompletionContentSource,\n  ChatCompletionParamsResult,\n  ContentAndReasoning,\n  ModelAdapterDefinition,\n} from '../model-adapter/types';\n\ntype GeminiContentPart = Record<string, unknown> & {\n  text?: string;\n  thought?: boolean;\n};\n\ntype GeminiContentExtension = {\n  content: string | null | GeminiContentPart[];\n  reasoning_content?: string;\n  extra_content?: unknown;\n};\n\ntype GeminiContentSource =\n  | ChatCompletionContentSource\n  | (Omit<OpenAI.Chat.Completions.ChatCompletionMessage, 'content'> &\n      GeminiContentExtension)\n  | (Omit<OpenAI.Chat.Completions.ChatCompletionChunk.Choice.Delta, 'content'> &\n      GeminiContentExtension);\n\nconst buildGeminiChatCompletionParams = (\n  input: ChatCompletionCallContext,\n): ChatCompletionParamsResult => {\n  const { midsceneDefaults, userConfig, intent } = input;\n  const { reasoningEnabled, reasoningEffort } = userConfig;\n  const commonOverrideConfig: Record<string, unknown> = {};\n\n  if (userConfig.temperature !== undefined) {\n    commonOverrideConfig.temperature = userConfig.temperature;\n  }\n\n  const modelSpecificConfig: {\n    extra_body?: {\n      google?: {\n        thinking_config: Record<string, unknown>;\n      };\n    };\n    reasoning_effort?: unknown;\n  } = {};\n\n  if (reasoningEnabled !== 'default') {\n    if (reasoningEffort) {\n      modelSpecificConfig.extra_body = {\n        google: {\n          thinking_config: {\n            thinking_level: reasoningEffort,\n            include_thoughts: true,\n          },\n        },\n      };\n    } else {\n      if (intent === 'insight') {\n        modelSpecificConfig.extra_body = {\n          google: {\n            thinking_config: {\n              // In real Gemini tests, insight calls need `include_thoughts` to get\n              // the model's thinking, and Gemini puts that thinking into `content`.\n              include_thoughts: true,\n            },\n          },\n        };\n      }\n\n      // Gemini 3.x cannot fully disable native thinking, so use the lowest\n      // supported effort unless the user explicitly requests another level.\n      modelSpecificConfig.reasoning_effort = 'minimal';\n    }\n  }\n\n  return {\n    config: {\n      ...midsceneDefaults,\n      ...commonOverrideConfig,\n      ...modelSpecificConfig,\n    },\n  };\n};\n\nconst extractInlineThought = (content: string): string | undefined => {\n  const match = content.match(/<thought>([\\s\\S]*?)<\\/thought>/i);\n  return match?.