shennian 0.2.78 → 0.2.84

package/dist/scripts/wechat-rpa-win.mjs

@@ -32,10 +32,13 @@ function printHelp() {
   node scripts/wechat-rpa-win.mjs send-files --group <name> --file <path> [--file <path> ...]
   node scripts/wechat-rpa-win.mjs listen --group <name> --seconds 60 --poll-ms 1000
   node scripts/wechat-rpa-win.mjs capture --region window --output C:\\tmp\\wechat.png
-  node scripts/wechat-rpa-win.mjs click --x 480 --y 470 --no-raise
+  node scripts/wechat-rpa-win.mjs resize-window --width 1200 --height 820 [--x 120 --y 80] [--include-offscreen]
+  node scripts/wechat-rpa-win.mjs click --x 480 --y 470 --no-raise [--right]
   node scripts/wechat-rpa-win.mjs paste-text --text "hello"
   node scripts/wechat-rpa-win.mjs paste-files --file C:\\tmp\\demo.png
+  node scripts/wechat-rpa-win.mjs read-clipboard
   node scripts/wechat-rpa-win.mjs press --keys "%s"
+  node scripts/wechat-rpa-win.mjs dismiss-menus --count 2
 
 Options:
   --helper <path>  Override helper exe path.
@@ -70,6 +73,33 @@ async function main() {
         'Set SHENNIAN_WECHAT_RPA_ALLOW_UIA_DIAGNOSTICS=1 in a local diagnostic shell to use it.',
     )
   }
+  if (command === 'restart-wechat-for-uia') {
+    throw new Error(
+      'restart-wechat-for-uia is disabled from this lab CLI because it closes the user\'s WeChat session. ' +
+        'Do not restart or kill WeChat during Windows RPA validation.',
+    )
+  }
+  if (command === 'resize-window') {
+    const width = Number(takeOption(argv, '--width'))
+    const height = Number(takeOption(argv, '--height'))
+    const x = takeOption(argv, '--x')
+    const y = takeOption(argv, '--y')
+    const includeOffscreen = argv.includes('--include-offscreen')
+    if (includeOffscreen) argv.splice(argv.indexOf('--include-offscreen'), 1)
+    if (!Number.isFinite(width) || !Number.isFinite(height)) throw new Error('resize-window requires --width and --height')
+    if (width < 760 || height < 600) throw new Error('WeChat RPA window must be at least 760x600')
+    if (x != null && !Number.isFinite(Number(x))) throw new Error('resize-window --x must be a number')
+    if (y != null && !Number.isFinite(Number(y))) throw new Error('resize-window --y must be a number')
+    const result = await resizeWindowWithPowerShell({
+      width: Math.round(width),
+      height: Math.round(height),
+      x: x == null ? null : Math.round(Number(x)),
+      y: y == null ? null : Math.round(Number(y)),
+      includeOffscreen,
+    })
+    process.stdout.write(`${JSON.stringify({ ok: true, command, payload: result }, null, 2)}\n`)
+    return
+  }
 
   const helper = takeOption(argv, '--helper') ?? process.env.WECHAT_RPA_WIN_HELPER ?? defaultHelper
   if (!fs.existsSync(helper)) {
@@ -120,6 +150,202 @@ async function main() {
   }
 }
 
