Install NVIDIA Ingest Locally and Use it with Thousands of Documents

 This video shares step-by-step instructions to install NVIDIA Ingest locally and use it with PDFs, Word, and PowerPoint.

Code:

Pre-requisites:

===============

-- Install docker

-- Get NGC api key from https://ngc.nvidia.com/

-- Get Early Access from https://developer.nvidia.com/nemo-microservices-early-access/join

Phase 1= Configure NV-INGEST Server:

====================================

Step 1:

git clone https://github.com/nvidia/nv-ingest && cd nv-ingest

Step 2:

docker login nvcr.io

Username: $oauthtoken

Password: <Your NGC API Key>

Step 3:

Make sure NVIDIA is set as your default container runtime before running the docker compose command:

sudo nvidia-ctk runtime configure --runtime=docker --set-as-default

Step 4:

docker compose up

Phase 2= Configure NV-INGEST client:

====================================

Step 1:

conda create --name nv-ingest-dev --file ./conda/environments/nv_ingest_environment.yml

conda activate nv-ingest-dev

cd client

pip install .

Step 2:

nv-ingest-cli \

  --doc ./data/multimodal_test.pdf \

  --output_directory ./processed_docs \

  --task='extract:{"document_type": "pdf", "extract_method": "pdfium", "extract_tables": "true", "extract_images": "true"}' \

  --client_host=localhost \

  --client_port=7670

 

Where to find output?

======================

After the ingestion steps above have completed, you should be able to find text and image subfolders inside your processed docs folder. Each will contain JSON formatted extracted content and metadata.

  ls -R processed_docs