shennian 0.2.76 → 0.2.78

package/dist/scripts/wechat-rpa-download-candidates.mjs

@@ -24,6 +24,27 @@ export function selectDownloadedAttachment(before, after, startedAt, attachment)
     .sort((a, b) => b.score - a.score || b.mtimeMs - a.mtimeMs)[0] || null
 }
 
+export function selectCachedAttachment(candidates, attachment) {
+  const expectedName = normalizeReplyText(attachment?.name || '')
+  if (expectedName.length < 4) return null
+  const expectedExt = path.extname(String(attachment?.name || '')).toLowerCase()
+  const expectedHead = expectedName.slice(0, Math.min(expectedName.length, 24))
+  const matched = Array.from(candidates.values())
+    .filter((file) => isPlausibleDownloadedAttachment(file, attachment))
+    .filter((file) => {
+      const base = normalizeReplyText(path.basename(file.path))
+      if (base === expectedName) return true
+      return expectedHead.length >= 8 && base.includes(expectedHead)
+    })
+  if (!matched.length) return null
+  return matched
+    .map((file) => ({
+      ...file,
+      score: (path.extname(file.path).toLowerCase() === expectedExt ? 10 : 0) + file.mtimeMs / 1_000_000_000_000,
+    }))
+    .sort((a, b) => b.score - a.score || b.mtimeMs - a.mtimeMs)[0] || null
+}
+
 const INTERNAL_EXTENSIONS = new Set([
   '.db',
   '.ini',

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

@@ -151,6 +151,11 @@ export function findTitleConfirmation(observations, targetName) {
   return rows.find(row => normalizeConversationName(observationText(row)) === target) || null
 }
 
+export function isRetryableOcrError(status, body) {
+  if ([408, 429, 500, 502, 503, 504].includes(Number(status))) return true
+  return /invalid model response|timeout|temporar/i.test(String(body || ''))
+}
+
 export function pointFromObservation(capturePayload, observation, imageSize) {
   if (!capturePayload?.bounds) throw new Error('capture payload missing bounds')
   if (!observation?.box) throw new Error('observation missing box')
@@ -286,22 +291,33 @@ async function recognizeScreenshot(capture, options, purpose) {
   const token = options.token || process.env.WECHAT_RPA_OCR_TOKEN || ''
   if (!token) throw new Error('Missing OCR bearer token. Pass --token or WECHAT_RPA_OCR_TOKEN.')
   const imageBase64 = fs.readFileSync(capture.file).toString('base64')
-  const response = await fetch(options.ocrUrl, {
-    method: 'POST',
-    headers: {
-      'Content-Type': 'application/json',
-      Authorization: `Bearer ${token}`,
-    },
-    body: JSON.stringify({
-      imageBase64,
-      mimeType: 'image/png',
-      purpose,
-      conversationName: options.group,
-      channelId: options.channelId,
-    }),
-    signal: AbortSignal.timeout(options.ocrTimeoutMs),
-  })
-  if (!response.ok) throw new Error(`OCR request failed: ${response.status} ${await response.text()}`)
+  const maxAttempts = Math.max(1, Number(options.ocrRetries || 2) + 1)
+  let lastError = null
+  let response = null
+  for (let attempt = 1; attempt <= maxAttempts; attempt++) {
+    response = await fetch(options.ocrUrl, {
+      method: 'POST',
+      headers: {
+        'Content-Type': 'application/json',
+        Authorization: `Bearer ${token}`,
+      },
+      body: JSON.stringify({
+        imageBase64,
+        mimeType: 'image/png',
+        purpose,
+        conversationName: options.group,
+        channelId: options.channelId,
+      }),
+      signal: AbortSignal.timeout(options.ocrTimeoutMs),
+    })
+    if (response.ok) break
+    const body = await response.text()
+    lastError = `OCR request failed: ${response.status} ${body}`
+    if (attempt >= maxAttempts || !isRetryableOcrError(response.status, body)) {
+      throw new Error(lastError)
+    }
+    await sleep(400 * attempt)
+  }
   const json = await response.json()
   const debugPath = capture.file.replace(/\.png$/i, `-${purpose}.ocr.json`)
   fs.writeFileSync(debugPath, `${JSON.stringify(json, null, 2)}\n`)
@@ -380,6 +396,13 @@ function attachmentTypeFromExt(ext) {
   return 'file'
 }
 
