OneKeyHQ · 0xchloe996 · Jul 17, 2026 · Jul 16, 2026 · Jul 17, 2026
diff --git a/.gitignore b/.gitignore
@@ -74,4 +74,6 @@ scripts/.venv/
 # Mnemonic secrets (use mnemonics.example.json as template)
 mnemonics.local.json
 
-./claude/worktrees/*
+./claude/worktrees/*
+# Python bytecode
+scripts/__pycache__/
diff --git a/README.md b/README.md
@@ -106,9 +106,8 @@ yarn setup:ocr
 # Download OCR models
 python3 - <<'PY'
 from huggingface_hub import snapshot_download
-snapshot_download('PaddlePaddle/en_PP-OCRv5_mobile_rec', local_dir='models/paddle_ocr/en_PP-OCRv5_mobile_rec')
-snapshot_download('PaddlePaddle/PP-OCRv5_mobile_det', local_dir='models/paddle_ocr/PP-OCRv5_mobile_det')
-snapshot_download('PaddlePaddle/PP-OCRv5_mobile_rec', local_dir='models/paddle_ocr/PP-OCRv5_mobile_rec')
+snapshot_download('PaddlePaddle/PP-OCRv6_medium_rec_onnx', local_dir='models/paddle_ocr/PP-OCRv6_medium_rec_onnx')
+snapshot_download('PaddlePaddle/PP-OCRv6_medium_det_onnx', local_dir='models/paddle_ocr/PP-OCRv6_medium_det_onnx')
 print('done')
 PY
 
@@ -131,8 +130,8 @@ yarn build:linux      # Linux
 
 QA Auto Hardware uses a dual-path OCR pipeline for hardware wallet mnemonic capture:
 
-- **Mnemonic display page & confirmation option area**: `en_PP-OCRv5_mobile_rec`
-- **Confirmation page question number area (`#N`)**: `PP-OCRv5_mobile_det` + `PP-OCRv5_mobile_rec`
+- **Mnemonic display page & confirmation option area**: `PP-OCRv6_medium_rec_onnx`
+- **Confirmation page question number area (`#N`)**: `PP-OCRv6_medium_det_onnx` + `PP-OCRv6_medium_rec_onnx`
 
 **Typical workflow:**
 
@@ -145,9 +144,16 @@ QA Auto Hardware uses a dual-path OCR pipeline for hardware wallet mnemonic capt
 
 | Model | Environment Variable Override |
 |-------|-------------------------------|
-| `en_PP-OCRv5_mobile_rec` | `QA_AUTO_HW_OCR_MODEL_DIR` |
-| `PP-OCRv5_mobile_det` | `QA_AUTO_HW_OCR_DET_MODEL_DIR` |
-| `PP-OCRv5_mobile_rec` | `QA_AUTO_HW_OCR_MULTI_REC_MODEL_DIR` |
+| `PP-OCRv6_medium_rec_onnx` | `QA_AUTO_HW_OCR_MODEL_DIR` |
+| `PP-OCRv6_medium_det_onnx` | `QA_AUTO_HW_OCR_DET_MODEL_DIR` |
+
+To try the smaller PP-OCRv6 small models:
+
+```bash
+QA_AUTO_HW_OCR_MODEL_DIR=models/paddle_ocr/PP-OCRv6_small_rec_onnx \
+QA_AUTO_HW_OCR_DET_MODEL_DIR=models/paddle_ocr/PP-OCRv6_small_det_onnx \
+yarn electron:dev
+```
 
 Optional tuning: `QA_AUTO_HW_OCR_MAX_IMAGE_SIDE` (default 1280), `QA_AUTO_HW_OCR_CPU_THREADS` (default 4)
 

diff --git a/electron/main.ts b/electron/main.ts
@@ -666,7 +666,7 @@ ipcMain.handle(
 );
 
