Nemo Retriever

NVIDIAOfficialPublished by NVIDIA

Skill
Tool integration
Data and AI
Apache-2.0
Medium risk

From the author

Use when searching, extracting, ingesting, or querying a document collection with the NeMo Retriever 26.8.1 CLI, including local LanceDB indexes and deployed Retriever services. Use for PDFs, images, Office files, HTML, text, audio, and video; not for editing documents or web search.

— the description by NVIDIA

AutoGineer summary

Written by AutoGineer from the skill’s own files, not by the author

The NeMo Retriever skill provides a standardized interface for interacting with the NeMo Retriever 26.8.1 CLI. It supports both local document indexing via LanceDB and connectivity to deployed Retriever services. The skill is designed to handle diverse file formats including PDFs, images, Office documents, HTML, text, audio, and video, enabling efficient retrieval-augmented generation workflows by allowing users to ingest data and query it for evidence-based answers.

What it does

Ingests local files and directories into LanceDB indexes.
Queries local indexes for specific information.
Connects to and interacts with remote, deployed Retriever services.
Supports multimedia ingestion including audio and video files.
Provides evidence-based responses with source and page metadata.

When to use it

When you need to build a local vector index from a collection of documents.
When querying an existing local LanceDB index for specific answers.
When interacting with a remote NeMo Retriever service via API.
When you need to extract information from diverse file types like PDFs or media.

Advantages

Standardizes retrieval workflows using a single CLI.
Supports both local and remote deployment architectures.
Enables evidence-based responses with citation support.
Flexible installation options based on specific ingestion needs.

Usage examples

“Ingest the documents in my ./data folder into a local LanceDB index.”: The agent runs the retriever ingest command on the specified directory, creating a local index at the defined LanceDB URI.
“Query the existing index for information about project timelines.”: The agent executes a retriever query against the local index and returns the top-k evidence snippets.

Install

Download the skill as a ZIP:
https://autogineer.com/agent-skills/nvidia/nemo-retriever/download

Unzip it into the cross-client skills folder of your project:

mkdir -p .agents/skills && unzip nemo-retriever.zip -d .agents/skills

Documentation by NVIDIA

NeMo Retriever

Use the retriever CLI. Prefer it over hand-built retrieval code.

Install only when missing

Create a project-local Python environment:

uv venv .venv --python 3.12
export PATH="$PWD/.venv/bin:$PATH"

Install the package variant required by the workflow:

# Remote NIM or service client
uv pip install --python .venv/bin/python "nemo-retriever==26.8.1"

# Local GPU ingestion
uv pip install --python .venv/bin/python "nemo-retriever[local]==26.8.1"

# Local service using Hugging Face models
uv pip install --python .venv/bin/python \
  "nemo-retriever[service,local]==26.8.1"

# Local audio or video ingestion
uv pip install --python .venv/bin/python \
  "nemo-retriever[local,multimedia]==26.8.1"

Do not clone NeMo Retriever or install from a Git URL. If retriever is already on PATH, use that installation.

Local workflow

Build a local index:

retriever ingest <file-or-directory> \
  --lancedb-uri lancedb --table-name nemo-retriever

Query it:

retriever query "<question>" \
  --lancedb-uri lancedb --table-name nemo-retriever \
  --top-k 5 --format evidence

Use retriever ingest batch only for an explicitly requested Ray batch run.

Service workflow

Use these forms for an already deployed Retriever service:

retriever ingest service <file-or-directory> \
  --service-url "$RETRIEVER_SERVICE_URL"

retriever query service "<question>" \
  --service-url "$RETRIEVER_SERVICE_URL" \
  --top-k 5 --format evidence

Set NEMO_RETRIEVER_API_TOKEN when the service requires Bearer authentication. Do not pass local LanceDB flags to the service commands.

Rules

Use the existing index or service when one is provided; do not rebuild it.
Use retriever ingest --help, retriever query --help, or the relevant batch / service help for options not shown here.
Answer only from retrieved evidence; preserve source and page metadata when the task requests citations.

Files

ATTRIBUTION.mdAdded by AutoGineer827 B
BENCHMARK.md7.5 KiB
LICENSELicence file added by AutoGineer11.2 KiB
SKILL.md2.3 KiB
evals/evals.json3.6 KiB
skill-card.md4.3 KiB
skill.oms.sig4.5 KiB

BENCHMARK.md

Skill Benchmark: nemo-retriever

✅ Overall verdict: PASS — Recommended for publication

Publication Recommendation

Recommended for publication based on the completed evaluation evidence in this report.

