@aiscene/shared 8.0.3 → 8.0.5

package/dist/es/cli/verbose.mjs

@@ -0,0 +1,404 @@
+import { aiActCliProgressRenderer, cliAiActProgressScope } from "./verbose-ai-act.mjs";
+import { collectScreenshotRefs, renderScreenshotList } from "./verbose-screenshot.mjs";
+const cliVerboseFlag = 'verbose';
+const progressRenderers = {
+    [cliAiActProgressScope]: aiActCliProgressRenderer
+};
+function progressRendererForEvent(event) {
+    if ('object' != typeof event || null === event) return;
+    const scope = event.scope;
+    if ('string' != typeof scope) return;
+    const renderer = progressRenderers[scope];
+    return renderer ? {
+        scope,
+        renderer
+    } : void 0;
+}
+const progressEventName = 'midscene_progress';
+const cliVerboseContextKey = '__midscene_cli_verbose_context__';
+const getGlobalContext = ()=>{
+    const globalWithContext = globalThis;
+    if (!globalWithContext[cliVerboseContextKey]) globalWithContext[cliVerboseContextKey] = {
+        current: {
+            enabled: false
+        }
+    };
+    return globalWithContext[cliVerboseContextKey];
+};
+function parseVerboseFormat(rawFormat) {
+    if ('text' === rawFormat || 'jsonl' === rawFormat) return rawFormat;
+    throw new Error(`Unsupported --${cliVerboseFlag} format "${rawFormat}". Use "--${cliVerboseFlag}" or "--${cliVerboseFlag}=jsonl".`);
+}
+function stripVerboseFlag(argv) {
+    const rawArgs = [];
+    let verbose = false;
+    let format = 'text';
+    for (const arg of argv){
+        if (arg === `--${cliVerboseFlag}`) {
+            verbose = true;
+            continue;
+        }
+        if (arg.startsWith(`--${cliVerboseFlag}=`)) {
+            verbose = true;
+            format = parseVerboseFormat(arg.slice(`--${cliVerboseFlag}=`.length));
+            continue;
+        }
+        rawArgs.push(arg);
+    }
+    return {
+        rawArgs,
+        verbose,
+        format
+    };
+}
+async function withCliVerboseContext(context, fn) {
+    const globalContext = getGlobalContext();
+    const previous = globalContext.current;
+    globalContext.current = {
+        ...context,
+        renderedLineKeys: context.renderedLineKeys ?? new Set()
+    };
+    try {
+        return await fn();
+    } finally{
+        globalContext.current = previous;
+    }
+}
+function getCliVerboseContext() {
+    return getGlobalContext().current;
+}
+function isCliVerboseEnabled() {
+    return getCliVerboseContext().enabled;
+}
+function emitCliVerboseEvent(event) {
+    const context = getCliVerboseContext();
+    if (!context.enabled) return;
+    const payload = {
+        type: progressEventName,
+        timestamp: new Date().toISOString(),
+        scriptName: context.scriptName,
+        command: context.commandName,
+        ...event
+    };
+    if ('jsonl' === context.format) return void console.log(JSON.stringify(payload));
+    const text = renderCliVerboseEventText(payload, context);
+    if (text) console.log(text);
+}
+function errorMessageOf(error) {
+    return error instanceof Error ? error.message : String(error);
+}
+function isRecord(value) {
+    return 'object' == typeof value && null !== value && !Array.isArray(value);
+}
+function compactText(value) {
+    if (null == value || '' === value) return '';
+    if ('string' == typeof value) return value.length > 180 ? `${value.slice(0, 177)}...` : value;
+    if ('number' == typeof value || 'boolean' == typeof value) return String(value);
+    try {
+        const json = JSON.stringify(value);
+        return json.length > 180 ? `${json.slice(0, 177)}...` : json;
+    } catch  {
+        return String(value);
+    }
+}
+function compactPrimitiveCliVerboseValue(value) {
+    if (null == value) return {
+        handled: true,
+        value
+    };
+    if ('string' == typeof value) return {
+        handled: true,
+        value: value.length > 180 ? `${value.slice(0, 177)}...` : value
+    };
+    if ('number' == typeof value || 'boolean' == typeof value) return {
+        handled: true,
+        value
+    };
+    if ('object' != typeof value) return {
+        handled: true,
+        value: String(value)
+    };
+    return {
+        handled: false
+    };
+}
+function compactStructuredCliVerboseValue(value) {
+    const primitive = compactPrimitiveCliVerboseValue(value);
+    if (primitive.handled) return primitive.value;