 /**
- * IPC handler: Runs en_PP-OCRv5_mobile_rec inference in a Python subprocess.
+ * IPC handler: Runs PP-OCRv6 small OCR inference in a Python subprocess.
  */
 ipcMain.handle(
   'paddleocr-en-recognize',
@@ -677,6 +677,7 @@ ipcMain.handle(
       layoutHint?: 'mnemonic' | 'verify-options' | 'verify-number' | 'generic';
       expectedWordCount?: number;
       recVariantCount?: number;
+      wordlistHint?: 'bip39' | 'slip39';
     }
   ) => runPaddleOcrEn(payload)
 );

diff --git a/electron/mcp/sequences.ts b/electron/mcp/sequences.ts
@@ -316,12 +316,13 @@ const ALL_PAGE_ACTIONS: PageAction[] = [
     name: '输入PIN码1111',
     group: '初始设置',
     steps: [
-      { label: '输入PIN码1', x: 25, y: 50, depth: 12 },
+      // Wait for the PIN pad to finish rendering before the first tap.
+      { label: '输入PIN码1', x: 25, y: 50, depth: 12, delayBefore: 500 },
       { label: '输入PIN码2', x: 25, y: 50, depth: 12 },
       { label: '输入PIN码3', x: 25, y: 50, depth: 12 },
       { label: '输入PIN码4', x: 25, y: 50, depth: 12 },
       { label: '点击确认', x: 55, y: 85, depth: 12 },
-      { label: '再次确认PIN码1', x: 25, y: 50, depth: 12 },
+      { label: '再次确认PIN码1', x: 25, y: 50, depth: 12, delayBefore: 500 },
       { label: '再次确认PIN码2', x: 25, y: 50, depth: 12 },
       { label: '再次确认PIN码3', x: 25, y: 50, depth: 12 },
       { label: '再次确认PIN码4', x: 25, y: 50, depth: 12 },

diff --git a/electron/mcp/state.ts b/electron/mcp/state.ts
@@ -70,6 +70,7 @@ export interface MnemonicOcrRequest {
   allowPartial?: boolean;
   requireBip39?: boolean;
   sceneConfig?: SequenceOcrSceneConfig;
+  deviceTestSetId?: 'pro' | 'pro2';
 }
 
 export interface MnemonicStoreMetadata {

diff --git a/electron/mcp/tools/executeSequence.ts b/electron/mcp/tools/executeSequence.ts
@@ -28,6 +28,7 @@ import {
   getPageAction,
   getSequence,
   normalizeDeviceTestSetId,
+  type DeviceTestSetId,
 } from '../sequenceSets';
 import { resolvePageActionStepsForDevice } from '../sequenceSetResolver';
 import { saveCaptureToDownloads } from '../../saveCapture';
@@ -90,6 +91,7 @@ function requiresPostSequenceCooldown(sequence: { category: string; actions: str
 async function executeStep(
   step: AutoStep,
   resourceHandle: number,
+  deviceTestSetId: DeviceTestSetId,
   httpRequest: (url: string) => Promise<string>
 ): Promise<void> {
   if ((step.delayBefore ?? 0) > 0) {
@@ -114,7 +116,10 @@ async function executeStep(
     // Give the device screen time to finish inertial scrolling before capture.
     await delay(1600);
 
-    const ocrResult = await runMnemonicOcr(ocrCaptureConfig);
+    const ocrResult = await runMnemonicOcr({
+      ...ocrCaptureConfig,
+      deviceTestSetId,
+    });
     if (!ocrResult) {
       throw new Error('Renderer did not return mnemonic OCR result');
     }
@@ -327,7 +332,7 @@ export async function executeExecuteSequence(
           };
         }
 
-        await executeStep(step, state.resourceHandle, httpRequest);
+        await executeStep(step, state.resourceHandle, deviceTestSetId, httpRequest);
         stepsCompleted++;
         console.log(`[execute-sequence] Step ${stepsCompleted}/${totalSteps}: ${step.label}`);
       }

diff --git a/electron/paddleOcrEn.ts b/electron/paddleOcrEn.ts
@@ -9,12 +9,14 @@ export interface PaddleOcrEnRequest {
   layoutHint?: 'mnemonic' | 'verify-options' | 'verify-number' | 'generic';
   expectedWordCount?: number;
   recVariantCount?: number;
+  /** Closed wordlist of the current flow (BIP39 vs SLIP39 — never mixed). */
+  wordlistHint?: 'bip39' | 'slip39';
 }
 
 export interface PaddleOcrEnResult {
   text: string;
   confidence: number;
-  backend: 'en_PP-OCRv5_mobile_rec';
+  backend: 'PP-OCRv6_medium_rec_onnx';
   elapsedMs: number;
 }
 
