Codeblaze.SemanticKernel.Plugins.Neo4j 1.3.0

Neo4j Cypher Codegen

This package provides semantic kernel plugin for Neo4j cyher codegen support, Enabling talking to knowledge graphs respecting it's schema

Important

• Plugin has been tested with gpt-3.5-turbo, other models may require tweaks to the prompts
• Make sure your neo4j instance has the apoc plugin
• On initialization, the plugin queries the database to generate a text representation of schema

Quick Start

• Install from nuget

  dotnet add package Codeblaze.SemanticKernel.Plugins.Neo4j
  

• Configure the kernel

  var builder = Kernel.CreateBuilder();
  
  builder.AddOpenAIChatCompletion(
      config["OpenAI:Model"], 
      config["OpenAI:Key"]
  );
  
  var kernel = builder.Build();
  
  kernel.AddNeo4jCypherGenPlugin(
      config["Neo4j:Url"], 
      config["Neo4j:Username"], 
      config["Neo4j:Password"]
  );
  

• Query the database using prompts

  var prompt = "Which movies came out in 1990?";
  
  var result = await _Kernel.InvokeAsync<Neo4jResult>(
      nameof(Neo4jCypherGenPlugin), "Query", 
      new KernelArguments
      {
          { "prompt", prompt }
      }
  );
  
  result.Cypher; // Generated cypher (string)
  result.Result; // Neo4j query execution result (List<IRecord>)
  

Semantic Kernel Functions

• Query

  • description

    Generates cypher code based on the provided prompt, executes it against the configured neo4j instance returns both cypher and the prompt

  • usage

      _Kernel.InvokeAsync<Neo4jResult>(
          nameof(Neo4jCypherGenPlugin), "Query", 
          new KernelArguments
          {
              { "prompt", prompt }
          }
      );
    

• GenerateCypher

  • description

    Generates cypher code based on the provided prompt

  • usage

        _Kernel.InvokeAsync<string>(
            nameof(Neo4jCypherGenPlugin), "GenerateCypher", 
            new KernelArguments
            {
                { "prompt", prompt }
            }
        );