[1]?.trim() || undefined;\n};\n\nfunction isRecord(value: unknown): value is Record<string, unknown> {\n  return !!value && typeof value === 'object' && !Array.isArray(value);\n}\n\ntype GeminiExtractedContent = {\n  content: string;\n  geminiReasoningContent: string;\n};\n\nconst formatReasoningContent = ({\n  geminiReasoningContent,\n  providerReasoningContent,\n}: {\n  geminiReasoningContent: string;\n  providerReasoningContent: string;\n}): string => {\n  if (geminiReasoningContent && providerReasoningContent) {\n    return `thoughtParts：${geminiReasoningContent}; reasoning_content：${providerReasoningContent}`;\n  }\n\n  return geminiReasoningContent || providerReasoningContent || '';\n};\n\nconst extractGeminiThoughtParts = (\n  content: GeminiContentPart[],\n): GeminiExtractedContent => {\n  const contentParts: string[] = [];\n  const reasoningParts: string[] = [];\n\n  for (const part of content) {\n    if (!isRecord(part)) {\n      continue;\n    }\n\n    const text = typeof part.text === 'string' ? part.text : undefined;\n    if (!text) {\n      continue;\n    }\n\n    if (part.thought === true) {\n      reasoningParts.push(text);\n    } else {\n      contentParts.push(text);\n    }\n  }\n\n  return {\n    content: contentParts.join(''),\n    geminiReasoningContent: reasoningParts.join(''),\n  };\n};\n\nexport const extractGeminiContentAndReasoning = (\n  message: GeminiContentSource | undefined,\n): ContentAndReasoning => {\n  // Gemini native API exposes thought summaries in\n  // `response.candidates[0].content.parts`, where each text part can carry\n  // `thought: true`. Some OpenAI-compatible adapters may pass through a\n  // native-like content-parts shape, even though the Gemini OpenAI\n  // compatibility docs do not guarantee it.\n  // Docs: https://ai.google.dev/gemini-api/docs/thinking#thought-summaries\n\n  if (!message) {\n    return {\n      content: '',\n      reasoning_content: '',\n    };\n  }\n\n  if (Array.isArray(message.content)) {\n    const extracted = extractGeminiThoughtParts(message.content);\n\n    return {\n      content: extracted.content,\n      reasoning_content: formatReasoningContent({\n        geminiReasoningContent: extracted.geminiReasoningContent,\n        providerReasoningContent:\n          typeof message.reasoning_content === 'string'\n            ? message.reasoning_content\n            : '',\n      }),\n    };\n  }\n\n  const result = defaultExtractContentAndReasoning(\n    message as ChatCompletionContentSource,\n  );\n  // In real Gemini OpenAI-compatible responses we observed that\n  // `include_thoughts` can still return a plain string `message.content`,\n  // with the thought summary prepended as `<thought>...</thought>` before the\n  // visible answer. Keep content unchanged, but extract that leading thought\n  // text for report display.