+async function resizeWindowWithPowerShell({ width, height, x, y, includeOffscreen = false }) {
+  const script = `
+$ErrorActionPreference = "Stop"
+$typeDefinition = '
+	using System;
+	using System.Collections.Generic;
+	using System.Runtime.InteropServices;
+	using System.Text;
+
+public static class Win32WechatResize {
+  public delegate bool EnumWindowsProc(IntPtr hWnd, IntPtr lParam);
+
+	  [StructLayout(LayoutKind.Sequential)]
+	  public struct RECT {
+	    public int Left;
+	    public int Top;
+	    public int Right;
+	    public int Bottom;
+	  }
+
+	  [StructLayout(LayoutKind.Sequential)]
+	  public struct MONITORINFO {
+	    public int cbSize;
+	    public RECT rcMonitor;
+	    public RECT rcWork;
+	    public uint dwFlags;
+	  }
+
+  [DllImport("user32.dll")]
+  public static extern bool EnumWindows(EnumWindowsProc callback, IntPtr lParam);
+
+  [DllImport("user32.dll")]
+  public static extern bool IsWindowVisible(IntPtr hWnd);
+
+  [DllImport("user32.dll")]
+  public static extern uint GetWindowThreadProcessId(IntPtr hWnd, out uint processId);
+
+  [DllImport("user32.dll", SetLastError = true)]
+  public static extern bool GetWindowRect(IntPtr hWnd, out RECT rect);
+
+  [DllImport("user32.dll", CharSet = CharSet.Unicode, SetLastError = true)]
+  public static extern int GetWindowText(IntPtr hWnd, StringBuilder text, int maxCount);
+
+  [DllImport("user32.dll", CharSet = CharSet.Unicode, SetLastError = true)]
+  public static extern int GetClassName(IntPtr hWnd, StringBuilder text, int maxCount);
+
+  [DllImport("user32.dll")]
+  public static extern bool ShowWindow(IntPtr hWnd, int nCmdShow);
+
+	  [DllImport("user32.dll")]
+	  public static extern bool SetWindowPos(IntPtr hWnd, IntPtr hWndInsertAfter, int x, int y, int cx, int cy, uint flags);
+
+	  [DllImport("user32.dll")]
+	  public static extern IntPtr MonitorFromWindow(IntPtr hWnd, uint dwFlags);
+
+	  [DllImport("user32.dll", SetLastError = true)]
+	  public static extern bool GetMonitorInfo(IntPtr hMonitor, ref MONITORINFO lpmi);
+	}
+'
+Add-Type -TypeDefinition $typeDefinition
+
+function Read-WindowText([IntPtr]$Handle) {
+  $builder = New-Object System.Text.StringBuilder 512
+  [void][Win32WechatResize]::GetWindowText($Handle, $builder, $builder.Capacity)
+  $builder.ToString()
+}
+
+function Read-ClassName([IntPtr]$Handle) {
+  $builder = New-Object System.Text.StringBuilder 256
+  [void][Win32WechatResize]::GetClassName($Handle, $builder, $builder.Capacity)
+  $builder.ToString()
+}
+
+$processIds = @(Get-Process Weixin,WeChat,WeChatAppEx -ErrorAction SilentlyContinue | ForEach-Object { [uint32]$_.Id })
+if ($processIds.Count -eq 0) { throw "Cannot find WeChat/Weixin process" }
+
+$windows = New-Object System.Collections.Generic.List[object]
+$callback = [Win32WechatResize+EnumWindowsProc]{
+  param([IntPtr]$Handle, [IntPtr]$Param)
+  $includeOffscreen = ${includeOffscreen ? '$true' : '$false'}
+  $visible = [Win32WechatResize]::IsWindowVisible($Handle)
+  if (-not $includeOffscreen -and -not $visible) { return $true }
+  [uint32]$windowPid = 0
+  [void][Win32WechatResize]::GetWindowThreadProcessId($Handle, [ref]$windowPid)
+  if ($processIds -notcontains $windowPid) { return $true }
+  $rect = New-Object Win32WechatResize+RECT
+  if (-not [Win32WechatResize]::GetWindowRect($Handle, [ref]$rect)) { return $true }
+  $w = $rect.Right - $rect.Left
+  $h = $rect.Bottom - $rect.Top
+  $title = Read-WindowText $Handle
+  $className = Read-ClassName $Handle
+  if ($title -ne "微信" -and $className -notmatch "Qt|WeChat") { return $true }
+  if (-not $includeOffscreen -and ($w -lt 760 -or $h -lt 600)) { return $true }
+  $windows.Add([pscustomobject]@{
+    handle = $Handle.ToInt64()
+    processId = $windowPid
+    title = $title
+    className = $className
+    visible = $visible
+    x = $rect.Left
+    y = $rect.Top
+    width = $w
+    height = $h
+    area = $w * $h
+    offscreen = (-not $visible -or $rect.Left -lt -1000 -or $rect.Top -lt -1000 -or $w -lt 760 -or $h -lt 600)
+  }) | Out-Null
+  return $true
+}
+[void][Win32WechatResize]::EnumWindows($callback, [IntPtr]::Zero)
