Remove synthetic training path and add fine-text OCR option

ocr_compare/server.py

@@ -132,7 +132,7 @@ def _prepare(data: bytes, header: str | None) -> tuple[bytes, tuple[int, int]]:
         raise InputError("invalid_image") from None
 
 
-def _classic(data: bytes, directory: Path) -> dict:
+def _classic(data: bytes, directory: Path, *, fine_text: bool = False) -> dict:
     started = time.monotonic()
     models = MODEL_HOME / "official_models"
     if any(not (models / name / "inference.pdiparams").is_file() for name in
@@ -147,8 +147,11 @@ def _classic(data: bytes, directory: Path) -> dict:
                TRANSFORMERS_OFFLINE="1", PADDLE_PDX_DISABLE_MODEL_SOURCE_CHECK="True",
                OMP_NUM_THREADS="1", OPENBLAS_NUM_THREADS="1", OMP_THREAD_LIMIT="1")
     try:
+        command = [sys.executable, "-m", "ocr_compare.classic_worker", str(image_path), str(output_path)]
+        if fine_text:
+            command.append("--fine-text")
         process = subprocess.run(
-            [sys.executable, "-m", "ocr_compare.classic_worker", str(image_path), str(output_path)],
+            command,
             cwd=PROJECT, env=env, stdout=subprocess.DEVNULL, stderr=subprocess.DEVNULL,
             timeout=120, check=False,
         )
@@ -165,7 +168,8 @@ def _classic(data: bytes, directory: Path) -> dict:
                  for row in raw["lines"] if row.get("bbox")]
     except (OSError, ValueError, KeyError, TypeError):
         return {"status": "error", "error": "classic_invalid", "elapsed_ms": elapsed}
-    return {"status": "ok", "model": "PP-OCRv5 mobile (Latin), lokal",
+    model = "PP-OCRv5 mobile (Latin), lokal" + (" · Feintext" if fine_text else "")
+    return {"status": "ok", "model": model,
             "elapsed_ms": elapsed, "lines": lines, "fields": assign_fields(lines),
             "regions": []}
 
@@ -269,6 +273,9 @@ async def compare(request: Request):
         return JSONResponse({"error": "busy"}, status_code=429)
     if request.headers.get("content-type", "").split(";", 1)[0] not in ("image/jpeg", "image/png"):
         return JSONResponse({"error": "unsupported_media_type"}, status_code=415)
+    detail = request.headers.get("x-ocr-detail", "standard")
+    if detail not in ("standard", "fine"):
+        return JSONResponse({"error": "invalid_detail"}, status_code=422)
     async with _slot:
         data = bytearray()
         try:
@@ -287,7 +294,7 @@ async def compare(request: Request):
         with tempfile.TemporaryDirectory(prefix="ocr-compare-") as path:
             directory = Path(path)
             classic, vision = await asyncio.gather(
-                asyncio.to_thread(_classic, image, directory),
+                asyncio.to_thread(_classic, image, directory, fine_text=detail == "fine"),
                 asyncio.to_thread(_vision, image),
             )
         # Exact JPEG bytes sent to both pipelines are returned for evidence