+function postPasteDelayMs(file) {
+  const type = classifyOutboundFile(file).type
+  if (type === 'image') return 8_000
+  if (type === 'video') return 15_000
+  return 2_000
+}
+
 const IMAGE_EXTENSIONS = new Set(['.apng', '.avif', '.bmp', '.gif', '.heic', '.heif', '.jpeg', '.jpg', '.png', '.tif', '.tiff', '.webp'])
 const VIDEO_EXTENSIONS = new Set(['.3g2', '.3gp', '.avi', '.m4v', '.mkv', '.mov', '.mp4', '.mpeg', '.mpg', '.webm', '.wmv'])
 
@@ -523,6 +546,14 @@ export function missingConfirmedFiles(observations, files) {
 async function openConversationBySearch(options, artifacts) {
   const initial = await capture('window', options, 'window-before-search')
   artifacts.push(initial.file)
+  try {
+    const initialTitleOcr = await recognizeScreenshot(initial, options, 'title-confirmation')
+    if (findTitleConfirmation(initialTitleOcr.observations, options.group)) {
+      return initial
+    }
+  } catch (error) {
+    process.stderr.write(`Initial title OCR was skipped: ${error instanceof Error ? error.message : String(error)}\n`)
+  }
 
   await click(geometryPoint(initial.payload, 'search'), options)
   await sleep(180)
@@ -583,9 +614,9 @@ export async function runVisualFlow(input) {
     await click(geometryPoint(opened.payload, 'input'), options)
     await pasteText(options.replyText, options)
     await sleep(250)
-    await click(geometryPoint(opened.payload, 'send'), options)
+    await press('{ENTER}', options)
     await sleep(900)
-    const confirmCapture = await capture('messages', options, 'after-text-send')
+    const confirmCapture = await capture('window', options, 'after-text-send')
     artifacts.push(confirmCapture.file)
     const confirmOcr = await recognizeScreenshot(confirmCapture, options, 'send-confirmation')
     if (!observationsContainText(confirmOcr.observations, options.replyText)) {
@@ -595,25 +626,33 @@ export async function runVisualFlow(input) {
   }
 
   if (!options.dryRun && options.files?.length) {
-    await click(geometryPoint(opened.payload, 'input'), options)
-    await pasteFiles(options.files, options)
-    await sleep(850)
-    const pendingCapture = await capture('input', options, 'pending-files')
-    artifacts.push(pendingCapture.file)
-    await click(geometryPoint(opened.payload, 'send'), options)
-    await sleep(1200)
-    const fileConfirmCapture = await capture('messages', options, 'after-file-send')
-    artifacts.push(fileConfirmCapture.file)
-    const fileConfirmOcr = await recognizeScreenshot(fileConfirmCapture, options, 'send-confirmation')
-    const missing = missingConfirmedFiles(fileConfirmOcr.observations, options.files)
-    if (missing.length > 0) {
-      throw new Error(`Sent files were not confirmed by OCR: ${missing.map(basenameForAnyPlatform).join(', ')}. See ${fileConfirmCapture.file}`)
+    const confirmedFiles = []
+    const sentAttachments = []
+    const confirmationObservations = []
+    for (const [index, file] of options.files.entries()) {
+      await click(geometryPoint(opened.payload, 'input'), options)
+      await pasteFiles([file], options)
+      await sleep(850)
+      const pendingCapture = await capture('input', options, `pending-file-${index + 1}`)
+      artifacts.push(pendingCapture.file)
+      await click(geometryPoint(opened.payload, 'send'), options)
+      await sleep(postPasteDelayMs(file))
+      const fileConfirmCapture = await capture('window', options, `after-file-${index + 1}-send`)
+      artifacts.push(fileConfirmCapture.file)
+      const fileConfirmOcr = await recognizeScreenshot(fileConfirmCapture, options, 'send-confirmation')
+      const missing = missingConfirmedFiles(fileConfirmOcr.observations, [file])
+      if (missing.length > 0) {
+        throw new Error(`Sent file was not confirmed by OCR: ${basenameForAnyPlatform(file)}. See ${fileConfirmCapture.file}`)
+      }
+      confirmedFiles.push(file)
+      sentAttachments.push(classifyOutboundFile(file))
+      confirmationObservations.push(...summarizeObservations(fileConfirmOcr.observations, options.recentLimit))
     }
     sent.push({
       type: 'files',
-      files: options.files,
-      attachments: options.files.map(classifyOutboundFile),
-      observations: summarizeObservations(fileConfirmOcr.observations, options.recentLimit),
+      files: confirmedFiles,
+      attachments: sentAttachments,
+      observations: confirmationObservations.slice(-options.recentLimit),
     })
   }
 