+    if (Array.isArray(value)) return value.slice(0, 5).map(compactStructuredCliVerboseValue);
+    if (!isRecord(value)) return String(value);
+    const entries = Object.entries(value).slice(0, 6);
+    return Object.fromEntries(entries.map(([key, entryValue])=>[
+            key,
+            compactStructuredCliVerboseValue(entryValue)
+        ]));
+}
+function compactCliVerboseValue(value) {
+    const primitive = compactPrimitiveCliVerboseValue(value);
+    if (primitive.handled) return primitive.value;
+    if (Array.isArray(value)) return value.slice(0, 5).map(compactCliVerboseValue);
+    if (isRecord(value) && 'string' == typeof value.prompt) return compactCliVerboseValue(value.prompt);
+    if (!isRecord(value)) return String(value);
+    const entries = Object.entries(value).slice(0, 6);
+    return Object.fromEntries(entries.map(([key, entryValue])=>[
+            key,
+            compactCliVerboseValue(entryValue)
+        ]));
+}
+function compactCliVerboseArgs(args) {
+    return Object.fromEntries(Object.entries(args).slice(0, 8).map(([key, value])=>[
+            key,
+            compactStructuredCliVerboseValue(value)
+        ]));
+}
+function summarizeParam(param) {
+    if (!param || 'object' != typeof param) return compactCliVerboseValue(param);
+    const record = param;
+    if ('string' == typeof record.prompt) return compactCliVerboseValue(record.prompt);
+    if (record.locate && 'object' == typeof record.locate && null !== record.locate) {
+        const locate = record.locate;
+        return {
+            locate: compactCliVerboseValue(locate.prompt),
+            ...Object.fromEntries(Object.entries(record).filter(([key])=>'locate' !== key).slice(0, 4).map(([key, value])=>[
+                    key,
+                    compactCliVerboseValue(value)
+                ]))
+        };
+    }
+    return compactCliVerboseValue(record);
+}
+function summarizeUserInstruction(value) {
+    if ('string' == typeof value) return compactText(value);
+    if (isRecord(value) && 'string' == typeof value.prompt) return compactText(value.prompt);
+    return compactText(value);
+}
+function summarizeTaskParamText(task) {
+    if (!isRecord(task.param)) return compactText(summarizeParam(task.param));
+    if ('Planning' === task.type && 'Locate' !== task.subType) {
+        if (task.param.userInstructionDisplay) return summarizeUserInstruction(task.param.userInstructionDisplay);
+        if (task.param.userInstruction) return summarizeUserInstruction(task.param.userInstruction);
+    }
+    if (task.param.dataDemand) return compactText(summarizeParam(task.param.dataDemand));
+    if (task.param.assertion) return compactText(summarizeParam(task.param.assertion));
+    return compactText(summarizeParam(task.param));
+}
+function summarizeSubGoals(value) {
+    if (!Array.isArray(value) || 0 === value.length) return '';
+    const goals = value.slice(0, 6).map((goal)=>{
+        if (!isRecord(goal)) return '';
+        const index = 'number' == typeof goal.index ? `${goal.index}. ` : '';
+        const status = 'string' == typeof goal.status ? `[${goal.status}] ` : '';
+        const description = 'string' == typeof goal.description ? goal.description : '';
+        return `${index}${status}${description}`.trim();
+    }).filter(Boolean);
+    return goals.length > 0 ? `sub-goals: ${goals.join('; ')}` : '';
+}
+function summarizeActions(value) {
+    if (!Array.isArray(value) || 0 === value.length) return '';
+    const actions = value.slice(0, 5).map((action)=>{
+        if (!isRecord(action)) return '';
+        const type = 'string' == typeof action.type ? action.type : 'Action';
+        const param = compactText(summarizeParam(action.param));
+        return param ? `${type}: ${param}` : type;
+    }).filter(Boolean);
+    return actions.length > 0 ? `actions: ${actions.join('; ')}` : '';
+}
+function summarizeTaskOutputText(task) {
+    if (!isRecord(task.output)) return compactText(task.output);
+    return summarizeSubGoals(task.output.updateSubGoals) || compactText(task.output.log) || compactText(task.output.output) || summarizeActions(task.output.actions);
+}
+function summarizeTaskText(task) {
+    const parts = [
+        task.type,
+        task.subType
+    ].filter((part)=>'string' == typeof part && part.length > 0);
+    const label = parts.length > 0 ? parts.join('/') : 'Task';