@@ -34,26 +36,32 @@ const HEALTH_CACHE_TTL_MS = 30 * 1000;
 const REQUIRED_PYTHON_MODULES = [
   'paddleocr',
   'paddle',
+  'onnxruntime',
   'cv2',
   'PIL',
   'yaml',
   'huggingface_hub',
 ] as const;
 const REQUIRED_MODEL_CONFIGS = [
   {
-    name: 'en_PP-OCRv5_mobile_rec',
-    envKeys: ['QA_AUTO_HW_OCR_MODEL_DIR', 'QA_AUTO_HW_EN_OCR_MODEL_DIR'],
+    name: 'PP-OCRv6_medium_rec_onnx',
+    envKeys: [
+      'QA_AUTO_HW_OCR_MODEL_DIR',
+      'QA_AUTO_HW_EN_OCR_MODEL_DIR',
+      'QA_AUTO_HW_OCR_MULTI_REC_MODEL_DIR',
+    ],
   },
   {
-    name: 'PP-OCRv5_mobile_rec',
-    envKeys: ['QA_AUTO_HW_OCR_MULTI_REC_MODEL_DIR'],
+    name: 'PP-OCRv6_medium_det_onnx',
+    envKeys: ['QA_AUTO_HW_OCR_DET_MODEL_DIR'],
   },
+] as const;
+const REQUIRED_MODEL_FILE_CONFIGS = [
   {
-    name: 'PP-OCRv5_mobile_det',
-    envKeys: ['QA_AUTO_HW_OCR_DET_MODEL_DIR'],
+    suffixes: ['inference.json', 'inference.onnx', 'inference.yml'],
   },
 ] as const;
-const REQUIRED_MODEL_FILES = ['inference.json', 'inference.pdiparams', 'inference.yml'] as const;
+const REQUIRED_MODEL_FILES = REQUIRED_MODEL_FILE_CONFIGS[0].suffixes;
 
 let cachedHealthStatus: { expiresAt: number; value: PaddleOcrEnHealthStatus } | null = null;
 
@@ -271,14 +279,14 @@ class PaddleOcrEnDaemon {
     await this.ensureStarted();
 
     if (!this.child) {
-      throw new Error('PP-OCRv5 rec daemon is not running');
+      throw new Error('PP-OCRv6 rec daemon is not running');
     }
 
     const id = this.nextId++;
     return new Promise((resolve, reject) => {
       const timer = setTimeout(() => {
         this.pending.delete(id);
-        reject(new Error('PP-OCRv5 rec request timed out after 2 minutes'));
+        reject(new Error('PP-OCRv6 rec request timed out after 2 minutes'));
       }, 2 * 60 * 1000);
 
       this.pending.set(id, { resolve, reject, timer });
@@ -294,7 +302,7 @@ class PaddleOcrEnDaemon {
     }
     this.startPromise = null;
     this.stdoutBuffer = '';
-    this.rejectAllPending(new Error('PP-OCRv5 rec daemon stopped'));
+    this.rejectAllPending(new Error('PP-OCRv6 rec daemon stopped'));
   }
 
   private async ensureStarted(): Promise<void> {
@@ -335,7 +343,7 @@ class PaddleOcrEnDaemon {
 
       const bootTimeout = setTimeout(() => {
         child.kill('SIGKILL');
-        reject(new Error('Timed out while starting PP-OCRv5 rec daemon'));
+        reject(new Error('Timed out while starting PP-OCRv6 rec daemon'));
       }, 45 * 1000);
 
       child.stdout.on('data', (chunk: Buffer) => {
@@ -359,14 +367,14 @@ class PaddleOcrEnDaemon {
       child.on('error', (err) => {
         clearTimeout(bootTimeout);
         this.child = null;
-        reject(new Error(`Failed to start PP-OCRv5 rec daemon: ${err.message}`));
+        reject(new Error(`Failed to start PP-OCRv6 rec daemon: ${err.message}`));
       });
 
