TL;DR

Discover how to leverage the new chat assistant in Dependify 1.5.1

dotnet tool install -g Dependify.Cli

Source code : https://github.com/NikiforovAll/dependify

Package	Version	Description
Dependify.Cli		Use Dependify directly from the CLI (supports plain, mermaidjs, and JSON formats) or from the browser.
Dependify.Aspire.Hosting		Aspire support
Dependify.Aspire.Hosting.Ollama		Ollama Aspire Component enables local chat completions, e.g., you can run the phi3:mini model and use it within Dependify.

Introduction

The AI chat feature enables you to inquire about dependencies. You can ask questions regarding the relationships between modules, identify potential issues, and gain a deeper understanding of your project structure.

💡Find my more about other features from my previous blog post: Explore .NET application dependencies by using Dependify tool

Run

All you need to do it to provide the API key and the endpoint to the chat assistant. You can do it via the command line:

dependify serve $dev/cap-aspire/ \
    --endpoint https://api.openai.azure.com/ \
    --deployment-name gpt-4o-mini \
    --api-key <api-key>

Configure

You can configure the chat assistant in two ways:

1. Environment variables (recommended way)
2. Command-line arguments

E.g:

export DEPENDIFY __AI__ ENDPOINT="https://api.openai.azure.com"
export DEPENDIFY __AI__ DEPLOYMENT_NAME="gpt-4o-mini"
export DEPENDIFY __AI__ API_KEY=""

Aspire support

You can use Dependify as a tool/extensions for Aspire.

Add the package to AppHost:

dotnet add package Dependify.Aspire.Hosting

In the code below, I’ve added the Dependify to the Aspire starter project. (dotnet new aspire-starter)

Register via IDistributedApplicationBuilder. Add the following code to your Program.cs:

var builder = DistributedApplication.CreateBuilder(args);

var endpointParam = builder.AddParameter("endpoint");
var deploymentNameParam = builder.AddParameter("deployment-name");
var apiKeyParam = builder.AddParameter("api-key", secret: true);

var apiService = builder.AddProject<Projects.aspire_project_ApiService>("apiservice");

builder.AddProject<Projects.aspire_project_Web>("webfrontend")
    .WithExternalHttpEndpoints()
    .WithReference(apiService);

builder.AddDependify().ServeFrom("../../aspire-project/")
    .WithAzureOpenAI(
        endpointParam.Resource.Value,
        deploymentNameParam.Resource.Value,
        apiKeyParam.Resource.Value
    );

builder.Build().Run();

Configure:

dotnet user-secrets set "Parameters:endpoint" "https://api.openai.azure.com"
dotnet user-secrets set "Parameters:deployment-name" "gpt-4o-mini"
dotnet user-secrets set "Parameters:api-key" ""

Run:

dotnet run --project ./AppHost

💡 It might take some time to download Dependify from GitHub Container Registry, but eventually, you will be able to run Dependify as Aspire Component.

Conclusion

In this post, we explored the Dependify tool, which can help you navigate the dependency graph of your .NET application. By using this tool, you can visualize the dependencies between modules, identify potential issues, and gain a better understanding of your project structure.

References

• https://github.com/NikiforovAll/dependify