@midscene/core 1.10.2-beta-20260706032158.0 → 1.10.3

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

@@ -1,6 +1,4 @@
-import { assert } from "@midscene/shared/utils";
-import { jsonrepair } from "jsonrepair";
-import { extractJSONFromCodeBlock, safeParseJson } from "../service-caller/json.mjs";
+import { parseModelResponseJson } from "../service-caller/json.mjs";
 import { createLocateResultValue, unwrapCoordinateListLikeInput } from "../shared/model-locate-result/index.mjs";
 const doubaoBboxCoordinatesMeta = {
     shape: 'bbox',
@@ -12,68 +10,48 @@ const doubaoPointCoordinatesMeta = {
     order: 'xy',
     normalizedBy: 1000
 };
-function normalizeDoubaoJsonObject(obj, context = {}) {
-    if (null == obj) return obj;
-    if (Array.isArray(obj)) return obj.map((item)=>normalizeDoubaoJsonObject(item, context));
-    if ('object' == typeof obj) {
-        const normalized = {};
-        for (const [key, value] of Object.entries(obj)){
-            const trimmedKey = key.trim();
-            const preserveStringValue = context.preserveStringValueKeys?.includes(trimmedKey) ?? false;
-            const normalizedValue = 'string' == typeof value ? preserveStringValue ? value : value.trim() : normalizeDoubaoJsonObject(value, context);
-            normalized[trimmedKey] = normalizedValue;
-        }
-        return normalized;
-    }
-    return 'string' == typeof obj ? obj.trim() : obj;
+function parseNumbersFromBboxString(input) {
+    return (input.match(/\d+/g) ?? []).map(Number).filter(Number.isFinite);
 }
-function shouldRepairDoubaoLocateJson(source) {
-    return 'locate' === source || 'section-locator' === source || 'planning-action-param' === source;
+function parseLeadingNumberFromString(input) {
+    const match = input.match(/^\s*(\d+)/);
+    return match ? Number(match[1]) : void 0;
 }
-function preprocessDoubaoLocateJson(input) {
-    if (input.includes('bbox')) while(/\d+\s+\d+/.test(input))input = input.replace(/(\d+)\s+(\d+)/g, '$1,$2');
-    return input;
+function isCleanCoordinateString(input) {
+    return /^\s*\d+(?:[\s,;]+\d+)*\s*[,;]?\s*$/.test(input);
 }
-const doubaoJsonParser = (raw, context = {
-    source: 'generic-object'
-})=>{
-    const { source } = context;
-    try {
-        return safeParseJson(raw, context);
-    } catch (firstError) {
-        if (!shouldRepairDoubaoLocateJson(source)) throw firstError;
-        const jsonString = preprocessDoubaoLocateJson(extractJSONFromCodeBlock(raw));
-        try {
-            return normalizeDoubaoJsonObject(JSON.parse(jsonrepair(jsonString)), context);
-        } catch (error) {
-            throw Error(`failed to parse LLM response into JSON. Error - ${String(error ?? firstError ?? 'unknown error')}. Response - \n ${raw}`);
+function parseNumbersFromBboxArray(input) {
+    const numbers = [];
+    for (const item of input){
+        if ('number' == typeof item) {
+            numbers.push(item);
+            continue;
+        }
+        if ('string' == typeof item) if (isCleanCoordinateString(item)) {
+            numbers.push(...parseNumbersFromBboxString(item));
+            continue;
+        } else {
+            const leadingNumber = parseLeadingNumberFromString(item);
+            if (void 0 !== leadingNumber) numbers.push(leadingNumber);
         }
+        break;
     }
-};
+    return numbers.filter(Number.isFinite);
+}
+function parseNumbersFromTaggedBboxArray(input) {
+    if (input.length < 2) return null;
+    const taggedBbox = input[1];
+    if ('string' == typeof taggedBbox) return parseNumbersFromBboxString(taggedBbox);
+    if (Array.isArray(taggedBbox)) return parseNumbersFromBboxArray(taggedBbox);
+    return null;
+}
 function parseDoubaoRawLocateValue(input) {
     const bbox = unwrapCoordinateListLikeInput(input);
-    if ('string' == typeof bbox) {
-        assert(/^(\d+)\s(\d+)\s(\d+)\s(\d+)$/.test(bbox.trim()), `invalid bbox data string for doubao-vision mode: ${bbox}`);
-        const splitted = bbox.split(' ');
-        if (4 === splitted.length) return createLocateResultValue(doubaoBboxCoordinatesMeta, [
-            Number(splitted[0]),
-            Number(splitted[1]),
-            Number(splitted[2]),
-            Number(splitted[3])
-        ]);
-        throw new Error(`invalid bbox data string for doubao-vision mode: ${bbox}`);
-    }
     let bboxList = [];
-    if (Array.isArray(bbox) && 'string' == typeof bbox[0]) bbox.forEach((item)=>{
-        if ('string' == typeof item && item.includes(',')) {
-            const [x, y] = item.split(',');
-            bboxList.push(Number(x.trim()), Number(y.trim()));
-        } else if ('string' == typeof item && item.includes(' ')) {
-            const [x, y] = item.split(' ');
-            bboxList.push(Number(x.trim()), Number(y.trim()));
-        } else bboxList.push(Number(item));
-    });
-    else bboxList = bbox;
+    if ('string' == typeof bbox) {
+        bboxList = parseNumbersFromBboxString(bbox);
+        if (4 !== bboxList.length) throw new Error(`invalid bbox data string for doubao-vision mode: ${bbox}`);
+    } else bboxList = Array.isArray(bbox) ? 'string' == typeof bbox[0] && 'bbox' === bbox[0].trim() ? parseNumbersFromTaggedBboxArray(bbox) ?? [] : parseNumbersFromBboxArray(bbox) : bbox;
     if (4 === bboxList.length || 5 === bboxList.length) return createLocateResultValue(doubaoBboxCoordinatesMeta, [
         bboxList[0],
         bboxList[1],
@@ -84,7 +62,7 @@ function parseDoubaoRawLocateValue(input) {
         bboxList[0],
         bboxList[1]
     ]);
-    if (8 === bbox.length) return createLocateResultValue(doubaoBboxCoordinatesMeta, [
+    if (8 === bboxList.length) return createLocateResultValue(doubaoBboxCoordinatesMeta, [
         bboxList[0],
         bboxList[1],
         bboxList[4],
@@ -114,7 +92,7 @@ const buildDoubaoChatCompletionParams = (input)=>{
     };
 };
 const doubaoVisionAdapter = {
-    jsonParser: doubaoJsonParser,
+    jsonParser: parseModelResponseJson,
     chatCompletion: {
         unsupportedUserConfig: [
             'reasoningBudget'
@@ -133,6 +111,6 @@ const doubaoAdapters = {
     'doubao-vision': doubaoVisionAdapter,
     'doubao-seed': doubaoVisionAdapter
 };
-export { doubaoAdapters, normalizeDoubaoJsonObject, parseDoubaoRawLocateValue, preprocessDoubaoLocateJson, shouldRepairDoubaoLocateJson };
+export { doubaoAdapters, parseDoubaoRawLocateValue };
 
