feat: switch hosted OCR CPU actor defaults to Nemotron OCR v2

.github/workflows/integration-test-library-mode.yml

@@ -37,7 +37,7 @@ jobs:
       # Hosted NIM / integrate.api auth (ExtractParams / EmbedParams read NVIDIA_API_KEY / NGC_API_KEY).
       NVIDIA_API_KEY: ${{ secrets.NVIDIA_BUILD_API_KEY }}
       PAGE_ELEMENTS_INVOKE_URL: "https://ai.api.nvidia.com/v1/cv/nvidia/nemotron-page-elements-v3"
-      OCR_INVOKE_URL: "https://ai.api.nvidia.com/v1/cv/nvidia/nemoretriever-ocr-v1"
+      OCR_INVOKE_URL: "https://ai.api.nvidia.com/v1/cv/nvidia/nemotron-ocr-v2"
       GRAPHIC_ELEMENTS_INVOKE_URL: "https://ai.api.nvidia.com/v1/cv/nvidia/nemotron-graphic-elements-v1"
       TABLE_STRUCTURE_INVOKE_URL: "https://ai.api.nvidia.com/v1/cv/nvidia/nemotron-table-structure-v1"
       EMBED_INVOKE_URL: "https://integrate.api.nvidia.com/v1"

nemo_retriever/README.md

@@ -178,7 +178,7 @@ python -m nemo_retriever.examples.graph_pipeline \
   /your-example-dir \
   --vdb-kwargs-json '{"uri":"lancedb","table_name":"nemo-retriever"}' \
   --page-elements-invoke-url https://ai.api.nvidia.com/v1/cv/nvidia/nemotron-page-elements-v3 \
-  --ocr-invoke-url https://ai.api.nvidia.com/v1/cv/nvidia/nemotron-ocr-v1 \
+  --ocr-invoke-url https://ai.api.nvidia.com/v1/cv/nvidia/nemotron-ocr-v2 \
   --table-structure-invoke-url https://ai.api.nvidia.com/v1/cv/nvidia/nemotron-table-structure-v1 \
   --embed-invoke-url https://integrate.api.nvidia.com/v1/embeddings \
   --embed-model-name nvidia/llama-nemotron-embed-1b-v2
@@ -537,7 +537,7 @@ ingestor = (
   .extract(
     # for self hosted NIMs, your URLs will depend on your NIM container DNS settings
     page_elements_invoke_url="https://ai.api.nvidia.com/v1/cv/nvidia/nemotron-page-elements-v3",
-    ocr_invoke_url="https://ai.api.nvidia.com/v1/cv/nvidia/nemotron-ocr-v1",
+    ocr_invoke_url="https://ai.api.nvidia.com/v1/cv/nvidia/nemotron-ocr-v2",
     table_structure_invoke_url="https://ai.api.nvidia.com/v1/cv/nvidia/nemotron-table-structure-v1",
   )
   .embed(

nemo_retriever/docs/cli/README.md

@@ -109,7 +109,7 @@ export NVIDIA_API_KEY=nvapi-...
 
 retriever ingest ./data/multimodal_test.pdf \
   --page-elements-invoke-url https://ai.api.nvidia.com/v1/cv/nvidia/nemotron-page-elements-v3 \
-  --ocr-invoke-url https://ai.api.nvidia.com/v1/cv/nvidia/nemotron-ocr-v1 \
+  --ocr-invoke-url https://ai.api.nvidia.com/v1/cv/nvidia/nemotron-ocr-v2 \
   --table-structure-invoke-url https://ai.api.nvidia.com/v1/cv/nvidia/nemotron-table-structure-v1 \
   --embed-invoke-url https://integrate.api.nvidia.com/v1/embeddings \
   --embed-model-name nvidia/llama-nemotron-embed-1b-v2

nemo_retriever/src/nemo_retriever/operators/extract/chart/cpu_actor.py

@@ -23,13 +23,12 @@
 class GraphicElementsCPUActor(AbstractOperator, CPUOperator):
     """CPU-only variant of :class:`GraphicElementsActor`.
 