+$window = $windows | Sort-Object @{ Expression = { if ($_.offscreen) { 1 } else { 0 } } }, @{ Expression = 'area'; Descending = $true } | Select-Object -First 1
+if (-not $window) { throw "Cannot locate WeChat main window via Win32" }
+$requestedWidth = ${width}
+$requestedHeight = ${height}
+$left = [int](${x == null ? '$window.x' : Number(x)})
+$top = [int](${y == null ? '$window.y' : Number(y)})
+$monitor = [Win32WechatResize]::MonitorFromWindow([IntPtr]$window.handle, 2)
+$info = New-Object Win32WechatResize+MONITORINFO
+$info.cbSize = [Runtime.InteropServices.Marshal]::SizeOf([type][Win32WechatResize+MONITORINFO])
+if (-not [Win32WechatResize]::GetMonitorInfo($monitor, [ref]$info)) { throw "Cannot read monitor work area" }
+$margin = 8
+$workLeft = [int]$info.rcWork.Left
+$workTop = [int]$info.rcWork.Top
+$workRight = [int]$info.rcWork.Right
+$workBottom = [int]$info.rcWork.Bottom
+$minWidth = 760
+$minHeight = 600
+$maxWidth = [Math]::Max($minWidth, $workRight - $workLeft - ($margin * 2))
+$maxHeight = [Math]::Max($minHeight, $workBottom - $workTop - ($margin * 2))
+$targetWidth = [Math]::Min($requestedWidth, $maxWidth)
+$targetHeight = [Math]::Min($requestedHeight, $maxHeight)
+if ($left -lt ($workLeft + $margin)) { $left = $workLeft + $margin }
+if ($top -lt ($workTop + $margin)) { $top = $workTop + $margin }
+if (($left + $targetWidth) -gt ($workRight - $margin)) { $left = [Math]::Max($workLeft + $margin, $workRight - $targetWidth - $margin) }
+if (($top + $targetHeight) -gt ($workBottom - $margin)) { $top = [Math]::Max($workTop + $margin, $workBottom - $targetHeight - $margin) }
+[void][Win32WechatResize]::ShowWindow([IntPtr]$window.handle, 9)
+[void][Win32WechatResize]::SetWindowPos([IntPtr]$window.handle, [intptr](-2), [int]$left, [int]$top, [int]$targetWidth, [int]$targetHeight, 0x0040)
+Start-Sleep -Milliseconds 250
+$afterRect = New-Object Win32WechatResize+RECT
+[void][Win32WechatResize]::GetWindowRect([IntPtr]$window.handle, [ref]$afterRect)
+$afterWidth = $afterRect.Right - $afterRect.Left
+$afterHeight = $afterRect.Bottom - $afterRect.Top
+if ($targetWidth -ge $minWidth -and $targetHeight -ge $minHeight -and ($afterWidth -lt ($minWidth - 20) -or $afterHeight -lt ($minHeight - 20))) {
+  throw ("resize-window did not produce a usable WeChat main window; after=" + $afterWidth + "x" + $afterHeight + " requested=" + $requestedWidth + "x" + $requestedHeight)
+}
+if ([Math]::Abs($afterWidth - $targetWidth) -gt 80 -or [Math]::Abs($afterHeight - $targetHeight) -gt 80) {
+  throw ("resize-window result diverged from applied target; after=" + $afterWidth + "x" + $afterHeight + " applied=" + $targetWidth + "x" + $targetHeight)
+}
+@{
+  action = "resize-window"
+  requested = @{
+    x = ${x == null ? '$null' : Number(x)}
+    y = ${y == null ? '$null' : Number(y)}
+    width = $requestedWidth
+    height = $requestedHeight
+  }
+  applied = @{
+    x = $left
+    y = $top
+    width = $targetWidth
+    height = $targetHeight
+  }
+  workArea = @{
+    x = $workLeft
+    y = $workTop
+    width = $workRight - $workLeft
+    height = $workBottom - $workTop
+  }
+  before = $window
+  after = @{
+    x = $afterRect.Left
+    y = $afterRect.Top
+    width = $afterWidth
+    height = $afterHeight
+  }
+} | ConvertTo-Json -Depth 5
+`
+  const encoded = Buffer.from(script, 'utf16le').toString('base64')
+  const child = spawn('powershell.exe', ['-NoProfile', '-NonInteractive', '-ExecutionPolicy', 'Bypass', '-EncodedCommand', encoded], {
+    cwd: repoRoot,
+    windowsHide: true,
+    stdio: ['ignore', 'pipe', 'pipe'],
+  })
+  let stdout = ''
+  let stderr = ''
+  child.stdout.setEncoding('utf8')
+  child.stderr.setEncoding('utf8')
+  child.stdout.on('data', chunk => { stdout += chunk })
+  child.stderr.on('data', chunk => { stderr += chunk })
+  const exitCode = await new Promise((resolve, reject) => {
+    child.on('error', reject)
+    child.on('close', resolve)
+  })
+  if (exitCode !== 0) throw new Error(stderr.trim() || stdout.trim() || `resize-window PowerShell exited ${exitCode}`)
+  return JSON.parse(stdout.trim())
+}
+
 main().catch(error => {
   console.error(error instanceof Error ? error.message : String(error))
   process.exit(1)