 //# sourceMappingURL=doubao.mjs.map

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 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  createLocateResultValue,\n  unwrapCoordinateListLikeInput,\n} from '../shared/model-locate-result';\n\nconst doubaoBboxCoordinatesMeta = {\n  shape: 'bbox',\n  order: 'xy',\n  normalizedBy: 1000,\n} as const;\nconst doubaoPointCoordinatesMeta = {\n  shape: 'point',\n  order: 'xy',\n  normalizedBy: 1000,\n} as const;\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 createLocateResultValue(doubaoBboxCoordinatesMeta, [\n        Number(splitted[0]),\n        Number(splitted[1]),\n        Number(splitted[2]),\n        Number(splitted[3]),\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 createLocateResultValue(doubaoBboxCoordinatesMeta, [\n      bboxList[0],\n      bboxList[1],\n      bboxList[2],\n      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 createLocateResultValue(doubaoPointCoordinatesMeta, [\n      bboxList[0],\n      bboxList[1],\n    ]);\n  }\n\n  if (bbox.length === 8) {\n    return createLocateResultValue(doubaoBboxCoordinatesMeta, [\n      bboxList[0],\n      bboxList[1],\n      bboxList[4],\n      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    useReasoningAsContentFallback: true,\n  },\n  locate: {\n    resultAdapter: {\n      coordinates: doubaoBboxCoordinatesMeta,\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":["doubaoBboxCoordinatesMeta","doubaoPointCoordinatesMeta","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","createLocateResultValue","Number","bboxList","x","y","msg","buildDoubaoChatCompletionParams","midsceneDefaults","userConfig","reasoningEnabled","reasoningEffort","commonOverrideConfig","undefined","modelSpecificConfig","doubaoVisionAdapter","doubaoAdapters"],"mappings":";;;;AAoBA,MAAMA,4BAA4B;IAChC,OAAO;IACP,OAAO;IACP,cAAc;AAChB;AACA,MAAMC,6BAA6B;IACjC,OAAO;IACP,OAAO;IACP,cAAc;AAChB;AAEO,SAASC,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,OAAOC,wBAAwBlC,2BAA2B;YACxDmC,OAAOF,QAAQ,CAAC,EAAE;YAClBE,OAAOF,QAAQ,CAAC,EAAE;YAClBE,OAAOF,QAAQ,CAAC,EAAE;YAClBE,OAAOF,QAAQ,CAAC,EAAE;SACnB;QAEH,MAAM,IAAIN,MAAM,CAAC,iDAAiD,EAAEG,MAAM;IAC5E;IAEA,IAAIM,WAAqB,EAAE;IAC3B,IAAI/B,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,CAAC+B,GAAGC,EAAE,GAAGhC,KAAK,KAAK,CAAC;YAC1B8B,SAAS,IAAI,CAACD,OAAOE,EAAE,IAAI,KAAKF,OAAOG,EAAE,IAAI;QAC/C,OAAO,IAAI,AAAgB,YAAhB,OAAOhC,QAAqBA,KAAK,QAAQ,CAAC,MAAM;YACzD,MAAM,CAAC+B,GAAGC,EAAE,GAAGhC,KAAK,KAAK,CAAC;YAC1B8B,SAAS,IAAI,CAACD,OAAOE,EAAE,IAAI,KAAKF,OAAOG,EAAE,IAAI;QAC/C,OACEF,SAAS,IAAI,CAACD,OAAO7B;IAEzB;SAEA8B,WAAWN;IAGb,IAAIM,AAAoB,MAApBA,SAAS,MAAM,IAAUA,AAAoB,MAApBA,SAAS,MAAM,EAC1C,OAAOF,wBAAwBlC,2BAA2B;QACxDoC,QAAQ,CAAC,EAAE;QACXA,QAAQ,CAAC,EAAE;QACXA,QAAQ,CAAC,EAAE;QACXA,QAAQ,CAAC,EAAE;KACZ;IAGH,IACEA,AAAoB,MAApBA,SAAS,MAAM,IACfA,AAAoB,MAApBA,SAAS,MAAM,IACfA,AAAoB,MAApBA,SAAS,MAAM,IACfA,AAAoB,MAApBA,SAAS,MAAM,EAEf,OAAOF,wBAAwBjC,4BAA4B;QACzDmC,QAAQ,CAAC,EAAE;QACXA,QAAQ,CAAC,EAAE;KACZ;IAGH,IAAIN,AAAgB,MAAhBA,KAAK,MAAM,EACb,OAAOI,wBAAwBlC,2BAA2B;QACxDoC,QAAQ,CAAC,EAAE;QACXA,QAAQ,CAAC,EAAE;QACXA,QAAQ,CAAC,EAAE;QACXA,QAAQ,CAAC,EAAE;KACZ;IAGH,MAAMG,MAAM,CAAC,0CAA0C,EAAEf,KAAK,SAAS,CAACM,MAAM,CAAC,CAAC;IAChF,MAAM,IAAIH,MAAMY;AAClB;AAEA,MAAMC,kCAAkC,CACtCvB;IAEA,MAAM,EAAEwB,gBAAgB,EAAEC,UAAU,EAAE,GAAGzB;IACzC,MAAM,EAAE0B,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,YAAY9B;IACZ,gBAAgB;QACd,uBAAuB;YAAC;SAAkB;QAC1C,2BAA2BsB;QAC3B,+BAA+B;IACjC;IACA,QAAQ;QACN,eAAe;YACb,aAAaxC;YACb,qBAAqB6B;QACvB;IACF;AACF;AAEO,MAAMoB,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 type {\n  ChatCompletionCallContext,\n  ChatCompletionParamsResult,\n  ModelAdapterDefinition,\n} from '../model-adapter/types';\nimport { parseModelResponseJson } from '../service-caller/json';\nimport {\n  type LocateResultValue,\n  createLocateResultValue,\n  unwrapCoordinateListLikeInput,\n} from '../shared/model-locate-result';\n\nconst doubaoBboxCoordinatesMeta = {\n  shape: 'bbox',\n  order: 'xy',\n  normalizedBy: 1000,\n} as const;\nconst doubaoPointCoordinatesMeta = {\n  shape: 'point',\n  order: 'xy',\n  normalizedBy: 1000,\n} as const;\n\nfunction parseNumbersFromBboxString(input: string): number[] {\n  return (input.match(/\\d+/g) ?? []).map(Number).filter(Number.isFinite);\n}\n\nfunction parseLeadingNumberFromString(input: string): number | undefined {\n  const match = input.match(/^\\s*(\\d+)/);\n  return match ? Number(match[1]) : undefined;\n}\n\n/**\n * Clean coordinate strings can contain multiple positive integers, e.g.\n * - \"123 100\"\n * - \"123,100\"\n * - \"277; 664 291;\"\n * Dirty strings like \"345<\" are handled by taking the leading number and\n * dropping the remaining array items.\n */\nfunction isCleanCoordinateString(input: string): boolean {\n  return /^\\s*\\d+(?:[\\s,;]+\\d+)*\\s*[,;]?\\s*$/.test(input);\n}\n\nfunction parseNumbersFromBboxArray(input: unknown[]): number[] {\n  const numbers: number[] = [];\n\n  for (const item of input) {\n    if (typeof item === 'number') {\n      numbers.push(item);\n      continue;\n    }\n\n    if (typeof item === 'string') {\n      if (isCleanCoordinateString(item)) {\n        numbers.push(...parseNumbersFromBboxString(item));\n        continue;\n      } else {\n        const leadingNumber = parseLeadingNumberFromString(item);\n        if (leadingNumber !== undefined) {\n          numbers.push(leadingNumber);\n        }\n        // Once a dirty string appears, the remaining array items usually belong\n        // to broken JSON repair output, e.g. [410, 295, 885, \"345<\", \"/bbox>...\"].\n        break;\n      }\n    }\n\n    break;\n  }\n\n  return numbers.filter(Number.isFinite);\n}\n\n/**\n * Some models return a tagged bbox tuple, e.g.\n * - [\"bbox\", [782, 541, 815, 559]]\n * - [\"bbox\", \"782, 541, 815, 559\"]\n */\nfunction parseNumbersFromTaggedBboxArray(input: unknown[]): number[] | null {\n  if (input.length < 2) {\n    return null;\n  }\n\n  const taggedBbox = input[1];\n  if (typeof taggedBbox === 'string') {\n    return parseNumbersFromBboxString(taggedBbox);\n  }\n\n  if (Array.isArray(taggedBbox)) {\n    return parseNumbersFromBboxArray(taggedBbox);\n  }\n\n  return null;\n}\n\nexport function parseDoubaoRawLocateValue(input: unknown): LocateResultValue {\n  const bbox = unwrapCoordinateListLikeInput(input as any);\n  let bboxList: number[] = [];\n\n  if (typeof bbox === 'string') {\n    /**\n     * Some models return bbox as a string, e.g.\n     * - { \"bbox\": \"[336, 163, 717, 200]\" }.\n     * - { \"bbox\": \"336, 163, 717, 200\" }.\n     * - { \"bbox\": \"336 163 717 200\" }.\n     */\n    bboxList = parseNumbersFromBboxString(bbox);\n    if (bboxList.length !== 4) {\n      throw new Error(\n        `invalid bbox data string for doubao-vision mode: ${bbox}`,\n      );\n    }\n  } else if (Array.isArray(bbox)) {\n    if (typeof bbox[0] === 'string' && bbox[0].trim() === 'bbox') {\n      bboxList = parseNumbersFromTaggedBboxArray(bbox) ?? [];\n    } else {\n      bboxList = parseNumbersFromBboxArray(bbox);\n    }\n  } else {\n    bboxList = bbox as number[];\n  }\n\n  if (bboxList.length === 4 || bboxList.length === 5) {\n    return createLocateResultValue(doubaoBboxCoordinatesMeta, [\n      bboxList[0],\n      bboxList[1],\n      bboxList[2],\n      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 createLocateResultValue(doubaoPointCoordinatesMeta, [\n      bboxList[0],\n      bboxList[1],\n    ]);\n  }\n\n  if (bboxList.length === 8) {\n    return createLocateResultValue(doubaoBboxCoordinatesMeta, [\n      bboxList[0],\n      bboxList[1],\n      bboxList[4],\n      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: parseModelResponseJson,\n  chatCompletion: {\n    unsupportedUserConfig: ['reasoningBudget'],\n    buildChatCompletionParams: buildDoubaoChatCompletionParams,\n    useReasoningAsContentFallback: true,\n  },\n  locate: {\n    resultAdapter: {\n      coordinates: doubaoBboxCoordinatesMeta,\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":["doubaoBboxCoordinatesMeta","doubaoPointCoordinatesMeta","parseNumbersFromBboxString","input","Number","parseLeadingNumberFromString","match","undefined","isCleanCoordinateString","parseNumbersFromBboxArray","numbers","item","leadingNumber","parseNumbersFromTaggedBboxArray","taggedBbox","Array","parseDoubaoRawLocateValue","bbox","unwrapCoordinateListLikeInput","bboxList","Error","createLocateResultValue","msg","JSON","buildDoubaoChatCompletionParams","midsceneDefaults","userConfig","reasoningEnabled","reasoningEffort","commonOverrideConfig","modelSpecificConfig","doubaoVisionAdapter","parseModelResponseJson","doubaoAdapters"],"mappings":";;AAaA,MAAMA,4BAA4B;IAChC,OAAO;IACP,OAAO;IACP,cAAc;AAChB;AACA,MAAMC,6BAA6B;IACjC,OAAO;IACP,OAAO;IACP,cAAc;AAChB;AAEA,SAASC,2BAA2BC,KAAa;IAC/C,OAAQA,AAAAA,CAAAA,MAAM,KAAK,CAAC,WAAW,EAAC,EAAG,GAAG,CAACC,QAAQ,MAAM,CAACA,OAAO,QAAQ;AACvE;AAEA,SAASC,6BAA6BF,KAAa;IACjD,MAAMG,QAAQH,MAAM,KAAK,CAAC;IAC1B,OAAOG,QAAQF,OAAOE,KAAK,CAAC,EAAE,IAAIC;AACpC;AAUA,SAASC,wBAAwBL,KAAa;IAC5C,OAAO,qCAAqC,IAAI,CAACA;AACnD;AAEA,SAASM,0BAA0BN,KAAgB;IACjD,MAAMO,UAAoB,EAAE;IAE5B,KAAK,MAAMC,QAAQR,MAAO;QACxB,IAAI,AAAgB,YAAhB,OAAOQ,MAAmB;YAC5BD,QAAQ,IAAI,CAACC;YACb;QACF;QAEA,IAAI,AAAgB,YAAhB,OAAOA,MACT,IAAIH,wBAAwBG,OAAO;YACjCD,QAAQ,IAAI,IAAIR,2BAA2BS;YAC3C;QACF,OAAO;YACL,MAAMC,gBAAgBP,6BAA6BM;YACnD,IAAIC,AAAkBL,WAAlBK,eACFF,QAAQ,IAAI,CAACE;QAKjB;QAGF;IACF;IAEA,OAAOF,QAAQ,MAAM,CAACN,OAAO,QAAQ;AACvC;AAOA,SAASS,gCAAgCV,KAAgB;IACvD,IAAIA,MAAM,MAAM,GAAG,GACjB,OAAO;IAGT,MAAMW,aAAaX,KAAK,CAAC,EAAE;IAC3B,IAAI,AAAsB,YAAtB,OAAOW,YACT,OAAOZ,2BAA2BY;IAGpC,IAAIC,MAAM,OAAO,CAACD,aAChB,OAAOL,0BAA0BK;IAGnC,OAAO;AACT;AAEO,SAASE,0BAA0Bb,KAAc;IACtD,MAAMc,OAAOC,8BAA8Bf;IAC3C,IAAIgB,WAAqB,EAAE;IAE3B,IAAI,AAAgB,YAAhB,OAAOF,MAAmB;QAO5BE,WAAWjB,2BAA2Be;QACtC,IAAIE,AAAoB,MAApBA,SAAS,MAAM,EACjB,MAAM,IAAIC,MACR,CAAC,iDAAiD,EAAEH,MAAM;IAGhE,OAEIE,WAFOJ,MAAM,OAAO,CAACE,QACnB,AAAmB,YAAnB,OAAOA,IAAI,CAAC,EAAE,IAAiBA,AAAmB,WAAnBA,IAAI,CAAC,EAAE,CAAC,IAAI,KAClCJ,gCAAgCI,SAAS,EAAE,GAE3CR,0BAA0BQ,QAG5BA;IAGb,IAAIE,AAAoB,MAApBA,SAAS,MAAM,IAAUA,AAAoB,MAApBA,SAAS,MAAM,EAC1C,OAAOE,wBAAwBrB,2BAA2B;QACxDmB,QAAQ,CAAC,EAAE;QACXA,QAAQ,CAAC,EAAE;QACXA,QAAQ,CAAC,EAAE;QACXA,QAAQ,CAAC,EAAE;KACZ;IAGH,IACEA,AAAoB,MAApBA,SAAS,MAAM,IACfA,AAAoB,MAApBA,SAAS,MAAM,IACfA,AAAoB,MAApBA,SAAS,MAAM,IACfA,AAAoB,MAApBA,SAAS,MAAM,EAEf,OAAOE,wBAAwBpB,4BAA4B;QACzDkB,QAAQ,CAAC,EAAE;QACXA,QAAQ,CAAC,EAAE;KACZ;IAGH,IAAIA,AAAoB,MAApBA,SAAS,MAAM,EACjB,OAAOE,wBAAwBrB,2BAA2B;QACxDmB,QAAQ,CAAC,EAAE;QACXA,QAAQ,CAAC,EAAE;QACXA,QAAQ,CAAC,EAAE;QACXA,QAAQ,CAAC,EAAE;KACZ;IAGH,MAAMG,MAAM,CAAC,0CAA0C,EAAEC,KAAK,SAAS,CAACN,MAAM,CAAC,CAAC;IAChF,MAAM,IAAIG,MAAME;AAClB;AAEA,MAAME,kCAAkC,CACtCrB;IAEA,MAAM,EAAEsB,gBAAgB,EAAEC,UAAU,EAAE,GAAGvB;IACzC,MAAM,EAAEwB,gBAAgB,EAAEC,eAAe,EAAE,GAAGF;IAC9C,MAAMG,uBAAgD,CAAC;IAEvD,IAAIH,AAA2BnB,WAA3BmB,WAAW,WAAW,EACxBG,qBAAqB,WAAW,GAAGH,WAAW,WAAW;IAG3D,MAAMI,sBAA+C,CAAC;IAEtD,IAAIH,AAAqB,cAArBA,kBAAgC;QAClCG,oBAAoB,QAAQ,GAAG;YAC7B,MAAOH,oBAAoB,QAAS,YAAY;QAClD;QACA,IAAIC,iBACFE,oBAAoB,gBAAgB,GAAGF;IAE3C;IAEA,OAAO;QACL,QAAQ;YACN,GAAGH,gBAAgB;YACnB,GAAGI,oBAAoB;YACvB,GAAGC,mBAAmB;QACxB;IACF;AACF;AAEA,MAAMC,sBAA8C;IAClD,YAAYC;IACZ,gBAAgB;QACd,uBAAuB;YAAC;SAAkB;QAC1C,2BAA2BR;QAC3B,+BAA+B;IACjC;IACA,QAAQ;QACN,eAAe;YACb,aAAaxB;YACb,qBAAqBgB;QACvB;IACF;AACF;AAEO,MAAMiB,iBAAiB;IAC5B,iBAAiBF;IACjB,eAAeA;AACjB"}

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

@@ -1,27 +1,17 @@
 const buildGeminiChatCompletionParams = (input)=>{
-    const { midsceneDefaults, userConfig, intent } = input;
+    const { midsceneDefaults, userConfig } = input;
     const { reasoningEnabled, reasoningEffort } = userConfig;
     const commonOverrideConfig = {};
     if (void 0 !== userConfig.temperature) commonOverrideConfig.temperature = userConfig.temperature;
     const modelSpecificConfig = {};
-    if ('default' !== reasoningEnabled) if (reasoningEffort) modelSpecificConfig.extra_body = {
+    if ('default' !== reasoningEnabled) modelSpecificConfig.extra_body = {
         google: {
             thinking_config: {
-                thinking_level: reasoningEffort,
-                include_thoughts: true
+                include_thoughts: true,
+                thinking_level: reasoningEnabled ? reasoningEffort ?? 'medium' : 'minimal'
             }
         }
     };
-    else {
-        if ('insight' === intent) modelSpecificConfig.extra_body = {
-            google: {
-                thinking_config: {
-                    include_thoughts: true
-                }
-            }
-        };
-        modelSpecificConfig.reasoning_effort = 'minimal';
-    }
     return {
         config: {
             ...midsceneDefaults,
@@ -84,7 +74,6 @@ const geminiAdapters = {
     gemini: {
         chatCompletion: {
             unsupportedUserConfig: [
-                'reasoningEnabled',
                 'reasoningBudget'
             ],
             buildChatCompletionParams: buildGeminiChatCompletionParams,

