ckg-nvidia-ai 0.2.0

NVIDIA AI developer stack as a traversable knowledge graph — 20 CKGs, 998 nodes, MCP-native

ckg-nvidia-ai

The NVIDIA AI developer stack as a traversable knowledge graph.

20 CKGs · 998 nodes · MCP-native · 4× F1 of RAG · 11× fewer tokens · auditable by design

---

Instead of sending an AI agent to scan thousands of pages of NVIDIA documentation, give it a graph it can traverse. Every concept, every dependency, every connection — declared, typed, and queryable in ~269 tokens per question.

NIM → TensorRT-LLM → quantization → FP8 precision → Hopper SM90 requirement

The agent calls query_ckg() and gets that chain. Not a summary. The actual dependency path.

---

What's inside

Domain	Description
nvidia-nim	NVIDIA Inference Microservices — deployment, scaling, speculative decoding
nvidia-nemo	NeMo framework — training, PEFT, guardrails, evaluation
nvidia-tensorrt-triton	TensorRT-LLM + Triton Inference Server — quantization, batching, KV cache
nvidia-cuda-toolkit	CUDA compiler, PTX, memory hierarchy, Hopper/Blackwell features
nvidia-cuda-x-libraries	cuBLAS, cuDNN, cuFFT, NCCL, Thrust — the acceleration layer
nvidia-hpc-sdk	OpenACC, OpenMP, CUDA Fortran, multi-GPU scaling
nvidia-omniverse	Universal Scene Description, simulation, digital twins
nvidia-isaac	Isaac Lab + Isaac Sim — robot learning, sensor simulation
nvidia-cosmos	Physical AI world foundation models — video generation, tokenization
nvidia-drive	Autonomous vehicle stack — perception, planning, safety validation
nvidia-jetson	Edge AI platform — Orin NX, AGX, DeepStream, Holoscan
nvidia-clara	Healthcare AI — MONAI, Parabricks genomics, BioNeMo, Holoscan SDK
nvidia-metropolis	Intelligent video analytics — VLMs, TAO Toolkit, DeepStream
nvidia-riva	Speech AI — ASR, TTS, NLP pipelines, streaming
nvidia-gameworks	Graphics R&D — DLSS, RTX, PhysX, Reflex
nvidia-developer-tools	Nsight, CUPTI, Compute Sanitizer, profiling stack
nvidia-graphics-research	Research graphics — neural rendering, path tracing, differentiable rendering
nvidia-ai-enterprise	Enterprise AI platform — NIM blueprints, governance, fleet management
nvidia-developer-ecosystem	Cross-cutting: NGC, DGX, Inception, AgentIQ, MCP integration
nvidia-openshell	Agent sandbox runtime — policy enforcement, CVEs, authorization gaps

---

Install

pip install ckg-nvidia-ai

Or run without installing:

uvx ckg-nvidia-ai

---

Use as MCP Server

Claude Desktop

{
  "mcpServers": {
    "nvidia-ai": {
      "command": "uvx",
      "args": ["ckg-nvidia-ai"]
    }
  }
}

Cursor / other MCP clients

Same config — substitute uvx with python -m ckg_nvidia_ai if you prefer a venv install.

---

Tools

list_domains()

Returns all 20 NVIDIA AI domains. Start here.

search_concepts(query, domain)

Find concepts by keyword within a domain.

search_concepts("speculative decoding", "nvidia-nim")
→ Speculative Decoding [Optimization]
   Draft Model [Component]
   KV Cache [Infrastructure]

query_ckg(concept, domain, depth=3)

Traverse the graph from a concept — see what it requires and what depends on it.

query_ckg("TensorRT-LLM", "nvidia-tensorrt-triton", 3)
→ ## TensorRT-LLM · nvidia-tensorrt-triton
   Type: Framework

   ### Prerequisites (what you need first)
     - CUDA Toolkit
       - CUDA Driver API
       - cuBLAS
     - Hopper SM90 Architecture
     - FP8 / FP4 Quantization

   ### Builds toward
     - Triton Inference Server
     - NIM Microservice Runtime

get_prerequisites(concept, domain)

Full ordered prerequisite chain — everything to understand or install first.

get_prerequisites("Isaac Lab", "nvidia-isaac")
→ Isaac Lab → Isaac Sim → USD Composer → Omniverse Kit → ...

---

How it works

Each domain is a dependency graph stored as CSV. Edges carry an explicit type:

ConceptID, ConceptLabel, Dependencies, TaxonomyID
1, TensorRT-LLM, "", Framework
2, CUDA Toolkit, "", Platform
3, FP8 Quantization, "2:REQUIRES", Optimization
4, Hopper SM90, "2:REQUIRES", Architecture
5, Speculative Decoding, "1:REQUIRES|4:REQUIRES", Optimization

Edge types:

Type	Meaning
REQUIRES	Hard prerequisite — cannot function without it
ENABLES	Unlocks a capability — useful but not strictly required
RELATES_TO	Conceptual proximity — not a dependency
IMPLEMENTS	Concrete instantiation of an abstraction

Untyped edges (from v0.1.0) default to REQUIRES — the conservative correct assumption. Edge types are refined over time by domain experts; PRs welcome.

When an agent queries a concept, the server runs BFS traversal over declared edges. The answer is composed entirely of traversed relationships — not probabilistic inference, not RAG retrieval, not token prediction over documentation.

The graph doesn't guess. It traverses.

---

Benchmark

Built on the KRB Benchmark v0.6.2:

System	F1	Tokens/query	Cost
CKG	0.471	269	$7.81/1K
RAG	0.123	2,982	$76.23/1K
GraphRAG	0.120	—	—

~4× F1 · 11× fewer tokens · auditable by design

---

Related

• ckg-mcp — 97 domains across all topics (includes NVIDIA + science, finance, law, healthcare, and more)
• KRB Benchmark — open benchmark dataset
• graphifymd.com — CKG catalog and Context-as-a-Service

---

Built by Graphify.md. Patent pending.