technology-selection

Selecting an AI/ML library for .NET 8 involves mapping task requirements like classification or LLM integration against deployment constraints like latency and offline needs. ML.NET suits local tasks, while Azure.AI.OpenAI handles cloud-based LLM integration.

Integrate LLM capabilities into .NET by utilizing Azure.AI.OpenAI, OpenAI, or the GitHub Copilot SDK. Selection depends on whether your project requires cloud-based inference, local execution via ONNX Runtime, or agentic workflows.

Run AI models offline in .NET applications using Microsoft.ML.OnnxRuntime or ML.NET. These libraries enable local model execution, satisfying strict offline requirements and reducing latency dependencies on cloud endpoints.

ML.NET supports recommendation, anomaly detection, classification, regression, and clustering tasks. It provides native .NET APIs to train, evaluate, and deploy models for these machine learning scenarios within .NET 8+ environments.

Use Microsoft.Agents.AI for complex, multi-step agentic workflows requiring autonomous decision-making, and use Azure.AI.OpenAI for direct LLM API calls. Your specific AI task complexity dictates the optimal library selection for .NET.

ONNX Runtime limitations in .NET include potential overhead during model conversion and strict dependency on compatible ONNX model formats. It is designed for cross-platform model execution rather than training models from scratch.