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 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 content = typeof message.content === 'string' ? message.content : '';\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(content) || '';\n\n  return {\n    content,\n    reasoning_content: formatReasoningContent({\n      geminiReasoningContent,\n      providerReasoningContent:\n        typeof message.reasoning_content === 'string'\n          ? message.reasoning_content\n          : '',\n    }),\n  };\n};\n\nexport const geminiAdapters = {\n  gemini: {\n    chatCompletion: {\n      unsupportedUserConfig: ['reasoningEnabled', 'reasoningBudget'],\n      buildChatCompletionParams: buildGeminiChatCompletionParams,\n      messageExtraction: {\n        kind: 'custom',\n        extractContentAndReasoning: extractGeminiContentAndReasoning,\n      },\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","geminiAdapters"],"mappings":"AA4BA,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,MAAMd,UAAU,AAA2B,YAA3B,OAAOc,QAAQ,OAAO,GAAgBA,QAAQ,OAAO,GAAG;IAMxE,MAAMR,yBAAyBP,qBAAqBC,YAAY;IAEhE,OAAO;QACLA;QACA,mBAAmBK,uBAAuB;YACxCC;YACA,0BACE,AAAqC,YAArC,OAAOQ,QAAQ,iBAAiB,GAC5BA,QAAQ,iBAAiB,GACzB;QACR;IACF;AACF;AAEO,MAAME,iBAAiB;IAC5B,QAAQ;QACN,gBAAgB;YACd,uBAAuB;gBAAC;gBAAoB;aAAkB;YAC9D,2BAA2B3B;YAC3B,mBAAmB;gBACjB,MAAM;gBACN,4BAA4BwB;YAC9B;QACF;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 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 } = 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: {\n          include_thoughts: true;\n          thinking_level: string;\n        };\n      };\n    };\n  } = {};\n\n  if (reasoningEnabled !== 'default') {\n    // Gemini cannot fully disable native thinking, so \"disabled\" maps to the\n    // lowest thinking level. `include_thoughts` only controls whether Gemini\n    // returns thought summaries; it does not enable thinking, and thought token\n    // usage is based on the full thoughts Gemini generates regardless of\n    // whether summaries are included.\n    modelSpecificConfig.extra_body = {\n      google: {\n        thinking_config: {\n          include_thoughts: true,\n          thinking_level: reasoningEnabled\n            ? (reasoningEffort ?? 'medium')\n            : 'minimal',\n        },\n      },\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 content = typeof message.content === 'string' ? message.content : '';\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(content) || '';\n\n  return {\n    content,\n    reasoning_content: formatReasoningContent({\n      geminiReasoningContent,\n      providerReasoningContent:\n        typeof message.reasoning_content === 'string'\n          ? message.reasoning_content\n          : '',\n    }),\n  };\n};\n\nexport const geminiAdapters = {\n  gemini: {\n    chatCompletion: {\n      unsupportedUserConfig: ['reasoningBudget'],\n      buildChatCompletionParams: buildGeminiChatCompletionParams,\n      messageExtraction: {\n        kind: 'custom',\n        extractContentAndReasoning: extractGeminiContentAndReasoning,\n      },\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","reasoningEnabled","reasoningEffort","commonOverrideConfig","undefined","modelSpecificConfig","extractInlineThought","content","match","isRecord","value","Array","formatReasoningContent","geminiReasoningContent","providerReasoningContent","extractGeminiThoughtParts","contentParts","reasoningParts","part","text","extractGeminiContentAndReasoning","message","extracted","geminiAdapters"],"mappings":"AA4BA,MAAMA,kCAAkC,CACtCC;IAEA,MAAM,EAAEC,gBAAgB,EAAEC,UAAU,EAAE,GAAGF;IACzC,MAAM,EAAEG,gBAAgB,EAAEC,eAAe,EAAE,GAAGF;IAC9C,MAAMG,uBAAgD,CAAC;IAEvD,IAAIH,AAA2BI,WAA3BJ,WAAW,WAAW,EACxBG,qBAAqB,WAAW,GAAGH,WAAW,WAAW;IAG3D,MAAMK,sBASF,CAAC;IAEL,IAAIJ,AAAqB,cAArBA,kBAMFI,oBAAoB,UAAU,GAAG;QAC/B,QAAQ;YACN,iBAAiB;gBACf,kBAAkB;gBAClB,gBAAgBJ,mBACXC,mBAAmB,WACpB;YACN;QACF;IACF;IAGF,OAAO;QACL,QAAQ;YACN,GAAGH,gBAAgB;YACnB,GAAGI,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,MAAMd,UAAU,AAA2B,YAA3B,OAAOc,QAAQ,OAAO,GAAgBA,QAAQ,OAAO,GAAG;IAMxE,MAAMR,yBAAyBP,qBAAqBC,YAAY;IAEhE,OAAO;QACLA;QACA,mBAAmBK,uBAAuB;YACxCC;YACA,0BACE,AAAqC,YAArC,OAAOQ,QAAQ,iBAAiB,GAC5BA,QAAQ,iBAAiB,GACzB;QACR;IACF;AACF;AAEO,MAAME,iBAAiB;IAC5B,QAAQ;QACN,gBAAgB;YACd,uBAAuB;gBAAC;aAAkB;YAC1C,2BAA2B1B;YAC3B,mBAAmB;gBACjB,MAAM;gBACN,4BAA4BuB;YAC9B;QACF;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

@@ -1,8 +1,12 @@
+import { isLocateIntent } from "./utils/intent.mjs";
 const buildGlmChatCompletionParams = (input)=>{
     const { midsceneDefaults, userConfig } = input;
     const { reasoningEnabled } = userConfig;
     const commonOverrideConfig = {};
     if (void 0 !== userConfig.temperature) commonOverrideConfig.temperature = userConfig.temperature;
+    if (isLocateIntent(input.intent)) commonOverrideConfig.response_format = {
+        type: 'json_object'
+    };
     const modelSpecificConfig = {};
     if ('default' !== reasoningEnabled) modelSpecificConfig.thinking = {
         type: reasoningEnabled ?? false ? 'enabled' : 'disabled'

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 '../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      useReasoningAsContentFallback: true,\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;YAC3B,+BAA+B;QACjC;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';\nimport { isLocateIntent } from './utils/intent';\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  // Zhipu structured output JSON mode:\n  // https://docs.bigmodel.cn/cn/guide/capabilities/struct-output\n  if (isLocateIntent(input.intent)) {\n    commonOverrideConfig.response_format = { type: 'json_object' };\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      useReasoningAsContentFallback: true,\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","isLocateIntent","modelSpecificConfig","glmAdapters"],"mappings":";AAQA,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;IAK3D,IAAII,eAAeN,MAAM,MAAM,GAC7BI,qBAAqB,eAAe,GAAG;QAAE,MAAM;IAAc;IAG/D,MAAMG,sBAA+C,CAAC;IAEtD,IAAIJ,AAAqB,cAArBA,kBACFI,oBAAoB,QAAQ,GAAG;QAC7B,MAAOJ,oBAAoB,QAAS,YAAY;IAClD;IAGF,OAAO;QACL,QAAQ;YACN,GAAGF,gBAAgB;YACnB,GAAGG,oBAAoB;YACvB,GAAGG,mBAAmB;QACxB;IACF;AACF;AAEO,MAAMC,cAAc;IACzB,SAAS;QACP,gBAAgB;YACd,uBAAuB;gBAAC;gBAAmB;aAAkB;YAC7D,2BAA2BT;YAC3B,+BAA+B;QACjC;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

@@ -1,9 +1,13 @@
-const originalImageDetailForDefaultIntent = (input)=>'default' === input.intent || input.requiresOriginalImageDetail ? 'original' : void 0;
+import { isLocateIntent } from "./utils/intent.mjs";
+const originalImageDetailForDefaultIntent = (input)=>isLocateIntent(input.intent) || input.requiresOriginalImageDetail ? 'original' : void 0;
 const buildGpt5ChatCompletionParams = (input)=>{
     const { midsceneDefaults, userConfig } = input;
     const { reasoningEnabled, reasoningEffort } = userConfig;
     const commonOverrideConfig = {};
     if (void 0 !== userConfig.temperature) commonOverrideConfig.temperature = userConfig.temperature;
+    if (isLocateIntent(input.intent)) commonOverrideConfig.response_format = {
+        type: 'json_object'
+    };
     const effectiveReasoningEffort = true === reasoningEnabled ? reasoningEffort ?? 'medium' : 'none';
     return {
         config: {

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 '../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"}
+{"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';\nimport { isLocateIntent } from './utils/intent';\n\nconst originalImageDetailForDefaultIntent = (\n  input: ChatCompletionCallContext,\n): ImageDetail | undefined =>\n  isLocateIntent(input.intent) || 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  // OpenAI Chat Completions JSON mode:\n  // https://platform.openai.com/docs/guides/structured-outputs?api-mode=chat#json-mode\n  if (isLocateIntent(input.intent)) {\n    commonOverrideConfig.response_format = { type: 'json_object' };\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","isLocateIntent","undefined","buildGpt5ChatCompletionParams","midsceneDefaults","userConfig","reasoningEnabled","reasoningEffort","commonOverrideConfig","effectiveReasoningEffort","gptAdapters"],"mappings":";AASA,MAAMA,sCAAsC,CAC1CC,QAEAC,eAAeD,MAAM,MAAM,KAAKA,MAAM,2BAA2B,GAC7D,aACAE;AAEN,MAAMC,gCAAgC,CACpCH;IAEA,MAAM,EAAEI,gBAAgB,EAAEC,UAAU,EAAE,GAAGL;IACzC,MAAM,EAAEM,gBAAgB,EAAEC,eAAe,EAAE,GAAGF;IAC9C,MAAMG,uBAAgD,CAAC;IAEvD,IAAIH,AAA2BH,WAA3BG,WAAW,WAAW,EACxBG,qBAAqB,WAAW,GAAGH,WAAW,WAAW;IAK3D,IAAIJ,eAAeD,MAAM,MAAM,GAC7BQ,qBAAqB,eAAe,GAAG;QAAE,MAAM;IAAc;IAG/D,MAAMC,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,oBAAoBJ;QACtB;QACA,QAAQ;YACN,eAAe;gBACb,aAAa;oBAAE,OAAO;oBAAQ,OAAO;gBAAK;YAC5C;QACF;IACF;AACF"}

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

@@ -1,4 +1,5 @@
 import { createLocateResultValue, parseCoordinateList } from "../shared/model-locate-result/index.mjs";
+import { isLocateIntent } from "./utils/intent.mjs";
 const kimiNormalizedPointCoordinatesMeta = {
     shape: 'point',
     order: 'xy',
@@ -24,6 +25,9 @@ const buildKimiChatCompletionParams = (input)=>{
     const effectiveReasoningEnabled = reasoningEnabled ?? false;
     const commonOverrideConfig = {};
     commonOverrideConfig.temperature = void 0;
+    if (isLocateIntent(input.intent)) commonOverrideConfig.response_format = {
+        type: 'json_object'
+    };
     const modelSpecificConfig = {
         thinking: {
             type: effectiveReasoningEnabled ? 'enabled' : 'disabled'