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

@@ -31,7 +31,6 @@ function printHelp() {
   node scripts/wechat-rpa-win.mjs send-text --group <name> --text <message>
   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 prepare-uia
   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 paste-text --text "hello"
@@ -46,8 +45,8 @@ Environment:
 
 Diagnostic note:
   probe does not raise WeChat by default. Use probe --raise only for diagnostics.
-  restart-wechat-for-uia is intentionally hidden from normal help because it closes
-  WeChat. It is guarded in the native helper and must never be used by product flows.`)
+  prepare-uia and restart-wechat-for-uia are intentionally hidden from normal help.
+  They can start accessibility diagnostics or close WeChat, and must never be used by product flows.`)
 }
 
 async function main() {
@@ -64,6 +63,14 @@ async function main() {
   const allowIndex = argv.indexOf('--allow-non-windows')
   if (allowIndex >= 0) argv.splice(allowIndex, 1)
 
+  const command = argv[0]
+  if (command === 'prepare-uia' && process.env.SHENNIAN_WECHAT_RPA_ALLOW_UIA_DIAGNOSTICS !== '1') {
+    throw new Error(
+      'prepare-uia is diagnostic-only because it can start Windows Narrator. ' +
+        'Set SHENNIAN_WECHAT_RPA_ALLOW_UIA_DIAGNOSTICS=1 in a local diagnostic shell to use it.',
+    )
+  }
+
   const helper = takeOption(argv, '--helper') ?? process.env.WECHAT_RPA_WIN_HELPER ?? defaultHelper
   if (!fs.existsSync(helper)) {
     throw new Error(

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

@@ -36,7 +36,10 @@ export type WeChatRpaNormalizedMessage = {
 export declare class WeChatRpaDeduper {
     private seen;
     private queue;
+    constructor(initialMessageIds?: string[]);
     accept(messageId: string): boolean;
+    snapshot(): string[];
+    private trim;
 }
 export declare function normalizeWeChatRpaMessage(input: WeChatRpaObservedMessage): WeChatRpaNormalizedMessage | null;
 export declare function weChatRpaConversationId(conversationName: string): string;

package/dist/src/channels/wechat-rpa/normalizer.js

@@ -5,17 +5,32 @@ const MAX_DEDUP_KEYS = 500;
 export class WeChatRpaDeduper {
     seen = new Set();
     queue = [];
+    constructor(initialMessageIds = []) {
+        for (const id of initialMessageIds) {
+            if (!id || this.seen.has(id))
+                continue;
+            this.seen.add(id);
+            this.queue.push(id);
+        }
+        this.trim();
+    }
     accept(messageId) {
         if (this.seen.has(messageId))
             return false;
         this.seen.add(messageId);
         this.queue.push(messageId);
+        this.trim();
+        return true;
+    }
+    snapshot() {
+        return [...this.queue];
+    }
+    trim() {
         while (this.queue.length > MAX_DEDUP_KEYS) {
             const old = this.queue.shift();
             if (old)
                 this.seen.delete(old);
         }
-        return true;
     }
 }
 export function normalizeWeChatRpaMessage(input) {