\n  const geminiReasoningContent = extractInlineThought(result.content) || '';\n\n  return {\n    content: result.content,\n    reasoning_content: formatReasoningContent({\n      geminiReasoningContent,\n      providerReasoningContent: result.reasoning_content,\n    }),\n  };\n};\n\nexport const geminiAdapters = {\n  gemini: {\n    chatCompletion: {\n      unsupportedUserConfig: ['reasoningEnabled', 'reasoningBudget'],\n      buildChatCompletionParams: buildGeminiChatCompletionParams,\n      extractContentAndReasoning: extractGeminiContentAndReasoning,\n    },\n    locate: {\n      resultAdapter: {\n        coordinates: { shape: 'bbox', order: 'yx', normalizedBy: 1000 },\n      },\n    },\n  },\n} satisfies Pick<Record<TModelFamily, ModelAdapterDefinition>, 'gemini'>;\n"],"names":["buildGeminiChatCompletionParams","input","midsceneDefaults","userConfig","intent","reasoningEnabled","reasoningEffort","commonOverrideConfig","undefined","modelSpecificConfig","extractInlineThought","content","match","isRecord","value","Array","formatReasoningContent","geminiReasoningContent","providerReasoningContent","extractGeminiThoughtParts","contentParts","reasoningParts","part","text","extractGeminiContentAndReasoning","message","extracted","result","defaultExtractContentAndReasoning","geminiAdapters"],"mappings":";AA6BA,MAAMA,kCAAkC,CACtCC;IAEA,MAAM,EAAEC,gBAAgB,EAAEC,UAAU,EAAEC,MAAM,EAAE,GAAGH;IACjD,MAAM,EAAEI,gBAAgB,EAAEC,eAAe,EAAE,GAAGH;IAC9C,MAAMI,uBAAgD,CAAC;IAEvD,IAAIJ,AAA2BK,WAA3BL,WAAW,WAAW,EACxBI,qBAAqB,WAAW,GAAGJ,WAAW,WAAW;IAG3D,MAAMM,sBAOF,CAAC;IAEL,IAAIJ,AAAqB,cAArBA,kBACF,IAAIC,iBACFG,oBAAoB,UAAU,GAAG;QAC/B,QAAQ;YACN,iBAAiB;gBACf,gBAAgBH;gBAChB,kBAAkB;YACpB;QACF;IACF;SACK;QACL,IAAIF,AAAW,cAAXA,QACFK,oBAAoB,UAAU,GAAG;YAC/B,QAAQ;gBACN,iBAAiB;oBAGf,kBAAkB;gBACpB;YACF;QACF;QAKFA,oBAAoB,gBAAgB,GAAG;IACzC;IAGF,OAAO;QACL,QAAQ;YACN,GAAGP,gBAAgB;YACnB,GAAGK,oBAAoB;YACvB,GAAGE,mBAAmB;QACxB;IACF;AACF;AAEA,MAAMC,uBAAuB,CAACC;IAC5B,MAAMC,QAAQD,QAAQ,KAAK,CAAC;IAC5B,OAAOC,OAAO,CAAC,EAAE,EAAE,UAAUJ;AAC/B;AAEA,SAASK,SAASC,KAAc;IAC9B,OAAO,CAAC,CAACA,SAAS,AAAiB,YAAjB,OAAOA,SAAsB,CAACC,MAAM,OAAO,CAACD;AAChE;AAOA,MAAME,yBAAyB,CAAC,EAC9BC,sBAAsB,EACtBC,wBAAwB,EAIzB;IACC,IAAID,0BAA0BC,0BAC5B,OAAO,CAAC,aAAa,EAAED,uBAAuB,oBAAoB,EAAEC,0BAA0B;IAGhG,OAAOD,0BAA0BC,4BAA4B;AAC/D;AAEA,MAAMC,4BAA4B,CAChCR;IAEA,MAAMS,eAAyB,EAAE;IACjC,MAAMC,iBAA2B,EAAE;IAEnC,KAAK,MAAMC,QAAQX,QAAS;QAC1B,IAAI,CAACE,SAASS,OACZ;QAGF,MAAMC,OAAO,AAAqB,YAArB,OAAOD,KAAK,IAAI,GAAgBA,KAAK,IAAI,GAAGd;QACzD,IAAKe,MAIL,IAAID,AAAiB,SAAjBA,KAAK,OAAO,EACdD,eAAe,IAAI,CAACE;aAEpBH,aAAa,IAAI,CAACG;IAEtB;IAEA,OAAO;QACL,SAASH,aAAa,IAAI,CAAC;QAC3B,wBAAwBC,eAAe,IAAI,CAAC;IAC9C;AACF;AAEO,MAAMG,mCAAmC,CAC9CC;IASA,IAAI,CAACA,SACH,OAAO;QACL,SAAS;QACT,mBAAmB;IACrB;IAGF,IAAIV,MAAM,OAAO,CAACU,QAAQ,OAAO,GAAG;QAClC,MAAMC,YAAYP,0BAA0BM,QAAQ,OAAO;QAE3D,OAAO;YACL,SAASC,UAAU,OAAO;YAC1B,mBAAmBV,uBAAuB;gBACxC,wBAAwBU,UAAU,sBAAsB;gBACxD,0BACE,AAAqC,YAArC,OAAOD,QAAQ,iBAAiB,GAC5BA,QAAQ,iBAAiB,GACzB;YACR;QACF;IACF;IAEA,MAAME,SAASC,kCACbH;IAOF,MAAMR,yBAAyBP,qBAAqBiB,OAAO,OAAO,KAAK;IAEvE,OAAO;QACL,SAASA,OAAO,OAAO;QACvB,mBAAmBX,uBAAuB;YACxCC;YACA,0BAA0BU,OAAO,iBAAiB;QACpD;IACF;AACF;AAEO,MAAME,iBAAiB;IAC5B,QAAQ;QACN,gBAAgB;YACd,uBAAuB;gBAAC;gBAAoB;aAAkB;YAC9D,2BAA2B7B;YAC3B,4BAA4BwB;QAC9B;QACA,QAAQ;YACN,eAAe;gBACb,aAAa;oBAAE,OAAO;oBAAQ,OAAO;oBAAM,cAAc;gBAAK;YAChE;QACF;IACF;AACF"}

