Loading...
Loading...
Build a real Graph RAG service that extracts entities from raw text, answers questions with generated Cypher, and fuses graph evidence with vector chunks. See exactly when graphs beat vectors and when vectors beat graphs on the same questions.
Message a mentor about fit, prerequisites, or where to start. Replies come on WhatsApp, usually within a day.
Engineers are learning here from
Extract entities and relationships from unstructured text with an LLM, build a Neo4j knowledge graph, and answer multi-hop questions by translating natural language into Cypher. Compare Graph RAG against vector RAG side by side, then fuse both into a hybrid retriever.
Build a knowledge graph from raw text, ask natural-language questions over it, and beat vector RAG on multi-hop queries.
What you'll ship
What you'll learn
Curriculum
Boot Neo4j
Start a Neo4j 5 container with Docker Compose, tour the browser, and write your first Cypher
Ingest and extract
Turn unstructured text into a knowledge graph with LangChain LLMGraphTransformer
Vector baseline
Index the same text into ChromaDB so every comparison against Graph RAG is apples to apples
Cypher answers
Turn natural-language questions into Cypher and grounded answers with GraphCypherQAChain
Hybrid RAG
Fuse graph context with vector chunks so structured and unstructured evidence land in the same answer
Side-by-side comparison
Run the same questions against Graph RAG and vector RAG and see where each one wins
Schema introspection
Expose the live graph schema so your Cypher prompts stay accurate as the graph grows
Who it's for
whose vector RAG keeps failing on questions that need two or three facts chained together
who want structured retrieval but have only shipped pgvector or ChromaDB so far
who need to turn unstructured docs into a queryable knowledge graph without hand-writing extractors
FAQ
No. A short Cypher 101 scene gets you reading and writing simple queries. The Cypher chain writes most queries for you; the goal is to read them and debug when they go wrong.
Neo4j has first-class LangChain integration via langchain-neo4j, LLMGraphTransformer writes directly to it, and Cypher is purpose-built for graph traversals. You can port the ideas to another graph later.
No. The workshop shows when graphs beat vectors, when vectors beat graphs, and how to combine them. You leave with a hybrid retriever, not a replacement.
Any OpenAI-compatible endpoint works. The workshop uses OpenRouter, Fireworks, Gemini, or OpenAI interchangeably. One API key is enough.
Pricing
Subscribe to Pro for every paid course, or buy just this one.
Unlock this course and every paid course plus workshop replays. One subscription.
You save 54% with regional pricing
One-time purchase. Lifetime access to every lesson, exercise, and update.
You save 47% with regional pricing
Still deciding? Ask Param a question
Graphs are not exotic. They are the retrieval layer your hardest questions need.
Graph RAG with LangChain and Neo4j
$79 one-time