Knowledge Graph Builder

Download SKILL.mdGet Murage How skills work

Open source. macOS, Windows and Linux.

At a glance

What it is

Knowledge graph engineering covering ontology design, graph database selection (Neo4j, Amazon Neptune, ArangoDB), entity and relation extraction from text, graph-based RAG (GraphRAG), knowledge graph embeddings, graph query optimization, and integration patterns with LLM systems.

A skill is a written procedure a bot loads when a job calls for it. This one is a single file, SKILL.md, and the whole file is on this page.

File: SKILL.md
Length: 334 lines1,589 words, about 8 min read
Category: AI and machine learningapplied AI
Level: Advanced
License

More AI and machine learning skills.

The full skill.

The whole file, exactly as it downloads. Read it before you rely on it.

SKILL.md

334 lines · 1,589 words

Download

Knowledge Graph Builder

Overview

Knowledge graphs represent structured relationships between entities, enabling powerful reasoning, contextual retrieval, and explainable AI. Building production knowledge graphs requires ontology design, entity extraction, graph database selection, query optimization, and integration with LLM systems via GraphRAG patterns.

Knowledge Graph Architecture

+------------------+     +-------------------+     +------------------+
| Data Sources     | --> | Extraction        | --> | Graph Database   |
| - Documents      |     | - NER             |     | (Person)-[:WORKS_AT]->(Company)
| - APIs           |     | - Relation extract|     | (Paper)-[:CITES]->(Paper)
| - Databases      |     | - Entity linking  |     | (Product)-[:HAS]->(Feature)
+------------------+     +-------------------+     +------------------+
                                                          |
+------------------+     +-------------------+     +------v-----------+
| Applications     | <-- | Query & Reasoning | <-- | Indexing &       |
| - GraphRAG       |     | - Cypher / SPARQL |     | Enrichment       |
| - Recommendations|     | - Path finding    |     | - Embeddings     |
| - Analytics      |     | - Subgraph match  |     | - Community det. |
+------------------+     +-------------------+     +------------------+

Put this skill to work.

Questions.

What is the Knowledge Graph Builder skill?

How much does it cost?

Nothing. It ships with Murage, which is open source. Your bots run on the AI plan you already have.

Can I change it?

Yes. Open it in Settings → Skills to read, edit or duplicate it. Skill Guard scans it again after every edit.

Give your first job to Murage.

Download the app, connect the AI you already pay for, and tell your Chief of Staff what needs doing. Plan on about ten minutes from install to a working team.

Download Murage

No account needed · Runs on the AI plan you already pay for

from neo4j import GraphDatabase class GraphRAGPipeline: def __init__(self, neo4j_uri: str, neo4j_auth: tuple, llm_model: str = "gpt-4o"): self.driver = GraphDatabase.driver(neo4j_uri, auth=neo4j_auth) self.client = OpenAI() self.llm_model = llm_model def query(self, question: str) -> dict: entities = self._extract_query_entities(question) subgraph = self._retrieve_subgraph(entities) answer = self._generate_answer(question, subgraph) return {"answer": answer, "entities_found": entities, "subgraph_nodes": len(subgraph["nodes"])} def _extract_query_entities(self, question: str) -> list[str]: resp = self.client.chat.completions.create( model="gpt-4o-mini", messages=[{"role": "user", "content": f"Extract entity names. JSON: {{\"entities\": [...]}}\nQuestion: {question}"}], response_format={"type": "json_object"}, temperature=0, ) return json.loads(resp.choices[0].message.content).get("entities", []) def _retrieve_subgraph(self, entities: list[str], hops: int = 2) -> dict: nodes, edges = [], [] with self.driver.session() as session: for entity in entities: result = session.run(""" CALL db.index.fulltext.queryNodes('entity_search', $name) YIELD node, score WHERE score > 0.5 WITH node LIMIT 5 CALL apoc.path.subgraphAll(node, {maxLevel: $hops, limit: 50}) YIELD nodes, relationships RETURN nodes, relationships """, name=entity, hops=hops) for record in result: nodes.extend([{"labels": list(n.labels), "props": dict(n)} for n in record["nodes"]]) edges.extend([{"type": r.type} for r in record["relationships"]]) return {"nodes": nodes, "edges": edges} def _generate_answer(self, question: str, subgraph: dict) -> str: context = "\n".join( f"[{'/'.join(n['labels'])}] {n['props']}" for n in subgraph["nodes"][:30] ) resp = self.client.chat.completions.create( model=self.llm_model, messages=[ {"role": "system", "content": "Answer using the knowledge graph context. Cite entities."}, {"role": "user", "content": f"Context:\n{context}\n\nQuestion: {question}"}, ], temperature=0, ) return resp.choices[0].message.content

Feature	Neo4j	Amazon Neptune	ArangoDB	TigerGraph	Dgraph
Query Lang	Cypher	Gremlin/SPARQL	AQL	GSQL	GraphQL+-
Hosting	Both	Managed (AWS)	Both	Both	Both
Vector Support	Since 5.13	No	No	No	No
RDF Support	Limited	Native	No	No	No
Community	Very Large	AWS ecosystem	Medium	Medium	Medium
Learning Curve	Low	Medium	Medium	High	Medium

Method	Speed	Quality	Cost	Best For
SpaCy NER	Very Fast	Good	Free	Standard entities, high volume
LLM extraction	Slow	Excellent	API cost	Complex relations, custom schema
Fine-tuned BERT	Fast	Very Good	Free (GPU)	Domain-specific entities
Hybrid (SpaCy + LLM)	Medium	Excellent	Moderate	Best quality/cost balance

Dimension	Vector RAG	GraphRAG	Hybrid
Retrieval	Semantic similarity	Structural traversal	Both
Reasoning	Limited	Multi-hop relationships	Strong
Explainability	Low	High (explicit paths)	High
Setup complexity	Low	High	High
Best for	Unstructured Q&A	Relational questions	Complex domains

Model	Approach	Best For
TransE	Translational (h + r = t)	Hierarchical relations
DistMult	Bilinear diagonal	Symmetric relations
ComplEx	Complex-valued bilinear	Asymmetric relations
RotatE	Rotational	Inferring new relations

Knowledge Graph Builder

More AI and machine learning skills.

The full skill.

Knowledge Graph Builder

Overview

Knowledge Graph Architecture

Put this skill to work.

Questions.

Give your first job to Murage.

Search murage.ai

Knowledge Graph Builder

More AI and machine learning skills.

The full skill.

Knowledge Graph Builder

Overview

Knowledge Graph Architecture

Put this skill to work.

Questions.

Give your first job to Murage.

Create a bot and use this skill.

Ask Ember to create the bot

Open the bot’s skills

Find Knowledge Graph Builder and switch it on

Ask the bot for the job

About this skill.

Graph Database Selection

Ontology Design

Design Process

Schema in Neo4j

Common Ontology Patterns

Entity and Relation Extraction

LLM-Based Extraction

Extraction Method Comparison

GraphRAG

GraphRAG vs. Vector RAG

Implementation

Knowledge Graph Embeddings

Query Optimization

Checklist

When to Use

Output Format

Example

Edge Cases