nv-ingest-api 2025.7.15.dev20250715__py3-none-any.whl → 2025.7.16.dev20250716__py3-none-any.whl

nv_ingest_api/internal/enums/common.py

@@ -52,6 +52,8 @@ class ContentDescriptionEnum(str, Enum):
         Description for image extracted from PDF document.
     PDF_INFOGRAPHIC : str
         Description for structured infographic extracted from PDF document.
+    PDF_PAGE_IMAGE : str
+        Description for a full-page image rendered from a PDF document.
     PDF_TABLE : str
         Description for structured table extracted from PDF document.
     PDF_TEXT : str
@@ -70,6 +72,7 @@ class ContentDescriptionEnum(str, Enum):
     PDF_CHART: str = "Structured chart extracted from PDF document."
     PDF_IMAGE: str = "Image extracted from PDF document."
     PDF_INFOGRAPHIC: str = "Structured infographic extracted from PDF document."
+    PDF_PAGE_IMAGE: str = "Full-page image rendered from a PDF document."
     PDF_TABLE: str = "Structured table extracted from PDF document."
     PDF_TEXT: str = "Unstructured text from PDF document."
     PPTX_IMAGE: str = "Image extracted from PPTX presentation."
@@ -94,6 +97,8 @@ class ContentTypeEnum(str, Enum):
         Represents image content.
     INFO_MSG : str
         Represents an informational message.
+    PAGE_IMAGE : str
+        Represents a full-page image rendered from a document.
     STRUCTURED : str
         Represents structured content.
     TEXT : str
@@ -111,6 +116,7 @@ class ContentTypeEnum(str, Enum):
     INFOGRAPHIC: str = "infographic"
     INFO_MSG: str = "info_message"
     NONE: str = "none"
+    PAGE_IMAGE: str = "page_image"
     STRUCTURED: str = "structured"
     TABLE: str = "table"
     TEXT: str = "text"

nv_ingest_api/internal/extract/pdf/engines/pdf_helpers/__init__.py

@@ -4,20 +4,21 @@
 # Copyright (c) 2024, NVIDIA CORPORATION.
 
 import base64
+import inspect
 import io
-
-import pandas as pd
-from typing import Any, Dict, List, Optional
 import logging
+from typing import Any
+from typing import Dict
+from typing import List
+from typing import Optional
 
-from nv_ingest_api.internal.extract.pdf.engines import (
-    adobe_extractor,
-    llama_parse_extractor,
-    nemoretriever_parse_extractor,
-    pdfium_extractor,
-    tika_extractor,
-    unstructured_io_extractor,
-)
+import pandas as pd
+from nv_ingest_api.internal.extract.pdf.engines import adobe_extractor
+from nv_ingest_api.internal.extract.pdf.engines import llama_parse_extractor
+from nv_ingest_api.internal.extract.pdf.engines import nemoretriever_parse_extractor
+from nv_ingest_api.internal.extract.pdf.engines import pdfium_extractor
+from nv_ingest_api.internal.extract.pdf.engines import tika_extractor
+from nv_ingest_api.internal.extract.pdf.engines import unstructured_io_extractor
 from nv_ingest_api.util.exception_handlers.decorators import unified_exception_handler
 
 # Import extraction functions for different engines.
@@ -43,6 +44,7 @@ def _work_extract_pdf(
     extract_infographics: bool,
     extract_tables: bool,
     extract_charts: bool,
+    extract_page_as_image: bool,
     extractor_config: dict,
     execution_trace_log=None,
 ) -> Any:
@@ -52,17 +54,25 @@ def _work_extract_pdf(
 
     extract_method = extractor_config["extract_method"]
     extractor_fn = EXTRACTOR_LOOKUP.get(extract_method, pdfium_extractor)
-    return extractor_fn(
-        pdf_stream,
-        extract_text,
-        extract_images,
-        extract_infographics,
-        extract_tables,
-        extract_charts,
-        extractor_config,
-        execution_trace_log,
+
+    extractor_fn_args = dict(
+        pdf_stream=pdf_stream,
+        extract_text=extract_text,
+        extract_images=extract_images,
+        extract_infographics=extract_infographics,
+        extract_tables=extract_tables,
+        extract_charts=extract_charts,
+        extractor_config=extractor_config,
+        execution_trace_log=execution_trace_log,
     )
 
