nv-ingest-api 2025.10.9.dev20251009__tar.gz → 2025.10.11.dev20251011__tar.gz

{nv_ingest_api-2025.10.9.dev20251009/src/nv_ingest_api.egg-info → nv_ingest_api-2025.10.11.dev20251011}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: nv-ingest-api
-Version: 2025.10.9.dev20251009
+Version: 2025.10.11.dev20251011
 Summary: Python module with core document ingestion functions.
 Author-email: Jeremy Dyer <jdyer@nvidia.com>
 License:                                  Apache License

{nv_ingest_api-2025.10.9.dev20251009 → nv_ingest_api-2025.10.11.dev20251011}/src/nv_ingest_api/internal/extract/image/chart_extractor.py

@@ -97,7 +97,7 @@ def _run_chart_inference(
             model_name="paddle",
             max_batch_size=1 if ocr_client.protocol == "grpc" else 2,
         )
-    elif ocr_model_name == "scene_text_ensemble":
+    elif ocr_model_name in {"scene_text_ensemble", "scene_text_wrapper", "scene_text_python"}:
         future_ocr_kwargs.update(
             model_name=ocr_model_name,
             input_names=["INPUT_IMAGE_URLS", "MERGE_LEVELS"],
@@ -237,7 +237,9 @@ def _create_ocr_client(
     auth_token: str,
 ) -> NimClient:
     ocr_model_interface = (
-        NemoRetrieverOCRModelInterface() if ocr_model_name == "scene_text_ensemble" else PaddleOCRModelInterface()
+        NemoRetrieverOCRModelInterface()
+        if ocr_model_name in {"scene_text_ensemble", "scene_text_wrapper", "scene_text_python"}
+        else PaddleOCRModelInterface()
     )
 
     ocr_client = create_inference_client(
@@ -245,7 +247,9 @@ def _create_ocr_client(
         model_interface=ocr_model_interface,
         auth_token=auth_token,
         infer_protocol=ocr_protocol,
-        enable_dynamic_batching=(True if ocr_model_name == "scene_text_ensemble" else False),
+        enable_dynamic_batching=(
+            True if ocr_model_name in {"scene_text_ensemble", "scene_text_wrapper", "scene_text_python"} else False
+        ),
         dynamic_batch_memory_budget_mb=32,
     )
 

{nv_ingest_api-2025.10.9.dev20251009 → nv_ingest_api-2025.10.11.dev20251011}/src/nv_ingest_api/internal/extract/image/infographic_extractor.py

@@ -107,7 +107,7 @@ def _update_infographic_metadata(
             model_name="paddle",
             max_batch_size=1 if ocr_client.protocol == "grpc" else 2,
         )
-    elif ocr_model_name == "scene_text_ensemble":
+    elif ocr_model_name in {"scene_text_ensemble", "scene_text_wrapper", "scene_text_python"}:
         infer_kwargs.update(
             model_name=ocr_model_name,
             input_names=["INPUT_IMAGE_URLS", "MERGE_LEVELS"],
@@ -152,7 +152,9 @@ def _create_ocr_client(
     auth_token: str,
 ) -> NimClient:
     ocr_model_interface = (
-        NemoRetrieverOCRModelInterface() if ocr_model_name == "scene_text_ensemble" else PaddleOCRModelInterface()
+        NemoRetrieverOCRModelInterface()
+        if ocr_model_name in {"scene_text_ensemble", "scene_text_wrapper", "scene_text_python"}
+        else PaddleOCRModelInterface()
     )
 
     ocr_client = create_inference_client(
@@ -160,7 +162,9 @@ def _create_ocr_client(
         model_interface=ocr_model_interface,
         auth_token=auth_token,
         infer_protocol=ocr_protocol,
-        enable_dynamic_batching=(True if ocr_model_name == "scene_text_ensemble" else False),
+        enable_dynamic_batching=(
+            True if ocr_model_name in {"scene_text_ensemble", "scene_text_wrapper", "scene_text_python"} else False
+        ),
         dynamic_batch_memory_budget_mb=32,
     )
 

{nv_ingest_api-2025.10.9.dev20251009 → nv_ingest_api-2025.10.11.dev20251011}/src/nv_ingest_api/internal/extract/image/table_extractor.py

@@ -99,7 +99,7 @@ def _run_inference(
             model_name="paddle",
             max_batch_size=1 if ocr_client.protocol == "grpc" else 2,
         )
-    elif ocr_model_name == "scene_text_ensemble":
+    elif ocr_model_name in {"scene_text_ensemble", "scene_text_wrapper", "scene_text_python"}:
         future_ocr_kwargs.update(
             model_name=ocr_model_name,
             input_names=["INPUT_IMAGE_URLS", "MERGE_LEVELS"],
@@ -246,7 +246,9 @@ def _create_ocr_client(
     auth_token: str,
 ) -> NimClient:
     ocr_model_interface = (
-        NemoRetrieverOCRModelInterface() if ocr_model_name == "scene_text_ensemble" else PaddleOCRModelInterface()
+        NemoRetrieverOCRModelInterface()
+        if ocr_model_name in {"scene_text_ensemble", "scene_text_wrapper", "scene_text_python"}
+        else PaddleOCRModelInterface()
     )
 
     ocr_client = create_inference_client(
@@ -254,7 +256,9 @@ def _create_ocr_client(
         model_interface=ocr_model_interface,
         auth_token=auth_token,
         infer_protocol=ocr_protocol,
-        enable_dynamic_batching=(True if ocr_model_name == "scene_text_ensemble" else False),
+        enable_dynamic_batching=(
+            True if ocr_model_name in {"scene_text_ensemble", "scene_text_wrapper", "scene_text_python"} else False
+        ),
         dynamic_batch_memory_budget_mb=32,
     )
 

