RKnowledge - Production-Grade Knowledge Graph CLI

How We Differ

rahulnyk/knowledge_graph is a brilliant ~300-line Python notebook proof of concept. RKnowledge turns that idea into a 15,500-line production CLI.

Dimension	Original (Python)	RKnowledge (Rust)
Interface	Jupyter notebook cells	Full CLI with 10 subcommands better
Language	Python	Rust — compiled binary, ~15x faster parsing better
LLM Providers	Ollama only (Mistral 7B)	Anthropic, OpenAI, Google, Ollama better
Concurrency	Sequential (1 chunk at a time)	Parallel LLM calls with `-j` flag new
Storage	In-memory Pandas DataFrames	Neo4j graph DB (Docker), persistent better
Incremental Builds	Always from scratch	`--append` merges into existing graph new
Input Formats	PDF only	PDF, Markdown, HTML, plain text better
Entity Typing	8 fixed categories	Free-form LLM classification better
Graph Algorithms	Degree + Louvain	PageRank, Label Propagation, Dijkstra shortest path, density new
Querying	None	`query`, `path`, `stats`, `communities` commands new
Visualization	Static Pyvis HTML	Interactive: click cards, search, type legend, toggle edges better
Export	None	JSON, CSV, GraphML, Cypher dump new
CI/CD	None	GitHub Actions: lint, test, multi-platform build new
Distribution	`docker build` + Jupyter	Single binary, curl install, skills.sh better
Tests	None	118 tests (107 unit + 11 integration) new

Why RKnowledge?

Everything you need to extract, store, analyze, and explore knowledge graphs from any text.

⚡

Blazing Fast

Compiled Rust binary. Parallel LLM extraction with -j concurrency. Zero runtime dependencies.

🔌

Multi-Provider LLM

Anthropic, OpenAI, Google Gemini, and Ollama for free local inference. Switch with one flag.

🏷️

Smart Entity Typing

LLM classifies each concept freely — "programming language", "database", "medical condition" — no fixed list.

📊

Graph Analytics

PageRank, community detection, shortest path, connected components, degree distribution. Built-in.

🔗

Neo4j + Incremental

Persistent Neo4j backend via Docker. --append merges new documents without rebuilding.

🌐

Interactive Visualization

Click-to-expand detail cards, entity type legend, search, toggle proximity edges, freeze physics.

Quick Start

Install

                            $ curl -fsSL https://raw.githubusercontent.com/Algiras/RKnowledge/main/install.sh | bash
                        

Initialize & configure

                            $ rknowledge init     # Start Neo4j + create config
$ rknowledge auth     # Configure LLM provider
                        

Build your knowledge graph

                            $ rknowledge build ./docs/ --provider ollama -j 8
# Extracted 40 relations from 4 chunks (8 concurrent)
# Graph: 54 nodes, 383 edges
                        

Explore

                            $ rknowledge query "machine learning" --depth 2
$ rknowledge path "docker" "kubernetes"
$ rknowledge stats
$ rknowledge viz
                        

Commands

Command	Description
`init`	Initialize config and start Neo4j via Docker
`auth`	Configure LLM provider API keys (interactive)
`build <path>`	Process documents and build knowledge graph. Flags: `-j`, `--append`, `--provider`
`query <q>`	Search the graph with variable `--depth` or raw `cypher:` queries
`path <from> <to>`	Find shortest path between two conceptsnew
`stats`	Graph analytics: PageRank, density, degree distribution, entity typesnew
`communities`	List detected communities and their membersnew
`export`	Export to JSON, CSV, GraphML, or Cypher
`viz`	Open interactive graph visualization in browser

🧠 RKnowledge