+    if "extract_page_as_image" in inspect.signature(extractor_fn).parameters:
+        extractor_fn_args["extract_page_as_image"] = extract_page_as_image
+    elif extract_page_as_image:
+        logger.warning(f"`extract_page_as_image` is set to True, but {extract_method} does not support it.")
+
+    return extractor_fn(**extractor_fn_args)
+
 
 @unified_exception_handler
 def _orchestrate_row_extraction(
@@ -97,6 +107,7 @@ def _orchestrate_row_extraction(
         extract_tables = params.pop("extract_tables", False)
         extract_charts = params.pop("extract_charts", False)
         extract_infographics = params.pop("extract_infographics", False)
+        extract_page_as_image = params.pop("extract_page_as_image", False)
         extract_method = params.get("extract_method", "pdfium")
     except KeyError as e:
         raise ValueError(f"Missing required extraction flag: {e}")
@@ -137,6 +148,7 @@ def _orchestrate_row_extraction(
         extract_text=extract_text,
         extract_images=extract_images,
         extract_infographics=extract_infographics,
+        extract_page_as_image=extract_page_as_image,
         extract_tables=extract_tables,
         extract_charts=extract_charts,
         extractor_config=extractor_config,

nv_ingest_api/internal/extract/pdf/engines/pdfium.py

@@ -24,6 +24,7 @@ import numpy as np
 import pandas as pd
 import pypdfium2 as libpdfium
 
+from nv_ingest_api.internal.enums.common import ContentTypeEnum
 from nv_ingest_api.internal.primitives.nim.default_values import YOLOX_MAX_BATCH_SIZE
 from nv_ingest_api.internal.primitives.nim.model_interface.yolox import (
     YOLOX_PAGE_IMAGE_PREPROC_WIDTH,
@@ -35,6 +36,7 @@ from nv_ingest_api.internal.primitives.nim.model_interface.yolox import (
 from nv_ingest_api.internal.schemas.extract.extract_pdf_schema import PDFiumConfigSchema
 from nv_ingest_api.internal.enums.common import TableFormatEnum, TextTypeEnum, AccessLevelEnum
 from nv_ingest_api.util.metadata.aggregators import (
+    construct_image_metadata_from_base64,
     construct_image_metadata_from_pdf_image,
     extract_pdf_metadata,
     construct_text_metadata,
@@ -47,6 +49,7 @@ from nv_ingest_api.util.pdf.pdfium import (
     extract_image_like_objects_from_pdfium_page,
 )
 from nv_ingest_api.util.pdf.pdfium import pdfium_pages_to_numpy
+from nv_ingest_api.util.image_processing import scale_image_to_encoding_size
 from nv_ingest_api.util.image_processing.transforms import numpy_to_base64, crop_image
 
 logger = logging.getLogger(__name__)
@@ -385,6 +388,7 @@ def pdfium_extractor(
     extract_infographics: bool,
     extract_tables: bool,
     extract_charts: bool,
+    extract_page_as_image: bool,
     extractor_config: dict,
     execution_trace_log: Optional[List[Any]] = None,
 ) -> pd.DataFrame:
@@ -525,6 +529,24 @@ def pdfium_extractor(
                 )
                 extracted_data.extend(image_data)
 
+            # Full page image extraction
+            if extract_page_as_image:
+                page_text = _extract_page_text(page)
+                image, _ = pdfium_pages_to_numpy([page], scale_tuple=(16384, 16384), trace_info=execution_trace_log)
+                base64_image = numpy_to_base64(image[0])
+                if len(base64_image) > 2**24 - 1:
+                    base64_image, _ = scale_image_to_encoding_size(base64_image, max_base64_size=2**24 - 1)
+                image_meta = construct_image_metadata_from_base64(
+                    base64_image,
+                    page_idx,
+                    page_count,
+                    source_metadata,
+                    base_unified_metadata,
+                    subtype=ContentTypeEnum.PAGE_IMAGE,
+                    text=page_text,
+                )
+                extracted_data.append(image_meta)
+
             # If we want tables or charts, rasterize the page and store it
             if extract_tables or extract_charts or extract_infographics:
                 image, padding_offsets = pdfium_pages_to_numpy(
@@ -575,6 +597,7 @@ def pdfium_extractor(
                 execution_trace_log=execution_trace_log,
             )
             futures.append(future)
+
             pages_for_tables.clear()
 