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 '../model-adapter/types';\nimport {\n  type LocateResultValue,\n  createLocateResultValue,\n  parseCoordinateList,\n} from '../shared/model-locate-result';\n\nconst kimiNormalizedPointCoordinatesMeta = {\n  shape: 'point',\n  order: 'xy',\n  normalizedBy: 1,\n} as const;\nconst kimiPixelPointCoordinatesMeta = {\n  shape: 'point',\n  order: 'xy',\n} as const;\n\nfunction parseKimiRawLocateValue(input: unknown): LocateResultValue {\n  const point = parseCoordinateList(input, 'point');\n  if (point.length < 2) {\n    throw new Error(`invalid point data: ${JSON.stringify(input)} `);\n  }\n  const [x, y] = point;\n  // Keep this compatible with OSWorld's Kimi adapter: values <= 1 are\n  // normalized coordinates, otherwise they are treated as screenshot pixels.\n  const coordinatesMeta =\n    x <= 1 && y <= 1\n      ? kimiNormalizedPointCoordinatesMeta\n      : kimiPixelPointCoordinatesMeta;\n  return createLocateResultValue(coordinatesMeta, [x, y]);\n}\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      useReasoningAsContentFallback: true,\n    },\n    locate: {\n      resultAdapter: {\n        coordinates: kimiNormalizedPointCoordinatesMeta,\n        parseRawLocateValue: parseKimiRawLocateValue,\n      },\n    },\n  },\n} satisfies Pick<Record<TModelFamily, ModelAdapterDefinition>, 'kimi'>;\n"],"names":["kimiNormalizedPointCoordinatesMeta","kimiPixelPointCoordinatesMeta","parseKimiRawLocateValue","input","point","parseCoordinateList","Error","JSON","x","y","coordinatesMeta","createLocateResultValue","buildKimiChatCompletionParams","midsceneDefaults","userConfig","reasoningEnabled","effectiveReasoningEnabled","commonOverrideConfig","undefined","modelSpecificConfig","kimiAdapters"],"mappings":";AAYA,MAAMA,qCAAqC;IACzC,OAAO;IACP,OAAO;IACP,cAAc;AAChB;AACA,MAAMC,gCAAgC;IACpC,OAAO;IACP,OAAO;AACT;AAEA,SAASC,wBAAwBC,KAAc;IAC7C,MAAMC,QAAQC,oBAAoBF,OAAO;IACzC,IAAIC,MAAM,MAAM,GAAG,GACjB,MAAM,IAAIE,MAAM,CAAC,oBAAoB,EAAEC,KAAK,SAAS,CAACJ,OAAO,CAAC,CAAC;IAEjE,MAAM,CAACK,GAAGC,EAAE,GAAGL;IAGf,MAAMM,kBACJF,KAAK,KAAKC,KAAK,IACXT,qCACAC;IACN,OAAOU,wBAAwBD,iBAAiB;QAACF;QAAGC;KAAE;AACxD;AAEA,MAAMG,gCAAgC,CACpCT;IAEA,MAAM,EAAEU,gBAAgB,EAAEC,UAAU,EAAE,GAAGX;IACzC,MAAM,EAAEY,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,2BAA2BR;YAC3B,+BAA+B;QACjC;QACA,QAAQ;YACN,eAAe;gBACb,aAAaZ;gBACb,qBAAqBE;YACvB;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';\nimport {\n  type LocateResultValue,\n  createLocateResultValue,\n  parseCoordinateList,\n} from '../shared/model-locate-result';\nimport { isLocateIntent } from './utils/intent';\n\nconst kimiNormalizedPointCoordinatesMeta = {\n  shape: 'point',\n  order: 'xy',\n  normalizedBy: 1,\n} as const;\nconst kimiPixelPointCoordinatesMeta = {\n  shape: 'point',\n  order: 'xy',\n} as const;\n\nfunction parseKimiRawLocateValue(input: unknown): LocateResultValue {\n  const point = parseCoordinateList(input, 'point');\n  if (point.length < 2) {\n    throw new Error(`invalid point data: ${JSON.stringify(input)} `);\n  }\n  const [x, y] = point;\n  // Keep this compatible with OSWorld's Kimi adapter: values <= 1 are\n  // normalized coordinates, otherwise they are treated as screenshot pixels.\n  const coordinatesMeta =\n    x <= 1 && y <= 1\n      ? kimiNormalizedPointCoordinatesMeta\n      : kimiPixelPointCoordinatesMeta;\n  return createLocateResultValue(coordinatesMeta, [x, y]);\n}\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  // Kimi Chat Completions response_format:\n  // https://platform.kimi.com/docs/api/chat\n  if (isLocateIntent(input.intent)) {\n    commonOverrideConfig.response_format = { type: 'json_object' };\n  }\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      useReasoningAsContentFallback: true,\n    },\n    locate: {\n      resultAdapter: {\n        coordinates: kimiNormalizedPointCoordinatesMeta,\n        parseRawLocateValue: parseKimiRawLocateValue,\n      },\n    },\n  },\n} satisfies Pick<Record<TModelFamily, ModelAdapterDefinition>, 'kimi'>;\n"],"names":["kimiNormalizedPointCoordinatesMeta","kimiPixelPointCoordinatesMeta","parseKimiRawLocateValue","input","point","parseCoordinateList","Error","JSON","x","y","coordinatesMeta","createLocateResultValue","buildKimiChatCompletionParams","midsceneDefaults","userConfig","reasoningEnabled","effectiveReasoningEnabled","commonOverrideConfig","undefined","isLocateIntent","modelSpecificConfig","kimiAdapters"],"mappings":";;AAaA,MAAMA,qCAAqC;IACzC,OAAO;IACP,OAAO;IACP,cAAc;AAChB;AACA,MAAMC,gCAAgC;IACpC,OAAO;IACP,OAAO;AACT;AAEA,SAASC,wBAAwBC,KAAc;IAC7C,MAAMC,QAAQC,oBAAoBF,OAAO;IACzC,IAAIC,MAAM,MAAM,GAAG,GACjB,MAAM,IAAIE,MAAM,CAAC,oBAAoB,EAAEC,KAAK,SAAS,CAACJ,OAAO,CAAC,CAAC;IAEjE,MAAM,CAACK,GAAGC,EAAE,GAAGL;IAGf,MAAMM,kBACJF,KAAK,KAAKC,KAAK,IACXT,qCACAC;IACN,OAAOU,wBAAwBD,iBAAiB;QAACF;QAAGC;KAAE;AACxD;AAEA,MAAMG,gCAAgC,CACpCT;IAEA,MAAM,EAAEU,gBAAgB,EAAEC,UAAU,EAAE,GAAGX;IACzC,MAAM,EAAEY,gBAAgB,EAAE,GAAGD;IAC7B,MAAME,4BAA4BD,oBAAoB;IACtD,MAAME,uBAAgD,CAAC;IAGvDA,qBAAqB,WAAW,GAAGC;IAInC,IAAIC,eAAehB,MAAM,MAAM,GAC7Bc,qBAAqB,eAAe,GAAG;QAAE,MAAM;IAAc;IAG/D,MAAMG,sBAA+C;QACnD,UAAU;YACR,MAAMJ,4BAA4B,YAAY;QAChD;IACF;IAEA,OAAO;QACL,QAAQ;YACN,GAAGH,gBAAgB;YACnB,GAAGI,oBAAoB;YACvB,GAAGG,mBAAmB;QACxB;IACF;AACF;AAEO,MAAMC,eAAe;IAC1B,MAAM;QACJ,gBAAgB;YACd,uBAAuB;gBAAC;gBAAmB;aAAkB;YAC7D,2BAA2BT;YAC3B,+BAA+B;QACjC;QACA,QAAQ;YACN,eAAe;gBACb,aAAaZ;gBACb,qBAAqBE;YACvB;QACF;IACF;AACF"}

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

@@ -1,9 +1,10 @@
+import { isLocateIntent } from "./utils/intent.mjs";
 const buildMimoChatCompletionParams = (input)=>{
     const { intent, midsceneDefaults, userConfig } = input;
     const { reasoningEnabled } = userConfig;
     const commonOverrideConfig = {};
-    commonOverrideConfig.response_format = {
-        type: 'default' === intent ? 'json_object' : 'text'
+    if (isLocateIntent(intent)) commonOverrideConfig.response_format = {
+        type: 'json_object'
     };
     if (void 0 !== userConfig.temperature) commonOverrideConfig.temperature = userConfig.temperature;
     const modelSpecificConfig = {

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

@@ -1 +1 @@
-{"version":3,"file":"ai-model/models/mimo.mjs","sources":["../../../../src/ai-model/models/mimo.ts"],"sourcesContent":["import type { TModelFamily } from '@midscene/shared/env';\nimport type {\n  ChatCompletionCallContext,\n  ChatCompletionParamsResult,\n  ModelAdapterDefinition,\n} from '../model-adapter/types';\n\nconst buildMimoChatCompletionParams = (\n  input: ChatCompletionCallContext,\n): ChatCompletionParamsResult => {\n  const { intent, midsceneDefaults, userConfig } = input;\n  const { reasoningEnabled } = userConfig;\n  const commonOverrideConfig: Record<string, unknown> = {};\n\n  // https://platform.xiaomimimo.com/docs/zh-CN/api/chat/openai-api\n  // Observed with thinking disabled: Mimo needs json_object to return JSON.\n  commonOverrideConfig.response_format = {\n    type: intent === 'default' ? 'json_object' : 'text',\n  };\n\n  if (userConfig.temperature !== undefined) {\n    commonOverrideConfig.temperature = userConfig.temperature;\n  }\n\n  const modelSpecificConfig: Record<string, unknown> = {\n    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 mimoAdapters = {\n  'xiaomi-mimo': {\n    chatCompletion: {\n      unsupportedUserConfig: ['reasoningEffort', 'reasoningBudget'],\n      buildChatCompletionParams: buildMimoChatCompletionParams,\n      useReasoningAsContentFallback: true,\n    },\n  },\n} satisfies Pick<Record<TModelFamily, ModelAdapterDefinition>, 'xiaomi-mimo'>;\n"],"names":["buildMimoChatCompletionParams","input","intent","midsceneDefaults","userConfig","reasoningEnabled","commonOverrideConfig","undefined","modelSpecificConfig","mimoAdapters"],"mappings":"AAOA,MAAMA,gCAAgC,CACpCC;IAEA,MAAM,EAAEC,MAAM,EAAEC,gBAAgB,EAAEC,UAAU,EAAE,GAAGH;IACjD,MAAM,EAAEI,gBAAgB,EAAE,GAAGD;IAC7B,MAAME,uBAAgD,CAAC;IAIvDA,qBAAqB,eAAe,GAAG;QACrC,MAAMJ,AAAW,cAAXA,SAAuB,gBAAgB;IAC/C;IAEA,IAAIE,AAA2BG,WAA3BH,WAAW,WAAW,EACxBE,qBAAqB,WAAW,GAAGF,WAAW,WAAW;IAG3D,MAAMI,sBAA+C;QACnD,UAAU;YACR,MAAOH,oBAAoB,QAAS,YAAY;QAClD;IACF;IAEA,OAAO;QACL,QAAQ;YACN,GAAGF,gBAAgB;YACnB,GAAGG,oBAAoB;YACvB,GAAGE,mBAAmB;QACxB;IACF;AACF;AAEO,MAAMC,eAAe;IAC1B,eAAe;QACb,gBAAgB;YACd,uBAAuB;gBAAC;gBAAmB;aAAkB;YAC7D,2BAA2BT;YAC3B,+BAA+B;QACjC;IACF;AACF"}
+{"version":3,"file":"ai-model/models/mimo.mjs","sources":["../../../../src/ai-model/models/mimo.ts"],"sourcesContent":["import type { TModelFamily } from '@midscene/shared/env';\nimport type {\n  ChatCompletionCallContext,\n  ChatCompletionParamsResult,\n  ModelAdapterDefinition,\n} from '../model-adapter/types';\nimport { isLocateIntent } from './utils/intent';\n\nconst buildMimoChatCompletionParams = (\n  input: ChatCompletionCallContext,\n): ChatCompletionParamsResult => {\n  const { intent, midsceneDefaults, userConfig } = input;\n  const { reasoningEnabled } = userConfig;\n  const commonOverrideConfig: Record<string, unknown> = {};\n\n  // https://platform.xiaomimimo.com/docs/zh-CN/api/chat/openai-api\n  // Observed with thinking disabled: Mimo needs json_object to return JSON.\n  if (isLocateIntent(intent)) {\n    commonOverrideConfig.response_format = { type: 'json_object' };\n  }\n\n  if (userConfig.temperature !== undefined) {\n    commonOverrideConfig.temperature = userConfig.temperature;\n  }\n\n  const modelSpecificConfig: Record<string, unknown> = {\n    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 mimoAdapters = {\n  'xiaomi-mimo': {\n    chatCompletion: {\n      unsupportedUserConfig: ['reasoningEffort', 'reasoningBudget'],\n      buildChatCompletionParams: buildMimoChatCompletionParams,\n      useReasoningAsContentFallback: true,\n    },\n  },\n} satisfies Pick<Record<TModelFamily, ModelAdapterDefinition>, 'xiaomi-mimo'>;\n"],"names":["buildMimoChatCompletionParams","input","intent","midsceneDefaults","userConfig","reasoningEnabled","commonOverrideConfig","isLocateIntent","undefined","modelSpecificConfig","mimoAdapters"],"mappings":";AAQA,MAAMA,gCAAgC,CACpCC;IAEA,MAAM,EAAEC,MAAM,EAAEC,gBAAgB,EAAEC,UAAU,EAAE,GAAGH;IACjD,MAAM,EAAEI,gBAAgB,EAAE,GAAGD;IAC7B,MAAME,uBAAgD,CAAC;IAIvD,IAAIC,eAAeL,SACjBI,qBAAqB,eAAe,GAAG;QAAE,MAAM;IAAc;IAG/D,IAAIF,AAA2BI,WAA3BJ,WAAW,WAAW,EACxBE,qBAAqB,WAAW,GAAGF,WAAW,WAAW;IAG3D,MAAMK,sBAA+C;QACnD,UAAU;YACR,MAAOJ,oBAAoB,QAAS,YAAY;QAClD;IACF;IAEA,OAAO;QACL,QAAQ;YACN,GAAGF,gBAAgB;YACnB,GAAGG,oBAAoB;YACvB,GAAGG,mBAAmB;QACxB;IACF;AACF;AAEO,MAAMC,eAAe;IAC1B,eAAe;QACb,gBAAgB;YACd,uBAAuB;gBAAC;gBAAmB;aAAkB;YAC7D,2BAA2BV;YAC3B,+BAA+B;QACjC;IACF;AACF"}