package/dist/es/ai-model/models/glm.mjs.map

@@ -1 +1 @@
-{"version":3,"file":"ai-model/models/glm.mjs","sources":["../../../../src/ai-model/models/glm.ts"],"sourcesContent":["import type { TModelFamily } from '@midscene/shared/env';\nimport type {\n  ChatCompletionCallContext,\n  ChatCompletionParamsResult,\n  ModelAdapterDefinition,\n} from './types';\n\nconst buildGlmChatCompletionParams = (\n  input: ChatCompletionCallContext,\n): ChatCompletionParamsResult => {\n  const { midsceneDefaults, userConfig } = input;\n  const { reasoningEnabled } = userConfig;\n  const commonOverrideConfig: Record<string, unknown> = {};\n\n  if (userConfig.temperature !== undefined) {\n    commonOverrideConfig.temperature = userConfig.temperature;\n  }\n\n  const modelSpecificConfig: Record<string, unknown> = {};\n\n  if (reasoningEnabled !== 'default') {\n    modelSpecificConfig.thinking = {\n      type: (reasoningEnabled ?? false) ? 'enabled' : 'disabled',\n    };\n  }\n\n  return {\n    config: {\n      ...midsceneDefaults,\n      ...commonOverrideConfig,\n      ...modelSpecificConfig,\n    },\n  };\n};\n\nexport const glmAdapters = {\n  'glm-v': {\n    chatCompletion: {\n      unsupportedUserConfig: ['reasoningEffort', 'reasoningBudget'],\n      buildChatCompletionParams: buildGlmChatCompletionParams,\n    },\n    locate: {\n      resultAdapter: {\n        coordinates: { shape: 'bbox', order: 'xy', normalizedBy: 1000 },\n      },\n    },\n  },\n} satisfies Pick<Record<TModelFamily, ModelAdapterDefinition>, 'glm-v'>;\n"],"names":["buildGlmChatCompletionParams","input","midsceneDefaults","userConfig","reasoningEnabled","commonOverrideConfig","undefined","modelSpecificConfig","glmAdapters"],"mappings":"AAOA,MAAMA,+BAA+B,CACnCC;IAEA,MAAM,EAAEC,gBAAgB,EAAEC,UAAU,EAAE,GAAGF;IACzC,MAAM,EAAEG,gBAAgB,EAAE,GAAGD;IAC7B,MAAME,uBAAgD,CAAC;IAEvD,IAAIF,AAA2BG,WAA3BH,WAAW,WAAW,EACxBE,qBAAqB,WAAW,GAAGF,WAAW,WAAW;IAG3D,MAAMI,sBAA+C,CAAC;IAEtD,IAAIH,AAAqB,cAArBA,kBACFG,oBAAoB,QAAQ,GAAG;QAC7B,MAAOH,oBAAoB,QAAS,YAAY;IAClD;IAGF,OAAO;QACL,QAAQ;YACN,GAAGF,gBAAgB;YACnB,GAAGG,oBAAoB;YACvB,GAAGE,mBAAmB;QACxB;IACF;AACF;AAEO,MAAMC,cAAc;IACzB,SAAS;QACP,gBAAgB;YACd,uBAAuB;gBAAC;gBAAmB;aAAkB;YAC7D,2BAA2BR;QAC7B;QACA,QAAQ;YACN,eAAe;gBACb,aAAa;oBAAE,OAAO;oBAAQ,OAAO;oBAAM,cAAc;gBAAK;YAChE;QACF;IACF;AACF"}