         # Wait for all asynchronous jobs to complete.

nv_ingest_api/internal/schemas/meta/ingest_job_schema.py

@@ -107,6 +107,10 @@ class IngestTaskEmbedSchema(BaseModelNoExt):
     model_name: Optional[str] = None
     api_key: Optional[str] = None
     filter_errors: bool = False
+    text_elements_modality: Optional[str] = None
+    image_elements_modality: Optional[str] = None
+    structured_elements_modality: Optional[str] = None
+    audio_elements_modality: Optional[str] = None
 
 
 class IngestTaskVdbUploadSchema(BaseModelNoExt):
@@ -195,6 +199,7 @@ class IngestTaskSchema(BaseModelNoExt):
         validated_task_properties = expected_schema_cls(**task_properties)
         values["type"] = task_type  # ensure type is now always the enum
         values["task_properties"] = validated_task_properties
+
         return values
 
     @field_validator("type", mode="before")

nv_ingest_api/internal/schemas/transform/transform_text_embedding_schema.py

@@ -22,5 +22,9 @@ class TextEmbeddingSchema(BaseModel):
     input_type: str = Field(default="passage")
     raise_on_failure: bool = Field(default=False)
     truncate: str = Field(default="END")
+    text_elements_modality: str = Field(default="text")
+    image_elements_modality: str = Field(default="text")
+    structured_elements_modality: str = Field(default="text")
+    audio_elements_modality: str = Field(default="text")
 
     model_config = ConfigDict(extra="forbid")

nv_ingest_api/internal/transform/embed_text.py

@@ -4,6 +4,7 @@
 
 import logging
 from concurrent.futures import ThreadPoolExecutor
+from functools import partial
 from typing import Any, Dict, Tuple, Optional, Iterable, List
 
 import pandas as pd
@@ -19,6 +20,9 @@ from nv_ingest_api.util.schema.schema_validator import validate_schema
 logger = logging.getLogger(__name__)
 
 
+MULTI_MODAL_MODELS = ["llama-3.2-nemoretriever-1b-vlm-embed-v1"]
+
+
 # ------------------------------------------------------------------------------
 # Asynchronous Embedding Requests
 # ------------------------------------------------------------------------------
@@ -33,6 +37,7 @@ def _make_async_request(
     input_type: str,
     truncate: str,
     filter_errors: bool,
+    modalities: Optional[List[str]] = None,
 ) -> list:
     """
     Interacts directly with the NIM embedding service to calculate embeddings for a batch of prompts.
@@ -74,11 +79,18 @@ def _make_async_request(
             base_url=embedding_nim_endpoint,
         )
 
+        extra_body = {
+            "input_type": input_type,
+            "truncate": truncate,
+        }
+        if modalities:
+            extra_body["modality"] = modalities
+
         resp = client.embeddings.create(
             input=prompts,
             model=embedding_model,
             encoding_format=encoding_format,
-            extra_body={"input_type": input_type, "truncate": truncate},
+            extra_body=extra_body,
         )
 
         response["embedding"] = resp.data
@@ -110,6 +122,7 @@ def _async_request_handler(
     input_type: str,
     truncate: str,
     filter_errors: bool,
+    modalities: Optional[List[str]] = None,
 ) -> List[dict]:
     """
     Gathers calculated embedding results from the NIM embedding service concurrently.
@@ -138,6 +151,9 @@ def _async_request_handler(
     List[dict]
         A list of response dictionaries from the embedding service.
     """
+    if modalities is None:
+        modalities = [None] * len(prompts)
+
     with ThreadPoolExecutor() as executor:
         futures = [
             executor.submit(
@@ -150,8 +166,9 @@ def _async_request_handler(
                 input_type=input_type,
                 truncate=truncate,
                 filter_errors=filter_errors,
+                modalities=modality_batch,
             )
-            for prompt_batch in prompts
+            for prompt_batch, modality_batch in zip(prompts, modalities)
         ]
         results = [future.result() for future in futures]
 
@@ -167,6 +184,7 @@ def _async_runner(
     input_type: str,
     truncate: str,
     filter_errors: bool,
+    modalities: Optional[List[str]] = None,
 ) -> dict:
     """
     Concurrently launches all NIM embedding requests and flattens the results.
@@ -204,6 +222,7 @@ def _async_runner(
         input_type,
         truncate,
         filter_errors,
+        modalities=modalities,
     )
 
     flat_results = {"embeddings": [], "info_msgs": []}
@@ -263,7 +282,19 @@ def _add_embeddings(row, embeddings, info_msgs):
     return row
 
 
-def _get_pandas_text_content(row):
+def _format_image_input_string(image_b64: Optional[str]) -> str:
+    if not image_b64:
+        return
+    return f"data:image/png;base64,{image_b64}"
+
+
+def _format_text_image_pair_input_string(text: Optional[str], image_b64: Optional[str]) -> str:
+    if (not text) or (not text.strip()) or (not image_b64):
+        return
+    return f"{text.strip()} {_format_image_input_string(image_b64)}"
+
+
+def _get_pandas_text_content(row, modality="text"):
     """
     Extracts text content from a DataFrame row.
 