package/dist/src/agents/external-channel-instructions.js

@@ -62,9 +62,7 @@ function weChatRpaRuntimeInstructions(channel) {
         '这是运行在绑定电脑上的个人微信 RPA 通道，不是云端托管企业微信账号。',
         groups.length ? `当前监听群：${groups.join('、')}。只把这些群里的消息当成此通道上下文。` : '',
         channel.forceForeground ? '本通道允许在用户空闲时短暂前台化微信。' : '本通道默认不强制抢前台；用户正在操作时可能延迟收发。',
-        channel.cloudOcrMode && channel.cloudOcrMode !== 'off'
-            ? `截图识别不足时会通过神念服务端调用云端 OCR fallback（模式：${channel.cloudOcrMode}），边缘端不直接持有千问 API key。`
-            : '默认只用本机 OCR/视觉结果；不要假设云端多模态已经理解图片内容。',
+        '运行时只用绑定电脑上的本机 OCR/视觉结果；不要假设服务端或云端多模态已经理解图片内容。',
     ].filter(Boolean).join('\n');
 }
 export function buildExternalChannelInstructions(channel, workDir, sessionId, mode = 'agent') {

package/dist/src/channels/base.d.ts

@@ -1,5 +1,5 @@
 import type { ExternalChannelSessionStatus } from '@shennian/wire';
-export type ExternalChannelRuntimeStatus = Pick<ExternalChannelSessionStatus, 'wechatRpaRuntimeState' | 'wechatRpaLastRunAt' | 'wechatRpaLastMessageAt' | 'wechatRpaLastInterruptedAt' | 'wechatRpaPendingReplyCount' | 'wechatRpaLastError' | 'wechatRpaLastCloudOcrAt' | 'wechatRpaLastCloudOcrPurpose' | 'wechatRpaLastCloudOcrRequestId' | 'wechatRpaLastCloudOcrImageHash' | 'wechatRpaLastCloudOcrUsage'>;
+export type ExternalChannelRuntimeStatus = Pick<ExternalChannelSessionStatus, 'wechatRpaRuntimeState' | 'wechatRpaLastRunAt' | 'wechatRpaLastMessageAt' | 'wechatRpaLastInterruptedAt' | 'wechatRpaPendingReplyCount' | 'wechatRpaLastError' | 'wechatRpaLastRunId' | 'wechatRpaLastTracePath' | 'wechatRpaLastTraceSummary' | 'wechatRpaRecentTaskSummaries' | 'wechatRpaLastCloudOcrAt' | 'wechatRpaLastCloudOcrPurpose' | 'wechatRpaLastCloudOcrRequestId' | 'wechatRpaLastCloudOcrImageHash' | 'wechatRpaLastCloudOcrUsage'>;
 export type ExternalChannelType = 'wecom' | 'websocket' | 'wechat-rpa';
 export type ExternalChannelConfig = {
     id: string;
@@ -41,6 +41,7 @@ export type ExternalChannelView = {
     noRestore?: boolean;
     downloadAttachments?: boolean;
     downloadAttachmentsDir?: string;
+    selfNickname?: string | null;
     cloudOcrUrl?: string;
     cloudOcrToken?: string;
     cloudOcrMode?: string;
@@ -49,6 +50,10 @@ export type ExternalChannelView = {
     wechatRpaLastMessageAt?: string | null;
     wechatRpaLastInterruptedAt?: string | null;
     wechatRpaLastError?: string | null;
+    wechatRpaLastRunId?: string | null;
+    wechatRpaLastTracePath?: string | null;
+    wechatRpaLastTraceSummary?: string | null;
+    wechatRpaRecentTaskSummaries?: ExternalChannelSessionStatus['wechatRpaRecentTaskSummaries'];
     wechatRpaLastCloudOcrAt?: string | null;
     wechatRpaLastCloudOcrPurpose?: string | null;
     wechatRpaLastCloudOcrRequestId?: string | null;
@@ -64,6 +69,7 @@ export type ExternalMessageEvent = {
     channelId: string;
     channelType: ExternalChannelType;
     conversationId: string;
+    conversationName?: string | null;
     messageId: string;
     sender: {
         id: string;
@@ -72,6 +78,7 @@ export type ExternalMessageEvent = {
     text: string;
     attachments: ExternalMessageAttachment[];
     receivedAt: string;
+    isMentioned?: boolean;
     replyTarget: string;
     rawRef?: string | null;
 };
@@ -107,7 +114,7 @@ export type ExternalReply = {
     idempotencyKey?: string;
 };
 export type ExternalChannelSendResult = void | {
-    status: 'sent' | 'queued';
+    status: 'sent' | 'queued' | 'manual-review';
     reason?: string;
 };
 export interface ExternalChannelAdapter {