       child.on('close', (code) => {
         clearTimeout(bootTimeout);
         this.child = null;
         this.startPromise = null;
-        const reason = `PP-OCRv5 rec daemon exited with code ${code}. stderr: ${this.stderrBuffer.trim() || '(empty)'}`;
+        const reason = `PP-OCRv6 rec daemon exited with code ${code}. stderr: ${this.stderrBuffer.trim() || '(empty)'}`;
         this.rejectAllPending(new Error(reason));
       });
     }).finally(() => {
@@ -391,6 +399,11 @@ class PaddleOcrEnDaemon {
       }
 
       if (parsed.type === 'ready') {
+        console.log(
+          `[paddleOcrEn] OCR server ready (script version: ${
+            typeof parsed.scriptVersion === 'string' ? parsed.scriptVersion : 'pre-versioning'
+          })`
+        );
         options?.onReady?.();
         continue;
       }
@@ -407,15 +420,15 @@ class PaddleOcrEnDaemon {
         const errorText =
           typeof parsed.error === 'string'
             ? parsed.error
-            : 'Unknown PP-OCRv5 rec daemon error';
+            : 'Unknown PP-OCRv6 rec daemon error';
         pending.reject(new Error(errorText));
         continue;
       }
 
       pending.resolve({
         text: typeof parsed.text === 'string' ? parsed.text : '',
         confidence: typeof parsed.confidence === 'number' ? parsed.confidence : 0,
-        backend: 'en_PP-OCRv5_mobile_rec',
+        backend: 'PP-OCRv6_medium_rec_onnx',
         elapsedMs: typeof parsed.elapsedMs === 'number' ? parsed.elapsedMs : 0,
       });
     }

diff --git a/electron/preload.ts b/electron/preload.ts
@@ -67,7 +67,7 @@ interface VerifyOcrPayload {
 interface PaddleOcrEnPayload {
   text: string;
   confidence: number;
-  backend: 'en_PP-OCRv5_mobile_rec';
+  backend: 'PP-OCRv6_medium_rec_onnx';
   elapsedMs: number;
 }
 
@@ -88,6 +88,12 @@ interface ResolvedSequenceStepPayload {
     mergeWithStored?: boolean;
     allowPartial?: boolean;
     requireBip39?: boolean;
+    sceneConfig?: {
+      roi: { x: number; y: number; width: number; height: number };
+      scale?: number;
+      useNearestNeighbor?: boolean;
+    };
+    deviceTestSetId?: 'pro' | 'pro2';
   };
   ocrVerify?: {
     options: { x: number; y: number; depth: number }[];
@@ -237,13 +243,15 @@ contextBridge.exposeInMainWorld('electronAPI', {
     imageDataUrl: string,
     layoutHint?: 'mnemonic' | 'verify-options' | 'verify-number' | 'generic',
     expectedWordCount?: number,
-    recVariantCount?: number
+    recVariantCount?: number,
+    wordlistHint?: 'bip39' | 'slip39'
   ) =>
     ipcRenderer.invoke('paddleocr-en-recognize', {
       imageDataUrl,
       layoutHint,
       expectedWordCount,
       recVariantCount,
+      wordlistHint,
     }) as Promise<PaddleOcrEnPayload>,
 
   // MCP Log: Receive log entries from main process
@@ -327,7 +335,8 @@ declare global {
         imageDataUrl: string,
         layoutHint?: 'mnemonic' | 'verify-options' | 'verify-number' | 'generic',
         expectedWordCount?: number,
-        recVariantCount?: number
+        recVariantCount?: number,
+        wordlistHint?: 'bip39' | 'slip39'
       ) => Promise<PaddleOcrEnPayload>;
       onMcpServerReady: (callback: (info: { port: number }) => void) => void;
       // MCP Logs

diff --git a/electron/saveCapture.ts b/electron/saveCapture.ts
@@ -30,11 +30,20 @@ function extractBase64(dataUrlOrBase64: string): string {
   return dataUrlOrBase64;
 }
 
