Skip to content

Release Notes for NeMo Retriever Library

This documentation contains the release notes for NeMo Retriever Library.

26.05 Release Notes (26.5.0)

NVIDIA® NeMo Retriever Library version 26.05 (PyPI 26.5.0 at GA) continues the 26.05 release line on the 26.05 branch. Pre-release builds are tagged 26.05-RC1, 26.05-RC2, and so on; install and deploy using the RC tag that matches your build.

To upgrade the Helm charts for this release, refer to the NeMo Retriever Helm chart README and pin chart version 26.05-RC1 (or the RC you are validating).

Highlights for the 26.05 release line include everything in 26.03 plus changes on main merged into the 26.05 branch. See the Git compare view for the full commit list.

Migration note: Direct Retriever(...) construction uses grouped configuration dictionaries. Replace flat lancedb_uri=, lancedb_table=, embedder=, embedding_endpoint=, local_query_embed_backend=, and reranker= arguments with vdb_kwargs={...}, embed_kwargs={...}, and rerank=.... For example, local_query_embed_backend="hf" maps to embed_kwargs={"local_ingest_embed_backend": "hf"}. Helper APIs that document their own flat kwargs keep their own compatibility layer.

Install (RC1 example):

uv pip install nemo-retriever==26.05-RC1

Use your organization's Artifactory or PyPI index URL when installing published wheels from CI (see the Perform Release workflow summary for the exact index).

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:

  • Legacy ingestion repository consolidated under 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 TTL for long-running pipeline job state increased from 1–2 hours to 48 hours so long-running jobs (for example, VLM captioning) do not expire before completion
  • NeMo Retriever Library currently does not support image captioning via VLM; this feature will be added in the next release
  • Documentation: multimodal extraction is covered on one page with an in-page table of contents and redirects from the former per-topic URLs
  • Container images built from this repository no longer install ffmpeg and ffprobe by default. Audio and video extraction require these binaries on PATH; for Helm deployments set service.installFfmpeg=true, or install system FFmpeg manually in non-container environments.

Release Notes for Previous Versions

| 26.03 | 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