FREE BOILERPLATE
Local LLM Starter for .NET
Call a model on your own machine from ASP.NET Core, and see what every answer really costs: tokens in, tokens out, and time. No cloud bill, no API key.
What you get
- A minimal API with IChatClient: one answer, streaming (SSE), and a health check that tells you why it’s not working
- A measuring layer that logs tokens, time and time to first token, and warns you when a model’s hidden “thinking” leaks into the answer
- A small bench tool: warm-up, 5 runs, p50, for any model you have
- Provider and model in settings, so you can switch without touching code
- Tests that run without Ollama, a GitHub Actions build, and a list of the real pitfalls I hit
Tested with .NET 10 · Ollama 0.40.3 · OllamaSharp 5.5.0 · qwen3:4b-instruct
FREE DOWNLOAD