@@ -280,7 +311,7 @@ def _get_pandas_text_content(row):
     return row["content"]
 
 
-def _get_pandas_table_content(row):
+def _get_pandas_table_content(row, modality="text"):
     """
     Extracts table/chart content from a DataFrame row.
 
@@ -294,10 +325,19 @@ def _get_pandas_table_content(row):
     str
         The table/chart content from the row.
     """
-    return row.get("table_metadata", {}).get("table_content")
+    if modality == "text":
+        content = row.get("table_metadata", {}).get("table_content")
+    elif modality == "image":
+        content = _format_image_input_string(row.get("content"))
+    elif modality == "text_image":
+        text = row.get("table_metadata", {}).get("table_content")
+        image = row.get("content")
+        content = _format_text_image_pair_input_string(text, image)
+
+    return content
 
 
-def _get_pandas_image_content(row):
+def _get_pandas_image_content(row, modality="text"):
     """
     Extracts image caption content from a DataFrame row.
 
@@ -311,10 +351,28 @@ def _get_pandas_image_content(row):
     str
         The image caption from the row.
     """
-    return row.get("image_metadata", {}).get("caption")
+    subtype = row.get("content_metadata", {}).get("subtype")
+    if modality == "text":
+        if subtype == "page_image":
+            content = row.get("image_metadata", {}).get("text")
+        else:
+            content = row.get("image_metadata", {}).get("caption")
+    elif modality == "image":
+        content = _format_image_input_string(row.get("content"))
+    elif modality == "text_image":
+        if subtype == "page_image":
+            text = row.get("image_metadata", {}).get("text")
+        else:
+            text = row.get("image_metadata", {}).get("caption")
+        image = row.get("content")
+        content = _format_text_image_pair_input_string(text, image)
 
+    # A workaround to save memory.
+    row["content"] = ""
+    return content
 
-def _get_pandas_audio_content(row):
+
+def _get_pandas_audio_content(row, modality="text"):
     """
     A pandas UDF used to select extracted audio transcription to be used to create embeddings.
     """
@@ -408,6 +466,23 @@ def _concatenate_extractions_pandas(
 # ------------------------------------------------------------------------------
 
 
+def does_model_support_multimodal_embeddings(model: str) -> bool:
+    """
+    Checks if a given model supports multi-modal embeddings.
+
+    Parameters
+    ----------
+    model : str
+        The name of the model.
+
+    Returns
+    -------
+    bool
+        True if the model supports multi-modal embeddings, False otherwise.
+    """
+    return model in MULTI_MODAL_MODELS
+
+
 def transform_create_text_embeddings_internal(
     df_transform_ledger: pd.DataFrame,
     task_config: Dict[str, Any],
@@ -460,6 +535,15 @@ def transform_create_text_embeddings_internal(
         ContentTypeEnum.AUDIO: _get_pandas_audio_content,
         ContentTypeEnum.VIDEO: lambda x: None,  # Not supported yet.
     }
+    task_type_to_modality = {
+        ContentTypeEnum.TEXT: task_config.get("text_elements_modality") or transform_config.text_elements_modality,
+        ContentTypeEnum.STRUCTURED: (
+            task_config.get("structured_elements_modality") or transform_config.structured_elements_modality
+        ),
+        ContentTypeEnum.IMAGE: task_config.get("image_elements_modality") or transform_config.image_elements_modality,
+        ContentTypeEnum.AUDIO: task_config.get("audio_elements_modality") or transform_config.audio_elements_modality,
+        ContentTypeEnum.VIDEO: lambda x: None,  # Not supported yet.
+    }
 