+    const status = 'string' == typeof task.status && task.status.length > 0 ? task.status : 'unknown';
+    const param = summarizeTaskParamText(task);
+    const output = summarizeTaskOutputText(task);
+    const thought = compactText(task.thought);
+    const detail = output || thought || param;
+    return detail ? `${label} ${status}: ${detail}` : `${label} ${status}`;
+}
+function isCliVerboseExecutionDumpLike(value) {
+    return isRecord(value) && Array.isArray(value.tasks);
+}
+function summarizeDumpUpdate(executionDump, screenshotOptions = {}) {
+    if (!isCliVerboseExecutionDumpLike(executionDump)) return;
+    const tasks = executionDump.tasks;
+    const latestTask = tasks[tasks.length - 1];
+    const steps = tasks.map((task, index)=>({
+            id: task.taskId,
+            index: index + 1,
+            total: tasks.length,
+            type: task.type,
+            subType: task.subType,
+            status: task.status,
+            param: summarizeParam(task.param),
+            message: summarizeTaskText(task),
+            error: task.errorMessage,
+            durationMs: task.timing?.cost,
+            screenshots: collectScreenshotRefs(task, screenshotOptions).slice(-3)
+        }));
+    return {
+        execution: {
+            id: executionDump.id,
+            name: executionDump.name,
+            description: compactCliVerboseValue(executionDump.description),
+            taskCount: tasks.length
+        },
+        steps,
+        task: latestTask ? {
+            id: latestTask.taskId,
+            index: tasks.length,
+            total: tasks.length,
+            type: latestTask.type,
+            subType: latestTask.subType,
+            status: latestTask.status,
+            param: summarizeParam(latestTask.param),
+            message: summarizeTaskText(latestTask),
+            error: latestTask.errorMessage,
+            durationMs: latestTask.timing?.cost
+        } : void 0,
+        screenshots: collectScreenshotRefs(tasks, screenshotOptions).slice(-5)
+    };
+}
+function stringifyArgs(args) {
+    if (!isRecord(args) || 0 === Object.keys(args).length) return '';
+    return Object.entries(args).map(([key, value])=>`${key}=${compactText(value)}`).join(', ');
+}
+function isActVerboseEvent(command, tool) {
+    return 'act' === command || 'act' === tool;
+}
+function renderLinesOnce(context, lines) {
+    if (0 === lines.length) return;
+    if (!context.renderedLineKeys) context.renderedLineKeys = new Set();
+    const renderedLineKeys = context.renderedLineKeys;
+    const pendingLines = lines.filter((line)=>{
+        if (renderedLineKeys.has(line.key)) return false;
+        renderedLineKeys.add(line.key);
+        return true;
+    });
+    return pendingLines.length > 0 ? pendingLines.map((line)=>line.text).join('\n') : void 0;
+}
+function renderCliVerboseEventText(event, context) {
+    const command = 'string' == typeof event.command ? event.command : 'command';
+    const tool = 'string' == typeof event.tool ? event.tool : void 0;
+    switch(event.event){
+        case 'agent_progress':
+            {
+                const scope = 'string' == typeof event.scope ? event.scope : void 0;
+                const renderer = scope ? progressRenderers[scope] : void 0;
+                const progressEvent = isRecord(event.progress) ? event.progress : void 0;
+                if (!renderer || !progressEvent) return;
+                return renderLinesOnce(context, renderer.buildLines(progressEvent));
+            }
+        case 'command_start':
+            {
+                if (isActVerboseEvent(command, tool)) return;
+                const args = stringifyArgs(event.args);
+                return args ? `[Midscene] ${command} started (${args})` : `[Midscene] ${command} started`;
+            }
+        case 'agent_ready':
+            if (isActVerboseEvent(command, tool)) return;
+            return `[Midscene] ${tool ?? command} ready`;
+        case 'recording_ready':
+            return '[Midscene] Recording. Press Ctrl+C to stop and save.';
+        case 'recording_stopping':
+            if ('watchdog' === event.reason) return '[Midscene] Recording watchdog reached; finalizing and saving.';
+            if ('sigterm' === event.reason) return '[Midscene] SIGTERM received; finalizing and saving.';
