EasyException.Core 0.1.0

EasyException (WIP)

EasyException is a set of NuGet packages that capture unhandled exceptions in ASP.NET Core and analyze the exception asynchronously with a local LLM (Ollama).

Design goals:

• Do not block the request while the LLM runs (analysis happens in a background queue).
• Keep the original exception; analysis is attached via Exception.Data["EasyException.Analysis"] when possible.

Packages (in this repo)

• EasyException.Core: core types (ExceptionReport, breadcrumbs, analyzer contracts)
• EasyException.Analyzers.Ollama: IExceptionAnalyzer implementation for Ollama
• EasyException.AspNetCore: middleware + DI + background analysis queue

ASP.NET Core usage

1) Register services

using EasyException.AspNetCore;

var builder = WebApplication.CreateBuilder(args);

builder.Services.AddEasyException(
    configure: options =>
    {
        options.BreadcrumbCapacity = 200;
        options.MaxQueueSize = 256;
        options.AnalysisTimeout = TimeSpan.FromSeconds(10);
    },
    configureOllama: ollama =>
    {
        ollama.BaseUrl = new Uri("http://localhost:11434/");
        ollama.Model = "llama3.2";
        ollama.Timeout = TimeSpan.FromSeconds(10);
    });

var app = builder.Build();

2) Add middleware

app.UseEasyException();

When an unhandled exception happens, the middleware will enqueue it for analysis and attach:

• Exception.Data["EasyException.Id"]
• Exception.Data["EasyException.Fingerprint"]
• Exception.Data["EasyException.AnalysisStatus"] (queued/completed/failed)
• Exception.Data["EasyException.Analysis"] (human-readable analysis, when completed)

Sample WebAPI (for testing)

This repo includes a runnable WebAPI sample:

• samples/EasyException.SampleWebApi

Run it:

dotnet run --project .\samples\EasyException.SampleWebApi\EasyException.SampleWebApi.csproj --urls http://localhost:5055

Trigger an exception:

Invoke-RestMethod http://localhost:5055/boom -Method Get -SkipHttpErrorCheck

You should see EasyException analysis completed... in the server logs.

Setting up Ollama locally (Windows)

Install

Option A (Winget):

winget install Ollama.Ollama

Option B: download from https://ollama.com/download

Start Ollama

Ollama typically runs a local server on http://localhost:11434/.

If it’s not already running, start it:

ollama serve

Pull a model

ollama pull llama3.2

(You can use any Ollama model name you have installed.)

Quick health check

Invoke-RestMethod http://localhost:11434/api/tags

If that returns JSON, the server is up.

Notes / caveats

• Breadcrumbs are what make “steps before the exception” possible. Without them, an LLM can only infer from stack traces/logs and may be wrong.
• This project is intentionally minimal; add redaction, throttling/deduping, and storage if you plan to ship it broadly.