AI/참고/neo4j-graphrag-python-main/examples

Examples Index

This folder contains examples usage for the different features supported by the neo4j-graphrag package:

• Automatic Schema Extraction from PDF or text
• Build Knowledge Graph from PDF or text
• Retrieve information from the graph
• Question Answering (Q&A)

Each of these steps have many customization options which are listed in the last section of this file.

Build Knowledge Graph

• End to end PDF to graph simple pipeline
• End to end text to graph simple pipeline
• Build KG pipeline from config file
• Build KG pipeline with PDF URL

Retrieve

• Retriever from an embedding vector
• Retriever from a text
• Graph-based retrieval with VectorCypherRetriever
• Hybrid retriever
• Hybrid Cypher retriever
• Text2Cypher retriever
• Cypher Template retriever

External Retrievers

Weaviate

• Vector search
• Text search with local embeder
• Text search with remote embeder

Pinecone

• Vector search
• Text search

Qdrant

• Vector search
• Text search

Answer: GraphRAG

• End to end GraphRAG
• GraphRAG with message history
• GraphRAG with Neo4j message history

Customize

Retriever

• Control result format for VectorRetriever
• Control result format for VectorCypherRetriever
• Use pre-filters
• Text2Cypher: use a custom prompt

LLMs

• OpenAI (GPT)

• Azure OpenAI

• VertexAI (Gemini)

• MistralAI

• Cohere

• Anthropic (Claude)

• Ollama

• Custom LLM

• Message history

• Message history with Neo4j

• System Instruction

• Tool Calling with OpenAI

• Tool Calling with VertexAI

• Tool Calling with Ollama

Prompts

• Using a custom prompt for RAG

Embedders

• OpenAI
• Azure OpenAI
• VertexAI
• MistralAI
• Cohere
• Ollama
• Custom LLM

KG Construction - Pipeline

• End to end example with explicit components and text input
• End to end example with explicit components and PDF input
• Process multiple documents
• Export lexical graph creation into another pipeline
• Build pipeline from config file
• Add event listener to get notification about Pipeline progress
• Use component context to send notifications about Component progress

Components

• Loaders:
  • Load PDF file
  • Custom
• Text Splitter:
  • Fixed size splitter
  • Splitter from LangChain
  • Splitter from LLamaIndex
  • Custom
• Chunk embedder
• Schema Builder:
  • User-defined
  • Automatic schema extraction
• Entity Relation Extractor:
  • LLM-based
  • LLM-based with custom prompt
  • Custom
• Graph Pruner
• Knowledge Graph Writer:
  • Neo4j writer
  • Custom
• Entity Resolver:
  • FuzzyMatchResolver
  • SinglePropertyExactMatchResolver with pre-filter
  • SpaCySemanticMatchResolver with pre-filter
  • Custom resolver
• Custom component

Answer: GraphRAG

• LangChain compatibility
• Use a custom prompt

Database Operations

• Create vector index
• Create full text index
• Populate vector index