+            return 'sighup' === event.reason ? '[Midscene] SIGHUP received; finalizing and saving.' : '[Midscene] Ctrl+C received; finalizing and saving. Press Ctrl+C again to force exit.';
+        case 'dump_update':
+            {
+                if (isActVerboseEvent(command, tool)) return;
+                const task = isRecord(event.task) ? event.task : void 0;
+                const steps = Array.isArray(event.steps) ? event.steps.filter(isRecord) : [];
+                const execution = isRecord(event.execution) ? event.execution : void 0;
+                const executionName = 'string' == typeof execution?.name ? execution.name : 'execution';
+                const lines = steps.length > 0 ? [
+                    `[Midscene] Progress: ${executionName} (${steps.length} steps)`,
+                    ...steps.map((step)=>{
+                        const index = 'number' == typeof step.index && 'number' == typeof step.total ? `Step ${step.index}/${step.total}` : 'Step';
+                        const message = 'string' == typeof step.message ? step.message : 'string' == typeof step.status ? step.status : 'updated';
+                        return `[Midscene] ${index}: ${message}`;
+                    })
+                ] : [
+                    `[Midscene] Step: ${'string' == typeof task?.message ? task.message : 'string' == typeof task?.status ? task.status : 'updated'}`
+                ];
+                const screenshots = renderScreenshotList(event.screenshots);
+                if (screenshots) lines.push(`[Midscene] Screenshot: ${screenshots}`);
+                if ('string' == typeof event.report && event.report.length > 0) lines.push(`[Midscene] Report: ${event.report}`);
+                return lines.join('\n');
+            }
+        case 'artifact':
+            {
+                if (isActVerboseEvent(command, tool)) return;
+                const kind = 'string' == typeof event.kind ? event.kind : 'artifact';
+                const path = 'string' == typeof event.path ? event.path : void 0;
+                return path ? `[Midscene] ${kind} saved: ${path}` : `[Midscene] ${kind} ready`;
+            }
+        case 'command_done':
+            {
+                if (isActVerboseEvent(command, tool)) {
+                    if ('error' === event.status) {
+                        const error = 'string' == typeof event.error && event.error.length > 0 ? `: ${event.error}` : '';
+                        return renderLinesOnce(context, [
+                            {
+                                key: `aiAct:command_failed:${event.error ?? 'unknown'}`,
+                                text: `[Midscene][aiAct] Failed${error}`
+                            }
+                        ]);
+                    }
+                    return;
+                }
+                const status = 'error' === event.status ? 'failed' : 'finished';
+                const duration = 'number' == typeof event.durationMs ? ` in ${event.durationMs}ms` : '';
+                const error = 'error' === event.status && 'string' == typeof event.error ? `: ${event.error}` : '';
+                return `[Midscene] ${command} ${status}${duration}${error}`;
+            }
+        default:
+            return;
+    }
+}
+function attachCliVerboseDumpListener(agent, options) {
+    if (!isCliVerboseEnabled()) return ()=>{};
+    const isActTool = options?.toolName === 'act';
+    const screenshotExportCache = new Map();
+    if (isActTool && 'function' == typeof agent.addProgressListener) return agent.addProgressListener((event)=>{
+        const match = progressRendererForEvent(event);
+        if (!match) return;
+        const context = getCliVerboseContext();
+        const isTextMode = 'jsonl' !== context.format;
+        const progress = match.renderer.normalize(event, {
+            reportFile: agent.reportFile,
+            exportMode: isTextMode ? 'report' : 'none',
+            cache: screenshotExportCache
+        });
+        if (!progress) return;
+        emitCliVerboseEvent({
+            event: 'agent_progress',
+            tool: options?.toolName,
+            scope: match.scope,
+            progress
+        });
+    });
+    if (isActTool) return ()=>{};
+    if ('function' != typeof agent.addDumpUpdateListener) return ()=>{};
+    return agent.addDumpUpdateListener((_dump, executionDump)=>{