package/dist/src/channels/runtime.d.ts

@@ -91,7 +91,7 @@ export declare class ChannelRuntime {
         }>;
         canReply?: boolean;
         systemPrompt?: string;
-        source?: 'macos-flow' | 'macos-probe' | 'windows-visual-flow' | 'fixture-jsonl';
+        source?: 'macos-flow' | 'macos-probe' | 'windows-visual-flow' | 'wechat-rpa-lab' | 'fixture-jsonl';
         pollIntervalMs?: number;
         recentLimit?: number;
         idleSeconds?: number;
@@ -99,6 +99,7 @@ export declare class ChannelRuntime {
         noRestore?: boolean;
         downloadAttachments?: boolean;
         downloadAttachmentsDir?: string;
+        selfNickname?: string;
         flowScriptPath?: string;
         cloudOcrUrl?: string;
         cloudOcrToken?: string;

package/dist/src/channels/runtime.js

@@ -9,8 +9,6 @@ import { WeComChannelAdapter } from './wecom.js';
 import { WeChatRpaChannelAdapter } from './wechat-rpa.js';
 import { ExternalWebSocketChannelAdapter } from './websocket.js';
 import { splitExternalReplyText } from './reply-split.js';
-import { loadConfig } from '../config/index.js';
-import { SERVERS } from '../region.js';
 export class ChannelRuntime {
     onExternalMessage;
     createReplyTarget;
@@ -82,6 +80,9 @@ export class ChannelRuntime {
                     pending = true;
                     continue;
                 }
+                if (result?.status === 'manual-review') {
+                    return { ok: false, error: result.reason || 'Reply requires manual review before retry' };
+                }
                 if (idempotencyKey)
                     this.markReplyCompleted(config, input.conversationId, idempotencyKey);
             }
@@ -152,7 +153,7 @@ export class ChannelRuntime {
             tokenConfigured: Boolean(secret?.token),
             canReply: Boolean(secret?.canReply),
             systemPrompt: typeof secret?.systemPrompt === 'string' ? secret.systemPrompt : '',
-            ...wechatRpaViewFields(secret, opts.includeSecret),
+            ...wechatRpaViewFields(secret),
             ...adapterStatus,
         };
     }
@@ -178,7 +179,7 @@ export class ChannelRuntime {
             tokenConfigured: Boolean(secret?.token),
             canReply: Boolean(secret?.canReply),
             systemPrompt: typeof secret?.systemPrompt === 'string' ? secret.systemPrompt : '',
-            ...wechatRpaViewFields(secret, opts.includeSecret),
+            ...wechatRpaViewFields(secret),
             ...adapterStatus,
         };
     }
@@ -332,6 +333,8 @@ export class ChannelRuntime {
         const groups = normalizeWeChatRpaGroups(input.groups);
         if (input.enabled && !groups.length)
             throw new Error('WeChat RPA 至少需要配置一个群');
+        if (input.enabled && groups.length > 1)
+            throw new Error('WeChat RPA 每个对话只能绑定一个群');
         const nextConfig = {
             id: input.id,
             type: 'wechat-rpa',
@@ -346,7 +349,7 @@ export class ChannelRuntime {
             secretRef: previous?.secretRef || `channel:${input.id}`,
         };
         const priorSecret = this.secrets.get(nextConfig.secretRef);
-        const source = input.source || (priorSecret?.source === 'macos-probe' || priorSecret?.source === 'fixture-jsonl' || priorSecret?.source === 'macos-flow' || priorSecret?.source === 'windows-visual-flow' ? priorSecret.source : defaultWeChatRpaSource());
+        const source = input.source || (priorSecret?.source === 'macos-probe' || priorSecret?.source === 'fixture-jsonl' || priorSecret?.source === 'macos-flow' || priorSecret?.source === 'windows-visual-flow' || priorSecret?.source === 'wechat-rpa-lab' ? priorSecret.source : defaultWeChatRpaSource());
         const configs = allConfigs
             .filter((channel) => channel.id !== nextConfig.id)
             .map((channel) => (channel.sessionId ?? channel.managerSessionId) === boundSessionId && channel.type === 'wechat-rpa'
@@ -354,8 +357,7 @@ export class ChannelRuntime {
             : channel);
         configs.push(nextConfig);
         this.configs.replaceAll(configs);
-        const cloudOcrMode = normalizeCloudOcrMode(input.cloudOcrMode ?? priorSecret?.cloudOcrMode);
-        const cloudOcrDefaults = cloudOcrMode === 'off' ? {} : defaultWeChatRpaCloudOcrConfig();
+        const cloudOcrMode = 'off';
         this.secrets.upsert(nextConfig.secretRef, {
