# NVIDIA/NeMo-Retriever

NeMo Retriever Library is a scalable, performance-oriented document content and metadata extraction microservice. NeMo Retriever Library uses specialized NVIDIA NIM microservices to find, contextualize, and extract text, tables, charts and images that you can use in downstream generative applications.

Repository: https://github.com/NVIDIA/NeMo-Retriever
Canonical: https://ross.abutalabs.com/products/nemo-retriever
Homepage: https://nvidia.github.io/NeMo-Retriever/
Language: Python
License: Apache-2.0
License Family: permissive
Last push: 2026-08-26T22:11:59+00:00

## Health v2 (maintenance only)
Score: 81/100 (v2, computed 2026-09-03T02:20:16.233290+00:00)
- activity 99, release rhythm 74, longevity 52
- inputs: {"age_days": 741, "days_push": 7, "days_rel": 95, "gap_med": 57, "n_releases_24m": 10}
- flags: none
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 2970, forks 344 (observed 2026-08-28T04:07:32.738591+00:00)

## What it is
NVIDIA's scalable document content and metadata extraction library (also known as NVIDIA Ingest) that splits documents, classifies and extracts text, tables, charts, and images via OCR, and computes embeddings for storage in LanceDB. It supports NVIDIA NIM microservices and HuggingFace models, with Helm-based Kubernetes deployment for production scale.

## Use cases
- extract text and tables from pdfs for rag pipelines
- parse charts and infographics out of documents
- build a document ingestion pipeline for generative ai apps
- compute embeddings from extracted document content
- deploy a scalable document extraction microservice on kubernetes
- ocr scanned pdfs into structured json
- ingest pdfs into a vector database

## When to choose
- you need high-performance, parallelized extraction of text, tables, charts, and images from documents
- you want to feed extracted content into retrieval-augmented generation pipelines
- you can use NVIDIA GPUs and NIM microservices for production-scale workloads
- you want embeddings computed and stored automatically (e.g., in LanceDB)

## When to avoid
- you only need simple text extraction from plain-text or HTML files
- you have no GPU access and need a lightweight CPU-only solution
- you need a fully managed SaaS document service rather than self-hosted infrastructure
- your documents are small in number and a simple parser library would suffice

## Facets
- artifact type: library
- maturity: active
- function: ocr, pdf, rag, etl, machine-learning, data-generation, sdk
- domain: artificial-intelligence, large-language-models, pdf, developer-tools
- platform: python, self-hosted
- tags: document-extraction, nvidia-nim, multimodal, embeddings, lancedb, helm-chart, microservice, ingestion-pipeline, retrieval-augmented-generation, data-engineering, docker, kubernetes, gpu, linux

## Member repositories
- NVIDIA/NeMo-Retriever (main) score 81

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:07:32.738591+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-30T07:31:43.989729+00:00, confidence not recorded.
  - readme: https://github.com/NVIDIA/NeMo-Retriever (fetched 2026-08-28T04:07:32.738591+00:00, sha 9a8eae0fcd0c)
  - homepage: https://nvidia.github.io/NeMo-Retriever/ (fetched 2026-08-29T09:47:01.168913+00:00, sha 926f96d142ee)
- Data as of 2026-08-30T08:39:29.467469+00:00.