+{"version":3,"file":"ai-model/models/glm.mjs","sources":["../../../../src/ai-model/models/glm.ts"],"sourcesContent":["import type { TModelFamily } from '@midscene/shared/env';\nimport type {\n  ChatCompletionCallContext,\n  ChatCompletionParamsResult,\n  ModelAdapterDefinition,\n} from '../model-adapter/types';\n\nconst buildGlmChatCompletionParams = (\n  input: ChatCompletionCallContext,\n): ChatCompletionParamsResult => {\n  const { midsceneDefaults, userConfig } = input;\n  const { reasoningEnabled } = userConfig;\n  const commonOverrideConfig: Record<string, unknown> = {};\n\n  if (userConfig.temperature !== undefined) {\n    commonOverrideConfig.temperature = userConfig.temperature;\n  }\n\n  const modelSpecificConfig: Record<string, unknown> = {};\n\n  if (reasoningEnabled !== 'default') {\n    modelSpecificConfig.thinking = {\n      type: (reasoningEnabled ?? false) ? 'enabled' : 'disabled',\n    };\n  }\n\n  return {\n    config: {\n      ...midsceneDefaults,\n      ...commonOverrideConfig,\n      ...modelSpecificConfig,\n    },\n  };\n};\n\nexport const glmAdapters = {\n  'glm-v': {\n    chatCompletion: {\n      unsupportedUserConfig: ['reasoningEffort', 'reasoningBudget'],\n      buildChatCompletionParams: buildGlmChatCompletionParams,\n    },\n    locate: {\n      resultAdapter: {\n        coordinates: { shape: 'bbox', order: 'xy', normalizedBy: 1000 },\n      },\n    },\n  },\n} satisfies Pick<Record<TModelFamily, ModelAdapterDefinition>, 'glm-v'>;\n"],"names":["buildGlmChatCompletionParams","input","midsceneDefaults","userConfig","reasoningEnabled","commonOverrideConfig","undefined","modelSpecificConfig","glmAdapters"],"mappings":"AAOA,MAAMA,+BAA+B,CACnCC;IAEA,MAAM,EAAEC,gBAAgB,EAAEC,UAAU,EAAE,GAAGF;IACzC,MAAM,EAAEG,gBAAgB,EAAE,GAAGD;IAC7B,MAAME,uBAAgD,CAAC;IAEvD,IAAIF,AAA2BG,WAA3BH,WAAW,WAAW,EACxBE,qBAAqB,WAAW,GAAGF,WAAW,WAAW;IAG3D,MAAMI,sBAA+C,CAAC;IAEtD,IAAIH,AAAqB,cAArBA,kBACFG,oBAAoB,QAAQ,GAAG;QAC7B,MAAOH,oBAAoB,QAAS,YAAY;IAClD;IAGF,OAAO;QACL,QAAQ;YACN,GAAGF,gBAAgB;YACnB,GAAGG,oBAAoB;YACvB,GAAGE,mBAAmB;QACxB;IACF;AACF;AAEO,MAAMC,cAAc;IACzB,SAAS;QACP,gBAAgB;YACd,uBAAuB;gBAAC;gBAAmB;aAAkB;YAC7D,2BAA2BR;QAC7B;QACA,QAAQ;YACN,eAAe;gBACb,aAAa;oBAAE,OAAO;oBAAQ,OAAO;oBAAM,cAAc;gBAAK;YAChE;QACF;IACF;AACF"}

package/dist/es/ai-model/models/gpt.mjs.map

@@ -1 +1 @@
-{"version":3,"file":"ai-model/models/gpt.mjs","sources":["../../../../src/ai-model/models/gpt.ts"],"sourcesContent":["import type { TModelFamily } from '@midscene/shared/env';\nimport type {\n  ChatCompletionCallContext,\n  ChatCompletionParamsResult,\n  ImageDetail,\n  ModelAdapterDefinition,\n} from './types';\n\nconst originalImageDetailForDefaultIntent = (\n  input: ChatCompletionCallContext,\n): ImageDetail | undefined =>\n  input.intent === 'default' || input.requiresOriginalImageDetail\n    ? 'original'\n    : undefined;\n\nconst buildGpt5ChatCompletionParams = (\n  input: ChatCompletionCallContext,\n): ChatCompletionParamsResult => {\n  const { midsceneDefaults, userConfig } = input;\n  const { reasoningEnabled, reasoningEffort } = userConfig;\n  const commonOverrideConfig: Record<string, unknown> = {};\n\n  if (userConfig.temperature !== undefined) {\n    commonOverrideConfig.temperature = userConfig.temperature;\n  }\n\n  const effectiveReasoningEffort =\n    reasoningEnabled === true ? (reasoningEffort ?? 