NVIDIA · kheiss-uwzoo · Apr 22, 2026 · Mar 13, 2026 · Apr 22, 2026 · greptile-apps
@@ -27,7 +27,7 @@ to transcribe speech to text, which is then embedded by using the NeMo Retriever
 
 !!! important
 
-    Due to limitations in available VRAM controls in the current release, the parakeet-1-1b-ctc-en-us ASR NIM microservice must run on a [dedicated additional GPU](support-matrix.md). For the full list of requirements, refer to [Support Matrix](support-matrix.md).
+    Due to limitations in available VRAM controls in the current release, the RIVA ASR NIM microservice must run on a [dedicated additional GPU](support-matrix.md). For the full list of requirements, refer to [Support Matrix](https://docs.nvidia.com/deeplearning/riva/user-guide/docs/support-matrix/support-matrix.html).
 
 This pipeline enables users to retrieve speech files at the segment level.
 

@@ -35,7 +35,7 @@ Before you use this documentation, you need the following:
 ### Run Your First Test
 
 ```bash
-# 1. Navigate to the nv-ingest-harness directory
+# 1. Navigate to the harness directory
 cd tools/harness
 
 # 2. Install dependencies
@@ -181,8 +181,8 @@ Example:
 # YAML active section has api_version: v2
 # Dataset bo767 has extract_images: false
 # Override via environment variable (highest priority)
-EXTRACT_IMAGES=true API_VERSION=v1 uv run nv-ingest-harness-run --case=e2e --dataset=bo767
-# Result: Uses bo767 path, but extract_images=true (env override) and api_version=v1 (env override)
+EXTRACT_IMAGES=true API_VERSION=v2 uv run nv-ingest-harness-run --case=e2e --dataset=bo767
+# Result: Uses bo767 path, but extract_images=true (env override) and api_version=v2 (env override)
 ```
 
 **Precedence Details:**
@@ -636,7 +636,7 @@ tools/harness/artifacts/<test_name>_<timestamp>_UTC/
 Enable per-document element breakdown:
 
 ```bash
-uv run nv-ingest-harness-run --case=e2e --dataset=bo767 --doc-analysis
+uv run nv-ingest-harness-run --case=e2e --doc-analysis
 ```
 
 **Sample Output:**

@@ -164,7 +164,7 @@ Describes the structural location of content within a document.
 | `span`           | `int`                 | `-1`                       | Span identifier within a line, for finer granularity.                                                   |
 | `nearby_objects` | `NearbyObjectsSchema` | `NearbyObjectsSchema()`    | Information about objects (text, images, structured data) near the current content. See [NearbyObjectsSchema](#nearbyobjectsschema). |
 
-### `NearbyObjectsSchema` (Currently Unused)
+### `NearbyObjectsSchema` (Currently Unused) {#nearbyobjectsschema}
 Container for different types of nearby objects.
 
 | Field        | Type                   | Default Value                | Description                                                              |
@@ -243,7 +243,7 @@ Specific metadata for audio content.
 | `audio_transcript` | `str` | `""`          | Transcript of the audio content.                |
 | `audio_type`       | `str` | `""`          | Type or format of the audio (e.g., `mp3`, `wav`). |
 
-### `ErrorMetadataSchema` (Currently Unused)
+### `ErrorMetadataSchema` (Currently Unused) {#errormetadataschema}
 Metadata describing errors encountered during processing.
 
 | Field       | Type           | Default Value | Description                                                              |
@@ -253,7 +253,7 @@ Metadata describing errors encountered during processing.
 | `source_id` | `str`          | `""`          | Identifier of the source item that caused the error, if applicable.        |
 | `error_msg` | `str`          | *Required*    | The error message.                                                       |
 
-### `InfoMessageMetadataSchema` (Currently Unused)
+### `InfoMessageMetadataSchema` (Currently Unused) {#infomessagemetadataschema}
 Informational messages related to processing.
 
 | Field     | Type           | Default Value | Description                                                              |

@@ -60,7 +60,7 @@ For more information about the `Ingestor` class, see [Use the Python API](nv-ing
 For more information about the `vdb_upload` method, see [Upload Data](data-store.md).
 
 ```python
-from nv_ingest_client.client import Ingestor
+from nv_ingest_client.client.interface import Ingestor
 
 hostname="localhost"
 collection_name = "nv_ingest_collection"
@@ -142,7 +142,7 @@ you can use the `content_metadata` field to filter search results.
 The following example uses a filter expression to narrow results by department.
 
