Skip to content
View Yarmoluk's full-sized avatar
✌️
At Peace
✌️
At Peace
29 followers · 44 following

Highlights

Block or report Yarmoluk

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. Markdown is supported. This note will only be visible to you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
Yarmoluk/README.md

Model for Language, Context for Knowledge

PyPI downloads MCP Registry CKG domains Benchmark F1 HuggingFace Patent

LangChain LlamaIndex CrewAI Zep ElevenLabs NVIDIA Meta Llama Smithery Anthropic

Agents need certainty. LLMs give probability. Compact Knowledge Graphs (CKGs) are the missing layer — structured, auditable domain knowledge served over MCP. 269 tokens per query vs 2,982 for RAG. F1 0.471, +283% vs RAG baseline. SHA-256 provenance on every node. Drop-in MCP servers, no infra required.

New: CKG Router — deterministic context + model routing from graph depth. route_query() returns the right subgraph AND the right model tier (haiku/sonnet/opus) derived from hop count. No heuristic — the graph decides. 4× more accurate, 90% cheaper, routed automatically.

Packages

Package Domain F1 Downloads
ckg-mcp 97 domains · full stack · CKG Router —
ckg-nvidia-ai NVIDIA AI · NIM · NeMo · 20 domains · 1,055 nodes · CKG Router —
ckg-nvidia-nemoclaw NemoClaw · 55 nodes / 74 edges · CKG Router 0.576
ckg-agentforce Salesforce AgentForce · billing + permissions · CKG Router —
ckg-nemotron-perplexity Nemotron + Perplexity Sonar · 83 nodes —
langchain-ckg LangChain retriever · CKGRetriever + CKGHostedRetriever — 
uvx ckg-mcp              # 97 domains — any agent stack
uvx ckg-nvidia-ai        # NVIDIA AI · NIM · NeMo · NemoClaw
uvx ckg-agentforce       # Salesforce AgentForce

# CKG Router — deterministic context + model routing
route_query("TensorRT-LLM", "nvidia-tensorrt-triton")
# → model_tier: opus · reasoning: sparql_cot · why: 4-hop chain
# → context subgraph: 847 tokens (not 10,000)

Benchmark

System Macro F1 Tokens/query Cost @ $10/1M
▸ CKG 0.471 269 $0.003
RAG 0.123 2,982 $0.030
GraphRAG 0.120 3,450 $0.035

ckg-benchmark v0.6.2 · 97 domains · dataset · paper · patent pending

F1 scales with hop depth: 0.374 → 0.772 at hop=5. RAG plateaus at hop=2.

Pricing

Tier Calls Price Best for
Free 50/day — Evaluation
Dev Unlimited $10/yr Solo devs, small teams
Pro Unlimited + priority $99/yr Production
Sealed On-prem Docker appliance $299 Enterprise / air-gap

Upgrade → graphifymd.com/pricing · Dev $10/yr

Popular repositories Loading

  1. ckg-benchmark ckg-benchmark Public

    Benchmarking RAG vs GraphRAG vs Compressed Knowledge Graphs on the McCreary Intelligent Textbook Corpus

    HTML 14

  2. ckg-nvidia-ai ckg-nvidia-ai Public

    MCP server — Compressed Knowledge Graph for the NVIDIA AI developer stack. 20 domains · 998 nodes · 4× F1 vs RAG · 11× fewer tokens · deterministic traversal.

    HTML 8

  3. ckg-mcp ckg-mcp Public

    Context optimization for AI agents — 97 domain knowledge graphs, MCP-native traversal. 4× F1 of RAG, 11× fewer tokens per query.

    Python 4

  4. skills-1 skills-1 Public

    Forked from anthropics/skills

    Public repository for Agent Skills

    Python 1 1

  5. awesome-mcp-servers awesome-mcp-servers Public

    Forked from punkpeye/awesome-mcp-servers

    A collection of MCP servers.

    1

  6. ckg-nemotron-perplexity ckg-nemotron-perplexity Public

    NVIDIA Nemotron + Perplexity Sonar as traversable knowledge graphs — 2 domains, 83 nodes, MCP-native agent traversal.

    Python 1