agent-graphrag

Name: agent-graphrag
Availability: InStock
Author: agenticgogol

Implements knowledge-graph-based retrieval using LLM-extracted triples and graph traversal.

1|9|Updated Jul 11, 2026

Search Tags:#knowledge graph #llm #rag #networkx #graphrag #graph traversal #triple extraction

Ready to run

npx skills add https://github.com/agenticgogol/Edureka_Coding_Agent_Enabled_Demo_11Jul --skill agent-graphrag-agenticgogol

💬 Prompt for ChatGPT / Claude Chat▼

Please help me install this Agent Skill.
Skill: agent-graphrag
Source: https://github.com/agenticgogol/Edureka_Coding_Agent_Enabled_Demo_11Jul/tree/main/.claude/skills/agent-graphrag
Command: npx skills add https://github.com/agenticgogol/Edureka_Coding_Agent_Enabled_Demo_11Jul --skill agent-graphrag-agenticgogol

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill requires networkx, and includes references (resource) components.

What problem does it solve? Standard vector-based RAG cannot answer multi-hop relationship questions, and teams often mislabel plain vector search as GraphRAG. This Skill builds a genuine knowledge-graph retrieval pipeline so answers come from entity-relationship traversal, not embedding similarity. ## Core Features & Use Cases - Triple Extraction: Uses an LLM to extract (subject, relation, object) triples from source text and builds a directed graph with networkx. - Graph Traversal Answering: Answers questions by finding shortest paths between entities, exposing relationship chains that vector search cannot produce. - Real Provider Enforcement: Routes all LLM calls through a shared llm_client with a verified API key — no mock mode. - Use Case: Given documents like "Alice founded Acme Corp" and "Acme Corp acquired Beta Inc", answer "How are Alice and Beta Inc related?" via multi-hop graph traversal. ## Quick Start Ask the agent to build a GraphRAG demo that extracts entity triples from my documents with networkx and answers a multi-hop relationship question using a real LLM provider key.

Frequently Asked Questions about agent-graphrag

High-intent search queries and answers about installing and using this skill.

FAQPage Schema

How do I build a GraphRAG pipeline in Python?▼

Extract (subject, relation, object) triples from your documents with an LLM prompt, then add them as edges to a networkx DiGraph. Answer questions by finding the shortest path between entities mentioned in the query and returning the relation chain.

What is the difference between GraphRAG and vector-based RAG?▼

GraphRAG retrieves answers by traversing a knowledge graph of entities and relationships, while vector RAG relies on embedding similarity over text chunks. GraphRAG can answer multi-hop questions like how two entities are related, which plain vector search cannot.

Which Python library should I use for knowledge graph retrieval?▼

networkx is used here for a lightweight local demo, providing DiGraph construction and shortest_path traversal. For larger deployments, a dedicated graph-RAG library can be substituted after confirming its extraction and query APIs.

Does GraphRAG require a real LLM API key?▼

Yes. Entity extraction and answer synthesis call an LLM through llm_client.complete, which requires a verified real provider key configured in the environment. There is no mock mode in this workflow.

When should I not use GraphRAG?▼

Avoid GraphRAG when questions are answerable by single-document lookup or plain semantic similarity, since graph construction adds LLM extraction cost without benefit. It fits cases needing multi-hop relational reasoning across entities.