Starstruk Logo MarkStarstruk

25.9.0

OpenMOSS/MOSS-TTS25.9.0Sep 17, 2025by sosahi

AI Summary

The NeMo Retriever extraction 25.09 release introduces expanded hardware support for the RTX Pro 6000 and DGX B200, alongside new software capabilities like OCR support, custom vector databases, and enhanced multimodal embedding features.

Key Highlights

  • Added functional support for RTX Pro 6000 and DGX B200 GPUs.
  • Introduced support for custom vector database implementations and Lambda stages.
  • Added new OCR model support and enhanced multimodal embedding capabilities.
  • Expanded metadata handling to support Integer, float, boolean, and array types.

New Features

  • RTX Pro 6000 functional support
  • DGX B200 functional support
  • nemoretriever-ocr-v1 support
  • llama-3.2-nemoretriever-1b-vlm-embed-v1 support
  • Llama Nemotron VLM 8b NIM for image captioning
  • Custom vector database implementations
  • Custom Lambda stages
  • Expanded documentation for Library Mode, Ray Logging, and Multimodal Embedding
  • Support for Integer, float, boolean, and array in custom metadata
  • Support for running multiple VLMs at once via Helm

Full Release Notes

The NeMo Retriever extraction 25.09 release adds new hardware and software support, and other improvements, including the following:

- Add functional support for [RTX Pro 6000](https://www.nvidia.com/en-us/products/workstations/professional-desktop-gpus/rtx-pro-6000/).
- Add functional support for [DGX B200](https://www.nvidia.com/en-us/data-center/dgx-b200/).
- Add support for [nemoretriever-ocr-v1](https://build.nvidia.com/nvidia/nemoretriever-ocr-v1). For details, refer to [Deploy With Docker Compose (Self-Hosted)](https://docs.nvidia.com/nemo/retriever/latest/extraction/quickstart-guide/) and [NV-Ingest Helm Charts](https://github.com/nkmcalli/nv-ingest/tree/main/helm).
- Add support for [llama-3.2-nemoretriever-1b-vlm-embed-v1](https://build.nvidia.com/nvidia/llama-3_2-nemoretriever-1b-vlm-embed-v1).
- Add support for Llama Nemotron VLM 8b NIM for image captioning. For details, refer to [Extract Captions from Images](https://docs.nvidia.com/nemo/retriever/latest/extraction/nv-ingest-python-api/#extract-captions-from-images).
- Add support for custom vector database implementations. For details, refer to [Build a Custom Vector Database Operator](https://github.com/NVIDIA/nv-ingest/blob/main/examples/building_vdb_operator.ipynb).
- Add support for custom Lambda stages. For details, refer to [Add User-defined Stages to Your NeMo Retriever Extraction Pipeline](https://docs.nvidia.com/nemo/retriever/latest/extraction/user-defined-stages/).
- Expanded documentation for [Library Mode](https://docs.nvidia.com/nemo/retriever/latest/extraction/quickstart-library-mode/).
- New documentation [Configure Ray Logging](https://docs.nvidia.com/nemo/retriever/latest/extraction/ray-logging/).
- New documentation [Use Multimodal Embedding](https://docs.nvidia.com/nemo/retriever/latest/extraction/vlm-embed/).
- Add support for Integer, float, boolean, and array in custom metadata during Milvus entity creation.
- Add support for running more than one VLM at a time by using Helm. For details, refer to [NV-Ingest Helm Charts](https://github.com/nkmcalli/nv-ingest/tree/main/helm).

**Known Issues**

The following are the known issues for this release:

- A10G and L40S are not supported. For details, refer to [Support Matrix](https://docs.nvidia.com/nemo/retriever/latest/extraction/support-matrix/).
- `nemoretriever-parse` is not supported on RTX Pro 6000 or B200. For details, refer to [Support Matrix](https://docs.nvidia.com/nemo/retriever/latest/extraction/support-matrix/).
- The NeMo Retriever extraction pipeline does not support ingestion of batches that include individual files greater than approximately 400MB.