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

@@ -1 +1 @@
-{"version":3,"file":"ai-model/models/qwen.mjs","sources":["../../../../src/ai-model/models/qwen.ts"],"sourcesContent":["import type { TModelFamily } from '@midscene/shared/env';\nimport type {\n  ChatCompletionCallContext,\n  ChatCompletionParamsResult,\n  ModelAdapterDefinition,\n} from '../model-adapter/types';\nimport {\n  type LocateResultValue,\n  type PixelBbox,\n  createLocateResultValue,\n  isBboxLocateResultValue,\n  unwrapCoordinateListLikeInput,\n} from '../shared/model-locate-result';\n\nconst defaultBboxSize = 20;\nconst qwen25BboxCoordinatesMeta = {\n  shape: 'bbox',\n  order: 'xy',\n} as const;\nconst qwen25PointCoordinatesMeta = {\n  shape: 'point',\n  order: 'xy',\n} as const;\nconst qwen3BboxCoordinatesMeta = {\n  shape: 'bbox',\n  order: 'xy',\n  normalizedBy: 1000,\n} as const;\n\nfunction topLeftPointToPixelBbox(x: number, y: number): PixelBbox {\n  return [\n    Math.round(x),\n    Math.round(y),\n    Math.round(x + defaultBboxSize),\n    Math.round(y + defaultBboxSize),\n  ];\n}\n\nfunction parseQwen25RawLocateValue(input: unknown): LocateResultValue {\n  const bbox = unwrapCoordinateListLikeInput(input as any) as number[];\n  if (bbox.length < 2) {\n    const msg = `invalid bbox data for qwen-vl mode: ${JSON.stringify(bbox)} `;\n    throw new Error(msg);\n  }\n\n  if (typeof bbox[2] === 'number' && typeof bbox[3] === 'number') {\n    return createLocateResultValue(qwen25BboxCoordinatesMeta, [\n      bbox[0],\n      bbox[1],\n      bbox[2],\n      bbox[3],\n    ]);\n  }\n\n  return createLocateResultValue(qwen25PointCoordinatesMeta, [\n    bbox[0],\n    bbox[1],\n  ]);\n}\n\nfunction normalizeQwen25ResultToPixelBbox(\n  result: LocateResultValue,\n): PixelBbox {\n  if (isBboxLocateResultValue(result)) {\n    const { coordinates } = result;\n    return [\n      Math.round(coordinates[0]),\n      Math.round(coordinates[1]),\n      Math.round(coordinates[2]),\n      Math.round(coordinates[3]),\n    ];\n  }\n\n  const { coordinates } = result;\n  return topLeftPointToPixelBbox(coordinates[0], coordinates[1]);\n}\n\nconst buildQwenChatCompletionParams = (\n  input: ChatCompletionCallContext,\n): ChatCompletionParamsResult => {\n  const { midsceneDefaults, userConfig } = input;\n  const { reasoningEnabled, reasoningBudget } = userConfig;\n\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.enable_thinking = reasoningEnabled ?? false;\n    if (reasoningBudget !== undefined) {\n      modelSpecificConfig.thinking_budget = reasoningBudget;\n    }\n  }\n\n  return {\n    config: {\n      ...midsceneDefaults,\n      ...commonOverrideConfig,\n      ...modelSpecificConfig,\n    },\n  };\n};\n\nconst buildQwen25ChatCompletionParams = (\n  input: ChatCompletionCallContext,\n): ChatCompletionParamsResult => {\n  const { midsceneDefaults, userConfig } = input;\n  const commonOverrideConfig: Record<string, unknown> = {};\n\n  if (userConfig.temperature !== undefined) {\n    commonOverrideConfig.temperature = userConfig.temperature;\n  }\n\n  return {\n    config: {\n      ...midsceneDefaults,\n      ...commonOverrideConfig,\n      vl_high_resolution_images: true,\n    },\n  };\n};\n\nconst qwen3Adapter: ModelAdapterDefinition = {\n  chatCompletion: {\n    unsupportedUserConfig: ['reasoningEffort'],\n    buildChatCompletionParams: buildQwenChatCompletionParams,\n    messageExtraction: {\n      kind: 'default',\n      // DashScope returns Qwen thinking as `reasoning_content`. vLLM's\n      // OpenAI-compatible server is migrating the recommended field to\n      // `reasoning`, so Qwen adapters accept both serving conventions:\n      // https://github.com/vllm-project/vllm/issues/27755\n      reasoningContentKeys: ['reasoning_content', 'reasoning'],\n    },\n    useReasoningAsContentFallback: true,\n  },\n  locate: {\n    resultAdapter: {\n      coordinates: qwen3BboxCoordinatesMeta,\n    },\n  },\n};\n\nexport const qwenAdapters = {\n  'qwen2.5-vl': {\n    chatCompletion: {\n      unsupportedUserConfig: [\n        'reasoningEnabled',\n        'reasoningEffort',\n        'reasoningBudget',\n      ],\n      buildChatCompletionParams: buildQwen25ChatCompletionParams,\n    },\n    imagePreprocess: {\n      padBlockSize: 28,\n    },\n    locate: {\n      resultAdapter: {\n        coordinates: qwen25BboxCoordinatesMeta,\n        parseRawLocateValue: parseQwen25RawLocateValue,\n        mapLocateResultToPixelBbox: normalizeQwen25ResultToPixelBbox,\n      },\n    },\n  },\n  'qwen3-vl': qwen3Adapter,\n  qwen3: qwen3Adapter,\n  'qwen3.5': qwen3Adapter,\n  'qwen3.6': qwen3Adapter,\n} satisfies Pick<\n  Record<TModelFamily, ModelAdapterDefinition>,\n  'qwen2.5-vl' | 'qwen3-vl' | 'qwen3' | 'qwen3.5' | 'qwen3.6'\n>;\n"],"names":["defaultBboxSize","qwen25BboxCoordinatesMeta","qwen25PointCoordinatesMeta","qwen3BboxCoordinatesMeta","topLeftPointToPixelBbox","x","y","Math","parseQwen25RawLocateValue","input","bbox","unwrapCoordinateListLikeInput","msg","JSON","Error","createLocateResultValue","normalizeQwen25ResultToPixelBbox","result","isBboxLocateResultValue","coordinates","buildQwenChatCompletionParams","midsceneDefaults","userConfig","reasoningEnabled","reasoningBudget","commonOverrideConfig","undefined","modelSpecificConfig","buildQwen25ChatCompletionParams","qwen3Adapter","qwenAdapters"],"mappings":";AAcA,MAAMA,kBAAkB;AACxB,MAAMC,4BAA4B;IAChC,OAAO;IACP,OAAO;AACT;AACA,MAAMC,6BAA6B;IACjC,OAAO;IACP,OAAO;AACT;AACA,MAAMC,2BAA2B;IAC/B,OAAO;IACP,OAAO;IACP,cAAc;AAChB;AAEA,SAASC,wBAAwBC,CAAS,EAAEC,CAAS;IACnD,OAAO;QACLC,KAAK,KAAK,CAACF;QACXE,KAAK,KAAK,CAACD;QACXC,KAAK,KAAK,CAACF,IAAIL;QACfO,KAAK,KAAK,CAACD,IAAIN;KAChB;AACH;AAEA,SAASQ,0BAA0BC,KAAc;IAC/C,MAAMC,OAAOC,8BAA8BF;IAC3C,IAAIC,KAAK,MAAM,GAAG,GAAG;QACnB,MAAME,MAAM,CAAC,oCAAoC,EAAEC,KAAK,SAAS,CAACH,MAAM,CAAC,CAAC;QAC1E,MAAM,IAAII,MAAMF;IAClB;IAEA,IAAI,AAAmB,YAAnB,OAAOF,IAAI,CAAC,EAAE,IAAiB,AAAmB,YAAnB,OAAOA,IAAI,CAAC,EAAE,EAC/C,OAAOK,wBAAwBd,2BAA2B;QACxDS,IAAI,CAAC,EAAE;QACPA,IAAI,CAAC,EAAE;QACPA,IAAI,CAAC,EAAE;QACPA,IAAI,CAAC,EAAE;KACR;IAGH,OAAOK,wBAAwBb,4BAA4B;QACzDQ,IAAI,CAAC,EAAE;QACPA,IAAI,CAAC,EAAE;KACR;AACH;AAEA,SAASM,iCACPC,MAAyB;IAEzB,IAAIC,wBAAwBD,SAAS;QACnC,MAAM,EAAEE,WAAW,EAAE,GAAGF;QACxB,OAAO;YACLV,KAAK,KAAK,CAACY,WAAW,CAAC,EAAE;YACzBZ,KAAK,KAAK,CAACY,WAAW,CAAC,EAAE;YACzBZ,KAAK,KAAK,CAACY,WAAW,CAAC,EAAE;YACzBZ,KAAK,KAAK,CAACY,WAAW,CAAC,EAAE;SAC1B;IACH;IAEA,MAAM,EAAEA,WAAW,EAAE,GAAGF;IACxB,OAAOb,wBAAwBe,WAAW,CAAC,EAAE,EAAEA,WAAW,CAAC,EAAE;AAC/D;AAEA,MAAMC,gCAAgC,CACpCX;IAEA,MAAM,EAAEY,gBAAgB,EAAEC,UAAU,EAAE,GAAGb;IACzC,MAAM,EAAEc,gBAAgB,EAAEC,eAAe,EAAE,GAAGF;IAE9C,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,eAAe,GAAGJ,oBAAoB;QAC1D,IAAIC,AAAoBE,WAApBF,iBACFG,oBAAoB,eAAe,GAAGH;IAE1C;IAEA,OAAO;QACL,QAAQ;YACN,GAAGH,gBAAgB;YACnB,GAAGI,oBAAoB;YACvB,GAAGE,mBAAmB;QACxB;IACF;AACF;AAEA,MAAMC,kCAAkC,CACtCnB;IAEA,MAAM,EAAEY,gBAAgB,EAAEC,UAAU,EAAE,GAAGb;IACzC,MAAMgB,uBAAgD,CAAC;IAEvD,IAAIH,AAA2BI,WAA3BJ,WAAW,WAAW,EACxBG,qBAAqB,WAAW,GAAGH,WAAW,WAAW;IAG3D,OAAO;QACL,QAAQ;YACN,GAAGD,gBAAgB;YACnB,GAAGI,oBAAoB;YACvB,2BAA2B;QAC7B;IACF;AACF;AAEA,MAAMI,eAAuC;IAC3C,gBAAgB;QACd,uBAAuB;YAAC;SAAkB;QAC1C,2BAA2BT;QAC3B,mBAAmB;YACjB,MAAM;YAKN,sBAAsB;gBAAC;gBAAqB;aAAY;QAC1D;QACA,+BAA+B;IACjC;IACA,QAAQ;QACN,eAAe;YACb,aAAajB;QACf;IACF;AACF;AAEO,MAAM2B,eAAe;IAC1B,cAAc;QACZ,gBAAgB;YACd,uBAAuB;gBACrB;gBACA;gBACA;aACD;YACD,2BAA2BF;QAC7B;QACA,iBAAiB;YACf,cAAc;QAChB;QACA,QAAQ;YACN,eAAe;gBACb,aAAa3B;gBACb,qBAAqBO;gBACrB,4BAA4BQ;YAC9B;QACF;IACF;IACA,YAAYa;IACZ,OAAOA;IACP,WAAWA;IACX,WAAWA;AACb"}
+{"version":3,"file":"ai-model/models/qwen.mjs","sources":["../../../../src/ai-model/models/qwen.ts"],"sourcesContent":["import type { TModelFamily } from '@midscene/shared/env';\nimport type {\n  ChatCompletionCallContext,\n  ChatCompletionParamsResult,\n  ModelAdapterDefinition,\n} from '../model-adapter/types';\nimport {\n  type LocateResultValue,\n  type PixelBbox,\n  createLocateResultValue,\n  isBboxLocateResultValue,\n  unwrapCoordinateListLikeInput,\n} from '../shared/model-locate-result';\n\nconst defaultBboxSize = 20;\nconst qwen25BboxCoordinatesMeta = {\n  shape: 'bbox',\n  order: 'xy',\n} as const;\nconst qwen25PointCoordinatesMeta = {\n  shape: 'point',\n  order: 'xy',\n} as const;\nconst qwen3BboxCoordinatesMeta = {\n  shape: 'bbox',\n  order: 'xy',\n  normalizedBy: 1000,\n} as const;\n\nfunction topLeftPointToPixelBbox(x: number, y: number): PixelBbox {\n  return [\n    Math.round(x),\n    Math.round(y),\n    Math.round(x + defaultBboxSize),\n    Math.round(y + defaultBboxSize),\n  ];\n}\n\nfunction parseQwen25RawLocateValue(input: unknown): LocateResultValue {\n  const bbox = unwrapCoordinateListLikeInput(input as any) as number[];\n  if (bbox.length < 2) {\n    const msg = `invalid bbox data for qwen-vl mode: ${JSON.stringify(bbox)} `;\n    throw new Error(msg);\n  }\n\n  if (typeof bbox[2] === 'number' && typeof bbox[3] === 'number') {\n    return createLocateResultValue(qwen25BboxCoordinatesMeta, [\n      bbox[0],\n      bbox[1],\n      bbox[2],\n      bbox[3],\n    ]);\n  }\n\n  return createLocateResultValue(qwen25PointCoordinatesMeta, [\n    bbox[0],\n    bbox[1],\n  ]);\n}\n\nfunction normalizeQwen25ResultToPixelBbox(\n  result: LocateResultValue,\n): PixelBbox {\n  if (isBboxLocateResultValue(result)) {\n    const { coordinates } = result;\n    return [\n      Math.round(coordinates[0]),\n      Math.round(coordinates[1]),\n      Math.round(coordinates[2]),\n      Math.round(coordinates[3]),\n    ];\n  }\n\n  const { coordinates } = result;\n  return topLeftPointToPixelBbox(coordinates[0], coordinates[1]);\n}\n\nconst buildQwenChatCompletionParams = (\n  input: ChatCompletionCallContext,\n): ChatCompletionParamsResult => {\n  const { midsceneDefaults, userConfig } = input;\n  const { reasoningEnabled, reasoningBudget } = userConfig;\n\n  const commonOverrideConfig: Record<string, unknown> = {};\n\n  if (userConfig.temperature !== undefined) {\n    commonOverrideConfig.temperature = userConfig.temperature;\n  }\n\n  // Alibaba Cloud Model Studio JSON mode:\n  // https://help.aliyun.com/zh/model-studio/json-mode\n  // Observed in qwen3.6 grounding runs: enabling this can make the model\n  // return only [\"bbox_2d\"] without coordinates.\n  // if (isLocateIntent(input.intent)) {\n  //   commonOverrideConfig.response_format = { type: 'json_object' };\n  // }\n\n  const modelSpecificConfig: Record<string, unknown> = {};\n\n  if (reasoningEnabled !== 'default') {\n    modelSpecificConfig.enable_thinking = reasoningEnabled ?? false;\n    if (reasoningBudget !== undefined) {\n      modelSpecificConfig.thinking_budget = reasoningBudget;\n    }\n  }\n\n  return {\n    config: {\n      ...midsceneDefaults,\n      ...commonOverrideConfig,\n      ...modelSpecificConfig,\n    },\n  };\n};\n\nconst buildQwen25ChatCompletionParams = (\n  input: ChatCompletionCallContext,\n): ChatCompletionParamsResult => {\n  const { midsceneDefaults, userConfig } = input;\n  const commonOverrideConfig: Record<string, unknown> = {};\n\n  if (userConfig.temperature !== undefined) {\n    commonOverrideConfig.temperature = userConfig.temperature;\n  }\n\n  return {\n    config: {\n      ...midsceneDefaults,\n      ...commonOverrideConfig,\n      vl_high_resolution_images: true,\n    },\n  };\n};\n\nconst qwen3Adapter: ModelAdapterDefinition = {\n  chatCompletion: {\n    unsupportedUserConfig: ['reasoningEffort'],\n    buildChatCompletionParams: buildQwenChatCompletionParams,\n    messageExtraction: {\n      kind: 'default',\n      // DashScope returns Qwen thinking as `reasoning_content`. vLLM's\n      // OpenAI-compatible server is migrating the recommended field to\n      // `reasoning`, so Qwen adapters accept both serving conventions:\n      // https://github.com/vllm-project/vllm/issues/27755\n      reasoningContentKeys: ['reasoning_content', 'reasoning'],\n    },\n    useReasoningAsContentFallback: true,\n  },\n  locate: {\n    resultAdapter: {\n      coordinates: qwen3BboxCoordinatesMeta,\n    },\n  },\n};\n\nexport const qwenAdapters = {\n  'qwen2.5-vl': {\n    chatCompletion: {\n      unsupportedUserConfig: [\n        'reasoningEnabled',\n        'reasoningEffort',\n        'reasoningBudget',\n      ],\n      buildChatCompletionParams: buildQwen25ChatCompletionParams,\n    },\n    imagePreprocess: {\n      padBlockSize: 28,\n    },\n    locate: {\n      resultAdapter: {\n        coordinates: qwen25BboxCoordinatesMeta,\n        parseRawLocateValue: parseQwen25RawLocateValue,\n        mapLocateResultToPixelBbox: normalizeQwen25ResultToPixelBbox,\n      },\n    },\n  },\n  'qwen3-vl': qwen3Adapter,\n  qwen3: qwen3Adapter,\n  'qwen3.5': qwen3Adapter,\n  'qwen3.6': qwen3Adapter,\n} satisfies Pick<\n  Record<TModelFamily, ModelAdapterDefinition>,\n  'qwen2.5-vl' | 'qwen3-vl' | 'qwen3' | 'qwen3.5' | 'qwen3.6'\n>;\n"],"names":["defaultBboxSize","qwen25BboxCoordinatesMeta","qwen25PointCoordinatesMeta","qwen3BboxCoordinatesMeta","topLeftPointToPixelBbox","x","y","Math","parseQwen25RawLocateValue","input","bbox","unwrapCoordinateListLikeInput","msg","JSON","Error","createLocateResultValue","normalizeQwen25ResultToPixelBbox","result","isBboxLocateResultValue","coordinates","buildQwenChatCompletionParams","midsceneDefaults","userConfig","reasoningEnabled","reasoningBudget","commonOverrideConfig","undefined","modelSpecificConfig","buildQwen25ChatCompletionParams","qwen3Adapter","qwenAdapters"],"mappings":";AAcA,MAAMA,kBAAkB;AACxB,MAAMC,4BAA4B;IAChC,OAAO;IACP,OAAO;AACT;AACA,MAAMC,6BAA6B;IACjC,OAAO;IACP,OAAO;AACT;AACA,MAAMC,2BAA2B;IAC/B,OAAO;IACP,OAAO;IACP,cAAc;AAChB;AAEA,SAASC,wBAAwBC,CAAS,EAAEC,CAAS;IACnD,OAAO;QACLC,KAAK,KAAK,CAACF;QACXE,KAAK,KAAK,CAACD;QACXC,KAAK,KAAK,CAACF,IAAIL;QACfO,KAAK,KAAK,CAACD,IAAIN;KAChB;AACH;AAEA,SAASQ,0BAA0BC,KAAc;IAC/C,MAAMC,OAAOC,8BAA8BF;IAC3C,IAAIC,KAAK,MAAM,GAAG,GAAG;QACnB,MAAME,MAAM,CAAC,oCAAoC,EAAEC,KAAK,SAAS,CAACH,MAAM,CAAC,CAAC;QAC1E,MAAM,IAAII,MAAMF;IAClB;IAEA,IAAI,AAAmB,YAAnB,OAAOF,IAAI,CAAC,EAAE,IAAiB,AAAmB,YAAnB,OAAOA,IAAI,CAAC,EAAE,EAC/C,OAAOK,wBAAwBd,2BAA2B;QACxDS,IAAI,CAAC,EAAE;QACPA,IAAI,CAAC,EAAE;QACPA,IAAI,CAAC,EAAE;QACPA,IAAI,CAAC,EAAE;KACR;IAGH,OAAOK,wBAAwBb,4BAA4B;QACzDQ,IAAI,CAAC,EAAE;QACPA,IAAI,CAAC,EAAE;KACR;AACH;AAEA,SAASM,iCACPC,MAAyB;IAEzB,IAAIC,wBAAwBD,SAAS;QACnC,MAAM,EAAEE,WAAW,EAAE,GAAGF;QACxB,OAAO;YACLV,KAAK,KAAK,CAACY,WAAW,CAAC,EAAE;YACzBZ,KAAK,KAAK,CAACY,WAAW,CAAC,EAAE;YACzBZ,KAAK,KAAK,CAACY,WAAW,CAAC,EAAE;YACzBZ,KAAK,KAAK,CAACY,WAAW,CAAC,EAAE;SAC1B;IACH;IAEA,MAAM,EAAEA,WAAW,EAAE,GAAGF;IACxB,OAAOb,wBAAwBe,WAAW,CAAC,EAAE,EAAEA,WAAW,CAAC,EAAE;AAC/D;AAEA,MAAMC,gCAAgC,CACpCX;IAEA,MAAM,EAAEY,gBAAgB,EAAEC,UAAU,EAAE,GAAGb;IACzC,MAAM,EAAEc,gBAAgB,EAAEC,eAAe,EAAE,GAAGF;IAE9C,MAAMG,uBAAgD,CAAC;IAEvD,IAAIH,AAA2BI,WAA3BJ,WAAW,WAAW,EACxBG,qBAAqB,WAAW,GAAGH,WAAW,WAAW;IAW3D,MAAMK,sBAA+C,CAAC;IAEtD,IAAIJ,AAAqB,cAArBA,kBAAgC;QAClCI,oBAAoB,eAAe,GAAGJ,oBAAoB;QAC1D,IAAIC,AAAoBE,WAApBF,iBACFG,oBAAoB,eAAe,GAAGH;IAE1C;IAEA,OAAO;QACL,QAAQ;YACN,GAAGH,gBAAgB;YACnB,GAAGI,oBAAoB;YACvB,GAAGE,mBAAmB;QACxB;IACF;AACF;AAEA,MAAMC,kCAAkC,CACtCnB;IAEA,MAAM,EAAEY,gBAAgB,EAAEC,UAAU,EAAE,GAAGb;IACzC,MAAMgB,uBAAgD,CAAC;IAEvD,IAAIH,AAA2BI,WAA3BJ,WAAW,WAAW,EACxBG,qBAAqB,WAAW,GAAGH,WAAW,WAAW;IAG3D,OAAO;QACL,QAAQ;YACN,GAAGD,gBAAgB;YACnB,GAAGI,oBAAoB;YACvB,2BAA2B;QAC7B;IACF;AACF;AAEA,MAAMI,eAAuC;IAC3C,gBAAgB;QACd,uBAAuB;YAAC;SAAkB;QAC1C,2BAA2BT;QAC3B,mBAAmB;YACjB,MAAM;YAKN,sBAAsB;gBAAC;gBAAqB;aAAY;QAC1D;QACA,+BAA+B;IACjC;IACA,QAAQ;QACN,eAAe;YACb,aAAajB;QACf;IACF;AACF;AAEO,MAAM2B,eAAe;IAC1B,cAAc;QACZ,gBAAgB;YACd,uBAAuB;gBACrB;gBACA;gBACA;aACD;YACD,2BAA2BF;QAC7B;QACA,iBAAiB;YACf,cAAc;QAChB;QACA,QAAQ;YACN,eAAe;gBACb,aAAa3B;gBACb,qBAAqBO;gBACrB,4BAA4BQ;YAC9B;QACF;IACF;IACA,YAAYa;IACZ,OAAOA;IACP,WAAWA;IACX,WAAWA;AACb"}