 ```python
-from nv_ingest_client.util.milvus import nvingest_retrieval
+from nv_ingest_client.util.vdb.milvus import nvingest_retrieval
 
 hostname="localhost"
 collection_name = "nv_ingest_collection"
@@ -157,14 +157,14 @@ q_results = []
 for que in queries:
     q_results.append(
         nvingest_retrieval(
-            [que], 
-            collection_name, 
-            milvus_uri=f"http://{hostname}:19530", 
-            embedding_endpoint=f"http://{hostname}:8012/v1",  
-            hybrid=sparse, 
-            top_k=top_k, 
-            model_name=model_name, 
-            gpu_search=False, 
+            [que],
+            collection_name=collection_name,
+            milvus_uri=f"http://{hostname}:19530",
+            embedding_endpoint=f"http://{hostname}:8012/v1",
+            hybrid=sparse,
+            top_k=top_k,
+            model_name=model_name,
+            gpu_search=False,
             _filter=filter_expr
         )
     )

@@ -82,7 +82,7 @@ For more information, refer to [Extract Specific Elements from PDFs](nv-ingest-p
 ```python
 Ingestor(client=client)
     .files("data/multimodal_test.pdf")
-    .extract(              
+    .extract(
         extract_text=True,
         extract_tables=True,
         extract_charts=True,

@@ -11,6 +11,7 @@ Before you begin using [NeMo Retriever Library](overview.md), ensure the followi
 ## Software Requirements
 
 - Linux operating systems (Ubuntu 22.04 or later recommended)
+- **Python 3.12 or later** (required for NeMo Retriever Library packages; see note below)
 - [Docker](https://docs.docker.com/engine/install/)
 - [Docker Compose](https://docs.docker.com/compose/install/)
 - [Docker Buildx](https://docs.docker.com/build/concepts/overview/#buildx) `>= 0.17` (Compose 2.40+ enforces this)
@@ -22,7 +23,7 @@ Before you begin using [NeMo Retriever Library](overview.md), ensure the followi
 
 !!! note
 
-    When you use UV, create the environment with Python 3.12 or later — for example, `uv venv --python 3.12`. This matches the `requires-python` metadata in the library packages.
+    Install **Python 3.12 or later** before creating your environment. Using Python 3.10 or 3.11 will cause dependency resolution failures when installing NeMo Retriever Library packages.
 
 
 

@@ -80,7 +80,7 @@ The caption task can call a vision-language model (VLM) with the following optio
 
 Example:
 ```python
-from nemo_retriever.client.interface import Ingestor
+from nv_ingest_client.client.interface import Ingestor
 
 ingestor = (
     Ingestor()
@@ -224,7 +224,7 @@ The `extract` method enables different types of data to be extracted.
 Use the following code to extract a single PDF file.
 
 ```python
-from nemo_retriever.client.interface import Ingestor
+from nv_ingest_client.client.interface import Ingestor
 
 # Initialize Ingestor with a local PDF file
 ingestor = Ingestor().files("path/to/document.pdf")
@@ -527,7 +527,7 @@ The caption task can call a VLM with optional prompt and reasoning overrides:
 
 Example:
 ```python
-from nemo_retriever.client.interface import Ingestor
+from nv_ingest_client.client.interface import Ingestor
 
 ingestor = (
     Ingestor()
@@ -662,7 +662,7 @@ For more information on environment variables, refer to [Environment Variables](
 Use the following code to extract mp3 audio content.
 
 ```python
-from nemo_retriever.client import Ingestor
+from nv_ingest_client.client.interface import Ingestor
 
 ingestor = Ingestor().files("audio_file.mp3")
 

@@ -184,7 +184,8 @@ The following examples demonstrate how to extract text, charts, tables, and imag
 <a id="ingest_python_example"></a>
 ```python
 import logging, os, time
-from nv_ingest_client.client import Ingestor, NvIngestClient
+from nv_ingest_client.client.interface import Ingestor
+from nv_ingest_client.client import NvIngestClient
 from nv_ingest_client.util.process_json_files import ingest_json_results_to_blob
 client = NvIngestClient(                                                                         
     message_client_port=7670,                                                               
@@ -535,7 +536,7 @@ docker compose \
 