             type: 'wechat-rpa',
             source,
@@ -367,9 +369,10 @@ export class ChannelRuntime {
             noRestore: input.noRestore ?? (priorSecret?.noRestore === undefined ? true : Boolean(priorSecret.noRestore)),
             downloadAttachments: input.downloadAttachments ?? (priorSecret?.downloadAttachments === undefined ? true : Boolean(priorSecret.downloadAttachments)),
             downloadAttachmentsDir: input.downloadAttachmentsDir?.trim() || stringOrUndefined(priorSecret?.downloadAttachmentsDir),
+            selfNickname: input.selfNickname?.trim() || stringOrUndefined(priorSecret?.selfNickname),
             flowScriptPath: input.flowScriptPath?.trim() || stringOrUndefined(priorSecret?.flowScriptPath),
-            cloudOcrUrl: input.cloudOcrUrl?.trim() || stringOrUndefined(priorSecret?.cloudOcrUrl) || cloudOcrDefaults.cloudOcrUrl,
-            cloudOcrToken: input.cloudOcrToken?.trim() || stringOrUndefined(priorSecret?.cloudOcrToken) || cloudOcrDefaults.cloudOcrToken,
+            cloudOcrUrl: '',
+            cloudOcrToken: '',
             cloudOcrMode,
             canReply: input.canReply ?? priorSecret?.canReply ?? false,
             systemPrompt: input.systemPrompt ?? (typeof priorSecret?.systemPrompt === 'string' ? priorSecret.systemPrompt : ''),
@@ -397,7 +400,7 @@ function isChannelSecretConfigured(config, secret) {
         return Boolean(secret.token || secret.botId || secret.secret);
     return Boolean(secret.token);
 }
-function wechatRpaViewFields(secret, includeSecret) {
+function wechatRpaViewFields(secret) {
     if (!secret || secret.type !== 'wechat-rpa')
         return {};
     return {
@@ -410,9 +413,7 @@ function wechatRpaViewFields(secret, includeSecret) {
         noRestore: secret.noRestore === undefined ? undefined : Boolean(secret.noRestore),
         downloadAttachments: secret.downloadAttachments === undefined ? true : Boolean(secret.downloadAttachments),
         downloadAttachmentsDir: typeof secret.downloadAttachmentsDir === 'string' ? secret.downloadAttachmentsDir : '',
-        cloudOcrUrl: typeof secret.cloudOcrUrl === 'string' ? secret.cloudOcrUrl : '',
-        cloudOcrToken: includeSecret && typeof secret.cloudOcrToken === 'string' ? secret.cloudOcrToken : '',
-        cloudOcrMode: typeof secret.cloudOcrMode === 'string' ? secret.cloudOcrMode : 'off',
+        selfNickname: typeof secret.selfNickname === 'string' ? secret.selfNickname : '',
     };
 }
 function wechatRpaStatusFields(secret) {
@@ -428,8 +429,7 @@ function wechatRpaStatusFields(secret) {
         noRestore: secret.noRestore === undefined ? null : Boolean(secret.noRestore),
         downloadAttachments: secret.downloadAttachments === undefined ? true : Boolean(secret.downloadAttachments),
         downloadAttachmentsDir: typeof secret.downloadAttachmentsDir === 'string' ? secret.downloadAttachmentsDir : null,
-        cloudOcrUrl: typeof secret.cloudOcrUrl === 'string' ? secret.cloudOcrUrl : null,
-        cloudOcrMode: typeof secret.cloudOcrMode === 'string' ? secret.cloudOcrMode : 'off',
+        selfNickname: typeof secret.selfNickname === 'string' ? secret.selfNickname : null,
     };
 }
 function normalizeWeChatRpaGroups(groups) {
@@ -444,24 +444,8 @@ function normalizeWeChatRpaGroups(groups) {
     }
     return result;
 }
-function normalizeCloudOcrMode(value) {
-    return value === 'fallback' || value === 'always' ? value : 'off';
-}
 function defaultWeChatRpaSource() {
-    return process.platform === 'win32' ? 'windows-visual-flow' : 'macos-flow';
-}
-function defaultWeChatRpaCloudOcrConfig() {
-    const config = loadConfig();
-    const serverUrl = (config.serverUrl || SERVERS.cn.url).replace(/\/+$/, '');
-    const token = config.machineToken?.trim() || config.accessToken?.trim();
-    return {
-        cloudOcrUrl: `${serverApiBaseUrl(serverUrl)}/integrations/wechat-rpa/ocr`,
-        ...(token ? { cloudOcrToken: token } : {}),
-    };
-}
-function serverApiBaseUrl(serverUrl) {
-    const normalized = serverUrl.replace(/\/+$/, '');
-    return normalized.endsWith('/api') ? normalized : `${normalized}/api`;
+    return process.platform === 'win32' ? 'windows-visual-flow' : 'wechat-rpa-lab';
 }
 export function planExternalReplySends(channelType, input) {
     const parts = splitExternalReplyText(input.text);

