Metadata-Version: 2.4
Name: ckg-nemotron-perplexity
Version: 0.1.0
Summary: NVIDIA Nemotron + Perplexity Sonar as traversable knowledge graphs — 2 domains, 83 nodes, MCP-native agent traversal.
Project-URL: Homepage, https://graphifymd.com
Project-URL: Repository, https://github.com/Yarmoluk/ckg-nemotron-perplexity
Author-email: Daniel Yarmoluk <daniel.yarmoluk@gmail.com>
License: MIT
Keywords: ai-agents,knowledge-graph,llm,mcp,nemotron,nvidia,perplexity,sonar
Classifier: Development Status :: 3 - Alpha
Classifier: Intended Audience :: Developers
Classifier: License :: OSI Approved :: MIT License
Classifier: Programming Language :: Python :: 3.10
Classifier: Programming Language :: Python :: 3.11
Classifier: Programming Language :: Python :: 3.12
Requires-Python: >=3.10
Requires-Dist: mcp>=1.0.0
Description-Content-Type: text/markdown

# ckg-nemotron-perplexity

**NVIDIA Nemotron + Perplexity Sonar as traversable knowledge graphs — MCP-native.**

[![PyPI version](https://img.shields.io/pypi/v/ckg-nemotron-perplexity?color=0f6e56&label=PyPI)](https://pypi.org/project/ckg-nemotron-perplexity/)
[![License: MIT](https://img.shields.io/badge/license-MIT-0f6e56)](LICENSE)
[![Built by Graphify.md](https://img.shields.io/badge/built_by-Graphify.md-0f6e56)](https://graphifymd.com)

Your agent asks about Nemotron's synthetic data pipeline or Perplexity's reasoning stack — and gets a dependency graph back, not a paragraph it has to parse.

---

## Domains

| Domain | Nodes | Coverage |
|---|---|---|
| `nemotron` | 48 | Architecture · Training · Synthetic Data · RLHF · Deployment · Pruning |
| `perplexity-sonar` | 35 | Sonar API · Models · Web Search · Citations · Output Modes |

---

## Install

```bash
pip install ckg-nemotron-perplexity
```

**Claude Desktop** — add to `claude_desktop_config.json`:
```json
{
  "mcpServers": {
    "ckg-nemotron-perplexity": {
      "command": "ckg-nemotron-perplexity"
    }
  }
}
```

**uvx (no install):**
```bash
uvx ckg-nemotron-perplexity
```

---

## Tools

| Tool | What it does |
|---|---|
| `list_domains` | Show available domains |
| `query_ckg(concept, domain, depth)` | Subgraph: prerequisites + what builds on this concept |
| `get_prerequisites(concept, domain)` | Full upstream chain to root concepts |
| `search_concepts(query, domain)` | Find exact concept labels by partial match |

---

## Example traversals

**Nemotron synthetic data pipeline:**
```
query_ckg("Synthetic Data Generation", "nemotron", 3)

→ Prerequisites:
  - Nemotron-4 as Judge
    - Nemotron-4 340B
      - Transformer Architecture
      - Group Query Attention
      - Rotary Position Embedding

→ Builds toward:
  - Topic Diversity Sampling
  - Response Diversity Sampling
  - Synthetic RLHF Data
```

**Perplexity deep research prerequisites:**
```
get_prerequisites("sonar-deep-research", "perplexity-sonar")

Chain: sonar-deep-research → sonar-pro → Sonar API
```

**Cross-domain: what Nemotron deployment requires:**
```
query_ckg("NVIDIA NIM", "nemotron", 3)

→ TensorRT-LLM Backend → Triton Inference Server → NVIDIA NIM
```

---

## Why traversal

Retrieval gives you documents. Traversal gives you the path.

Asking "what does sonar-reasoning-pro require?" from a vector index returns a paragraph about Perplexity. Asking the same question via CKG returns the exact dependency chain: `sonar-reasoning-pro → Sonar API → API Key Authentication → Rate Limiting`.

No inference. No hallucinated edges. Every relationship was declared, not retrieved.

---

## EVAL

```
benchmark: ckg-benchmark v0.6.2
dataset: huggingface.co/datasets/danyarm/ckg-benchmark
benchmarked: false
rag_baseline_f1: 0.123
graphrag_baseline_f1: 0.120
mean_tokens: 269
paper: github.com/Yarmoluk/ckg-benchmark/blob/main/paper/main.pdf
```

---

<div align="center">

[graphifymd.com](https://graphifymd.com) · patent pending

</div>