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

@@ -30,6 +30,7 @@ export class WeChatRpaChannelAdapter {
         conn.stopped = false;
         conn.config = config;
         hydratePendingReplyState(conn, config);
+        hydrateMessageState(conn, config);
         this.seedConfiguredConversations(conn, secret);
         if (conn.timer)
             return;
@@ -56,6 +57,7 @@ export class WeChatRpaChannelAdapter {
         conn.stopped = false;
         conn.config = config;
         hydratePendingReplyState(conn, config);
+        hydrateMessageState(conn, config);
         this.seedConfiguredConversations(conn, secret);
         return this.enqueueOperation(conn, () => this.pollOnce(conn, this.readSecret(conn.config)));
     }
@@ -69,6 +71,7 @@ export class WeChatRpaChannelAdapter {
         const conn = this.ensureConnection(config);
         conn.config = config;
         hydratePendingReplyState(conn, config);
+        hydrateMessageState(conn, config);
         this.seedConfiguredConversations(conn, secret);
         const conversationName = this.resolveConversationName(config, secret, reply.conversationId);
         if (!conversationName)
@@ -201,6 +204,7 @@ export class WeChatRpaChannelAdapter {
                 pendingReplies: new Map(),
                 completedPendingReplyKeys: new Set(),
                 pendingStatePath: undefined,
+                messageStatePath: undefined,
                 operation: Promise.resolve(),
                 runtimeState: 'idle_waiting',
                 consecutiveInterruptions: 0,
@@ -248,6 +252,7 @@ export class WeChatRpaChannelAdapter {
                 const message = normalizeWeChatRpaMessage(item);
                 if (!message || !conn.deduper.accept(message.messageId))
                     continue;
+                persistMessageState(conn);
                 conn.conversations.set(message.conversationId, message.conversationName);
                 conn.lastMessageAt = message.receivedAt;
                 const event = {
@@ -568,6 +573,38 @@ function pendingReplyStatePath(config) {
     const id = crypto.createHash('sha256').update(config.id).digest('hex').slice(0, 16);
     return path.join(config.workDir, '.shennian', 'wechat-rpa-pending-replies', `${id}.json`);
 }
+function hydrateMessageState(conn, config) {
+    const filePath = messageStatePath(config);
+    if (conn.messageStatePath === filePath)
+        return;
+    conn.messageStatePath = filePath;
+    const store = readMessageSeenStore(filePath);
+    conn.deduper = new WeChatRpaDeduper((store.messageIds ?? []).filter((id) => typeof id === 'string' && id.length > 0));
+}
+function persistMessageState(conn) {
+    if (!conn.messageStatePath)
+        return;
+    try {
+        fs.mkdirSync(path.dirname(conn.messageStatePath), { recursive: true });
+        fs.writeFileSync(conn.messageStatePath, JSON.stringify({ version: 1, messageIds: conn.deduper.snapshot() }, null, 2));
+    }
+    catch {
+        // Best-effort only; in-memory dedupe still protects the current daemon.
+    }
+}
+function readMessageSeenStore(filePath) {
+    try {
+        const parsed = JSON.parse(fs.readFileSync(filePath, 'utf8'));
+        return parsed && parsed.version === 1 ? parsed : { version: 1, messageIds: [] };
+    }
+    catch {
+        return { version: 1, messageIds: [] };
+    }
+}
+function messageStatePath(config) {
+    const id = crypto.createHash('sha256').update(config.id).digest('hex').slice(0, 16);
+    return path.join(config.workDir, '.shennian', 'wechat-rpa-seen-messages', `${id}.json`);
+}
 function noteInterruption(conn, flow, reason) {
     conn.lastInterruptedAt = new Date().toISOString();
     conn.lastError = null;

package/dist/src/upgrade/engine.d.ts

@@ -54,6 +54,7 @@ export declare function fetchLatestVersion(): Promise<string>;
  *   'major' — major differs
  */
 export declare function compareVersions(current: string, latest: string): 'none' | 'patch' | 'minor' | 'major';
+export declare function copyPackageRuntimeFiles(src: string, dest: string): void;
 export declare function readUpgradeAttempt(): UpgradeAttempt | null;
 export declare function writeUpgradeAttempt(attempt: UpgradeAttempt): void;
 export declare function clearUpgradeAttempt(): void;

