Codeblaze.SemanticKernel.Connectors.Memory.Neo4j 1.3.1

Neo4j Memory Store

Supports

• Customizable queries
• Runtime properties
• Vector embedding Update, Get, Remove and Search
• TextMemoryPlugin support see example 15 of Semantic Kernel

⚠️ Kernel Memory: Is experimental in the semantic kernel.

Quick Start

• Install from nuget

  dotnet add package Codeblaze.SemanticKernel.Connectors.Memory.Neo4j
  

• Configure the memory store

  var builder = new MemoryBuilder();
  
  builder.WithHttpClient(new HttpClient());
  builder.WithOllamaTextEmbeddingGeneration(config["Ollama:Model"], config["Ollama:BaseUrlEmbeddings"]);
  builder.WithNeo4jMemoryStore(config["Neo4j:Url"], config["Neo4j:Username"], config["Neo4j:Password"],
      new Neo4jVectorIndexQueryFactory(
          "<index-name>",
          "<node-name>",
          "<index-property-name>",
          "<text-property-name>",
          384 // Dimensions
      )
  );
  
  memory = builder.Build();
  

• Store embedding

  memory.SaveReferenceAsync(
      collection: "<index-name>",
      externalSourceName: "<node-name>",
      externalId: "<node-id>",
      description: "not used",
      text: "<actual-text>"
  );
  

• Search embedding

  var results = memory.SearchAsync("<index-name>", "<search-text>", limit: 3);
  

• Retrieve node

  var results = await memory.GetAsync("<index-name>", "<node-id>");
  

Query Factory

You can implement you own INeo4jQueryFactory or Extend Neo4jVectorIndexQueryFactory to modify any of the queries (Default queries found in Neo4jVectorQueries).