Add docs for segment_audio param (#1818)

Co-authored-by: Kurt Heiss <kheiss@nvidia.com>
Co-authored-by: greptile-apps[bot] <165735046+greptile-apps[bot]@users.noreply.github.com>

Author: 3 people

client/src/nv_ingest_client/client/interface.py

@@ -1070,6 +1070,9 @@ def extract(self, **kwargs: Any) -> "Ingestor":
             - extract_charts: bool, extract charts (default True)
             - extract_infographics: bool, extract infographics (default False)
             - extract_page_as_image: bool, extract full page as image (default False)
+            - extract_audio_params: dict, audio extraction options such as
+              endpoint settings and `segment_audio` for sentence-like ASR
+              segmentation when using a hosted Parakeet service
             - table_output_format: str, format for table output (default "markdown")
             - auto_dedup: bool, auto-enable bbox deduplication when extracting both
               structured elements and images (default True). Set to False to disable.

docs/docs/extraction/audio.md

@@ -82,10 +82,13 @@ Use the following procedure to run the NIM locally.
         .extract(
             document_type="wav",  # Ingestor should detect type automatically in most cases
             extract_method="audio",
+            extract_audio_params={
+                "segment_audio": True,
+            },
         )
     )
     ```
-
+To generate one extracted element for each sentence-like ASR segment, include `extract_audio_params={"segment_audio": True}` when calling `.extract(...)`. This option applies when audio extraction runs with a Parakeet NIM (either locally through Docker or remotely via NVCF) but has no effect when using the local Hugging Face Parakeet model.
 
     !!! tip
 
@@ -117,6 +120,7 @@ Instead of running the pipeline locally, you can use NVCF to perform inference b
                 "auth_token": "<API key>",
                 "function_id": "<function ID>",
                 "use_ssl": True,
+                "segment_audio": True,
             },
         )
     )

docs/docs/extraction/nv-ingest-python-api.md

@@ -549,11 +549,13 @@ ingestor = Ingestor().files("audio_file.mp3")
 
 ingestor = ingestor.extract(
         document_type="mp3",
+        extract_method="audio",
         extract_text=True,
         extract_tables=False,
         extract_charts=False,
         extract_images=False,
         extract_infographics=False,
+        extract_audio_params={"segment_audio": True},
     ).split(
         tokenizer="meta-llama/Llama-3.2-1B",
         chunk_size=150,
@@ -563,8 +565,7 @@ ingestor = ingestor.extract(
 
 results = ingestor.ingest()
 ```
-
-
+Set extract_audio_params={"segment_audio": True} to output sentence-like audio segments as distinct extracted elements. This setting applies only when audio extraction runs through a hosted Parakeet endpoint—such as the Parakeet ASR NIM or NVCF—and has no effect when using the local Hugging Face Parakeet model.
 
 ## Related Topics
 

docs/docs/extraction/python-api-reference.md

@@ -641,11 +641,13 @@ ingestor = Ingestor().files("audio_file.mp3")
 
 ingestor = ingestor.extract(
         document_type="mp3",
+        extract_method="audio",
         extract_text=True,
         extract_tables=False,
         extract_charts=False,
         extract_images=False,
         extract_infographics=False,
+        extract_audio_params={"segment_audio": True},
     ).split(
         tokenizer="meta-llama/Llama-3.2-1B",
         chunk_size=150,
@@ -656,6 +658,12 @@ ingestor = ingestor.extract(
 results = ingestor.ingest()
 ```
 
+Set `extract_audio_params={"segment_audio": True}` to emit sentence-like
+audio segments as separate extracted elements. This option only takes effect
+when audio extraction is performed through a hosted Parakeet endpoint--such as the
+Parakeet ASR NIM or NVCF--and does not affect behavior when using the local Hugging
+Face Parakeet model.
+
 
 
 ## Related Topics