'medium') : 'none';\n\n  return {\n    config: {\n      ...midsceneDefaults,\n      ...commonOverrideConfig,\n      reasoning_effort: effectiveReasoningEffort,\n    },\n  };\n};\n\nexport const gptAdapters = {\n  'gpt-5': {\n    chatCompletion: {\n      unsupportedUserConfig: ['reasoningBudget'],\n      buildChatCompletionParams: buildGpt5ChatCompletionParams,\n      resolveImageDetail: originalImageDetailForDefaultIntent,\n    },\n    locate: {\n      resultAdapter: {\n        coordinates: { shape: 'bbox', order: 'xy' },\n      },\n    },\n  },\n} satisfies Pick<Record<TModelFamily, ModelAdapterDefinition>, 'gpt-5'>;\n"],"names":["originalImageDetailForDefaultIntent","input","undefined","buildGpt5ChatCompletionParams","midsceneDefaults","userConfig","reasoningEnabled","reasoningEffort","commonOverrideConfig","effectiveReasoningEffort","gptAdapters"],"mappings":"AAQA,MAAMA,sCAAsC,CAC1CC,QAEAA,AAAiB,cAAjBA,MAAM,MAAM,IAAkBA,MAAM,2BAA2B,GAC3D,aACAC;AAEN,MAAMC,gCAAgC,CACpCF;IAEA,MAAM,EAAEG,gBAAgB,EAAEC,UAAU,EAAE,GAAGJ;IACzC,MAAM,EAAEK,gBAAgB,EAAEC,eAAe,EAAE,GAAGF;IAC9C,MAAMG,uBAAgD,CAAC;IAEvD,IAAIH,AAA2BH,WAA3BG,WAAW,WAAW,EACxBG,qBAAqB,WAAW,GAAGH,WAAW,WAAW;IAG3D,MAAMI,2BACJH,AAAqB,SAArBA,mBAA6BC,mBAAmB,WAAY;IAE9D,OAAO;QACL,QAAQ;YACN,GAAGH,gBAAgB;YACnB,GAAGI,oBAAoB;YACvB,kBAAkBC;QACpB;IACF;AACF;AAEO,MAAMC,cAAc;IACzB,SAAS;QACP,gBAAgB;YACd,uBAAuB;gBAAC;aAAkB;YAC1C,2BAA2BP;YAC3B,oBAAoBH;QACtB;QACA,QAAQ;YACN,eAAe;gBACb,aAAa;oBAAE,OAAO;oBAAQ,OAAO;gBAAK;YAC5C;QACF;IACF;AACF"}
+{"version":3,"file":"ai-model/models/gpt.mjs","sources":["../../../../src/ai-model/models/gpt.ts"],"sourcesContent":["import type { TModelFamily } from '@midscene/shared/env';\nimport type {\n  ChatCompletionCallContext,\n  ChatCompletionParamsResult,\n  ImageDetail,\n  ModelAdapterDefinition,\n} from '../model-adapter/types';\n\nconst originalImageDetailForDefaultIntent = (\n  input: ChatCompletionCallContext,\n): ImageDetail | undefined =>\n  input.intent === 'default' || input.requiresOriginalImageDetail\n    ? 'original'\n    : undefined;\n\nconst buildGpt5ChatCompletionParams = (\n  input: ChatCompletionCallContext,\n): ChatCompletionParamsResult => {\n  const { midsceneDefaults, userConfig } = input;\n  const { reasoningEnabled, reasoningEffort } = userConfig;\n  const commonOverrideConfig: Record<string, unknown> = {};\n\n  if (userConfig.temperature !== undefined) {\n    commonOverrideConfig.temperature = userConfig.temperature;\n  }\n\n  const effectiveReasoningEffort =\n    reasoningEnabled === true ? (reasoningEffort ?? 