Skip to content

Release Notes for NeMo Retriever Library

This documentation contains the release notes for NeMo Retriever Library.

Note

NVIDIA Ingest (nv-ingest) has been renamed NeMo Retriever Library.

26.03 Release Notes (26.3.0)

NVIDIA® NeMo Retriever Library version 26.03 adds broader hardware and software support along with many pipeline, evaluation, and deployment enhancements.

To upgrade the Helm charts for this release, refer to the NeMo Retriever Library Helm Charts.

Highlights for the 26.03 release include:

  • NV-Ingest GitHub repo renamed to NeMo-Retriever
  • NeMo Retriever Extraction pipeline renamed to NeMo Retriever Library
  • NeMo Retriever Library now supports two deployment options:
  • A new no-container, pip-installable in-process library for development (available on PyPI)
  • Existing production-ready Helm chart with NIMs
  • Added documentation notes on Air-gapped deployment support
  • Added documentation notes on OpenShift support
  • Added support for RTX4500 Pro Blackwell SKU
  • Added support for llama-nemotron-embed-vl-v2 in text and text+image modes
  • New extract methods pdfium_hybrid and ocr target scanned PDFs to improve text and layout extraction from image-based pages
  • VLM-based image caption enhancements:
  • Infographics can be captioned
  • Reasoning mode is configurable
  • Enabled hybrid search with Lancedb
  • Added retrieval_bench subfolder with generalizable agentic retrieval pipeline
  • The project now uses UV as the primary environment and package manager instead of Conda, resulting in faster installs and simpler dependency handling
  • Default Redis TTL increased from 1–2 hours to 48 hours so long-running jobs (e.g., VLM captioning) don’t expire before completion
  • NeMo Retriever Library currently does not support image captioning via VLM; this feature will be added in the next release

Release Notes for Previous Versions

| 26.1.2 | 26.1.1 | 25.9.0 | 25.6.3 | 25.6.2 | 25.4.2 | 25.3.0 | 24.12.1 | 24.12.0