+        const context = getCliVerboseContext();
+        const isTextMode = 'jsonl' !== context.format;
+        const summaryExportMode = isTextMode && !isActTool ? 'tmp' : 'none';
+        const summary = summarizeDumpUpdate(executionDump, {
+            exportMode: summaryExportMode,
+            cache: screenshotExportCache
+        });
+        if (!summary) return;
+        emitCliVerboseEvent({
+            event: 'dump_update',
+            tool: options?.toolName,
+            report: agent.reportFile || void 0,
+            ...summary
+        });
+    });
+}
+export { attachCliVerboseDumpListener, errorMessageOf as cliVerboseErrorMessage, cliVerboseFlag, compactCliVerboseArgs, compactCliVerboseValue, emitCliVerboseEvent, getCliVerboseContext, isCliVerboseEnabled, stripVerboseFlag, withCliVerboseContext };

package/dist/es/env/parse-model-config.mjs

@@ -5,7 +5,7 @@ import { assert } from "../utils.mjs";
 import { maskConfig, parseJson } from "./helper.mjs";
 import { initDebugConfig } from "./init-debug.mjs";
 const MODEL_CONFIG_DOC_URL = 'https://midscenejs.com/model-common-config.html';
-const getCurrentVersion = ()=>"8.0.3";
+const getCurrentVersion = ()=>"8.0.5";
 const getInvalidModelFamilyMessage = (modelFamily)=>`Invalid MIDSCENE_MODEL_FAMILY value: ${modelFamily}. Current version v${getCurrentVersion()} accepts the following model families: ${MODEL_FAMILY_VALUES.join(', ')}. You can also visit ${MODEL_CONFIG_DOC_URL} for the latest configuration information.`;
 const KEYS_MAP = {
     insight: INSIGHT_MODEL_CONFIG_KEYS,

package/dist/es/env/types.mjs

@@ -6,6 +6,7 @@ const MIDSCENE_DEBUG_MODEL_RESPONSE = 'MIDSCENE_DEBUG_MODEL_RESPONSE';
 const MIDSCENE_DANGEROUSLY_PRINT_ALL_CONFIG = 'MIDSCENE_DANGEROUSLY_PRINT_ALL_CONFIG';
 const MIDSCENE_DEBUG_MODE = 'MIDSCENE_DEBUG_MODE';
 const MIDSCENE_MCP_USE_PUPPETEER_MODE = 'MIDSCENE_MCP_USE_PUPPETEER_MODE';
+const MIDSCENE_CHROME_PATH = 'MIDSCENE_CHROME_PATH';
 const MIDSCENE_MCP_CHROME_PATH = 'MIDSCENE_MCP_CHROME_PATH';
 const MIDSCENE_MCP_ANDROID_MODE = 'MIDSCENE_MCP_ANDROID_MODE';
 const DOCKER_CONTAINER = 'DOCKER_CONTAINER';
@@ -23,6 +24,7 @@ const MIDSCENE_MODEL_RETRY_INTERVAL = 'MIDSCENE_MODEL_RETRY_INTERVAL';
 const MIDSCENE_MODEL_REASONING_EFFORT = 'MIDSCENE_MODEL_REASONING_EFFORT';
 const MIDSCENE_MODEL_REASONING_ENABLED = 'MIDSCENE_MODEL_REASONING_ENABLED';
 const MIDSCENE_MODEL_REASONING_BUDGET = 'MIDSCENE_MODEL_REASONING_BUDGET';
+const MIDSCENE_MODEL_RESPONSE_FORMAT = 'MIDSCENE_MODEL_RESPONSE_FORMAT';
 const OPENAI_API_KEY = 'OPENAI_API_KEY';
 const OPENAI_BASE_URL = 'OPENAI_BASE_URL';
 const MIDSCENE_OPENAI_INIT_CONFIG_JSON = 'MIDSCENE_OPENAI_INIT_CONFIG_JSON';
@@ -50,6 +52,7 @@ const MIDSCENE_PREFERRED_LANGUAGE = 'MIDSCENE_PREFERRED_LANGUAGE';
 const MIDSCENE_CACHE_MAX_FILENAME_LENGTH = 'MIDSCENE_CACHE_MAX_FILENAME_LENGTH';
 const MIDSCENE_REPLANNING_CYCLE_LIMIT = 'MIDSCENE_REPLANNING_CYCLE_LIMIT';
 const MIDSCENE_RUN_DIR = 'MIDSCENE_RUN_DIR';
+const MIDSCENE_RECORD_MODEL_CALL = 'MIDSCENE_RECORD_MODEL_CALL';
 const MIDSCENE_INSIGHT_MODEL_NAME = 'MIDSCENE_INSIGHT_MODEL_NAME';
 const MIDSCENE_INSIGHT_MODEL_SOCKS_PROXY = 'MIDSCENE_INSIGHT_MODEL_SOCKS_PROXY';
 const MIDSCENE_INSIGHT_MODEL_HTTP_PROXY = 'MIDSCENE_INSIGHT_MODEL_HTTP_PROXY';
@@ -65,6 +68,7 @@ const MIDSCENE_INSIGHT_MODEL_FAMILY = 'MIDSCENE_INSIGHT_MODEL_FAMILY';
 const MIDSCENE_INSIGHT_MODEL_REASONING_EFFORT = 'MIDSCENE_INSIGHT_MODEL_REASONING_EFFORT';
 const MIDSCENE_INSIGHT_MODEL_REASONING_ENABLED = 'MIDSCENE_INSIGHT_MODEL_REASONING_ENABLED';
 const MIDSCENE_INSIGHT_MODEL_REASONING_BUDGET = 'MIDSCENE_INSIGHT_MODEL_REASONING_BUDGET';
+const MIDSCENE_INSIGHT_MODEL_RESPONSE_FORMAT = 'MIDSCENE_INSIGHT_MODEL_RESPONSE_FORMAT';
 const MIDSCENE_PLANNING_MODEL_NAME = 'MIDSCENE_PLANNING_MODEL_NAME';
 const MIDSCENE_PLANNING_MODEL_SOCKS_PROXY = 'MIDSCENE_PLANNING_MODEL_SOCKS_PROXY';
 const MIDSCENE_PLANNING_MODEL_HTTP_PROXY = 'MIDSCENE_PLANNING_MODEL_HTTP_PROXY';
@@ -80,6 +84,7 @@ const MIDSCENE_PLANNING_MODEL_FAMILY = 'MIDSCENE_PLANNING_MODEL_FAMILY';
 const MIDSCENE_PLANNING_MODEL_REASONING_EFFORT = 'MIDSCENE_PLANNING_MODEL_REASONING_EFFORT';
 const MIDSCENE_PLANNING_MODEL_REASONING_ENABLED = 'MIDSCENE_PLANNING_MODEL_REASONING_ENABLED';
 const MIDSCENE_PLANNING_MODEL_REASONING_BUDGET = 'MIDSCENE_PLANNING_MODEL_REASONING_BUDGET';
