fix(ocr): update PaddleOCR for v3 API and add Tesseract fallback

Fix PaddleOCR crash by removing deprecated parameters (use_angle_cls,
show_log, cls) that were removed in PaddleOCR v3. Add graceful
fallback to Tesseract when PaddleOCR fails at runtime, so users
always get OCR results without errors.

packages/ai/python/ocr.py

@@ -1,6 +1,7 @@
 """Text extraction from images using Tesseract or PaddleOCR."""
 import sys
 import json
+import os
 
 
 def emit_progress(percent, stage):
@@ -8,6 +9,49 @@ def emit_progress(percent, stage):
     print(json.dumps({"progress": percent, "stage": stage}), file=sys.stderr, flush=True)
 
 
+def run_tesseract(input_path, language):
+    """Run Tesseract OCR."""
+    import subprocess
+
+    lang_map = {"en": "eng", "de": "deu", "fr": "fra", "es": "spa", "zh": "chi_sim", "ja": "jpn", "ko": "kor"}
+    tess_lang = lang_map.get(language, "eng")
+
+    emit_progress(30, "Scanning")
+    result = subprocess.run(
+        ["tesseract", input_path, "stdout", "-l", tess_lang],
+        capture_output=True,
+        text=True,
+        timeout=120,
+    )
+    emit_progress(70, "Extracting text")
+    text = result.stdout.strip()
+    if result.returncode != 0 and not text:
+        raise RuntimeError(result.stderr.strip() or "Tesseract failed")
+    return text, "tesseract"
+
+
+def run_paddleocr(input_path, language):
+    """Run PaddleOCR."""
+    os.environ["PADDLE_PDX_DISABLE_MODEL_SOURCE_CHECK"] = "True"
+    from paddleocr import PaddleOCR
+
+    emit_progress(20, "Loading")
+    ocr = PaddleOCR(lang=language)
+    emit_progress(30, "Scanning")
+    result = ocr.ocr(input_path)
+    emit_progress(70, "Extracting text")
+    text = "\n".join(
+        [
+            line[1][0]
+            for res in result
+            if res
+            for line in res
+            if line and line[1]
+        ]
+    )
+    return text, "paddleocr"
+
+
 def main():
     input_path = sys.argv[1]
     settings = json.loads(sys.argv[2]) if len(sys.argv) > 2 else {}
@@ -16,74 +60,48 @@ def main():
     language = settings.get("language", "en")
 
     try:
-        emit_progress(10, "Loading OCR engine")
+        emit_progress(10, "Preparing")
+
         if engine == "paddleocr":
             try:
-                from paddleocr import PaddleOCR
-
-                ocr = PaddleOCR(use_angle_cls=True, lang=language)
-                emit_progress(30, "Analyzing text regions")
-                result = ocr.ocr(input_path, cls=True)
-                emit_progress(70, "Extracting text")
-                text = "\n".join(
-                    [
-                        line[1][0]
-                        for res in result
-                        if res
-                        for line in res
-                        if line and line[1]
-                    ]
-                )
-                emit_progress(95, "Formatting results")
-                print(
-                    json.dumps({"success": True, "text": text, "engine": "paddleocr"})
-                )
+                text, used_engine = run_paddleocr(input_path, language)
             except ImportError:
                 print(
                     json.dumps(
                         {
                             "success": False,
-                            "error": "PaddleOCR is not installed. Install with: pip install paddleocr paddlepaddle",
+                            "error": "PaddleOCR is not installed",
                         }
                     )
                 )
                 sys.exit(1)
-        else:
-            # Tesseract via subprocess
-            import subprocess
-
-            lang_map = {"en": "eng", "de": "deu", "fr": "fra", "es": "spa", "zh": "chi_sim", "ja": "jpn", "ko": "kor"}
-            tess_lang = lang_map.get(language, "eng")
-
-            try:
-                emit_progress(30, "Running Tesseract")
-                result = subprocess.run(
-                    ["tesseract", input_path, "stdout", "-l", tess_lang],
-                    capture_output=True,
-                    text=True,
-                    timeout=120,
-                )
-                emit_progress(70, "Extracting text")
-                text = result.stdout.strip()
-                if result.returncode != 0 and not text:
-                    raise RuntimeError(result.stderr.strip() or "Tesseract failed")
-                emit_progress(95, "Formatting results")
-                print(
-                    json.dumps(
-                        {"success": True, "text": text, "engine": "tesseract"}
+            except Exception:
+                # PaddleOCR failed at runtime — fall back to Tesseract
+                emit_progress(25, "Falling back")
+                try:
+                    text, used_engine = run_tesseract(input_path, language)
+                except FileNotFoundError:
+                    print(
+                        json.dumps({"success": False, "error": "OCR engines unavailable"})
                     )
-                )
+                    sys.exit(1)
+        else:
+            try:
+                text, used_engine = run_tesseract(input_path, language)
             except FileNotFoundError:
                 print(
                     json.dumps(
                         {
                             "success": False,
-                            "error": "Tesseract is not installed. Install with: apt-get install tesseract-ocr",
+                            "error": "Tesseract is not installed",
                         }
                     )
                 )
                 sys.exit(1)
 
+        emit_progress(95, "Done")
+        print(json.dumps({"success": True, "text": text, "engine": used_engine}))
+
     except Exception as e:
         print(json.dumps({"success": False, "error": str(e)}))
         sys.exit(1)