package/dist/src/channels/secret-registry.d.ts

@@ -6,7 +6,7 @@ type ChannelSecretRecord = {
     token?: string;
     canReply?: boolean;
     systemPrompt?: string;
-    source?: 'macos-probe' | 'macos-flow' | 'windows-visual-flow' | 'fixture-jsonl';
+    source?: 'macos-probe' | 'macos-flow' | 'windows-visual-flow' | 'wechat-rpa-lab' | 'fixture-jsonl';
     fixturePath?: string;
     pollIntervalMs?: number;
     groups?: Array<{
@@ -16,8 +16,10 @@ type ChannelSecretRecord = {
     noRestore?: boolean;
     downloadAttachments?: boolean;
     downloadAttachmentsDir?: string;
+    selfNickname?: string;
     idleSeconds?: number;
     recentLimit?: number;
+    readMode?: 'local-ocr' | 'hybrid-vlm';
     flowScriptPath?: string;
     cloudOcrUrl?: string;
     cloudOcrToken?: string;

package/dist/src/channels/wechat-rpa/macos-flow.d.ts

@@ -20,6 +20,8 @@ export type MacWeChatRpaFlowResult = {
     groupName: string;
     interrupted?: boolean;
     reason?: string;
+    rpaRunId?: string;
+    rpaTraceSummary?: string;
     screenshotPath?: string;
     recentMessages?: MacWeChatRpaFlowMessage[];
     newMessages?: MacWeChatRpaFlowMessage[];
@@ -40,6 +42,10 @@ export type MacWeChatRpaFlowMessage = {
     id?: string;
     text?: string;
     confidence?: number;
+    senderName?: string | null;
+    observedAt?: string | null;
+    timestampIso?: string | null;
+    isMentioned?: boolean | null;
     attachments?: MacWeChatRpaAttachment[];
 };
 export type MacWeChatRpaAttachment = {
@@ -68,11 +74,6 @@ export type MacWeChatRpaCloudOcrUsage = {
     totalTokens?: number;
 };
 export declare function runMacWeChatRpaFlow(options: MacWeChatRpaFlowOptions): Promise<MacWeChatRpaFlowResult>;
-export declare function buildCloudOcrImagePayload(screenshotPath: string, options: Pick<MacWeChatRpaFlowOptions, 'cloudOcrUrl' | 'cloudOcrToken'>): Promise<{
-    imageUrl: string;
-    imageObjectKey?: string;
-} | null>;
-export declare function cloudOcrImageUploadUrl(cloudOcrUrl?: string): string | null;
 export declare function selectCloudOcrRequest(result: Pick<MacWeChatRpaFlowResult, 'newMessages' | 'recentMessages' | 'screenshotPath' | 'postSendScreenshotPath' | 'sentReply' | 'sentAttachment'>, mode: 'off' | 'fallback' | 'always'): {
     screenshotPath: string;
     purpose: MacWeChatRpaCloudOcrPurpose;

