ckg-nvidia-ai logo

ckg-nvidia-ai

by GraphifyMDOfficialGitHubWebsiteUpdated Aug 10, 2026

NVIDIA AI developer stack as a traversable knowledge graph — 20 domains, 1,006 nodes, declared edges only. 4× RAG F1, 11× fewer tokens.

nvidia
knowledge-graph
mcp
+7
|

The NVIDIA developer stack is large, versioned fast, and full of dependencies that docs don't make explicit. ckg-nvidia-ai gives your agent a structured map — 20 domains, 1,006 nodes, 1,055+ typed edges — so it traverses declared relationships instead of inferring them from chunks.

Every edge traces to an authoritative NVIDIA source document. Every source is SHA-256 pinned. A mismatch is deterministic — no judgment required.

20 domains covered: NIM · NeMo · AgentIQ · Isaac · Cosmos · Omniverse · CUDA Toolkit · TensorRT-LLM/Triton · Jetson · DRIVE · Clara · Metropolis · Riva · GameWorks · HPC SDK · CUDA-X Libraries · Developer Tools · Graphics Research · AI Enterprise · OpenShell

Benchmark (KRB v0.6.2 — locked): F1 0.471 · 4× RAG · 11× fewer tokens (269 vs 2,982) · $7.81 vs $76.23 per 1k queries

Need a custom domain or enterprise deployment? Graphify.md builds CKGs for specific platforms, internal knowledge bases, and regulated stacks. Reach out at daniel.yarmoluk@gmail.com or graphifymd.com.