What the blueprint is
Build a RAG Pipeline Blueprint is a retrieval-augmented generation blueprint published in NVIDIA's blueprint catalogue. It is designed to power fast, accurate semantic search across multimodal enterprise data, and to connect agents to trusted authoritative sources of knowledge. It is built on NeMo Retriever and Nemotron models and is distributed as a launchable developer example. The catalogue entry was last updated on September 22, 2026.
- Category: General
- Format: launchable developer example
- Tags: NVIDIA AI, NIM, Nemotron, Retrieval-Augmented Generation, NeMo Retriever
- Last updated on September 22, 2026
Underlying model families
The blueprint draws on the NeMo Retriever and Nemotron model families offered as NVIDIA NIM microservices. The retrieval catalogue on build.nvidia.com groups these into reranking, embedding and extraction models, including Nemotron Retriever reranking and embedding models and document parsing models for text and table extraction. Extraction models cover chart, object and table detection for data ingestion, which supports the multimodal aspect of the pipeline.
- Reranking models such as nemotron-3.5-rerank-8b and llama-nemotron-rerank-vl-1b-v2
- Embedding models such as nemotron-3.5-embed-1b, nemotron-3.5-embed-8b and nemotron-3-embed-1b
- Extraction and document parsing models such as nemotron-3.5-parse
- Capabilities including table extraction, chart detection and object detection
Place in the NVIDIA Blueprints catalogue
The blueprint sits alongside other NVIDIA Blueprints on build.nvidia.com, where entries can be sorted by most recent or alphabetically and filtered by categories such as Agentic AI, Developer Productivity and NVIDIA AI. Neighbouring entries include DocsFab, NVIDIA NemoClaw for Agent Optimization, Agent For Life Sciences, NVIDIA Deep Researcher Agent, AI Factory Operations Agent and GPU Query Engine. Blueprints are also surfaced from model exploration pages, where an "Explore NVIDIA Blueprints" section links catalogue entries to related model families.
- Listed at build.nvidia.com/blueprints
- Sortable by most recent or alphabetical order
- Grouped with other NVIDIA AI and General blueprints
Access and deployment
The blueprint and its component models are reached through the NVIDIA NIM API pages on build.nvidia.com, which require a login to use. NVIDIA presents NIM models for self-hosting on the user's own GPU infrastructure, with continuous vulnerability fixes. As a launchable developer example, the blueprint is intended to be started from the catalogue listing rather than requested through an application process.
- Accessed via build.nvidia.com
- Login required for NIM API access
- Self-hosting on GPU infrastructure offered through NVIDIA NIM
Sources
-
Explore Vision Models | Try NVIDIA NIM APIs https://build.nvidia.com/explore/vision Verified 03 Nov 2026
-
Explore Retrieval Models | Try NVIDIA NIM APIs https://build.nvidia.com/explore/retrieval Verified 03 Nov 2026
Last verified 03 Nov 2026. This entry is compiled from the public web pages listed above. Nothing here is stated that those pages do not, and each of them was read on the date shown.