package/dist/es/ai-model/models/ui-tars/adapter.mjs

@@ -1,7 +1,6 @@
 import { UITarsModelVersion } from "@midscene/shared/env";
 import { assert } from "@midscene/shared/utils";
-import { jsonrepair } from "jsonrepair";
-import { extractJSONFromCodeBlock, safeParseJson } from "../../service-caller/json.mjs";
+import { parseModelResponseJson } from "../../service-caller/json.mjs";
 import { createLocateResultValue, unwrapCoordinateListLikeInput } from "../../shared/model-locate-result/index.mjs";
 import { createUiTarsPlanner } from "./planning.mjs";
 const defaultVlmUiTarsReplanningCycleLimit = 40;
@@ -15,44 +14,6 @@ const uiTarsPointCoordinatesMeta = {
     order: 'xy',
     normalizedBy: 1000
 };
-function normalizeJsonObject(obj, context = {}) {
-    if (null == obj) return obj;
-    if (Array.isArray(obj)) return obj.map((item)=>normalizeJsonObject(item, context));
-    if ('object' == typeof obj) {
-        const normalized = {};
-        for (const [key, value] of Object.entries(obj)){
-            const trimmedKey = key.trim();
-            const preserveStringValue = context.preserveStringValueKeys?.includes(trimmedKey) ?? false;
-            const normalizedValue = 'string' == typeof value ? preserveStringValue ? value : value.trim() : normalizeJsonObject(value, context);
-            normalized[trimmedKey] = normalizedValue;
-        }
-        return normalized;
-    }
-    return 'string' == typeof obj ? obj.trim() : obj;
-}
-function shouldRepairUiTarsLocateJson(source) {
-    return 'locate' === source || 'section-locator' === source || 'planning-action-param' === source;
-}
-function preprocessUiTarsLocateJson(input) {
-    if (input.includes('bbox')) while(/\d+\s+\d+/.test(input))input = input.replace(/(\d+)\s+(\d+)/g, '$1,$2');
-    return input;
-}
-const uiTarsJsonParser = (raw, context = {
-    source: 'generic-object'
-})=>{
-    const { source } = context;
-    try {
-        return safeParseJson(raw, context);
-    } catch (firstError) {
-        if (!shouldRepairUiTarsLocateJson(source)) throw firstError;
-        const jsonString = preprocessUiTarsLocateJson(extractJSONFromCodeBlock(raw));
-        try {
-            return normalizeJsonObject(JSON.parse(jsonrepair(jsonString)), context);
-        } catch (error) {
-            throw Error(`failed to parse LLM response into JSON. Error - ${String(error ?? firstError ?? 'unknown error')}. Response - \n ${raw}`);
-        }
-    }
-};
 function parseUiTarsRawLocateValue(input) {
     const bbox = unwrapCoordinateListLikeInput(input);
     if ('string' == typeof bbox) {
@@ -98,7 +59,7 @@ function parseUiTarsRawLocateValue(input) {
 }
 function createUiTarsAdapter(uiTarsModelVersion) {
     return {
-        jsonParser: uiTarsJsonParser,
+        jsonParser: parseModelResponseJson,
         chatCompletion: {
             unsupportedUserConfig: [
                 'reasoningEnabled',