+const MIDSCENE_PLANNING_MODEL_RESPONSE_FORMAT = 'MIDSCENE_PLANNING_MODEL_RESPONSE_FORMAT';
 const MIDSCENE_MODEL_FAMILY = 'MIDSCENE_MODEL_FAMILY';
 const UNUSED_ENV_KEYS = [
     MIDSCENE_DANGEROUSLY_PRINT_ALL_CONFIG
@@ -88,7 +93,8 @@ const BASIC_ENV_KEYS = [
     MIDSCENE_DEBUG_MODE,
     MIDSCENE_DEBUG_MODEL_PROFILE,
     MIDSCENE_DEBUG_MODEL_RESPONSE,
-    MIDSCENE_RUN_DIR
+    MIDSCENE_RUN_DIR,
+    MIDSCENE_RECORD_MODEL_CALL
 ];
 const BOOLEAN_ENV_KEYS = [
     MIDSCENE_CACHE,
@@ -115,6 +121,7 @@ const STRING_ENV_KEYS = [
     MIDSCENE_REPORT_TAG_NAME,
     MIDSCENE_PREFERRED_LANGUAGE,
     MATCH_BY_POSITION,
+    MIDSCENE_CHROME_PATH,
     MIDSCENE_MCP_CHROME_PATH,
     DOCKER_CONTAINER
 ];
@@ -138,6 +145,7 @@ const MODEL_ENV_KEYS = [
     MIDSCENE_MODEL_REASONING_EFFORT,
     MIDSCENE_MODEL_REASONING_ENABLED,
     MIDSCENE_MODEL_REASONING_BUDGET,
+    MIDSCENE_MODEL_RESPONSE_FORMAT,
     MIDSCENE_USE_VLM_UI_TARS,
     MIDSCENE_USE_QWEN_VL,
     MIDSCENE_USE_QWEN3_VL,
@@ -164,6 +172,7 @@ const MODEL_ENV_KEYS = [
     MIDSCENE_INSIGHT_MODEL_REASONING_EFFORT,
     MIDSCENE_INSIGHT_MODEL_REASONING_ENABLED,
     MIDSCENE_INSIGHT_MODEL_REASONING_BUDGET,
+    MIDSCENE_INSIGHT_MODEL_RESPONSE_FORMAT,
     MIDSCENE_PLANNING_MODEL_NAME,
     MIDSCENE_PLANNING_MODEL_SOCKS_PROXY,
     MIDSCENE_PLANNING_MODEL_HTTP_PROXY,
@@ -179,6 +188,7 @@ const MODEL_ENV_KEYS = [
     MIDSCENE_PLANNING_MODEL_REASONING_EFFORT,
     MIDSCENE_PLANNING_MODEL_REASONING_ENABLED,
     MIDSCENE_PLANNING_MODEL_REASONING_BUDGET,
+    MIDSCENE_PLANNING_MODEL_RESPONSE_FORMAT,
     MIDSCENE_MODEL_FAMILY
 ];
 const ALL_ENV_KEYS = [
@@ -193,6 +203,7 @@ const MODEL_FAMILY_VALUES = [
     'gemini',
     'qwen2.5-vl',
     'qwen3-vl',
+    'qwen3',
     'qwen3.5',
     'qwen3.6',
     'vlm-ui-tars',
@@ -201,7 +212,11 @@ const MODEL_FAMILY_VALUES = [
     'glm-v',
     'auto-glm',
     'auto-glm-multilingual',
-    'gpt-5'
+    'gpt-5',
+    'deepseek',
+    'kimi',
+    'kimi3',
+    'xiaomi-mimo'
 ];
 var types_UITarsModelVersion = /*#__PURE__*/ function(UITarsModelVersion) {
     UITarsModelVersion["V1_0"] = "1.0";
@@ -210,4 +225,4 @@ var types_UITarsModelVersion = /*#__PURE__*/ function(UITarsModelVersion) {
     UITarsModelVersion["DOUBAO_1_5_20B"] = "doubao-1.5-20B";
     return UITarsModelVersion;
 }({});
-export { ALL_ENV_KEYS, BASIC_ENV_KEYS, BOOLEAN_ENV_KEYS, DOCKER_CONTAINER, GLOBAL_ENV_KEYS, MATCH_BY_POSITION, MIDSCENE_ADB_PATH, MIDSCENE_ADB_REMOTE_HOST, MIDSCENE_ADB_REMOTE_PORT, MIDSCENE_ANDROID_IME_STRATEGY, MIDSCENE_CACHE, MIDSCENE_CACHE_MAX_FILENAME_LENGTH, MIDSCENE_DANGEROUSLY_PRINT_ALL_CONFIG, MIDSCENE_DEBUG_MODE, MIDSCENE_DEBUG_MODEL_PROFILE, MIDSCENE_DEBUG_MODEL_RESPONSE, MIDSCENE_INSIGHT_MODEL_API_KEY, MIDSCENE_INSIGHT_MODEL_BASE_URL, MIDSCENE_INSIGHT_MODEL_EXTRA_BODY_JSON, MIDSCENE_INSIGHT_MODEL_FAMILY, MIDSCENE_INSIGHT_MODEL_HTTP_PROXY, MIDSCENE_INSIGHT_MODEL_INIT_CONFIG_JSON, MIDSCENE_INSIGHT_MODEL_NAME, MIDSCENE_INSIGHT_MODEL_REASONING_BUDGET, MIDSCENE_INSIGHT_MODEL_REASONING_EFFORT, MIDSCENE_INSIGHT_MODEL_REASONING_ENABLED, MIDSCENE_INSIGHT_MODEL_RETRY_COUNT, MIDSCENE_INSIGHT_MODEL_RETRY_INTERVAL, MIDSCENE_INSIGHT_MODEL_SOCKS_PROXY, MIDSCENE_INSIGHT_MODEL_TEMPERATURE, MIDSCENE_INSIGHT_MODEL_TIMEOUT, MIDSCENE_IOS_DEVICE_CLASS_OVERRIDE, MIDSCENE_IOS_DEVICE_UDID, MIDSCENE_IOS_SIMULATOR_UDID, MIDSCENE_LANGFUSE_DEBUG, MIDSCENE_LANGSMITH_DEBUG, MIDSCENE_MCP_ANDROID_MODE, MIDSCENE_MCP_CHROME_PATH, MIDSCENE_MCP_USE_PUPPETEER_MODE, MIDSCENE_MODEL_API_KEY, MIDSCENE_MODEL_BASE_URL, MIDSCENE_MODEL_EXTRA_BODY_JSON, MIDSCENE_MODEL_FAMILY, MIDSCENE_MODEL_HTTP_PROXY, MIDSCENE_MODEL_INIT_CONFIG_JSON, MIDSCENE_MODEL_MAX_TOKENS, MIDSCENE_MODEL_NAME, MIDSCENE_MODEL_REASONING_BUDGET, MIDSCENE_MODEL_REASONING_EFFORT, MIDSCENE_MODEL_REASONING_ENABLED, MIDSCENE_MODEL_RETRY_COUNT, MIDSCENE_MODEL_RETRY_INTERVAL, MIDSCENE_MODEL_SOCKS_PROXY, MIDSCENE_MODEL_TEMPERATURE, MIDSCENE_MODEL_TIMEOUT, MIDSCENE_OPENAI_HTTP_PROXY, MIDSCENE_OPENAI_INIT_CONFIG_JSON, MIDSCENE_OPENAI_SOCKS_PROXY, MIDSCENE_PLANNING_MODEL_API_KEY, MIDSCENE_PLANNING_MODEL_BASE_URL, MIDSCENE_PLANNING_MODEL_EXTRA_BODY_JSON, MIDSCENE_PLANNING_MODEL_FAMILY, MIDSCENE_PLANNING_MODEL_HTTP_PROXY, MIDSCENE_PLANNING_MODEL_INIT_CONFIG_JSON, MIDSCENE_PLANNING_MODEL_NAME, MIDSCENE_PLANNING_MODEL_REASONING_BUDGET, MIDSCENE_PLANNING_MODEL_REASONING_EFFORT, MIDSCENE_PLANNING_MODEL_REASONING_ENABLED, MIDSCENE_PLANNING_MODEL_RETRY_COUNT, MIDSCENE_PLANNING_MODEL_RETRY_INTERVAL, MIDSCENE_PLANNING_MODEL_SOCKS_PROXY, MIDSCENE_PLANNING_MODEL_TEMPERATURE, MIDSCENE_PLANNING_MODEL_TIMEOUT, MIDSCENE_PREFERRED_LANGUAGE, MIDSCENE_REPLANNING_CYCLE_LIMIT, MIDSCENE_REPORT_QUIET, MIDSCENE_REPORT_TAG_NAME, MIDSCENE_RUN_DIR, MIDSCENE_USE_DOUBAO_VISION, MIDSCENE_USE_GEMINI, MIDSCENE_USE_QWEN3_VL, MIDSCENE_USE_QWEN_VL, MIDSCENE_USE_VLM_UI_TARS, MIDSCENE_USE_VL_MODEL, MODEL_ENV_KEYS, MODEL_FAMILY_VALUES, NUMBER_ENV_KEYS, OPENAI_API_KEY, OPENAI_BASE_URL, OPENAI_MAX_TOKENS, STRING_ENV_KEYS, types_UITarsModelVersion as UITarsModelVersion, UNUSED_ENV_KEYS };