 ## Specify MIG slices for NIM models
 
-When you deploy the pipeline with NIM models on MIG‑enabled GPUs, MIG device slices are requested and scheduled through the `values.yaml` file for the corresponding NIM microservice. For IBM Content-Aware Storage (CAS) deployments, this allows NIM pods to land only on nodes that expose the desired MIG profiles [raw.githubusercontent](https://raw.githubusercontent.com/NVIDIA/nv-ingest/main/helm/README.md).
+When you deploy NeMo Retriever Library with NIM models on MIG‑enabled GPUs, MIG device slices are requested and scheduled through the `values.yaml` file for the corresponding NIM microservice. For IBM Content-Aware Storage (CAS) deployments, this allows NeMo Retriever Library NIM pods to land only on nodes that expose the desired MIG profiles [raw.githubusercontent](https://raw.githubusercontent.com/NVIDIA/NeMo-Retriever/main/helm/README.md).
 
 To target a specific MIG profile—for example, a 3g.20gb slice on an A100, which is a hardware-partitioned virtual GPU instance that gives your workload a fixed mid-sized share of the A100’s compute plus 20 GB of dedicated GPU memory and behaves like a smaller independent GPU—for a given NIM, configure the `resources` and `nodeSelector` under that NIM’s values path in `values.yaml`.
 

@@ -13,10 +13,11 @@ The primary Python entry point for launching the Ray-based ingestion pipeline in
 ```python
 import time
 
-from nemo_retriever.framework.orchestration.ray.util.pipeline.pipeline_runners import run_pipeline
-from nemo_retriever.client import Ingestor, NemoRetrieverClient
-from nemo_retriever.util.message_brokers.simple_message_broker import SimpleClient
-from nemo_retriever.util.process_json_files import ingest_json_results_to_blob
+from nv_ingest.framework.orchestration.ray.util.pipeline.pipeline_runners import run_pipeline
+from nv_ingest_client.client.interface import Ingestor
+from nv_ingest_client.client import NvIngestClient
+from nv_ingest_api.util.message_brokers.simple_message_broker import SimpleClient
+from nv_ingest_client.util.process_json_files import ingest_json_results_to_blob
 
 def main():
     # Start the pipeline subprocess for library mode
@@ -122,24 +123,24 @@ To query for relevant snippets of the ingested content, and use them with an LLM
 ```python
 import os
 from openai import OpenAI
-from nemo_retriever.util.milvus import query
+from nv_ingest_client.util.vdb.milvus import nvingest_retrieval
 
 milvus_uri = "milvus.db"
 collection_name = "test"
 sparse=False
 
 queries = ["Which animal is responsible for the typos?"]
 
-retrieved_docs = query(
+retrieved_docs = nvingest_retrieval(
     queries,
-    collection_name,
+    collection_name=collection_name,
     milvus_uri=milvus_uri,
     hybrid=sparse,
     top_k=1,
 )
 
 # simple generation example
-extract = retrieved_docs[0][0]["entity"]["text"]
+extract = retrieved_docs[0][0].get("entity", retrieved_docs[0][0]).get("text", "")
 client = OpenAI(
   base_url = "https://integrate.api.nvidia.com/v1",
   api_key = os.environ["NVIDIA_API_KEY"]
@@ -239,8 +240,8 @@ It listens for ingestion requests on port `7671` from an external client.
 import logging
 import os
 
