Use cases

This section shows how to build real applications with Aspose.LLM for .NET. Each use case is a complete, runnable scenario — paste it into a .NET 8 or .NET 10 project, apply a license, and run.

For compact snippets that illustrate a single call, see Quick wins. For the first runnable example, see Hello, world!.

We assume familiarity with C# and async/await. If you are new to the SDK, read Getting started first.

Core scenarios

• Simple chat — one or a few messages without managing sessions yourself.
• Multi-turn chat — an explicit session with several exchanges in one conversation.
• Save and restore session — persist conversation state to disk and resume later.
• Custom preset — patterns for customizing a built-in preset or building one from scratch.

Vision

• Vision question answering — ask questions about images, compare images, transcribe documents.

Throughput and scaling

• Batch processing — run many prompts through one loaded model.
• Streaming-like responses — the no-streaming limitation and workarounds.
• Multiple concurrent sessions — serve many users or workflows from a single instance.

Deployment

• Offline deployment — air-gapped, firewalled, or no-internet targets.
• CPU-only deployment — no GPU, tune threads, set realistic expectations.
• GPU deployment with CUDA — NVIDIA GPUs, single and multi-GPU.
• Integration with ASP.NET Core — host behind HTTP via Minimal API.

Tuning

• Long context tuning — 128K-262K contexts with flash attention and KV quantization.
• Low memory tuning — fit into tight memory budgets.

Advanced

• Bring your own GGUF — custom models from Hugging Face or disk.
• System prompt recipes — effective system-prompt patterns.

What’s next

• Quick wins — compact recipes for single tasks.
• Developer’s reference — conceptual reference for presets, sessions, and the API.
• Supported presets — pick a preset for your scenario.