[dotnet-ai] ONNX Runtime + local LLM inference

[luisquintanilla]

Fixes #231
Part of #225

Dependencies

Merge order: This PR should be merged after #237 Plugin scaffold.

Summary

Adds two inference skills covering model execution at the edge:

• onnx-runtime-inference Running pre-trained ONNX models via standalone InferenceSession or ML.NET's ApplyOnnxModel(), with execution provider selection (CPU, CUDA, DirectML). Includes references/tensors.md for tensor I/O patterns and a text preprocessing section using Microsoft.ML.Tokenizers for BERT/WordPiece models.
• local-llm-inference Running LLMs locally via Ollama and Foundry Local, both surfacing models through MEAI's IChatClient abstraction (provider-agnostic).

Changes

File	Description
plugins/dotnet-ai/skills/onnx-runtime-inference/SKILL.md	Standalone ONNX + ML.NET ONNX, execution providers, text preprocessing with tokenizers
plugins/dotnet-ai/skills/onnx-runtime-inference/references/tensors.md	Tensor I/O patterns: shape management, named inputs/outputs, batch processing, OrtValue
plugins/dotnet-ai/skills/local-llm-inference/SKILL.md	Ollama + Foundry Local setup, model selection, IChatClient abstraction
tests/dotnet-ai/onnx-runtime-inference/eval.yaml	Eval: ONNX model inference scenario
tests/dotnet-ai/local-llm-inference/eval.yaml	Eval: local LLM setup + production rejection scenarios

Key Packages

• Microsoft.ML.OnnxRuntime / .Gpu / .DirectML
• Microsoft.ML.OnnxTransformer
• OllamaSharp
• Microsoft.AI.Foundry.Local

Validation

• onnx-runtime-inference covers both standalone and ML.NET approaches
• Execution providers (CPU, CUDA, DirectML) documented
• tensors.md provides tensor shape and memory management guidance
• local-llm-inference presents Ollama and Foundry Local neutrally
• Both eval.yaml files have scenarios

[github-actions]

This PR has been automatically marked as stale because it has no activity for 30 days. It will be closed if no further activity occurs within another 7 days of this comment. If it is closed, you may reopen it anytime when you're ready again.

Generated by Close Stale Pull Requests · ◷