{nv_ingest_api-2025.10.9.dev20251009 → nv_ingest_api-2025.10.11.dev20251011}/src/nv_ingest_api/internal/primitives/nim/model_interface/ocr.py

@@ -21,7 +21,10 @@ from nv_ingest_api.internal.primitives.nim.model_interface.helpers import prepro
 from nv_ingest_api.util.image_processing.transforms import base64_to_numpy
 
 DEFAULT_OCR_MODEL_NAME = "paddle"
-NEMORETRIEVER_OCR_MODEL_NAME = "scene_text_ensemble"
+NEMORETRIEVER_OCR_MODEL_NAME = "scene_text_wrapper"
+NEMORETRIEVER_OCR_ENSEMBLE_MODEL_NAME = "scene_text_ensemble"
+NEMORETRIEVER_OCR_BLS_MODEL_NAME = "scene_text_python"
+
 
 logger = logging.getLogger(__name__)
 
@@ -231,7 +234,11 @@ class OCRModelInterfaceBase(ModelInterface):
         if not isinstance(response, np.ndarray):
             raise ValueError("Unexpected response format: response is not a NumPy array.")
 
-        if model_name == NEMORETRIEVER_OCR_MODEL_NAME:
+        if model_name in [
+            NEMORETRIEVER_OCR_MODEL_NAME,
+            NEMORETRIEVER_OCR_ENSEMBLE_MODEL_NAME,
+            NEMORETRIEVER_OCR_BLS_MODEL_NAME,
+        ]:
             response = response.transpose((1, 0))
 
         # If we have shape (3,), convert to (3, 1)

{nv_ingest_api-2025.10.9.dev20251009 → nv_ingest_api-2025.10.11.dev20251011}/src/nv_ingest_api/internal/primitives/nim/nim_client.py

@@ -121,9 +121,6 @@ class NimClient:
         if model_name == "yolox_ensemble":
             model_name = "yolox"
 
-        if model_name == "scene_text_ensemble":
-            model_name = "scene_text_pre"
-
         if model_name in self._max_batch_sizes:
             return self._max_batch_sizes[model_name]
 

{nv_ingest_api-2025.10.9.dev20251009 → nv_ingest_api-2025.10.11.dev20251011}/src/nv_ingest_api/internal/schemas/meta/ingest_job_schema.py

@@ -24,8 +24,12 @@ logger = logging.getLogger(__name__)
 # Tracing Options Schema
 class TracingOptionsSchema(BaseModelNoExt):
     trace: bool = False
-    ts_send: int
+    ts_send: Optional[int] = None
     trace_id: Optional[str] = None
+    # V2 PDF splitting support
+    parent_job_id: Optional[str] = None
+    page_num: Optional[int] = None
+    total_pages: Optional[int] = None
 
 
 # Ingest Task Schemas

{nv_ingest_api-2025.10.9.dev20251009 → nv_ingest_api-2025.10.11.dev20251011}/src/nv_ingest_api/util/service_clients/rest/rest_client.py

@@ -3,6 +3,7 @@
 # SPDX-License-Identifier: Apache-2.0
 
 import logging
+import os
 import re
 import time
 from typing import Any, Union, Tuple, Optional, Dict, Callable
@@ -137,13 +138,19 @@ class RestClient(MessageBrokerClientBase):
                 )
                 self._client = requests.Session()
 
-        self._submit_endpoint: str = "/v1/submit_job"
-        self._fetch_endpoint: str = "/v1/fetch_job"
+        # Allow API version override via environment variable or kwargs
+        api_version = kwargs.get("api_version") or os.getenv("NV_INGEST_API_VERSION", "v1")
+        self._api_version = api_version
+        self._submit_endpoint: str = f"/{api_version}/submit_job"
+        self._fetch_endpoint: str = f"/{api_version}/fetch_job"
         self._base_url: str = kwargs.get("base_url") or self._generate_url(self._host, self._port)
         self._headers = kwargs.get("headers", {})
         self._auth = kwargs.get("auth", None)
 
         logger.debug(f"RestClient base URL set to: {self._base_url}")
+        logger.info(
+            f"RestClient using API version: {api_version} (endpoints: {self._submit_endpoint}, {self._fetch_endpoint})"
+        )
 
     @staticmethod
     def _generate_url(host: str, port: int) -> str:

{nv_ingest_api-2025.10.9.dev20251009 → nv_ingest_api-2025.10.11.dev20251011/src/nv_ingest_api.egg-info}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: nv-ingest-api
-Version: 2025.10.9.dev20251009
+Version: 2025.10.11.dev20251011
 Summary: Python module with core document ingestion functions.
 Author-email: Jeremy Dyer <jdyer@nvidia.com>
 License:                                  Apache License