     def _content_type_getter(row):
         return row["content_metadata"]["type"]
@@ -480,7 +564,7 @@ def transform_create_text_embeddings_internal(
         # Extract content and normalize empty or non-str to None
         extracted_content = (
             df_content["metadata"]
-            .apply(content_getter)
+            .apply(partial(content_getter, modality=task_type_to_modality[content_type]))
             .apply(lambda x: x.strip() if isinstance(x, str) and x.strip() else None)
         )
         df_content["_content"] = extracted_content
@@ -488,9 +572,15 @@ def transform_create_text_embeddings_internal(
         # Prepare batches for only valid (non-None) content
         valid_content_mask = df_content["_content"].notna()
         if valid_content_mask.any():
-            filtered_content_batches = _generate_batches(
-                df_content.loc[valid_content_mask, "_content"].tolist(), batch_size=transform_config.batch_size
-            )
+            filtered_content_list = df_content.loc[valid_content_mask, "_content"].tolist()
+            filtered_content_batches = _generate_batches(filtered_content_list, batch_size=transform_config.batch_size)
+
+            if model_name in MULTI_MODAL_MODELS:
+                modality_list = [task_type_to_modality[content_type]] * len(filtered_content_list)
+                modality_batches = _generate_batches(modality_list, batch_size=transform_config.batch_size)
+            else:
+                modality_batches = None
+
             content_embeddings = _async_runner(
                 filtered_content_batches,
                 api_key,
@@ -500,6 +590,7 @@ def transform_create_text_embeddings_internal(
                 transform_config.input_type,
                 transform_config.truncate,
                 False,
+                modalities=modality_batches,
             )
             # Build a simple row index -> embedding map
             embeddings_dict = dict(

nv_ingest_api/util/image_processing/transforms.py

@@ -20,6 +20,9 @@ cv2.setNumThreads(1)
 DEFAULT_MAX_WIDTH = 1024
 DEFAULT_MAX_HEIGHT = 1280
 
+# Workaround for PIL.Image.DecompressionBombError
+Image.MAX_IMAGE_PIXELS = None
+
 logger = logging.getLogger(__name__)
 
 

nv_ingest_api/util/metadata/aggregators.py

@@ -201,6 +201,8 @@ def construct_image_metadata_from_base64(
     page_count: int,
     source_metadata: Dict[str, Any],
     base_unified_metadata: Dict[str, Any],
+    subtype: None | ContentTypeEnum | str = "",
+    text: str = "",
 ) -> List[Any]:
     """
     Extracts image data from a base64-encoded image string, decodes the image to get
@@ -252,6 +254,7 @@ def construct_image_metadata_from_base64(
             "line": -1,
             "span": -1,
         },
+        "subtype": subtype or "",
     }
 
     # Construct image metadata
@@ -259,7 +262,7 @@ def construct_image_metadata_from_base64(
         "image_type": DocumentTypeEnum.PNG,
         "structured_image_type": ContentTypeEnum.UNKNOWN,
         "caption": "",
-        "text": "",
+        "text": text,
         "image_location": bbox,
         "image_location_max_dimensions": (width, height),
         "height": height,

{nv_ingest_api-2025.7.15.dev20250715.dist-info → nv_ingest_api-2025.7.16.dev20250716.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: nv-ingest-api
-Version: 2025.7.15.dev20250715
+Version: 2025.7.16.dev20250716
 Summary: Python module with core document ingestion functions.
 Author-email: Jeremy Dyer <jdyer@nvidia.com>
 License:                                  Apache License

{nv_ingest_api-2025.7.15.dev20250715.dist-info → nv_ingest_api-2025.7.16.dev20250716.dist-info}/RECORD

@@ -7,7 +7,7 @@ nv_ingest_api/interface/transform.py,sha256=g6YnFR7TpEU0xNtzCvv6kqnFbuCwQ6vRMjjB
 nv_ingest_api/interface/utility.py,sha256=AL4l0cJNvTjG1MAe1YNTk1jbbPED3g4HCewzx6Ffcio,7296
 nv_ingest_api/internal/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
 nv_ingest_api/internal/enums/__init__.py,sha256=wQSlVx3T14ZgQAt-EPzEczQusXVW0W8yynnUaFFGE3s,143
-nv_ingest_api/internal/enums/common.py,sha256=HSj7qqNr6KXu_FIyK_Wvel24R-r8lV7dLA173z5XFBc,12321
+nv_ingest_api/internal/enums/common.py,sha256=lzDJ35VWfIwlL_Lx_q0dfHUuwEB7CXudHIQAilpjoRw,12611
 nv_ingest_api/internal/extract/__init__.py,sha256=uLsBITo_XfgbwpzqXUm1IYX6XlZrTfx6T1cIhdILwG8,140
 nv_ingest_api/internal/extract/audio/__init__.py,sha256=wQSlVx3T14ZgQAt-EPzEczQusXVW0W8yynnUaFFGE3s,143
 nv_ingest_api/internal/extract/audio/audio_extraction.py,sha256=_jf_UC_FTqZr-xEpwG8edwBzdDjM01gGhqm9ulOsDcY,6973
@@ -32,10 +32,10 @@ nv_ingest_api/internal/extract/pdf/engines/__init__.py,sha256=u4GnAZmDKRl0RwYGIR
 nv_ingest_api/internal/extract/pdf/engines/adobe.py,sha256=VT0dEqkU-y2uGkaCqxtKYov_Q8R1028UQVBchgMLca4,17466
 nv_ingest_api/internal/extract/pdf/engines/llama.py,sha256=PpKTqS8jGHBV6mKLGZWwjpfT8ga6Fy8ffrvL-gPAf2c,8182
 nv_ingest_api/internal/extract/pdf/engines/nemoretriever.py,sha256=XNYz4S2tMFBv0KFzXNERrVs-1raxJ_iIIXpBGlJFcD0,22987
-nv_ingest_api/internal/extract/pdf/engines/pdfium.py,sha256=vtdBue1EEQJsHcBuX3NdPutbLfyKPIzily6JOK6yV0w,22421
+nv_ingest_api/internal/extract/pdf/engines/pdfium.py,sha256=8hUJUdpx6FhOBgabFmGhJiAQdl12kR8YoSbUfN-geOk,23506
 nv_ingest_api/internal/extract/pdf/engines/tika.py,sha256=6GyR2l6EsgNZl9jnYDXLeKNK9Fj2Mw9y2UWDq-eSkOc,3169
 nv_ingest_api/internal/extract/pdf/engines/unstructured_io.py,sha256=jrv2B4VZAH4PevAQrFz965qz8UyXq3rViiOTbGLejec,14908
-nv_ingest_api/internal/extract/pdf/engines/pdf_helpers/__init__.py,sha256=Jk3wrQ2CZs167juvEZ-uV6qXWQjR08hhIu8otk2MWj4,4931
+nv_ingest_api/internal/extract/pdf/engines/pdf_helpers/__init__.py,sha256=4bvN6LsPksLicI6jM0JqbJFiOZNHEcuc8MVVW4XfgV8,5875
 nv_ingest_api/internal/extract/pptx/__init__.py,sha256=HIHfzSig66GT0Uk8qsGBm_f13fKYcPtItBicRUWOOVA,183
 nv_ingest_api/internal/extract/pptx/pptx_extractor.py,sha256=o-0P2dDyRFW37uQi_lKk6-eFozTcZvbq-2Y4I0EBMIY,7749
 nv_ingest_api/internal/extract/pptx/engines/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
@@ -82,7 +82,7 @@ nv_ingest_api/internal/schemas/message_brokers/request_schema.py,sha256=LZX_wXDx
 nv_ingest_api/internal/schemas/message_brokers/response_schema.py,sha256=4b275HlzBSzpmuE2wdoeaGKPCdKki3wuWldtRIfrj8w,727
 nv_ingest_api/internal/schemas/meta/__init__.py,sha256=wQSlVx3T14ZgQAt-EPzEczQusXVW0W8yynnUaFFGE3s,143
 nv_ingest_api/internal/schemas/meta/base_model_noext.py,sha256=8hXU1uuiqZ6t8EsoZ8vlC5EFf2zSZrKEX133FcfZMwI,316
-nv_ingest_api/internal/schemas/meta/ingest_job_schema.py,sha256=szDvgc2A_JetD2Jyewyl4ac4lwpy3NiLxD9dOYz42sM,8116
+nv_ingest_api/internal/schemas/meta/ingest_job_schema.py,sha256=ceYQjRjhBSDbbZ6q-Db7Y6GHVOvWPdGAMb3TX1vMWfY,8321
 nv_ingest_api/internal/schemas/meta/metadata_schema.py,sha256=VnAzkSFat_ckI19mlwQTlFrvP6EZVCwyNl9bt51b8oU,7193
 nv_ingest_api/internal/schemas/mutate/__init__.py,sha256=wQSlVx3T14ZgQAt-EPzEczQusXVW0W8yynnUaFFGE3s,143
 nv_ingest_api/internal/schemas/mutate/mutate_image_dedup_schema.py,sha256=k1JOdlPPpsipc0XhHf-9YxJ_-W0HvpVE1ZhYmr7fzj0,395
@@ -92,14 +92,14 @@ nv_ingest_api/internal/schemas/store/store_image_schema.py,sha256=p2LGij9i6sG6RY
 nv_ingest_api/internal/schemas/transform/__init__.py,sha256=wQSlVx3T14ZgQAt-EPzEczQusXVW0W8yynnUaFFGE3s,143
 nv_ingest_api/internal/schemas/transform/transform_image_caption_schema.py,sha256=OtM1iPw26uioC3mghbOJQurKGg641uQfhASH462VqOY,578
 nv_ingest_api/internal/schemas/transform/transform_image_filter_schema.py,sha256=31ThI5fr0yyENeJeE1xMAA-pxk1QVJLwM842zMate_k,429
-nv_ingest_api/internal/schemas/transform/transform_text_embedding_schema.py,sha256=ongmHkJA2953f9_RI7ZYzf5BUnFzVL6Al5E8WKyfgw4,885
+nv_ingest_api/internal/schemas/transform/transform_text_embedding_schema.py,sha256=RZCISA8CUqKiY8eJuk4uWxzo4PZ-fuYdzMO7_LYFkoM,1117
 nv_ingest_api/internal/schemas/transform/transform_text_splitter_schema.py,sha256=D9K8tvu-tkEBQkZo7uuRzgrHdGyM3ZcNycHbHy5HV2E,791
 nv_ingest_api/internal/store/__init__.py,sha256=wQSlVx3T14ZgQAt-EPzEczQusXVW0W8yynnUaFFGE3s,143
 nv_ingest_api/internal/store/embed_text_upload.py,sha256=maxb4FPsBvWgvlrjAPEBlRZEFdJX5NxPG-p8kUbzV7I,9898
 nv_ingest_api/internal/store/image_upload.py,sha256=GNlY4k3pfcHv3lzXxkbmGLeHFsf9PI25bkBn6Xn9h3I,9654
 nv_ingest_api/internal/transform/__init__.py,sha256=wQSlVx3T14ZgQAt-EPzEczQusXVW0W8yynnUaFFGE3s,143
 nv_ingest_api/internal/transform/caption_image.py,sha256=0ILCG2F8ESqKtZiPUM-6F1BHUflFZ76Dzi2GNzkE-lU,8517
-nv_ingest_api/internal/transform/embed_text.py,sha256=A8JMotTkC8KQ0pmz4AIJhaKebza6JzhQ0aEnHX2oHY8,16539
+nv_ingest_api/internal/transform/embed_text.py,sha256=kvVGlNH1S91UENXWLD31uh3KzlfJYOlYitpIFMsyowU,20033
 nv_ingest_api/internal/transform/split_text.py,sha256=-kwpRWSVZrPldm1hn3-tVz_TkzuKM-kPvNU3HTp9zOY,7476
 nv_ingest_api/util/__init__.py,sha256=wQSlVx3T14ZgQAt-EPzEczQusXVW0W8yynnUaFFGE3s,143
 nv_ingest_api/util/control_message/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
@@ -123,7 +123,7 @@ nv_ingest_api/util/image_processing/__init__.py,sha256=Jiy8C1ZuSrNb_eBM1ZTV9IKFI
 nv_ingest_api/util/image_processing/clustering.py,sha256=sUGlZI4cx1q8h4Pns1N9JVpdfSM2BOH8zRmn9QFCtzI,9236
 nv_ingest_api/util/image_processing/processing.py,sha256=LSoDDEmahr7a-qSS12McVcowRe3dOrAZwa1h-PD_JPQ,6554
 nv_ingest_api/util/image_processing/table_and_chart.py,sha256=bxOu9PZYkG_WFCDGw_JLaO60S2pDSN8EOWK3xkIwr2A,14376
-nv_ingest_api/util/image_processing/transforms.py,sha256=CJVGQgUvHk_mzihR8ZZrvwJUBgUYcgFAKzXyRTmKdCE,23371
+nv_ingest_api/util/image_processing/transforms.py,sha256=3-xeUerc2AaXJTYuR23EjwdtjRQ8F85pS5D9zxR4cLA,23452
 nv_ingest_api/util/imports/__init__.py,sha256=wQSlVx3T14ZgQAt-EPzEczQusXVW0W8yynnUaFFGE3s,143
 nv_ingest_api/util/imports/callable_signatures.py,sha256=e2bJB1pmkN4Ee-Bf-VggOSBaQ4RXofWF5eKkWXgIj2U,1855
 nv_ingest_api/util/imports/dynamic_resolvers.py,sha256=7GByV_-8z2X0tnVoabCxVioxOP3sYMros3ZllVAW-wY,4343
@@ -135,7 +135,7 @@ nv_ingest_api/util/message_brokers/simple_message_broker/broker.py,sha256=h9Q4q_
 nv_ingest_api/util/message_brokers/simple_message_broker/ordered_message_queue.py,sha256=3p-LRqG8qLnsfEhBNf73_DG22C08JKahTqUvPLS2Apg,2554
 nv_ingest_api/util/message_brokers/simple_message_broker/simple_client.py,sha256=fh7Q0wO5H_FtrHV1VdT6V66aZNqglOh_2XdkfLt8hgg,15722
 nv_ingest_api/util/metadata/__init__.py,sha256=HIHfzSig66GT0Uk8qsGBm_f13fKYcPtItBicRUWOOVA,183
-nv_ingest_api/util/metadata/aggregators.py,sha256=Y5JSKuLhhk_ldpzT3eRIcVg7QM7cTNhfQZn4g5bcbq4,15884
+nv_ingest_api/util/metadata/aggregators.py,sha256=YYdvJ1E04eGFZKKHUxXoH6mzLg8nor9Smvnv0qzqK5w,15988
 nv_ingest_api/util/multi_processing/__init__.py,sha256=4fojP8Rp_5Hu1YAkqGylqTyEZ-HBVVEunn5Z9I99swA,242
 nv_ingest_api/util/multi_processing/mp_pool_singleton.py,sha256=dTfP82DgGPaXEJH3jywTO8rNlLZUniD4FFzwv84_giE,7372
 nv_ingest_api/util/nim/__init__.py,sha256=UqbiXFCqjWcjNvoduXd_0gOUOGBT8JvppiYHOmMyneA,1775
@@ -153,8 +153,8 @@ nv_ingest_api/util/service_clients/rest/rest_client.py,sha256=dZ-jrk7IK7oNtHoXFS
 nv_ingest_api/util/string_processing/__init__.py,sha256=mkwHthyS-IILcLcL1tJYeF6mpqX3pxEw5aUzDGjTSeU,1411
 nv_ingest_api/util/system/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
 nv_ingest_api/util/system/hardware_info.py,sha256=ORZeKpH9kSGU_vuPhyBwkIiMyCViKUX2CP__MCjrfbU,19463
-nv_ingest_api-2025.7.15.dev20250715.dist-info/licenses/LICENSE,sha256=xx0jnfkXJvxRnG63LTGOxlggYnIysveWIZ6H3PNdCrQ,11357
-nv_ingest_api-2025.7.15.dev20250715.dist-info/METADATA,sha256=OWZyeCR9DZ23SdT0RcMdodCkxR508CZZaVczdM3qXPE,13947
-nv_ingest_api-2025.7.15.dev20250715.dist-info/WHEEL,sha256=_zCd3N1l69ArxyTb8rzEoP9TpbYXkqRFSNOD5OuxnTs,91
-nv_ingest_api-2025.7.15.dev20250715.dist-info/top_level.txt,sha256=abjYMlTJGoG5tOdfIB-IWvLyKclw6HLaRSc8MxX4X6I,14
-nv_ingest_api-2025.7.15.dev20250715.dist-info/RECORD,,
+nv_ingest_api-2025.7.16.dev20250716.dist-info/licenses/LICENSE,sha256=xx0jnfkXJvxRnG63LTGOxlggYnIysveWIZ6H3PNdCrQ,11357
+nv_ingest_api-2025.7.16.dev20250716.dist-info/METADATA,sha256=RaPAkQ4Dtkkrn6hi9Va1t2XDpDgRbe-bFqmCVL3IlEA,13947
+nv_ingest_api-2025.7.16.dev20250716.dist-info/WHEEL,sha256=_zCd3N1l69ArxyTb8rzEoP9TpbYXkqRFSNOD5OuxnTs,91
+nv_ingest_api-2025.7.16.dev20250716.dist-info/top_level.txt,sha256=abjYMlTJGoG5tOdfIB-IWvLyKclw6HLaRSc8MxX4X6I,14
+nv_ingest_api-2025.7.16.dev20250716.dist-info/RECORD,,