'medium') : 'none';\n\n  return {\n    config: {\n      ...midsceneDefaults,\n      ...commonOverrideConfig,\n      reasoning_effort: effectiveReasoningEffort,\n    },\n  };\n};\n\nexport const gptAdapters = {\n  'gpt-5': {\n    chatCompletion: {\n      unsupportedUserConfig: ['reasoningBudget'],\n      buildChatCompletionParams: buildGpt5ChatCompletionParams,\n      resolveImageDetail: originalImageDetailForDefaultIntent,\n    },\n    locate: {\n      resultAdapter: {\n        coordinates: { shape: 'bbox', order: 'xy' },\n      },\n    },\n  },\n} satisfies Pick<Record<TModelFamily, ModelAdapterDefinition>, 'gpt-5'>;\n"],"names":["originalImageDetailForDefaultIntent","input","undefined","buildGpt5ChatCompletionParams","midsceneDefaults","userConfig","reasoningEnabled","reasoningEffort","commonOverrideConfig","effectiveReasoningEffort","gptAdapters"],"mappings":"AAQA,MAAMA,sCAAsC,CAC1CC,QAEAA,AAAiB,cAAjBA,MAAM,MAAM,IAAkBA,MAAM,2BAA2B,GAC3D,aACAC;AAEN,MAAMC,gCAAgC,CACpCF;IAEA,MAAM,EAAEG,gBAAgB,EAAEC,UAAU,EAAE,GAAGJ;IACzC,MAAM,EAAEK,gBAAgB,EAAEC,eAAe,EAAE,GAAGF;IAC9C,MAAMG,uBAAgD,CAAC;IAEvD,IAAIH,AAA2BH,WAA3BG,WAAW,WAAW,EACxBG,qBAAqB,WAAW,GAAGH,WAAW,WAAW;IAG3D,MAAMI,2BACJH,AAAqB,SAArBA,mBAA6BC,mBAAmB,WAAY;IAE9D,OAAO;QACL,QAAQ;YACN,GAAGH,gBAAgB;YACnB,GAAGI,oBAAoB;YACvB,kBAAkBC;QACpB;IACF;AACF;AAEO,MAAMC,cAAc;IACzB,SAAS;QACP,gBAAgB;YACd,uBAAuB;gBAAC;aAAkB;YAC1C,2BAA2BP;YAC3B,oBAAoBH;QACtB;QACA,QAAQ;YACN,eAAe;gBACb,aAAa;oBAAE,OAAO;oBAAQ,OAAO;gBAAK;YAC5C;QACF;IACF;AACF"}

package/dist/es/ai-model/models/kimi.mjs.map

@@ -1 +1 @@
-{"version":3,"file":"ai-model/models/kimi.mjs","sources":["../../../../src/ai-model/models/kimi.ts"],"sourcesContent":["import type { TModelFamily } from '@midscene/shared/env';\nimport type {\n  ChatCompletionCallContext,\n  ChatCompletionParamsResult,\n  ModelAdapterDefinition,\n} from './types';\n\nconst buildKimiChatCompletionParams = (\n  input: ChatCompletionCallContext,\n): ChatCompletionParamsResult => {\n  const { midsceneDefaults, userConfig } = input;\n  const { reasoningEnabled } = userConfig;\n  const effectiveReasoningEnabled = reasoningEnabled ?? false;\n  const commonOverrideConfig: Record<string, unknown> = {};\n\n  // kimi disallow custom temperature\n  commonOverrideConfig.temperature = undefined;\n\n  const modelSpecificConfig: Record<string, unknown> = {\n    thinking: {\n      type: effectiveReasoningEnabled ? 'enabled' : 'disabled',\n    },\n  };\n\n  return {\n    config: {\n      ...midsceneDefaults,\n      ...commonOverrideConfig,\n      ...modelSpecificConfig,\n    },\n  };\n};\n\nexport const kimiAdapters = {\n  kimi: {\n    chatCompletion: {\n      unsupportedUserConfig: ['reasoningEffort', 'reasoningBudget'],\n      buildChatCompletionParams: buildKimiChatCompletionParams,\n    },\n    locate: {\n      resultAdapter: {\n        coordinates: { shape: 'point', order: 'xy', normalizedBy: 1 },\n      },\n    },\n  },\n} satisfies Pick<Record<TModelFamily, ModelAdapterDefinition>, 