Build GenAI Apps
for the Enterprise

Return complete answers with LLMs backed by the only knowledge graph with native vector search.

Get Started
Generative AI video thumbnail

Contact Us

POWER LLMS WITH ENTERPRISE KNOWLEDGE

GenAI, Meet Knowledge Graphs

Completeness of Answers

Produce accurate and relevant responses: always explainable.

Lightning-Fast Responses

Return answers with queries that don’t require joins. Ever.

Dev-Friendly Schema

Easily add new and different data without rebuilding from scratch. 

Privacy 
Safeguards

Integrate access policies and add rules for granular control.

Nearly 40% of organizations struggle to validate and trust GenAI content.

Learn how knowledge graph + native vector search takes GenAI from test case to enterprise-ready solution.

Completeness of Answers.
Every Time

Accurate

Return True Facts

Neo4j’s knowledge graph grounds LLM responses in validated facts using retrieval augmented generation (RAG). Easily add and update data, indexes, and RAG sources with our developer friendly schema that never requires redesign. Graph structure makes finding factual answers to multi-hop questions possible through traversals that follow a breadcrumb trail of connections.

Download O'Reilly's Knowledge Graph Book

Query Knowledge graph with LLM application diagram
Expand Diagram
vector search diagram
Expand Diagram
Contextual

Essence of Meaning

Knowledge graphs provide explicit accuracy from your data, while vector search offers implicit responses using semantic meaning. Neo4j offers explicit and implicit responses from one database to deliver the best possible answers to the user. Storing embeddings as node properties gives users full context in addition to explicit responses native to knowledge graphs.

Watch: Vector Search Explained

Relevant

Individualized Results

Ground your LLM in a knowledge graph that represents your company-specific data so responses match your business. Apply data governance policies to control information flow and integrate with your identity and access management provider for continuous governance. Define access policies by role or user and add constraints on nodes, labels, relationships, or properties – you can even wall off specific parts of the graph or limit traversal depth.

Learn More: Add Constraints to Queries

LLM knowledge graph diagram
Expand Diagram
LLM linked data sources and origins diagram
Expand Diagram
Explainable

Traceable Data Sources

Verify enriched LLM responses using linked data sources and origins. Neo4j’s knowledge graphs can capture metadata like sources and organizing principles, simplifying source identification. Users can view and interact with data sources through graph visualizations to better understand how data is connected.

Learn More: Explainable Graph Models

Ground LLMs with Knowledge Graphs:
Step By Step

  1. step 1

    Use Neo4j directly in orchestration frameworks like LangChain, LlamaIndex, and others

  2. step 2

    Add and index vector embeddings in the Neo4j knowledge graph

  3. step 3

    Generate embeddings for user inputs with all model-providers both cloud & local

  4. step 4

    Find most relevant nodes with similarity search in the vector index & retrieve contextual information from the knowledge graph

  5. step 5

    Prompt any LLM, cloud or local, with the user question for natural language searches

  6. step 6

    Ground the LLM with that contextual information with retrieval augmented generation

Cypher

import neo4j
import langchain.embeddings
import langchain.chat_models
import langchain.prompts.chat

emb = OpenAIEmbeddings() # VertexAIEmbeddings() or BedrockEmbeddings() or ...
llm = ChatOpenAI() # ChatVertexAI() or BedrockChat() or ChatOllama() ...

vector = emb.embed_query(user_input)

vectory_query = """
// find products by similarity search in vector index
CALL db.index.vector.queryNodes('products', 5, $embedding) yield node as product, score

// enrich with additional explicit relationships from the knowledge graph
MATCH (product)-[:HAS_CATEGORY]->(cat), (product)-[:BY_BRAND]->(brand)
MATCH (product)-[:HAS_REVIEW]->(review {rating:5})<-[:WROTE]-(customer) 

// return relevant contextual information
RETURN product.Name, product.Description, brand.Name, cat.Name, 
       collect(review { .Date, .Text })[0..5] as reviews, score
"""

records = neo4j.driver.execute_query(vectory_query, embedding = vector)
context = format_context(records)

template = """
You are a helpful assistant that helps users find information for their shopping needs.
Only use the context provided, do not add any additional information.
Context:  {context}
User question: {question}
"""

chain = prompt(template) | llm

answer = chain.invoke({"question":user_input, "context":context}).content

Deep Partnerships with Generative AI Providers

Neo4j Integrates with LLMs and Data Ecosystems

Neo4j’s scalable and flexible database and analytics technologies seamlessly integrate with generative AI tools like LangChain, LlamaIndex, Haystack, Ollama, and beyond. This unlocks a universe of information, while setting a new standard for AI accuracy, transparency, and explainability.

What people are saying about LLMs with Knowledge Graphs

"Before you can get the value of your AI, you have to fix the data. Our journey with Neo4j began several years ago with a data liberalization journey across more than 250+ entities that did not share data… Our enterprise intelligence hub today is trained on knowledge graphs powered Neo4j to co-exist and perform based on how our different departments and teams actually think and work. We have multiple deployments in production now…and it’s enabling us to do generative AI at scale."

Chief Digital Officer
Fortune 500 energy leader