Evaluation Metadata

Skill: nemo-retriever
Evaluation date: 2026-09-11
Evaluator version: 1.5.6
Agents: Claude Code (aws/anthropic/bedrock-claude-opus-4-8), Codex (openai/openai/gpt-5.5)
Tasks: 4 evaluation tasks (3 positive, 1 negative)
Dataset digest: sha256:97b3fe3650a2b9d3c8ec824a7f21e6da3551ae7057847765e68213f4d5a72cfa (skill-evaluator-dataset-snapshot/1)
Attempts per task: 3
Environment: k8s-sandbox
Tier 2 evidence: required for publication
Tier 3 evidence: required for publication

Each task attempt ran in its own isolated sandbox pod.

What This Report Answers

The three-tier evaluation checks whether the skill:

is safe to use;
produces correct answers;
is discovered and activated when needed;
helps the agent complete the user's goal and expected workflow; and
avoids wasted skill and tool usage.

Results at a Glance

Measure	Claude Code (Baseline → Skill Uplift)	Codex (Baseline → Skill Uplift)
Overall	78.3% — baseline ran, but no comparable score was available; uplift unavailable	72.2% — baseline ran, but no comparable score was available; uplift unavailable
Security	100.0% → 100.0% (±0.0 points)	100.0% → 100.0% (±0.0 points)
Correctness	27.5% → 90.0% (+62.5 points)	26.0% → 70.0% (+44.0 points)
Discoverability	78.3% — baseline ran, but no comparable score was available; uplift unavailable	73.3% — baseline ran, but no comparable score was available; uplift unavailable
Effectiveness	31.9% → 41.3% (+9.4 points)	20.5% → 40.0% (+19.5 points)
Efficiency	82.0% — baseline ran, but no comparable score was available; uplift unavailable	77.5% — baseline ran, but no comparable score was available; uplift unavailable

How to read this table: baseline is the same task attempted without the target skill. Scores are rounded to one decimal; threshold-adjacent values use additional precision so their displayed band matches the verdict. Uplift is derived from those displayed scores and shown in percentage points.

Example: 47.0% → 92.0% (+45.0 points) means the skill-assisted run scored 92.0%, 45.0 percentage points above its 47.0% no-skill baseline.

A partial dimension was calculated from only the available configured signals; review the detailed report before relying on it.

Token Usage

Actual Tier 3 execution usage is reported for every observed agent/case pair and both conditions.

Agent	Dataset case	With skill	Without skill	Delta	Change	Coverage
claude-code	All cases	768,939	1,148,438	N/A	N/A	skill 4/4; base 8/8
claude-code	nemo-retriever-001	218,824	91,210	+127,614	+139.91%	skill 1/1; base 1/1
claude-code	nemo-retriever-002	155,704	499,675	N/A	N/A	skill 1/1; base 3/3
claude-code	nemo-retriever-003	364,539	527,862	N/A	N/A	skill 1/1; base 3/3
claude-code	nemo-retriever-004	29,872	29,691	+181	+0.61%	skill 1/1; base 1/1
codex	All cases	263,555	698,864	N/A	N/A	skill 4/4; base 10/10
codex	nemo-retriever-001	72,778	283,792	N/A	N/A	skill 1/1; base 3/3
codex	nemo-retriever-002	119,807	165,626	N/A	N/A	skill 1/1; base 3/3
codex	nemo-retriever-003	57,326	235,924	N/A	N/A	skill 1/1; base 3/3
codex	nemo-retriever-004	13,644	13,522	+122	+0.90%	skill 1/1; base 1/1
ALL AGENTS	Dataset aggregate	1,032,494	1,847,302	N/A	N/A	skill 8/8; base 18/18

Prompt tokens include cached reads, so total tokens are prompt + completion (cached is not added twice). The Efficiency score uses (prompt - cached) + completion. N/A means the relevant trajectory counters were not available; coverage is never estimated.