+export { ALL_ENV_KEYS, BASIC_ENV_KEYS, BOOLEAN_ENV_KEYS, DOCKER_CONTAINER, GLOBAL_ENV_KEYS, MATCH_BY_POSITION, MIDSCENE_ADB_PATH, MIDSCENE_ADB_REMOTE_HOST, MIDSCENE_ADB_REMOTE_PORT, MIDSCENE_ANDROID_IME_STRATEGY, MIDSCENE_CACHE, MIDSCENE_CACHE_MAX_FILENAME_LENGTH, MIDSCENE_CHROME_PATH, MIDSCENE_DANGEROUSLY_PRINT_ALL_CONFIG, MIDSCENE_DEBUG_MODE, MIDSCENE_DEBUG_MODEL_PROFILE, MIDSCENE_DEBUG_MODEL_RESPONSE, MIDSCENE_INSIGHT_MODEL_API_KEY, MIDSCENE_INSIGHT_MODEL_BASE_URL, MIDSCENE_INSIGHT_MODEL_EXTRA_BODY_JSON, MIDSCENE_INSIGHT_MODEL_FAMILY, MIDSCENE_INSIGHT_MODEL_HTTP_PROXY, MIDSCENE_INSIGHT_MODEL_INIT_CONFIG_JSON, MIDSCENE_INSIGHT_MODEL_NAME, MIDSCENE_INSIGHT_MODEL_REASONING_BUDGET, MIDSCENE_INSIGHT_MODEL_REASONING_EFFORT, MIDSCENE_INSIGHT_MODEL_REASONING_ENABLED, MIDSCENE_INSIGHT_MODEL_RESPONSE_FORMAT, MIDSCENE_INSIGHT_MODEL_RETRY_COUNT, MIDSCENE_INSIGHT_MODEL_RETRY_INTERVAL, MIDSCENE_INSIGHT_MODEL_SOCKS_PROXY, MIDSCENE_INSIGHT_MODEL_TEMPERATURE, MIDSCENE_INSIGHT_MODEL_TIMEOUT, MIDSCENE_IOS_DEVICE_CLASS_OVERRIDE, MIDSCENE_IOS_DEVICE_UDID, MIDSCENE_IOS_SIMULATOR_UDID, MIDSCENE_LANGFUSE_DEBUG, MIDSCENE_LANGSMITH_DEBUG, MIDSCENE_MCP_ANDROID_MODE, MIDSCENE_MCP_CHROME_PATH, MIDSCENE_MCP_USE_PUPPETEER_MODE, MIDSCENE_MODEL_API_KEY, MIDSCENE_MODEL_BASE_URL, MIDSCENE_MODEL_EXTRA_BODY_JSON, MIDSCENE_MODEL_FAMILY, MIDSCENE_MODEL_HTTP_PROXY, MIDSCENE_MODEL_INIT_CONFIG_JSON, MIDSCENE_MODEL_MAX_TOKENS, MIDSCENE_MODEL_NAME, MIDSCENE_MODEL_REASONING_BUDGET, MIDSCENE_MODEL_REASONING_EFFORT, MIDSCENE_MODEL_REASONING_ENABLED, MIDSCENE_MODEL_RESPONSE_FORMAT, MIDSCENE_MODEL_RETRY_COUNT, MIDSCENE_MODEL_RETRY_INTERVAL, MIDSCENE_MODEL_SOCKS_PROXY, MIDSCENE_MODEL_TEMPERATURE, MIDSCENE_MODEL_TIMEOUT, MIDSCENE_OPENAI_HTTP_PROXY, MIDSCENE_OPENAI_INIT_CONFIG_JSON, MIDSCENE_OPENAI_SOCKS_PROXY, MIDSCENE_PLANNING_MODEL_API_KEY, MIDSCENE_PLANNING_MODEL_BASE_URL, MIDSCENE_PLANNING_MODEL_EXTRA_BODY_JSON, MIDSCENE_PLANNING_MODEL_FAMILY, MIDSCENE_PLANNING_MODEL_HTTP_PROXY, MIDSCENE_PLANNING_MODEL_INIT_CONFIG_JSON, MIDSCENE_PLANNING_MODEL_NAME, MIDSCENE_PLANNING_MODEL_REASONING_BUDGET, MIDSCENE_PLANNING_MODEL_REASONING_EFFORT, MIDSCENE_PLANNING_MODEL_REASONING_ENABLED, MIDSCENE_PLANNING_MODEL_RESPONSE_FORMAT, MIDSCENE_PLANNING_MODEL_RETRY_COUNT, MIDSCENE_PLANNING_MODEL_RETRY_INTERVAL, MIDSCENE_PLANNING_MODEL_SOCKS_PROXY, MIDSCENE_PLANNING_MODEL_TEMPERATURE, MIDSCENE_PLANNING_MODEL_TIMEOUT, MIDSCENE_PREFERRED_LANGUAGE, MIDSCENE_RECORD_MODEL_CALL, MIDSCENE_REPLANNING_CYCLE_LIMIT, MIDSCENE_REPORT_QUIET, MIDSCENE_REPORT_TAG_NAME, MIDSCENE_RUN_DIR, MIDSCENE_USE_DOUBAO_VISION, MIDSCENE_USE_GEMINI, MIDSCENE_USE_QWEN3_VL, MIDSCENE_USE_QWEN_VL, MIDSCENE_USE_VLM_UI_TARS, MIDSCENE_USE_VL_MODEL, MODEL_ENV_KEYS, MODEL_FAMILY_VALUES, NUMBER_ENV_KEYS, OPENAI_API_KEY, OPENAI_BASE_URL, OPENAI_MAX_TOKENS, STRING_ENV_KEYS, types_UITarsModelVersion as UITarsModelVersion, UNUSED_ENV_KEYS };

package/dist/es/img/index.mjs

@@ -1,4 +1,4 @@
-import { imageInfoOfBase64, isValidImageBuffer, isValidJPEGImageBuffer, isValidPNGImageBuffer, validateScreenshotBuffer } from "./info.mjs";
-import { createImgBase64ByFormat, cropByRect, httpImg2Base64, localImg2Base64, paddingToMatchBlockByBase64, parseBase64, preProcessImageUrl, resizeAndConvertImgBuffer, resizeImgBase64, saveBase64Image, scaleImage, zoomForGPT4o } from "./transform.mjs";
+import { encodedImageInfoOfBuffer, imageInfoOfBase64, isValidImageBuffer, isValidJPEGImageBuffer, isValidPNGImageBuffer, validateScreenshotBuffer } from "./info.mjs";
+import { constrainBase64ImageToMaxSize, convertBase64ImageToJpeg, convertImgBufferToJpeg, createImgBase64ByFormat, cropByRect, httpImg2Base64, inferBase64ImageFormat, localImg2Base64, normalizeBase64Image, normalizeScreenshotBase64, paddingToMatchBlockByBase64, parseBase64, preProcessImageUrl, resizeAndConvertImgBuffer, resizeBase64ImageToJpeg, resizeImgBase64, saveBase64Image, scaleImage, zoomForGPT4o } from "./transform.mjs";