package/dist/src/upgrade/engine.js

@@ -10,6 +10,7 @@ const BACKUP_DIR = resolveShennianPath('backup');
 const UPGRADE_ATTEMPT_FILE = resolveShennianPath('upgrade-attempt.json');
 const MAX_CRASH_COUNT = 3;
 const BACKUP_TTL_DAYS = 7;
+const NPM_REGISTRY_FALLBACK = 'https://registry.npmjs.org';
 const RETRY_DELAYS_MS = [
     5 * 60_000,
     30 * 60_000,
@@ -86,8 +87,9 @@ function getGlobalBinScript() {
 function backupVersion(version) {
     const pkgDir = getGlobalPkgDir();
     const dest = path.join(BACKUP_DIR, version);
+    fs.rmSync(dest, { recursive: true, force: true });
     fs.mkdirSync(dest, { recursive: true });
-    fs.cpSync(pkgDir, dest, { recursive: true });
+    copyPackageRuntimeFiles(pkgDir, dest);
 }
 function restoreVersion(version) {
     const src = path.join(BACKUP_DIR, version);
@@ -96,6 +98,14 @@ function restoreVersion(version) {
     const pkgDir = getGlobalPkgDir();
     fs.cpSync(src, pkgDir, { recursive: true, force: true });
 }
+export function copyPackageRuntimeFiles(src, dest) {
+    for (const entry of ['package.json', 'README.md', 'dist']) {
+        const from = path.join(src, entry);
+        if (!fs.existsSync(from))
+            continue;
+        fs.cpSync(from, path.join(dest, entry), { recursive: true, force: true });
+    }
+}
 function cleanOldBackups() {
     try {
         if (!fs.existsSync(BACKUP_DIR))
@@ -263,7 +273,7 @@ export async function performUpgrade(targetVersion, onProgress, opts = {}) {
     // Step 3: npm install new version
     onProgress({ step: 'installing', version: targetVersion });
     try {
-        await exec(`npm install -g shennian@${targetVersion}`, { timeout: 120_000, windowsHide: true });
+        await installShennianVersion(targetVersion);
     }
     catch (err) {
         // Restore backup and abort
@@ -309,6 +319,24 @@ export async function performUpgrade(targetVersion, onProgress, opts = {}) {
     onProgress({ step: 'restarting', from: currentVersion, to: targetVersion });
     return { ok: true, from: currentVersion, to: targetVersion };
 }
+async function installShennianVersion(targetVersion) {
+    try {
+        await exec(`npm install -g shennian@${targetVersion}`, { timeout: 120_000, windowsHide: true });
+    }
+    catch (error) {
+        const firstMessage = error instanceof Error ? error.message : String(error);
+        try {
+            await exec(`npm install -g shennian@${targetVersion} --registry ${NPM_REGISTRY_FALLBACK}`, {
+                timeout: 120_000,
+                windowsHide: true,
+            });
+        }
+        catch (fallbackError) {
+            const secondMessage = fallbackError instanceof Error ? fallbackError.message : String(fallbackError);
+            throw new Error(`${firstMessage}\n--- npmjs fallback ---\n${secondMessage}`);
+        }
+    }
+}
 export async function checkForUpdate(currentVersion) {
     const current = currentVersion ?? getCurrentVersion();
     const latest = await fetchLatestVersion();

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "shennian",
-  "version": "0.2.76",
+  "version": "0.2.78",
   "description": "Shennian — AI Agent Control Plane CLI",
   "type": "module",
   "bin": {
@@ -33,16 +33,10 @@
   "engines": {
     "node": ">=18"
   },
-  "scripts": {
-    "build": "tsc && node scripts/copy-wechat-rpa-assets.mjs && node -e \"require('node:fs').chmodSync('dist/bin/shennian.js', 0o755)\"",
-    "build:publish": "node -e \"const fs=require('node:fs'); fs.rmSync('dist', { recursive: true, force: true }); fs.rmSync('.tsbuildinfo.publish', { force: true })\" && tsc -p tsconfig.publish.json && node scripts/copy-wechat-rpa-assets.mjs && node -e \"require('node:fs').chmodSync('dist/bin/shennian.js', 0o755)\"",
-    "dev": "tsc --watch",
-    "prepublishOnly": "pnpm build:publish"
-  },
   "dependencies": {
     "@mariozechner/pi-agent-core": "^0.64.0",
     "@sinclair/typebox": "^0.34.49",
-    "@shennian/wire": "workspace:*",
+    "@shennian/wire": "^0.1.5",
     "chalk": "^5.4.1",
     "commander": "^13.1.0",
     "qrcode-terminal": "^0.12.0",
@@ -54,5 +48,10 @@
     "@types/ws": "^8.18.1",
     "tsx": "^4.19.4",
     "typescript": "^5.9.3"
+  },
+  "scripts": {
+    "build": "tsc && node scripts/copy-wechat-rpa-assets.mjs && node -e \"require('node:fs').chmodSync('dist/bin/shennian.js', 0o755)\"",
+    "build:publish": "node -e \"const fs=require('node:fs'); fs.rmSync('dist', { recursive: true, force: true }); fs.rmSync('.tsbuildinfo.publish', { force: true })\" && tsc -p tsconfig.publish.json && node scripts/copy-wechat-rpa-assets.mjs && node -e \"require('node:fs').chmodSync('dist/bin/shennian.js', 0o755)\"",
+    "dev": "tsc --watch"
   }
-}
+}