-    Defaults to build.nvidia.com endpoints for ``nemotron-graphic-elements-v1``
-    and hosted ``nemotron-ocr-v1``. Remote OCR remains URL-driven until OCR v2
-    NIM is released.
+    Defaults to the build.nvidia.com endpoint for
+    ``nemotron-graphic-elements-v1`` and hosted ``nemotron-ocr-v2``.
     """
 
     DEFAULT_GRAPHIC_ELEMENTS_INVOKE_URL = "https://ai.api.nvidia.com/v1/cv/nvidia/nemotron-graphic-elements-v1"
-    DEFAULT_OCR_INVOKE_URL = "https://ai.api.nvidia.com/v1/cv/nvidia/nemotron-ocr-v1"
+    DEFAULT_OCR_INVOKE_URL = "https://ai.api.nvidia.com/v1/cv/nvidia/nemotron-ocr-v2"
 
     def __init__(
         self,

nemo_retriever/src/nemo_retriever/operators/extract/ocr/cpu_ocr.py

@@ -19,11 +19,10 @@
 class OCRCPUActor(AbstractOperator, CPUOperator):
     """Remote OCR variant of :class:`OCRActor`.
 
-    The hosted NIM endpoint is still v1-only until OCR v2 is released remotely.
+    Defaults to the hosted Nemotron OCR v2 NIM endpoint.
     """
 
-    # Keep the remote default on v1 until a hosted OCR v2 NIM is available.
-    DEFAULT_INVOKE_URL = "https://ai.api.nvidia.com/v1/cv/nvidia/nemotron-ocr-v1"
+    DEFAULT_INVOKE_URL = "https://ai.api.nvidia.com/v1/cv/nvidia/nemotron-ocr-v2"
 
     def __init__(self, **ocr_kwargs: Any) -> None:
         super().__init__(**ocr_kwargs)

nemo_retriever/src/nemo_retriever/operators/extract/table/cpu_actor.py

@@ -23,9 +23,8 @@
 class TableStructureCPUActor(AbstractOperator, CPUOperator):
     """CPU-only variant of :class:`TableStructureActor`.
 
-    Defaults to the build.nvidia.com endpoint for
-    ``nemotron-table-structure-v1`` and hosted ``nemotron-ocr-v1``. Remote OCR
-    remains URL-driven until OCR v2 NIM is released.
+    Uses NVIDIA's hosted table-structure service by default. OCR is not turned on
+    unless you provide an ``ocr_invoke_url``.
     """
 
     DEFAULT_TABLE_STRUCTURE_INVOKE_URL = "https://ai.api.nvidia.com/v1/cv/nvidia/nemotron-table-structure-v1"

nemo_retriever/src/nemo_retriever/operators/extract/video/ocr_actor.py

@@ -87,11 +87,11 @@ def postprocess(self, data: Any, **kwargs: Any) -> Any:
 class VideoFrameOCRCPUActor(AbstractOperator, CPUOperator):
     """Remote Nemotron OCR on full video frames, batched per call.
 
-    Hosted OCR NIM is currently v1-only; the local GPU variant uses
+    Hosted OCR NIM defaults to Nemotron OCR v2; the local GPU variant uses
     ``NemotronOCRV2`` with the selected OCR mode.
     """
 
-    DEFAULT_INVOKE_URL = "https://ai.api.nvidia.com/v1/cv/nvidia/nemotron-ocr-v1"
+    DEFAULT_INVOKE_URL = "https://ai.api.nvidia.com/v1/cv/nvidia/nemotron-ocr-v2"
 
     def __init__(self, **ocr_kwargs: Any) -> None:
         super().__init__(**ocr_kwargs)

nemo_retriever/tests/test_actor_operators.py

@@ -391,9 +391,9 @@ def test_init_with_ocr_invoke_url_skips_local_ocr(self, monkeypatch):
         monkeypatch.setitem(local_models.__dict__, "NemotronTableStructureV1", mock_ts)
         monkeypatch.setitem(local_models.__dict__, "NemotronOCRV2", mock_ocr)
 
-        actor = GPUActor(ocr_invoke_url="http://ocr.example/v1/cv/nvidia/nemotron-ocr-v1")
+        actor = GPUActor(ocr_invoke_url="http://ocr.example/v1/cv/nvidia/nemotron-ocr-v2")
 