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

@@ -1 +1 @@
-{"version":3,"file":"ai-model/models/ui-tars/adapter.mjs","sources":["../../../../../src/ai-model/models/ui-tars/adapter.ts"],"sourcesContent":["import { type TModelFamily, UITarsModelVersion } from '@midscene/shared/env';\nimport { assert } from '@midscene/shared/utils';\nimport { jsonrepair } from 'jsonrepair';\nimport type {\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  createLocateResultValue,\n  unwrapCoordinateListLikeInput,\n} from '../../shared/model-locate-result';\nimport { createUiTarsPlanner } from './planning';\n\nconst defaultVlmUiTarsReplanningCycleLimit = 40;\nconst uiTarsBboxCoordinatesMeta = {\n  shape: 'bbox',\n  order: 'xy',\n  normalizedBy: 1000,\n} as const;\nconst uiTarsPointCoordinatesMeta = {\n  shape: 'point',\n  order: 'xy',\n  normalizedBy: 1000,\n} as const;\n\nfunction normalizeJsonObject(\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) => normalizeJsonObject(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          : normalizeJsonObject(value, context);\n      normalized[trimmedKey] = normalizedValue;\n    }\n    return normalized;\n  }\n\n  return typeof obj === 'string' ? obj.trim() : obj;\n}\n\nfunction shouldRepairUiTarsLocateJson(source: JsonParserSource) {\n  return (\n    source === 'locate' ||\n    source === 'section-locator' ||\n    source === 'planning-action-param'\n  );\n}\n\nfunction preprocessUiTarsLocateJson(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 uiTarsJsonParser: 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 (!shouldRepairUiTarsLocateJson(source)) {\n      throw firstError;\n    }\n\n    const jsonString = preprocessUiTarsLocateJson(\n      extractJSONFromCodeBlock(raw),\n    );\n    try {\n      return normalizeJsonObject(JSON.parse(jsonrepair(jsonString)), context);\n    } catch (error) {\n      throw Error(\n        `failed to parse LLM response into JSON. Error - ${String(\n          error ?? firstError ?? 'unknown error',\n        )}. Response - \\n ${raw}`,\n      );\n    }\n  }\n};\n\n// UI-TARS has not received active updates for a long time, so this parser is\n// intentionally kept separate from Doubao even though the current logic is the\n// same. This avoids coupling UI-TARS behavior to future Doubao adapter changes.\nfunction parseUiTarsRawLocateValue(input: unknown): LocateResultValue {\n  const bbox = unwrapCoordinateListLikeInput(input as any);\n  if (typeof bbox === 'string') {\n    assert(\n      /^(\\d+)\\s(\\d+)\\s(\\d+)\\s(\\d+)$/.test(bbox.trim()),\n      `invalid bbox data string for ui-tars mode: ${bbox}`,\n    );\n    const splitted = bbox.split(' ');\n    if (splitted.length === 4) {\n      return createLocateResultValue(uiTarsBboxCoordinatesMeta, [\n        Number(splitted[0]),\n        Number(splitted[1]),\n        Number(splitted[2]),\n        Number(splitted[3]),\n      ]);\n    }\n    throw new Error(`invalid bbox data string for ui-tars 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 createLocateResultValue(uiTarsBboxCoordinatesMeta, [\n      bboxList[0],\n      bboxList[1],\n      bboxList[2],\n      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 createLocateResultValue(uiTarsPointCoordinatesMeta, [\n      bboxList[0],\n      bboxList[1],\n    ]);\n  }\n\n  if (bbox.length === 8) {\n    return createLocateResultValue(uiTarsBboxCoordinatesMeta, [\n      bboxList[0],\n      bboxList[1],\n      bboxList[4],\n      bboxList[5],\n    ]);\n  }\n\n  const msg = `invalid bbox data for ui-tars mode: ${JSON.stringify(bbox)} `;\n  throw new Error(msg);\n}\n\nfunction createUiTarsAdapter(\n  uiTarsModelVersion: UITarsModelVersion,\n): ModelAdapterDefinition {\n  return {\n    jsonParser: uiTarsJsonParser,\n    chatCompletion: {\n      unsupportedUserConfig: [\n        'reasoningEnabled',\n        'reasoningEffort',\n        'reasoningBudget',\n      ],\n      buildChatCompletionParams: ({ midsceneDefaults, userConfig }) => {\n        const commonOverrideConfig: Record<string, unknown> = {};\n\n        if (userConfig.temperature !== undefined) {\n          commonOverrideConfig.temperature = userConfig.temperature;\n        }\n\n        return {\n          config: {\n            ...midsceneDefaults,\n            ...commonOverrideConfig,\n          },\n        };\n      },\n    },\n    planning: {\n      kind: 'custom',\n      cacheEnabled: false,\n      defaultReplanningCycleLimit: defaultVlmUiTarsReplanningCycleLimit,\n      planner: createUiTarsPlanner(uiTarsModelVersion),\n    },\n    locate: {\n      resultAdapter: {\n        coordinates: uiTarsBboxCoordinatesMeta,\n        parseRawLocateValue: parseUiTarsRawLocateValue,\n      },\n    },\n  };\n}\n\nconst uiTarsDoubao15Adapter = createUiTarsAdapter(\n  UITarsModelVersion.DOUBAO_1_5_20B,\n);\n\nexport const uiTarsAdapters = {\n  'vlm-ui-tars': createUiTarsAdapter(UITarsModelVersion.V1_0),\n  'vlm-ui-tars-doubao': uiTarsDoubao15Adapter,\n  'vlm-ui-tars-doubao-1.5': uiTarsDoubao15Adapter,\n} satisfies Pick<\n  Record<TModelFamily, ModelAdapterDefinition>,\n  'vlm-ui-tars' | 'vlm-ui-tars-doubao' | 'vlm-ui-tars-doubao-1.5'\n>;\n"],"names":["defaultVlmUiTarsReplanningCycleLimit","uiTarsBboxCoordinatesMeta","uiTarsPointCoordinatesMeta","normalizeJsonObject","obj","context","Array","item","normalized","key","value","Object","trimmedKey","preserveStringValue","normalizedValue","shouldRepairUiTarsLocateJson","source","preprocessUiTarsLocateJson","input","uiTarsJsonParser","raw","safeParseJson","firstError","jsonString","extractJSONFromCodeBlock","JSON","jsonrepair","error","Error","String","parseUiTarsRawLocateValue","bbox","unwrapCoordinateListLikeInput","assert","splitted","createLocateResultValue","Number","bboxList","x","y","msg","createUiTarsAdapter","uiTarsModelVersion","midsceneDefaults","userConfig","commonOverrideConfig","undefined","createUiTarsPlanner","uiTarsDoubao15Adapter","UITarsModelVersion","uiTarsAdapters"],"mappings":";;;;;;AAmBA,MAAMA,uCAAuC;AAC7C,MAAMC,4BAA4B;IAChC,OAAO;IACP,OAAO;IACP,cAAc;AAChB;AACA,MAAMC,6BAA6B;IACjC,OAAO;IACP,OAAO;IACP,cAAc;AAChB;AAEA,SAASC,oBACPC,GAAQ,EACRC,UAA8D,CAAC,CAAC;IAEhE,IAAID,QAAAA,KACF,OAAOA;IAGT,IAAIE,MAAM,OAAO,CAACF,MAChB,OAAOA,IAAI,GAAG,CAAC,CAACG,OAASJ,oBAAoBI,MAAMF;IAGrD,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,oBAAoBO,OAAOL;YACjCG,UAAU,CAACI,WAAW,GAAGE;QAC3B;QACA,OAAON;IACT;IAEA,OAAO,AAAe,YAAf,OAAOJ,MAAmBA,IAAI,IAAI,KAAKA;AAChD;AAEA,SAASW,6BAA6BC,MAAwB;IAC5D,OACEA,AAAW,aAAXA,UACAA,AAAW,sBAAXA,UACAA,AAAW,4BAAXA;AAEJ;AAEA,SAASC,2BAA2BC,KAAa;IAC/C,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,oBAAoBsB,KAAK,KAAK,CAACC,WAAWH,cAAclB;QACjE,EAAE,OAAOsB,OAAO;YACd,MAAMC,MACJ,CAAC,gDAAgD,EAAEC,OACjDF,SAASL,cAAc,iBACvB,gBAAgB,EAAEF,KAAK;QAE7B;IACF;AACF;AAKA,SAASU,0BAA0BZ,KAAc;IAC/C,MAAMa,OAAOC,8BAA8Bd;IAC3C,IAAI,AAAgB,YAAhB,OAAOa,MAAmB;QAC5BE,OACE,+BAA+B,IAAI,CAACF,KAAK,IAAI,KAC7C,CAAC,2CAA2C,EAAEA,MAAM;QAEtD,MAAMG,WAAWH,KAAK,KAAK,CAAC;QAC5B,IAAIG,AAAoB,MAApBA,SAAS,MAAM,EACjB,OAAOC,wBAAwBlC,2BAA2B;YACxDmC,OAAOF,QAAQ,CAAC,EAAE;YAClBE,OAAOF,QAAQ,CAAC,EAAE;YAClBE,OAAOF,QAAQ,CAAC,EAAE;YAClBE,OAAOF,QAAQ,CAAC,EAAE;SACnB;QAEH,MAAM,IAAIN,MAAM,CAAC,2CAA2C,EAAEG,MAAM;IACtE;IAEA,IAAIM,WAAqB,EAAE;IAC3B,IAAI/B,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,CAAC+B,GAAGC,EAAE,GAAGhC,KAAK,KAAK,CAAC;YAC1B8B,SAAS,IAAI,CAACD,OAAOE,EAAE,IAAI,KAAKF,OAAOG,EAAE,IAAI;QAC/C,OAAO,IAAI,AAAgB,YAAhB,OAAOhC,QAAqBA,KAAK,QAAQ,CAAC,MAAM;YACzD,MAAM,CAAC+B,GAAGC,EAAE,GAAGhC,KAAK,KAAK,CAAC;YAC1B8B,SAAS,IAAI,CAACD,OAAOE,EAAE,IAAI,KAAKF,OAAOG,EAAE,IAAI;QAC/C,OACEF,SAAS,IAAI,CAACD,OAAO7B;IAEzB;SAEA8B,WAAWN;IAGb,IAAIM,AAAoB,MAApBA,SAAS,MAAM,IAAUA,AAAoB,MAApBA,SAAS,MAAM,EAC1C,OAAOF,wBAAwBlC,2BAA2B;QACxDoC,QAAQ,CAAC,EAAE;QACXA,QAAQ,CAAC,EAAE;QACXA,QAAQ,CAAC,EAAE;QACXA,QAAQ,CAAC,EAAE;KACZ;IAGH,IACEA,AAAoB,MAApBA,SAAS,MAAM,IACfA,AAAoB,MAApBA,SAAS,MAAM,IACfA,AAAoB,MAApBA,SAAS,MAAM,IACfA,AAAoB,MAApBA,SAAS,MAAM,EAEf,OAAOF,wBAAwBjC,4BAA4B;QACzDmC,QAAQ,CAAC,EAAE;QACXA,QAAQ,CAAC,EAAE;KACZ;IAGH,IAAIN,AAAgB,MAAhBA,KAAK,MAAM,EACb,OAAOI,wBAAwBlC,2BAA2B;QACxDoC,QAAQ,CAAC,EAAE;QACXA,QAAQ,CAAC,EAAE;QACXA,QAAQ,CAAC,EAAE;QACXA,QAAQ,CAAC,EAAE;KACZ;IAGH,MAAMG,MAAM,CAAC,oCAAoC,EAAEf,KAAK,SAAS,CAACM,MAAM,CAAC,CAAC;IAC1E,MAAM,IAAIH,MAAMY;AAClB;AAEA,SAASC,oBACPC,kBAAsC;IAEtC,OAAO;QACL,YAAYvB;QACZ,gBAAgB;YACd,uBAAuB;gBACrB;gBACA;gBACA;aACD;YACD,2BAA2B,CAAC,EAAEwB,gBAAgB,EAAEC,UAAU,EAAE;gBAC1D,MAAMC,uBAAgD,CAAC;gBAEvD,IAAID,AAA2BE,WAA3BF,WAAW,WAAW,EACxBC,qBAAqB,WAAW,GAAGD,WAAW,WAAW;gBAG3D,OAAO;oBACL,QAAQ;wBACN,GAAGD,gBAAgB;wBACnB,GAAGE,oBAAoB;oBACzB;gBACF;YACF;QACF;QACA,UAAU;YACR,MAAM;YACN,cAAc;YACd,6BAA6B7C;YAC7B,SAAS+C,oBAAoBL;QAC/B;QACA,QAAQ;YACN,eAAe;gBACb,aAAazC;gBACb,qBAAqB6B;YACvB;QACF;IACF;AACF;AAEA,MAAMkB,wBAAwBP,oBAC5BQ,mBAAmB,cAAc;AAG5B,MAAMC,iBAAiB;IAC5B,eAAeT,oBAAoBQ,mBAAmB,IAAI;IAC1D,sBAAsBD;IACtB,0BAA0BA;AAC5B"}
