Neo4j LLM 知识图谱构建器 - 从非结构化文本中提取节点和关系

The Neo4j LLM Knowledge Graph Builder is an 在线应用 for turning unstructured text into a knowledge graph, it provides a magical text to graph experience.

It uses ML models (LLM - OpenAI, Gemini, Llama3, Diffbot, Claude, Qwen) to transform PDFs, documents, images, web pages, and YouTube video transcripts. The extraction turns them into a lexical graph of documents and chunks (with embeddings) and an entity graph with nodes and their relationships, which are both stored in your Neo4j database. You can configure the extraction schema and apply clean-up operations after the extraction.

Afterwards you can use different RAG approaches (GraphRAG, Vector, Text2Cypher) to ask questions of your data and see how the extracted data is used to construct the answers.

best results for files with long-form text in English
less well suited for tabular data like Excel or CSV or images/diagrams/slides
higher quality data extraction if you configure the graph schema for nodes and relationship types

The front-end is a React Application and the back-end a Python FastAPI application running on Google Cloud Run, but you can deploy it locally using docker compose. It uses the llm-graph-transformer module that Neo4j contributed to LangChain and other langchain integrations (e.g. for GraphRAG search).

All Features are documented in detail here: Features documentation

Here is a quick demo

分步说明

Open the LLM 知识图谱构建器
Connect to a Neo4j（Aura） instance
Provide your PDFs, Documents, URLs or S3/GCS buckets
Construct Graph with the selected LLM
Visualize Knowledge Graph in App
Chat with your data using GraphRAG
Open Neo4j Bloom for further visual exploration
Use the constructed knowledge graph in your applications

Login, or create an account at https://console.neo4j.io
Under Instances, create a new AuraDB Free Database
Downloads the credentials file
Wait until the instance is running
Drop the credentials file on the connect dialog for the LLM Graph Builder

工作原理

Uploaded Sources are stored as Document nodes in the graph
Each document (type) is loaded with the LangChain Loaders
The content is split into Chunks
Chunks are stored in the graph and connected to the Document and to each other for advanced RAG patterns
Highly similar Chunks are connected with a SIMILAR relationship to form a kNN Graph
Embeddings are computed and stored in the Chunks and Vector index
Using the llm-graph-transformer or diffbot-graph-transformer entities and relationships are extracted from the text
Entities and relationships are stored in the graph and connected to the originating Chunks