-from nemo_retriever.framework.orchestration.ray.util.pipeline.pipeline_runners import run_pipeline
-from nemo_retriever.util.logging.configuration import configure_logging as configure_local_logging
+from nv_ingest.framework.orchestration.ray.util.pipeline.pipeline_runners import run_pipeline
+from nv_ingest_api.util.logging.configuration import configure_logging as configure_local_logging
 
 # Configure the logger
 logger = logging.getLogger(__name__)
@@ -285,11 +286,11 @@ import logging
 import os
 import time
 
-from nemo_retriever.framework.orchestration.ray.util.pipeline.pipeline_runners import run_pipeline
-from nemo_retriever.util.logging.configuration import configure_logging as configure_local_logging
-from nemo_retriever.util.message_brokers.simple_message_broker import SimpleClient
-from nemo_retriever.client import Ingestor
-from nemo_retriever.client import NemoRetrieverClient
+from nv_ingest.framework.orchestration.ray.util.pipeline.pipeline_runners import run_pipeline
+from nv_ingest_api.util.logging.configuration import configure_logging as configure_local_logging
+from nv_ingest_api.util.message_brokers.simple_message_broker import SimpleClient
+from nv_ingest_client.client.interface import Ingestor
+from nv_ingest_client.client import NvIngestClient
 
 # Configure the logger
 logger = logging.getLogger(__name__)

@@ -7,6 +7,11 @@ Before you begin using [NeMo Retriever Library](overview.md), ensure that you ha
     NVIDIA Ingest (nv-ingest) has been renamed NeMo Retriever Library.
 
 
+## Software Requirements
+
+- **Python**: 3.12 or later. The NeMo Retriever Library core and harness require Python 3.12+; the client supports Python 3.11+. Using Python 3.10 or earlier will cause dependency resolution failures. For details, see [Prerequisites](prerequisites.md).
+
+
 ## Core and Advanced Pipeline Features
 
 The Nemo Retriever Library extraction core pipeline features run on a single A10G or better GPU. 
@@ -39,6 +44,7 @@ This includes the following:
 NeMo Retriever Library supports the following GPU hardware.
 
 - [RTX Pro 6000 Blackwell Server Edition](https://www.nvidia.com/en-us/data-center/rtx-pro-6000-blackwell-server-edition/)
+- [RTX PRO 4500 Blackwell](https://www.nvidia.com/en-us/products/workstations/professional-desktop-gpus/rtx-pro-4500/)
 - [DGX B200](https://www.nvidia.com/en-us/data-center/dgx-b200/)
 - [H200 NVL](https://www.nvidia.com/en-us/data-center/h200/)
 - [H100 Tensor Core GPU](https://www.nvidia.com/en-us/data-center/h100/)

@@ -304,7 +304,7 @@ UDFs can be executed at different stages of the pipeline by specifying the `targ
 - `broker_response` - Response message handling
 - `otel_tracer` - OpenTelemetry tracing
 
-> **Note:** For the complete and up-to-date list of pipeline stages, see the [default_pipeline.yaml](../../../config/default_pipeline.yaml) configuration file.
+> **Note:** For the complete and up-to-date list of pipeline stages, see the [default_pipeline.yaml](https://github.com/NVIDIA/nv-ingest/blob/main/config/default_pipeline.yaml) configuration file.
 
 #### Target Stage Selection Examples
 

@@ -1,7 +1,6 @@
 # Use Multimodal Embedding with NeMo Retriever Library
 
-This documentation describes how to use [NeMo Retriever Library](overview.md) 
-with the multimodal embedding model [Llama 3.2 NeMo Retriever Multimodal Embedding 1B](https://build.nvidia.com/nvidia/llama-3_2-nemoretriever-1b-vlm-embed-v1).
+This guide explains how to use the [NeMo Retriever Library](overview.md) with the multimodal embedding model [Llama Nemotron Embed VL 1B v2](https://build.nvidia.com/nvidia/llama-nemotron-embed-vl-1b-v2).
 
 The `Llama 3.2 NeMo Retriever Multimodal Embedding 1B` model is optimized for multimodal question-answering retrieval. 
 The model can embed documents in the form of an image, text, or a combination of image and text.