+{"version":3,"file":"ai-model/models/ui-tars/adapter.mjs","sources":["../../../../../src/ai-model/models/ui-tars/adapter.ts"],"sourcesContent":["import { type TModelFamily, UITarsModelVersion } from '@midscene/shared/env';\nimport { assert } from '@midscene/shared/utils';\nimport type { ModelAdapterDefinition } from '../../model-adapter/types';\nimport { parseModelResponseJson } from '../../service-caller/json';\nimport {\n  type LocateResultValue,\n  createLocateResultValue,\n  unwrapCoordinateListLikeInput,\n} from '../../shared/model-locate-result';\nimport { createUiTarsPlanner } from './planning';\n\nconst defaultVlmUiTarsReplanningCycleLimit = 40;\nconst uiTarsBboxCoordinatesMeta = {\n  shape: 'bbox',\n  order: 'xy',\n  normalizedBy: 1000,\n} as const;\nconst uiTarsPointCoordinatesMeta = {\n  shape: 'point',\n  order: 'xy',\n  normalizedBy: 1000,\n} as const;\n\n// UI-TARS has not received active updates for a long time, so this parser is\n// intentionally kept separate from Doubao even though the current logic is the\n// same. This avoids coupling UI-TARS behavior to future Doubao adapter changes.\nfunction parseUiTarsRawLocateValue(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 ui-tars mode: ${bbox}`,\n    );\n    const splitted = bbox.split(' ');\n    if (splitted.length === 4) {\n      return createLocateResultValue(uiTarsBboxCoordinatesMeta, [\n        Number(splitted[0]),\n        Number(splitted[1]),\n        Number(splitted[2]),\n        Number(splitted[3]),\n      ]);\n    }\n    throw new Error(`invalid bbox data string for ui-tars 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 createLocateResultValue(uiTarsBboxCoordinatesMeta, [\n      bboxList[0],\n      bboxList[1],\n      bboxList[2],\n      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 createLocateResultValue(uiTarsPointCoordinatesMeta, [\n      bboxList[0],\n      bboxList[1],\n    ]);\n  }\n\n  if (bbox.length === 8) {\n    return createLocateResultValue(uiTarsBboxCoordinatesMeta, [\n      bboxList[0],\n      bboxList[1],\n      bboxList[4],\n      bboxList[5],\n    ]);\n  }\n\n  const msg = `invalid bbox data for ui-tars mode: ${JSON.stringify(bbox)} `;\n  throw new Error(msg);\n}\n\nfunction createUiTarsAdapter(\n  uiTarsModelVersion: UITarsModelVersion,\n): ModelAdapterDefinition {\n  return {\n    jsonParser: parseModelResponseJson,\n    chatCompletion: {\n      unsupportedUserConfig: [\n        'reasoningEnabled',\n        'reasoningEffort',\n        'reasoningBudget',\n      ],\n      buildChatCompletionParams: ({ midsceneDefaults, userConfig }) => {\n        const commonOverrideConfig: Record<string, unknown> = {};\n\n        if (userConfig.temperature !== undefined) {\n          commonOverrideConfig.temperature = userConfig.temperature;\n        }\n\n        return {\n          config: {\n            ...midsceneDefaults,\n            ...commonOverrideConfig,\n          },\n        };\n      },\n    },\n    planning: {\n      kind: 'custom',\n      cacheEnabled: false,\n      defaultReplanningCycleLimit: defaultVlmUiTarsReplanningCycleLimit,\n      planner: createUiTarsPlanner(uiTarsModelVersion),\n    },\n    locate: {\n      resultAdapter: {\n        coordinates: uiTarsBboxCoordinatesMeta,\n        parseRawLocateValue: parseUiTarsRawLocateValue,\n      },\n    },\n  };\n}\n\nconst uiTarsDoubao15Adapter = createUiTarsAdapter(\n  UITarsModelVersion.DOUBAO_1_5_20B,\n);\n\nexport const uiTarsAdapters = {\n  'vlm-ui-tars': createUiTarsAdapter(UITarsModelVersion.V1_0),\n  'vlm-ui-tars-doubao': uiTarsDoubao15Adapter,\n  'vlm-ui-tars-doubao-1.5': uiTarsDoubao15Adapter,\n} satisfies Pick<\n  Record<TModelFamily, ModelAdapterDefinition>,\n  'vlm-ui-tars' | 'vlm-ui-tars-doubao' | 'vlm-ui-tars-doubao-1.5'\n>;\n"],"names":["defaultVlmUiTarsReplanningCycleLimit","uiTarsBboxCoordinatesMeta","uiTarsPointCoordinatesMeta","parseUiTarsRawLocateValue","input","bbox","unwrapCoordinateListLikeInput","assert","splitted","createLocateResultValue","Number","Error","bboxList","Array","item","x","y","msg","JSON","createUiTarsAdapter","uiTarsModelVersion","parseModelResponseJson","midsceneDefaults","userConfig","commonOverrideConfig","undefined","createUiTarsPlanner","uiTarsDoubao15Adapter","UITarsModelVersion","uiTarsAdapters"],"mappings":";;;;;AAWA,MAAMA,uCAAuC;AAC7C,MAAMC,4BAA4B;IAChC,OAAO;IACP,OAAO;IACP,cAAc;AAChB;AACA,MAAMC,6BAA6B;IACjC,OAAO;IACP,OAAO;IACP,cAAc;AAChB;AAKA,SAASC,0BAA0BC,KAAc;IAC/C,MAAMC,OAAOC,8BAA8BF;IAC3C,IAAI,AAAgB,YAAhB,OAAOC,MAAmB;QAC5BE,OACE,+BAA+B,IAAI,CAACF,KAAK,IAAI,KAC7C,CAAC,2CAA2C,EAAEA,MAAM;QAEtD,MAAMG,WAAWH,KAAK,KAAK,CAAC;QAC5B,IAAIG,AAAoB,MAApBA,SAAS,MAAM,EACjB,OAAOC,wBAAwBR,2BAA2B;YACxDS,OAAOF,QAAQ,CAAC,EAAE;YAClBE,OAAOF,QAAQ,CAAC,EAAE;YAClBE,OAAOF,QAAQ,CAAC,EAAE;YAClBE,OAAOF,QAAQ,CAAC,EAAE;SACnB;QAEH,MAAM,IAAIG,MAAM,CAAC,2CAA2C,EAAEN,MAAM;IACtE;IAEA,IAAIO,WAAqB,EAAE;IAC3B,IAAIC,MAAM,OAAO,CAACR,SAAS,AAAmB,YAAnB,OAAOA,IAAI,CAAC,EAAE,EACvCA,KAAK,OAAO,CAAC,CAACS;QACZ,IAAI,AAAgB,YAAhB,OAAOA,QAAqBA,KAAK,QAAQ,CAAC,MAAM;YAClD,MAAM,CAACC,GAAGC,EAAE,GAAGF,KAAK,KAAK,CAAC;YAC1BF,SAAS,IAAI,CAACF,OAAOK,EAAE,IAAI,KAAKL,OAAOM,EAAE,IAAI;QAC/C,OAAO,IAAI,AAAgB,YAAhB,OAAOF,QAAqBA,KAAK,QAAQ,CAAC,MAAM;YACzD,MAAM,CAACC,GAAGC,EAAE,GAAGF,KAAK,KAAK,CAAC;YAC1BF,SAAS,IAAI,CAACF,OAAOK,EAAE,IAAI,KAAKL,OAAOM,EAAE,IAAI;QAC/C,OACEJ,SAAS,IAAI,CAACF,OAAOI;IAEzB;SAEAF,WAAWP;IAGb,IAAIO,AAAoB,MAApBA,SAAS,MAAM,IAAUA,AAAoB,MAApBA,SAAS,MAAM,EAC1C,OAAOH,wBAAwBR,2BAA2B;QACxDW,QAAQ,CAAC,EAAE;QACXA,QAAQ,CAAC,EAAE;QACXA,QAAQ,CAAC,EAAE;QACXA,QAAQ,CAAC,EAAE;KACZ;IAGH,IACEA,AAAoB,MAApBA,SAAS,MAAM,IACfA,AAAoB,MAApBA,SAAS,MAAM,IACfA,AAAoB,MAApBA,SAAS,MAAM,IACfA,AAAoB,MAApBA,SAAS,MAAM,EAEf,OAAOH,wBAAwBP,4BAA4B;QACzDU,QAAQ,CAAC,EAAE;QACXA,QAAQ,CAAC,EAAE;KACZ;IAGH,IAAIP,AAAgB,MAAhBA,KAAK,MAAM,EACb,OAAOI,wBAAwBR,2BAA2B;QACxDW,QAAQ,CAAC,EAAE;QACXA,QAAQ,CAAC,EAAE;QACXA,QAAQ,CAAC,EAAE;QACXA,QAAQ,CAAC,EAAE;KACZ;IAGH,MAAMK,MAAM,CAAC,oCAAoC,EAAEC,KAAK,SAAS,CAACb,MAAM,CAAC,CAAC;IAC1E,MAAM,IAAIM,MAAMM;AAClB;AAEA,SAASE,oBACPC,kBAAsC;IAEtC,OAAO;QACL,YAAYC;QACZ,gBAAgB;YACd,uBAAuB;gBACrB;gBACA;gBACA;aACD;YACD,2BAA2B,CAAC,EAAEC,gBAAgB,EAAEC,UAAU,EAAE;gBAC1D,MAAMC,uBAAgD,CAAC;gBAEvD,IAAID,AAA2BE,WAA3BF,WAAW,WAAW,EACxBC,qBAAqB,WAAW,GAAGD,WAAW,WAAW;gBAG3D,OAAO;oBACL,QAAQ;wBACN,GAAGD,gBAAgB;wBACnB,GAAGE,oBAAoB;oBACzB;gBACF;YACF;QACF;QACA,UAAU;YACR,MAAM;YACN,cAAc;YACd,6BAA6BxB;YAC7B,SAAS0B,oBAAoBN;QAC/B;QACA,QAAQ;YACN,eAAe;gBACb,aAAanB;gBACb,qBAAqBE;YACvB;QACF;IACF;AACF;AAEA,MAAMwB,wBAAwBR,oBAC5BS,mBAAmB,cAAc;AAG5B,MAAMC,iBAAiB;IAC5B,eAAeV,oBAAoBS,mBAAmB,IAAI;IAC1D,sBAAsBD;IACtB,0BAA0BA;AAC5B"}

package/dist/es/ai-model/models/utils/intent.mjs

@@ -0,0 +1,6 @@
+function isLocateIntent(intent) {
+    return 'default' === intent;
+}
+export { isLocateIntent };
+
+//# sourceMappingURL=intent.mjs.map

package/dist/es/ai-model/models/utils/intent.mjs.map

@@ -0,0 +1 @@
+{"version":3,"file":"ai-model/models/utils/intent.mjs","sources":["../../../../../src/ai-model/models/utils/intent.ts"],"sourcesContent":["import type { TIntent } from '@midscene/shared/env';\n\nexport function isLocateIntent(intent?: TIntent): boolean {\n  return intent === 'default';\n}\n"],"names":["isLocateIntent","intent"],"mappings":"AAEO,SAASA,eAAeC,MAAgB;IAC7C,OAAOA,AAAW,cAAXA;AACT"}