Nvidia-nim-unified-skill nvidia-nim-unified-skill

name: nvidia-nim-unified

install

source · Clone the upstream repo

git clone https://github.com/JithendraNara/nvidia-nim-unified-skill

manifest: nvidia-nim-unified-skill.yaml

safety · automated scan (low risk)

This is a pattern-based risk scan, not a security review. Our crawler flagged:

references API keys

Always read a skill's source content before installing. Patterns alone don't mean the skill is malicious — but they warrant attention.

source content

name: nvidia-nim-unified version: 1.0.0 description: Unified routing skill for NVIDIA NIM OCR, layout, table/chart detection, and passage reranking author: Jithendra Nara license: MIT

requires:

python3
outbound HTTP access for invoke mode

config:

name: NVIDIA_API_KEY description: NVIDIA API key for managed reranking and CV endpoints required: false secret: true
name: NVIDIA_NIM_BEARER_TOKEN description: Optional bearer token for self-hosted endpoint overrides required: false secret: true
name: NVIDIA_NIM_OCR_URL description: Optional override for the OCR endpoint URL required: false example: https://ai.api.nvidia.com/v1/cv/nvidia/nemotron-ocr-v1
name: NVIDIA_NIM_PAGE_ELEMENTS_URL description: Optional override for the page elements endpoint URL required: false example: https://ai.api.nvidia.com/v1/cv/nvidia/nemotron-page-elements-v3
name: NVIDIA_NIM_TABLE_STRUCTURE_URL description: Optional override for the table structure endpoint URL required: false example: https://ai.api.nvidia.com/v1/cv/nvidia/nemoretriever-table-structure-v1
name: NVIDIA_NIM_GRAPHIC_ELEMENTS_URL description: Optional override for the graphic elements endpoint URL required: false example: https://ai.api.nvidia.com/v1/cv/nvidia/nemoretriever-graphic-elements-v1

capabilities:

OCR text extraction from document images
Page layout detection for titles, paragraphs, tables, charts, and header/footer
Table structure detection for border, cell, row, column, and header boxes
Graphic element detection for chart labels, legends, axes, and value labels
Passage reranking against a user query
Multi-step workflow planning across the above capabilities

entrypoints: plan: python3 scripts/nim_router.py plan --task-query "<task>" build_request: python3 scripts/nim_router.py build-request --capability <capability> ... invoke: python3 scripts/nim_router.py invoke --capability <capability> ...

workflows:

name: document-qa sequence:
- ocr
- rerank
name: layout-aware-table-extraction sequence:
- page_elements
- table_structure
- ocr
name: chart-aware-extraction sequence:
- page_elements
- graphic_elements
- ocr

notes:

This skill is a unified router, not a fake merged OpenAPI endpoint.
Use references/nim-capabilities.json for exact normalized request and response shapes.
Managed CV endpoints require base64 data-image inputs; the router normalizes incoming image sources automatically.