package/dist/src/channels/wechat-rpa/macos-flow.js

@@ -57,6 +57,8 @@ export async function runMacWeChatRpaFlow(options) {
         const parsed = JSON.parse(stdout);
         if (parsed.interrupted)
             return parsed;
+        if (!parsed.ok && hasSendAttempt(parsed))
+            return parsed;
         if (!parsed.ok)
             throw new Error(parsed.error || 'WeChat RPA flow failed');
         return await maybeEnrichWithCloudOcr(parsed, options);
@@ -77,87 +79,12 @@ function parseFlowStdout(stdout) {
 function hasPartialSendProgress(result) {
     return Boolean(result.sentReplyObserved || result.sentAttachmentObserved);
 }
-async function maybeEnrichWithCloudOcr(result, options) {
-    const mode = options.cloudOcrMode ?? 'off';
-    const request = selectCloudOcrRequest(result, mode);
-    if (!request || !options.cloudOcrUrl || !options.cloudOcrToken)
-        return result;
-    const imagePayload = await buildCloudOcrImagePayload(request.screenshotPath, options);
-    if (!imagePayload)
-        return result;
-    const response = await fetch(options.cloudOcrUrl, {
-        method: 'POST',
-        headers: {
-            authorization: `Bearer ${options.cloudOcrToken}`,
-            'content-type': 'application/json',
-        },
-        body: JSON.stringify({
-            ...imagePayload,
-            mimeType: 'image/png',
-            conversationName: options.groupName,
-            purpose: request.purpose,
-            channelId: options.cloudOcrChannelId,
-        }),
-    });
-    if (!response.ok)
-        return result;
-    const payload = await response.json().catch(() => null);
-    const observations = Array.isArray(payload?.observations) ? payload.observations : [];
-    if (!observations.length)
-        return result;
-    const cloudMessages = messagesFromCloudOcrObservations(options.groupName, observations);
-    const recentLimit = Math.max(1, options.recentLimit || 5);
-    return {
-        ...result,
-        cloudOcrPurpose: request.purpose,
-        cloudOcrObservations: observations,
-        ...(payload?.requestId ? { cloudOcrRequestId: payload.requestId } : {}),
-        ...(payload?.imageHash ? { cloudOcrImageHash: payload.imageHash } : {}),
-        ...(payload?.usage ? { cloudOcrUsage: payload.usage } : {}),
-        newMessages: mergeCloudMessages(result.newMessages ?? [], cloudMessages),
-        recentMessages: mergeCloudMessages(result.recentMessages ?? [], cloudMessages).slice(-recentLimit),
-    };
-}
-export async function buildCloudOcrImagePayload(screenshotPath, options) {
-    const buffer = fs.readFileSync(screenshotPath);
-    const uploadUrl = cloudOcrImageUploadUrl(options.cloudOcrUrl);
-    if (uploadUrl && options.cloudOcrToken) {
-        try {
-            const response = await fetch(uploadUrl, {
-                method: 'POST',
-                headers: {
-                    authorization: `Bearer ${options.cloudOcrToken}`,
-                    'content-type': 'image/png',
-                },
-                body: buffer,
-            });
-            if (response.ok) {
-                const payload = await response.json().catch(() => null);
-                if (payload?.imageUrl) {
-                    return {
-                        imageUrl: payload.imageUrl,
-                        ...(payload.imageObjectKey ? { imageObjectKey: payload.imageObjectKey } : {}),
-                    };
-                }
-            }
-        }
-        catch {
-            // Cloud OCR is an optional fallback; if image upload is unavailable, keep the local result.
-        }
-    }
-    return null;
+function hasSendAttempt(result) {
+    return Boolean(result.sentReply || result.sentAttachment || result.sentReplyObserved || result.sentAttachmentObserved);
 }
-export function cloudOcrImageUploadUrl(cloudOcrUrl) {
-    if (!cloudOcrUrl)
-        return null;
-    try {
-        const url = new URL(cloudOcrUrl);
-        url.pathname = url.pathname.replace(/\/ocr\/?$/, '/ocr/image');
-        return url.toString();
-    }
-    catch {
-        return null;
-    }
+async function maybeEnrichWithCloudOcr(result, options) {
+    void options;
+    return result;
 }
 export function selectCloudOcrRequest(result, mode) {
     if (!shouldUseCloudOcr(result, mode))

package/dist/src/channels/wechat-rpa/normalizer.d.ts

@@ -4,6 +4,7 @@ export type WeChatRpaObservedMessage = {
     text: string;
     attachments?: WeChatRpaObservedAttachment[];
     observedAt?: string | null;
+    isMentioned?: boolean | null;
     rawId?: string | null;
 };
 export type WeChatRpaObservedAttachment = {
@@ -31,6 +32,7 @@ export type WeChatRpaNormalizedMessage = {
     text: string;
     attachments: WeChatRpaObservedAttachment[];
     receivedAt: string;
+    isMentioned: boolean;
     rawRef: string | null;
 };
 export declare class WeChatRpaDeduper {
@@ -41,5 +43,7 @@ export declare class WeChatRpaDeduper {
     snapshot(): string[];
     private trim;
 }
-export declare function normalizeWeChatRpaMessage(input: WeChatRpaObservedMessage): WeChatRpaNormalizedMessage | null;
+export declare function normalizeWeChatRpaMessage(input: WeChatRpaObservedMessage, options?: {
+    selfNicknames?: string[];
+}): WeChatRpaNormalizedMessage | null;
 export declare function weChatRpaConversationId(conversationName: string): string;