+/**
+ * Picks the file extension matching the data URL mime type (default jpg).
+ */
+function extensionForDataUrl(dataUrlOrBase64: string): string {
+  const match = dataUrlOrBase64.match(/^data:image\/(png|jpeg|webp)/);
+  if (!match) return 'jpg';
+  return match[1] === 'jpeg' ? 'jpg' : match[1];
+}
+
 /**
  * Saves a capture (data URL or raw base64) to the user's Downloads folder.
- * Filename: qa-auto-hw-YYYYMMDD-HHmmss-sss-{contentHint}.jpg
+ * Filename: qa-auto-hw-YYYYMMDD-HHmmss-sss-{contentHint}.{ext}
  *
- * @param dataUrlOrBase64 - JPEG as data URL or raw base64 string
+ * @param dataUrlOrBase64 - image as data URL or raw base64 string (assumed JPEG)
  * @param contentHint - Short label for the file (e.g. ocr-x85-y0, ocr-trigger)
  * @returns Full path of the saved file
  */
@@ -48,7 +57,7 @@ export async function saveCaptureToDownloads(
   const hhmmss = now.toTimeString().slice(0, 8).replace(/:/g, '');
   const ms = now.getMilliseconds().toString().padStart(3, '0');
   const hint = sanitizeHint(contentHint);
-  const filename = `qa-auto-hw-${dateStr}-${hhmmss}-${ms}-${hint}.jpg`;
+  const filename = `qa-auto-hw-${dateStr}-${hhmmss}-${ms}-${hint}.${extensionForDataUrl(dataUrlOrBase64)}`;
   const fullPath = path.join(downloadsDir, filename);
 
   const base64 = extractBase64(dataUrlOrBase64);

diff --git a/models/paddle_ocr/PP-OCRv6_medium_det_onnx/.gitattributes b/models/paddle_ocr/PP-OCRv6_medium_det_onnx/.gitattributes
@@ -0,0 +1,35 @@
+*.7z filter=lfs diff=lfs merge=lfs -text
+*.arrow filter=lfs diff=lfs merge=lfs -text
+*.bin filter=lfs diff=lfs merge=lfs -text
+*.bz2 filter=lfs diff=lfs merge=lfs -text
+*.ckpt filter=lfs diff=lfs merge=lfs -text
+*.ftz filter=lfs diff=lfs merge=lfs -text
+*.gz filter=lfs diff=lfs merge=lfs -text
+*.h5 filter=lfs diff=lfs merge=lfs -text
+*.joblib filter=lfs diff=lfs merge=lfs -text
+*.lfs.* filter=lfs diff=lfs merge=lfs -text
+*.mlmodel filter=lfs diff=lfs merge=lfs -text
+*.model filter=lfs diff=lfs merge=lfs -text
+*.msgpack filter=lfs diff=lfs merge=lfs -text
+*.npy filter=lfs diff=lfs merge=lfs -text
+*.npz filter=lfs diff=lfs merge=lfs -text
+*.onnx filter=lfs diff=lfs merge=lfs -text
+*.ot filter=lfs diff=lfs merge=lfs -text
+*.parquet filter=lfs diff=lfs merge=lfs -text
+*.pb filter=lfs diff=lfs merge=lfs -text
+*.pickle filter=lfs diff=lfs merge=lfs -text
+*.pkl filter=lfs diff=lfs merge=lfs -text
+*.pt filter=lfs diff=lfs merge=lfs -text
+*.pth filter=lfs diff=lfs merge=lfs -text
+*.rar filter=lfs diff=lfs merge=lfs -text
+*.safetensors filter=lfs diff=lfs merge=lfs -text
+saved_model/**/* filter=lfs diff=lfs merge=lfs -text
+*.tar.* filter=lfs diff=lfs merge=lfs -text
+*.tar filter=lfs diff=lfs merge=lfs -text
+*.tflite filter=lfs diff=lfs merge=lfs -text
+*.tgz filter=lfs diff=lfs merge=lfs -text
+*.wasm filter=lfs diff=lfs merge=lfs -text
+*.xz filter=lfs diff=lfs merge=lfs -text
+*.zip filter=lfs diff=lfs merge=lfs -text
+*.zst filter=lfs diff=lfs merge=lfs -text
+*tfevents* filter=lfs diff=lfs merge=lfs -text