 import { annotateRects, compositeElementInfoImg, processImageElementInfo } from "./box-select.mjs";
-export { annotateRects, compositeElementInfoImg, createImgBase64ByFormat, cropByRect, httpImg2Base64, imageInfoOfBase64, isValidImageBuffer, isValidJPEGImageBuffer, isValidPNGImageBuffer, localImg2Base64, paddingToMatchBlockByBase64, parseBase64, preProcessImageUrl, processImageElementInfo, resizeAndConvertImgBuffer, resizeImgBase64, saveBase64Image, scaleImage, validateScreenshotBuffer, zoomForGPT4o };
+export { annotateRects, compositeElementInfoImg, constrainBase64ImageToMaxSize, convertBase64ImageToJpeg, convertImgBufferToJpeg, createImgBase64ByFormat, cropByRect, encodedImageInfoOfBuffer, httpImg2Base64, imageInfoOfBase64, inferBase64ImageFormat, isValidImageBuffer, isValidJPEGImageBuffer, isValidPNGImageBuffer, localImg2Base64, normalizeBase64Image, normalizeScreenshotBase64, paddingToMatchBlockByBase64, parseBase64, preProcessImageUrl, processImageElementInfo, resizeAndConvertImgBuffer, resizeBase64ImageToJpeg, resizeImgBase64, saveBase64Image, scaleImage, validateScreenshotBuffer, zoomForGPT4o };

package/dist/es/img/info.mjs

@@ -1,5 +1,49 @@
 import node_assert from "node:assert";
 import get_photon from "./get-photon.mjs";
+const isJpegStartOfFrameMarker = (marker)=>marker >= 0xc0 && marker <= 0xcf && 0xc4 !== marker && 0xc8 !== marker && 0xcc !== marker;
+function jpegInfoFromBuffer(imageBuffer) {
+    let offset = 2;
+    while(offset < imageBuffer.length){
+        if (0xff !== imageBuffer[offset]) {
+            offset += 1;
+            continue;
+        }
+        while(offset < imageBuffer.length && 0xff === imageBuffer[offset])offset += 1;
+        if (offset >= imageBuffer.length) break;
+        const marker = imageBuffer[offset];
+        offset += 1;
+        if (0xd9 === marker || 0xda === marker) break;
+        if (0x01 === marker || 0xd8 === marker || marker >= 0xd0 && marker <= 0xd7) continue;
+        if (offset + 2 > imageBuffer.length) break;
+        const segmentLength = imageBuffer.readUInt16BE(offset);
+        if (segmentLength < 2 || offset + segmentLength > imageBuffer.length) break;
+        if (isJpegStartOfFrameMarker(marker)) {
+            if (segmentLength < 7) break;
+            const height = imageBuffer.readUInt16BE(offset + 3);
+            const width = imageBuffer.readUInt16BE(offset + 5);
+            node_assert(width && height, 'Invalid image: cannot get width or height');
+            return {
+                width,
+                height
+            };
+        }
+        offset += segmentLength;
+    }
+    throw new Error('Invalid image: cannot get JPEG width or height');
+}
+function encodedImageInfoOfBuffer(imageBuffer) {
+    if (isValidPNGImageBuffer(imageBuffer)) {
+        const width = imageBuffer.readUInt32BE(16);
+        const height = imageBuffer.readUInt32BE(20);
+        node_assert(width && height, 'Invalid image: cannot get width or height');
+        return {
+            width,
+            height
+        };
+    }
+    if (isValidJPEGImageBuffer(imageBuffer)) return jpegInfoFromBuffer(imageBuffer);
+    throw new Error('Invalid image: unsupported format');
+}
 async function imageInfoOfBase64(imageBase64) {
     const { PhotonImage } = await get_photon();
     const base64Data = imageBase64.replace(/^data:image\/\w+;base64,/, '');
@@ -32,4 +76,4 @@ function validateScreenshotBuffer(screenshotBuffer, { label, minBufferSize = 0 }
     if (!isValidImageBuffer(screenshotBuffer)) throw new Error(`${label} buffer has invalid image format`);
     if (minBufferSize > 0 && bufferSize < minBufferSize) throw new Error(`${label} validation failed: buffer size ${bufferSize} bytes (minimum: ${minBufferSize})`);
 }
-export { imageInfoOfBase64, isValidImageBuffer, isValidJPEGImageBuffer, isValidPNGImageBuffer, validateScreenshotBuffer };
+export { encodedImageInfoOfBuffer, imageInfoOfBase64, isValidImageBuffer, isValidJPEGImageBuffer, isValidPNGImageBuffer, validateScreenshotBuffer };

package/dist/es/img/photon-loader.mjs

@@ -0,0 +1,4 @@
+async function loadPhotonModule() {
+    return import("@silvia-odwyer/photon");
+}
+export { loadPhotonModule };