Local AI for .NET.
Zero sidecars. In your process.

using OpenTail.Stingray;
using OpenTail.Stingray.Core.Catalog;
using OpenTail.Stingray.Executors;

// 1. Resolve model path from ModelHome inventory or direct GGUF:
string modelPath = ModelHome.ResolveModelPath("qwen2.5-0.5b");

// 2. Memory-map model, allocate context, and initialize chat session:
using var model = Model.Load(modelPath);
using var context = model.CreateContext(new ContextParams { ContextSize = 4096 });
var executor = new InteractiveExecutor(context);
var session = new ChatSession(executor);

// 3. Stream generated tokens directly to your UI or console:
await foreach (string piece in session.ChatAsync("Why write unit tests?"))
{
    Console.Write(piece);
}

using System.Text;
using System.Text.Json;
using OpenTail.Stingray;
using OpenTail.Stingray.Core.Grammar;
using OpenTail.Stingray.Engine;
using OpenTail.Stingray.Executors;

using var model = Model.Load(args[0]);
using var context = model.CreateContext(new ContextParams { ContextSize = 2048 });
var session = new ChatSession(new InteractiveExecutor(context));

// The schema becomes a token constraint: every token the model emits keeps the output valid for it.
string schema = """
    { "type": "object",
      "properties": { "vendor": { "type": "string" }, "invoice_number": { "type": "string" }, "total": { "type": "number" } },
      "required": ["vendor", "invoice_number", "total"] }
    """;
var constraint = JsonConstraint.FromSchemaJson(new GrammarVocabulary(context.Tokenizer), schema, orderedProperties: true);

var json = new StringBuilder();
await foreach (var chunk in session.ChatChunksAsync(
    "Extract the invoice details from this message: invoice INV-2026-0417 from Northwind Traders for 1,284.50 EUR.",
    new InferenceParams { MaxTokens = 200, Temperature = 0f, EnableThinking = false, Constraint = constraint }))
    if (chunk.Kind == GenerateChunkKind.Text) json.Append(chunk.Text);

// The constraint guarantees the SHAPE, not that the values are true: check them against the source.
using var doc = JsonDocument.Parse(json.ToString());
Console.WriteLine(json);

using OpenTail.Stingray.Audio;
using OpenTail.Stingray.Audio.Piper;
using OpenTail.Stingray.Audio.Whisper;

// Text to speech (Piper voice, runs through ONNX Runtime):
using var tts = PiperPipeline.FromConfigFile("en_US-lessac-medium.onnx.json");
tts.Generate(new AudioGenerationRequest { Text = "Hello from a voice that never left this computer.", OutputPath = "hello.wav" });

// Speech to text (Whisper, ggml-base.bin):
using var stt = WhisperPipeline.Load("ggml-base.bin");
var (samples, sampleRate, _) = WavReader.ReadWav("hello.wav");
Console.WriteLine(stt.Transcribe(new SpeechToTextRequest { AudioSamples = samples, SampleRate = sampleRate }).Text);

using OpenTail.Stingray.Vision;

// Open a vision projector (mmproj GGUF) and turn an image into visual-token embeddings.
// This is the building block for image understanding, not a finished "describe this picture" call.
using var embedder = UnifiedVisionPipeline.Open("models/mmproj-deepseek-ocr-2-q8_0.gguf");
float[] visualTokens = embedder.EmbedImageFile("document.png", out int tokenCount);
Console.WriteLine($"Generated {tokenCount} visual tokens with embedding dim {embedder.EmbeddingDim}");

using OpenTail.Stingray.Server;

var builder = WebApplication.CreateBuilder(args);
builder.Services.AddOpenTailStingray(builder.Configuration, o => o.ModelPath = "models/qwen2.5-0.5b-instruct-q4_k_m.gguf");

var app = builder.Build();
app.MapOpenTailStingray();   // /v1/chat/completions, /v1/models, Anthropic and Responses APIs
app.Run("http://localhost:5080");