-        assert actor._ocr_invoke_url == "http://ocr.example/v1/cv/nvidia/nemotron-ocr-v1"
+        assert actor._ocr_invoke_url == "http://ocr.example/v1/cv/nvidia/nemotron-ocr-v2"
         assert actor._ocr_model is None
         mock_ocr.assert_not_called()
         mock_ts.assert_called_once_with()

nemo_retriever/tests/test_operator_flags_and_cpu_actors.py

@@ -182,7 +182,7 @@ def test_uses_default_urls(self):
         assert actor._graphic_elements_model is None
         assert actor._ocr_model is None
         assert "nemotron-graphic-elements-v1" in actor._graphic_elements_invoke_url
-        assert "nemotron-ocr-v1" in actor._ocr_invoke_url
+        assert "nemotron-ocr-v2" in actor._ocr_invoke_url
 
     def test_creates_with_custom_urls(self):
         from nemo_retriever.operators.extract.chart.cpu_actor import GraphicElementsCPUActor
@@ -317,7 +317,7 @@ def test_uses_default_invoke_url(self):
 
         actor = OCRCPUActor()
         assert actor._model is None
-        assert "nemotron-ocr-v1" in actor.ocr_kwargs["invoke_url"]
+        assert "nemotron-ocr-v2" in actor.ocr_kwargs["invoke_url"]
 
     def test_creates_with_custom_invoke_url(self):
         from nemo_retriever.operators.extract.ocr.cpu_ocr import OCRCPUActor

nemo_retriever/tests/test_readme_video_pipeline_example.py

@@ -94,7 +94,7 @@ def test_readme_video_pipeline_build_graph_chain() -> None:
     graph = build_graph(
         extraction_mode="auto",
         extract_params=ExtractParams(
-            ocr_invoke_url="https://ai.api.nvidia.com/v1/cv/nvidia/nemoretriever-ocr-v1",
+            ocr_invoke_url="https://ai.api.nvidia.com/v1/cv/nvidia/nemotron-ocr-v2",
         ),
         audio_chunk_params=AudioChunkParams(
             enabled=True,
@@ -134,7 +134,7 @@ def test_audio_only_excludes_visual_branch_from_graph() -> None:
     graph = build_graph(
         extraction_mode="auto",
         extract_params=ExtractParams(
-            ocr_invoke_url="https://ai.api.nvidia.com/v1/cv/nvidia/nemoretriever-ocr-v1",
+            ocr_invoke_url="https://ai.api.nvidia.com/v1/cv/nvidia/nemotron-ocr-v2",
         ),
         audio_chunk_params=AudioChunkParams(
             enabled=True,

nemo_retriever/tests/test_table_structure_nim_empty_bbox.py

@@ -300,7 +300,7 @@ def test_empty_nim_response_does_not_raise_in_torchless_image(self) -> None:
                 df,
                 table_structure_invoke_url="http://nemotron-table-structure-v1:8000/v1/infer",
                 page_elements_invoke_url="http://nemotron-page-elements-v3:8000/v1/infer",
-                ocr_invoke_url="http://nemotron-ocr-v1:8000/v1/infer",
+                ocr_invoke_url="http://nemotron-ocr-v2:8000/v1/infer",
             )
 
         # Stage finished successfully, no row-level error recorded.
@@ -336,7 +336,7 @@ def test_empty_nim_response_does_not_call_prediction_to_detections(self) -> None
                 df,
                 table_structure_invoke_url="http://nemotron-table-structure-v1:8000/v1/infer",
                 page_elements_invoke_url="http://nemotron-page-elements-v3:8000/v1/infer",
-                ocr_invoke_url="http://nemotron-ocr-v1:8000/v1/infer",
+                ocr_invoke_url="http://nemotron-ocr-v2:8000/v1/infer",
             )
 
     def test_non_bbox_response_still_falls_through_to_legacy_parser(self) -> None:
@@ -364,7 +364,7 @@ def test_non_bbox_response_still_falls_through_to_legacy_parser(self) -> None:
                 df,
                 table_structure_invoke_url="http://nemotron-table-structure-v1:8000/v1/infer",
                 page_elements_invoke_url="http://nemotron-page-elements-v3:8000/v1/infer",
-                ocr_invoke_url="http://nemotron-ocr-v1:8000/v1/infer",
+                ocr_invoke_url="http://nemotron-ocr-v2:8000/v1/infer",
             )
 
         # The non-bbox legacy response triggers the fallback exactly once.

nemo_retriever/tests/test_video_frame_ocr_actor.py

@@ -72,10 +72,10 @@ def test_cpu_actor_calls_remote_batched_with_b64_list() -> None:
     assert call_kwargs["invoke_url"] == "https://example/ocr"
 
 
-def test_cpu_actor_defaults_to_hosted_ocr_v1() -> None:
+def test_cpu_actor_defaults_to_hosted_ocr_v2() -> None:
     actor = VideoFrameOCRCPUActor()
 
-    assert "nemotron-ocr-v1" in actor._invoke_url
+    assert "nemotron-ocr-v2" in actor._invoke_url
 
 
 def test_gpu_actor_invokes_local_model_per_frame() -> None: