nv-ingest-api 2025.10.14.dev20251014__py3-none-any.whl → 2025.10.16.dev20251016__py3-none-any.whl

nv_ingest_api/internal/primitives/nim/model_interface/ocr.py

@@ -758,8 +758,8 @@ def get_ocr_model_name(ocr_grpc_endpoint=None, default_model_name=DEFAULT_OCR_MO
     if ocr_model_name is not None:
         return ocr_model_name
 
-    # 2. If no gRPC endpoint is provided, fall back to the default immediately.
-    if not ocr_grpc_endpoint:
+    # 2. If no gRPC endpoint is provided or the endpoint is a NVCF endpoint, fall back to the default immediately.
+    if (not ocr_grpc_endpoint) or ("grpc.nvcf.nvidia.com" in ocr_grpc_endpoint):
         logger.debug(f"No OCR gRPC endpoint provided. Falling back to default model name '{default_model_name}'.")
         return default_model_name
 

nv_ingest_api/internal/schemas/meta/ingest_job_schema.py

@@ -32,6 +32,17 @@ class TracingOptionsSchema(BaseModelNoExt):
     total_pages: Optional[int] = None
 
 
+# PDF Configuration Schema
+class PdfConfigSchema(BaseModelNoExt):
+    """PDF-specific configuration options for job submission.
+
+    Note: split_page_count accepts any positive integer but will be clamped
+    to [1, 128] range by the server at runtime.
+    """
+
+    split_page_count: Annotated[int, Field(ge=1)] = 32
+
+
 # Ingest Task Schemas
 
 
@@ -270,6 +281,7 @@ class IngestJobSchema(BaseModelNoExt):
     job_id: Union[str, int]
     tasks: List[IngestTaskSchema]
     tracing_options: Optional[TracingOptionsSchema] = None
+    pdf_config: Optional[PdfConfigSchema] = None
 
 
 # ------------------------------------------------------------------------------

nv_ingest_api/util/service_clients/rest/rest_client.py

@@ -3,7 +3,6 @@
 # SPDX-License-Identifier: Apache-2.0
 
 import logging
-import os
 import re
 import time
 from typing import Any, Union, Tuple, Optional, Dict, Callable
@@ -104,6 +103,17 @@ class RestClient(MessageBrokerClientBase):
             Default timeout in seconds for waiting for data after connection. Default is None.
         http_allocator : Optional[Callable[[], Any]], optional
             A callable that returns an HTTP client instance. If None, `requests.Session()` is used.
+        **kwargs : dict
+            Additional keyword arguments. Supported keys:
+            - api_version : str, optional
+                API version to use ('v1' or 'v2'). Defaults to 'v1' if not specified.
+                Invalid versions will log a warning and fall back to 'v1'.
+            - base_url : str, optional
+                Override the generated base URL.
+            - headers : dict, optional
+                Additional headers to include in requests.
+            - auth : optional
+                Authentication configuration for requests.
 
         Returns
         -------
@@ -138,8 +148,19 @@ class RestClient(MessageBrokerClientBase):
                 )
                 self._client = requests.Session()
 
-        # Allow API version override via environment variable or kwargs
-        api_version = kwargs.get("api_version") or os.getenv("NV_INGEST_API_VERSION", "v1")
+        # Validate and normalize API version to prevent misconfiguration
+        # Default to v1 for backwards compatibility if not explicitly provided
+        VALID_API_VERSIONS = {"v1", "v2"}
+        raw_api_version = kwargs.get("api_version", "v1")
+        api_version = str(raw_api_version).strip().lower()
+
+        if api_version not in VALID_API_VERSIONS:
+            logger.warning(
+                f"Invalid API version '{raw_api_version}' specified. "
+                f"Valid versions are: {VALID_API_VERSIONS}. Falling back to 'v1'."
+            )
+            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"

nv_ingest_api/util/string_processing/yaml.py

@@ -1,5 +1,6 @@
 import os
 import re
+from typing import Optional
 
 # This regex finds all forms of environment variables:
 # $VAR, ${VAR}, $VAR|default, and ${VAR|default}
@@ -20,12 +21,46 @@ def _replacer(match: re.Match) -> str:
     var_name = match.group("braced") or match.group("named")
     default_val = match.group("braced_default") or match.group("named_default")
 
-    # Get value from environment, or use default.
-    value = os.environ.get(var_name, default_val)
+    # First try the primary env var
+    value = os.environ.get(var_name)
+    if value is not None:
+        return value
 
-    if value is None:
+    # If primary is missing, try the default.
+    resolved_default = _resolve_default_with_single_fallback(default_val)
+
+    if resolved_default is None:
         return ""
-    return value
+
+    return resolved_default
+
+
+def _is_var_ref(token: str) -> Optional[str]:
+    """If token is a $VAR or ${VAR} reference, return VAR name; else None."""
+    if not token:
+        return None
+    if token.startswith("${") and token.endswith("}"):
+        inner = token[2:-1]
+        return inner if re.fullmatch(r"\w+", inner) else None
+    if token.startswith("$"):
+        inner = token[1:]
+        return inner if re.fullmatch(r"\w+", inner) else None
+    return None
+
+
+def _resolve_default_with_single_fallback(default_val: Optional[str]) -> Optional[str]:
+    """
+    Support a single-level fallback where the default itself can be another env var.
+    For example, in $A|$B or ${A|$B}, we try B if A missing.
+    """
+    if default_val is None:
+        return None
+
+    var = _is_var_ref(default_val)
+    if var is not None:
+        return os.environ.get(var, None)
+
+    return default_val
 
 
 def substitute_env_vars_in_yaml_content(raw_content: str) -> str:
@@ -35,6 +70,8 @@ def substitute_env_vars_in_yaml_content(raw_content: str) -> str:
     This function finds all occurrences of environment variable placeholders
     ($VAR, ${VAR}, $VAR|default, ${VAR|default}) in the input string
     and replaces them with their corresponding environment variable values.
+    Also supports a single fallback to another env var: $VAR|$OTHER, ${VAR|$OTHER}
+    Quoted defaults are preserved EXACTLY as written (e.g., 'a,b' keeps quotes).
 
     Args:
         raw_content: The raw string content of a YAML file.

{nv_ingest_api-2025.10.14.dev20251014.dist-info → nv_ingest_api-2025.10.16.dev20251016.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: nv-ingest-api
-Version: 2025.10.14.dev20251014
+Version: 2025.10.16.dev20251016
 Summary: Python module with core document ingestion functions.
 Author-email: Jeremy Dyer <jdyer@nvidia.com>
 License:                                  Apache License

{nv_ingest_api-2025.10.14.dev20251014.dist-info → nv_ingest_api-2025.10.16.dev20251016.dist-info}/RECORD

@@ -58,7 +58,7 @@ nv_ingest_api/internal/primitives/nim/model_interface/decorators.py,sha256=qwubk
 nv_ingest_api/internal/primitives/nim/model_interface/deplot.py,sha256=TvKdk6PTuI1WNhRmNNrvygaI_DIutkJkDL-XdtLZQac,10787
 nv_ingest_api/internal/primitives/nim/model_interface/helpers.py,sha256=jqbEbavvr9giODpzsGQSRDu5yZ4YfNfKAQfqUm9yUDI,11698
 nv_ingest_api/internal/primitives/nim/model_interface/nemoretriever_parse.py,sha256=WysjDZeegclO3mZgVcGOwzWbr8wSI4pWRiYD4iC2EXo,7098
-nv_ingest_api/internal/primitives/nim/model_interface/ocr.py,sha256=jFPB0h2SAq4RrZqiHdh28bNHvrkm0RWKZEx-ei-ahlU,29521
+nv_ingest_api/internal/primitives/nim/model_interface/ocr.py,sha256=WuX-veTC510TbvMWYGYP6WCzjYCbUBAUc5ovJUWCrFU,29607
 nv_ingest_api/internal/primitives/nim/model_interface/parakeet.py,sha256=3XXJkeJaVFe_iIfNn_bDYn79JN20besjZHiNZ5dEnZQ,12778
 nv_ingest_api/internal/primitives/nim/model_interface/text_embedding.py,sha256=lFhppNqrq5X_fzbCWKphvZQMzaJd3gHrkWsyJORzFrU,5010
 nv_ingest_api/internal/primitives/nim/model_interface/vlm.py,sha256=sUDKKlVqKjiHEGr2D04I7S4pDfnLR8b-NplV1pf5GVQ,6240
@@ -84,7 +84,7 @@ 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=23k85wUArC9J1g0_4f1KRXYHikAB2hgsH8C5Gw13uVQ,10414
+nv_ingest_api/internal/schemas/meta/ingest_job_schema.py,sha256=oPBoukRAnLW8BH6iKB0A_WIdewi_Go0NlxrakBwnswo,10782
 nv_ingest_api/internal/schemas/meta/metadata_schema.py,sha256=FDD6yq-QxW8yDwn0Bq6bmWakX41ABMn3cytrvCbT-Po,11961
 nv_ingest_api/internal/schemas/meta/udf.py,sha256=GgzqbZOlipQgMpDhbXLqbF8xrHenj_hMNqhR_P-1ynw,779
 nv_ingest_api/internal/schemas/mutate/__init__.py,sha256=wQSlVx3T14ZgQAt-EPzEczQusXVW0W8yynnUaFFGE3s,143
@@ -158,16 +158,16 @@ nv_ingest_api/util/service_clients/kafka/__init__.py,sha256=uLsBITo_XfgbwpzqXUm1
 nv_ingest_api/util/service_clients/redis/__init__.py,sha256=wQSlVx3T14ZgQAt-EPzEczQusXVW0W8yynnUaFFGE3s,143
 nv_ingest_api/util/service_clients/redis/redis_client.py,sha256=3NLecvIvVN1v-sA7d7G-_f6qJVZyfJE2H8Iu5KG3Aew,37417
 nv_ingest_api/util/service_clients/rest/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
-nv_ingest_api/util/service_clients/rest/rest_client.py,sha256=laFVWLBFPlycLZnj6DP8zTABfSgr-FyRS2O_EbpzW_Y,22708
+nv_ingest_api/util/service_clients/rest/rest_client.py,sha256=7ymPxhuN9SP8nPSVepqqbvUxXPaTVunq2aC2bDbg98g,23684
 nv_ingest_api/util/string_processing/__init__.py,sha256=mkwHthyS-IILcLcL1tJYeF6mpqX3pxEw5aUzDGjTSeU,1411
 nv_ingest_api/util/string_processing/configuration.py,sha256=2HS08msccuPCT0fn_jfXRo9_M6hCZ59OxKLxG_47HRY,29888
-nv_ingest_api/util/string_processing/yaml.py,sha256=6SW2O6wbXRhGbhETMbtXjYCZn53HeCNOP6a96AaxlHs,1454
+nv_ingest_api/util/string_processing/yaml.py,sha256=4Zdmc4474lUZn6kznqaNTlQJwsmRnnJQZ-DvAWLu-zo,2678
 nv_ingest_api/util/system/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
 nv_ingest_api/util/system/hardware_info.py,sha256=1UFM8XE6M3pgQcpbVsCsqDQ7Dj-zzptL-XRE-DEu9UA,27213
-nv_ingest_api-2025.10.14.dev20251014.dist-info/licenses/LICENSE,sha256=xx0jnfkXJvxRnG63LTGOxlggYnIysveWIZ6H3PNdCrQ,11357
+nv_ingest_api-2025.10.16.dev20251016.dist-info/licenses/LICENSE,sha256=xx0jnfkXJvxRnG63LTGOxlggYnIysveWIZ6H3PNdCrQ,11357
 udfs/__init__.py,sha256=pXFqPgXIUqHDfj7SAR1Q19tt8KwGv_iMvhHyziz4AYM,205
-udfs/llm_summarizer_udf.py,sha256=t_ZFoz0e03uECYcRw4IabRj0GBlwAoJkJn13NL2wbsI,7217
-nv_ingest_api-2025.10.14.dev20251014.dist-info/METADATA,sha256=rcnea6vVn8B008pZg9on2O8nBSuwFRIM25KZHJ8Lb2E,14086
-nv_ingest_api-2025.10.14.dev20251014.dist-info/WHEEL,sha256=_zCd3N1l69ArxyTb8rzEoP9TpbYXkqRFSNOD5OuxnTs,91
-nv_ingest_api-2025.10.14.dev20251014.dist-info/top_level.txt,sha256=I1lseG9FF0CH93SPx4kFblsxFuv190cfzaas_CLNIiw,19
-nv_ingest_api-2025.10.14.dev20251014.dist-info/RECORD,,
+udfs/llm_summarizer_udf.py,sha256=9kM3W08FVnNHIRrDDZEuQRnPJmOdU2N-YTcnaBy8NdU,7267
+nv_ingest_api-2025.10.16.dev20251016.dist-info/METADATA,sha256=fsVHTpz4y-WxO1gDJOu87u8fSfAxOFrabEkcYf-Wvg8,14086
+nv_ingest_api-2025.10.16.dev20251016.dist-info/WHEEL,sha256=_zCd3N1l69ArxyTb8rzEoP9TpbYXkqRFSNOD5OuxnTs,91
+nv_ingest_api-2025.10.16.dev20251016.dist-info/top_level.txt,sha256=I1lseG9FF0CH93SPx4kFblsxFuv190cfzaas_CLNIiw,19
+nv_ingest_api-2025.10.16.dev20251016.dist-info/RECORD,,

udfs/llm_summarizer_udf.py

@@ -2,8 +2,9 @@
 """
 LLM Content Summarizer UDF for NV-Ingest Pipeline
 
-This UDF uses an LLM to generate concise summaries of text content chunks. These summaries are added to the metadata
-for enhanced downstream processing and search capabilities.
+Generates document summaries using NVIDIA-hosted LLMs. This production UDF demonstrates how to extract the pipeline
+payload, run custom code (summarization), and inject results into the metadata for downstream usecases (such as
+retrieval).
 
 These variables can be set in the environment before running the pipeline. These can be treated as kwargs.
 - NVIDIA_API_KEY: API key for NVIDIA NIM endpoints (required)
@@ -14,16 +15,14 @@ These variables can be set in the environment before running the pipeline. These
 - MAX_CONTENT_LENGTH: Maximum content length to send to API (default: 12000)
 TODO: Implement this
 - NUM_CHUNKS: (Optional) Number of first and last pages to summarize. default=1
+
+More info can be found in `examples/udfs/README.md`
 """
 
 import logging
 import os
 import time
 
-# REMOVE BEFORE MERGING
-# import yaml
-# from pathlib import Path
-
 
 logger = logging.getLogger(__name__)