Tier Status

Tier	Purpose	Status	Evidence
Tier 1	Static validation	PASSED WITH OBSERVATIONS	11 validator(s); 12 finding(s)
Tier 2	Semantic deduplication	PASSED	2 validator(s); 0 finding(s)
Tier 3	Live agent evaluation	PASS	2 agent(s); 4 task(s)

Findings and Observations

<details> <summary>Show detailed findings and successful checks</summary>

MEDIUM QUALITY/quality_correctness: SKILL_SPEC recommended field missing: 'metadata.author' (skills/nemo-retriever/SKILL.md)
MEDIUM QUALITY/quality_correctness: SKILL_SPEC recommended field missing: 'metadata.tags' (skills/nemo-retriever/SKILL.md)
MEDIUM SCHEMA/body_recommended_section: Missing recommended section: '## Instructions' (skills/nemo-retriever/SKILL.md)
MEDIUM SCHEMA/body_recommended_section: Missing recommended section: '## Examples' (skills/nemo-retriever/SKILL.md)
MEDIUM SCHEMA/author_missing: Author not specified in metadata (skills/nemo-retriever/SKILL.md)
7 additional finding(s) are available in the full evaluation artifacts.

</details>

Scoring Methodology

<details> <summary>Show dimension definitions, source signals, and thresholds</summary>

Dimension	Question	Scored signals
Security	Is it safe to use?	`security` (100%)
Correctness	Is the answer correct?	`accuracy` (100%)
Discoverability	Was the right skill loaded when needed?	`skill_execution` (100%)
Effectiveness	Did the skill help complete the task?	`goal_accuracy` (50%) + `behavior_check` (50%)
Efficiency	Did it avoid wasted tool calls and token usage?	`skill_efficiency` (50%) + `token_efficiency` (50%)

Dimension bands: PASS at 50% or above; NEUTRAL from 40% to below 50%; FAIL below 40%.
Overall Tier 3 lift: PASS at +5 points or more; FAIL at -10 points or less; values between those bands are NEUTRAL.
Overall verdict: PASS only when every configured dimension passes for at least one supported agent. Lift is reported as diagnostic evidence and does not override this gate.
The 50% attempt pass threshold is a separate per-task gate; it is not the dimension pass threshold.
Effectiveness is the equal-weight mean of goal completion (goal_accuracy) and expected workflow adherence (behavior_check).
Efficiency is 50% tool-call productivity (the backward-compatible skill_efficiency wire id) and 50% token_efficiency. Positive-case skill routing is scored under Discoverability, not Efficiency; a negative case without a routing target is N/A. N/A sources are omitted, remaining weights are renormalized, and the dimension is marked partial.

Signals present in this run:

security (Security): unsafe operations, secret leakage, and unauthorized access.
skill_execution (Skill Execution): whether the expected skill was selected, decoys were avoided, and the workflow executed.
skill_efficiency (Tool Productivity): tool-call productivity (legacy wire id; routing is scored under Discoverability).
accuracy (Accuracy): final-answer correctness against the reference answer.
goal_accuracy (Goal Accuracy): whether the user's goal was achieved.
behavior_check (Behavior Check): whether the expected workflow behavior was followed.
token_efficiency (Token Efficiency): actual uncached prompt plus completion usage (50% of Efficiency).

</details>

Freshness

Regenerate this benchmark when the skill, evaluation dataset, target agent/model, evaluator version, environment, or scoring policy changes.

Requirements and compatibility

Works with: Claude Code, Claude apps, OpenAI Codex, GitHub Copilot, Cursor, Gemini CLI, Kiro, OpenCode, Goose, Amp, Roo Code, JetBrains Junie, Factory Droid, Devin (ex-Windsurf), Mistral Vibe, OpenHands, Letta

“Works with” means the tool reads the open Agent Skills format; a skill written for one tool may use features others lack.

Needs: Needs Python

Security

What our scanner noticed· Medium risk

Large base64-encoded blobs
Instructions that use the network

A static scan is not a guarantee. Skills are third-party content: review a skill before you let an agent use it.

Attribution and licence

Published by

NVIDIA

Publisher

NVIDIA (Official)

Source

NVIDIA/skills · skills/nemo-retriever · commit fce3cca

Licence

Apache License 2.0 (redistribution allowed)

LICENSE (added by AutoGineer from LICENSE-APACHE)
ATTRIBUTION.md (written by AutoGineer)

Synced

2026-11-09

AutoGineer is not affiliated with or endorsed by NVIDIA or NVIDIA.

Download ZIP

The ZIP holds the skill’s files, its licence files and an ATTRIBUTION.md. Download nemo-retriever.zip