'kimi'>;\n"],"names":["buildKimiChatCompletionParams","input","midsceneDefaults","userConfig","reasoningEnabled","effectiveReasoningEnabled","commonOverrideConfig","undefined","modelSpecificConfig","kimiAdapters"],"mappings":"AAOA,MAAMA,gCAAgC,CACpCC;IAEA,MAAM,EAAEC,gBAAgB,EAAEC,UAAU,EAAE,GAAGF;IACzC,MAAM,EAAEG,gBAAgB,EAAE,GAAGD;IAC7B,MAAME,4BAA4BD,oBAAoB;IACtD,MAAME,uBAAgD,CAAC;IAGvDA,qBAAqB,WAAW,GAAGC;IAEnC,MAAMC,sBAA+C;QACnD,UAAU;YACR,MAAMH,4BAA4B,YAAY;QAChD;IACF;IAEA,OAAO;QACL,QAAQ;YACN,GAAGH,gBAAgB;YACnB,GAAGI,oBAAoB;YACvB,GAAGE,mBAAmB;QACxB;IACF;AACF;AAEO,MAAMC,eAAe;IAC1B,MAAM;QACJ,gBAAgB;YACd,uBAAuB;gBAAC;gBAAmB;aAAkB;YAC7D,2BAA2BT;QAC7B;QACA,QAAQ;YACN,eAAe;gBACb,aAAa;oBAAE,OAAO;oBAAS,OAAO;oBAAM,cAAc;gBAAE;YAC9D;QACF;IACF;AACF"}
+{"version":3,"file":"ai-model/models/kimi.mjs","sources":["../../../../src/ai-model/models/kimi.ts"],"sourcesContent":["import type { TModelFamily } from '@midscene/shared/env';\nimport type {\n  ChatCompletionCallContext,\n  ChatCompletionParamsResult,\n  ModelAdapterDefinition,\n} from '../model-adapter/types';\n\nconst buildKimiChatCompletionParams = (\n  input: ChatCompletionCallContext,\n): ChatCompletionParamsResult => {\n  const { midsceneDefaults, userConfig } = input;\n  const { reasoningEnabled } = userConfig;\n  const effectiveReasoningEnabled = reasoningEnabled ?? false;\n  const commonOverrideConfig: Record<string, unknown> = {};\n\n  // kimi disallow custom temperature\n  commonOverrideConfig.temperature = undefined;\n\n  const modelSpecificConfig: Record<string, unknown> = {\n    thinking: {\n      type: effectiveReasoningEnabled ? 'enabled' : 'disabled',\n    },\n  };\n\n  return {\n    config: {\n      ...midsceneDefaults,\n      ...commonOverrideConfig,\n      ...modelSpecificConfig,\n    },\n  };\n};\n\nexport const kimiAdapters = {\n  kimi: {\n    chatCompletion: {\n      unsupportedUserConfig: ['reasoningEffort', 'reasoningBudget'],\n      buildChatCompletionParams: buildKimiChatCompletionParams,\n    },\n    locate: {\n      resultAdapter: {\n        coordinates: { shape: 'point', order: 'xy', normalizedBy: 1 },\n      },\n    },\n  },\n} satisfies Pick<Record<TModelFamily, ModelAdapterDefinition>, 'kimi'>;\n"],"names":["buildKimiChatCompletionParams","input","midsceneDefaults","userConfig","reasoningEnabled","effectiveReasoningEnabled","commonOverrideConfig","undefined","modelSpecificConfig","kimiAdapters"],"mappings":"AAOA,MAAMA,gCAAgC,CACpCC;IAEA,MAAM,EAAEC,gBAAgB,EAAEC,UAAU,EAAE,GAAGF;IACzC,MAAM,EAAEG,gBAAgB,EAAE,GAAGD;IAC7B,MAAME,4BAA4BD,oBAAoB;IACtD,MAAME,uBAAgD,CAAC;IAGvDA,qBAAqB,WAAW,GAAGC;IAEnC,MAAMC,sBAA+C;QACnD,UAAU;YACR,MAAMH,4BAA4B,YAAY;QAChD;IACF;IAEA,OAAO;QACL,QAAQ;YACN,GAAGH,gBAAgB;YACnB,GAAGI,oBAAoB;YACvB,GAAGE,mBAAmB;QACxB;IACF;AACF;AAEO,MAAMC,eAAe;IAC1B,MAAM;QACJ,gBAAgB;YACd,uBAAuB;gBAAC;gBAAmB;aAAkB;YAC7D,2BAA2BT;QAC7B;QACA,QAAQ;YACN,eAAe;gBACb,aAAa;oBAAE,OAAO;oBAAS,OAAO;oBAAM,cAAc;gBAAE;YAC9D;QACF;IACF;AACF"}