Starstruk Logo MarkStarstruk

25.6.2

NVIDIA/NeMo-Retriever25.6.2Jun 12, 2025by sosahi

AI Summary

The NeMo Retriever extraction 25.06 release focuses on accuracy improvements, Python version upgrade, and feature expansions including MIG and time-slicing support for GPU utilization.

Key Highlights

  • Improved reranker accuracy
  • Upgraded Python from 3.10 to 3.12
  • Helm deployment throughput matching docker deployment
  • Added MIG support for GPUs
  • Added GPU time-slicing support
  • Added RIVA NIM support for audio extraction
  • Added latest OpenAI API support

New Features

  • Reranker accuracy improvements
  • Python 3.12 upgrade
  • MIG (Multi-Instance GPU) support
  • GPU time-slicing support
  • RIVA NIM audio extraction integration
  • OpenAI API v1 support
  • Metadata and filtered search notebook

Full Release Notes

The NeMo Retriever extraction 25.06 release focuses on accuracy improvements and feature expansions, including the following:

- Improve reranker accuracy.
- Upgrade Python version from 3.10 to 3.12
- Helm deployment now has similar throughput performance to docker deployment.
- Add support for the latest version of the OpenAI API.
- Add MIG support. For details, see [Enable NVIDIA GPU MIG](https://github.com/NVIDIA/nv-ingest/blob/main/helm/README.md#enable-nvidia-gpu-mig).
- Add time slicing support. For details, see [Enable GPU time-slicing](https://github.com/NVIDIA/nv-ingest/blob/main/helm/README.md#enabling-gpu-time-slicing).
- Add support for RIVA NIM for optional audio extraction. For details, see [helm/values.yaml](https://github.com/NVIDIA/nv-ingest/blob/main/helm/values.yaml).
- New notebook for [How to add metadata to your documents and filter searches](https://github.com/NVIDIA/nv-ingest/blob/main/examples/metadata_and_filtered_search.ipynb).

**Breaking Changes**
There are no breaking changes in this version.

**Upgrade**
To upgrade the Helm Charts for this version, refer to [NV-Ingest Helm Charts](https://github.com/NVIDIA/nv-ingest/blob/main/helm/values.yaml).