Kjarni.Extensions.AI 0.1.3

Kjarni.Extensions.AI

Local embeddings for .NET through Microsoft.Extensions.AI — no Python, no ONNX, no cloud.

Implements IEmbeddingGenerator<string, Embedding<float>> on top of the Kjarni inference engine, so anything built against the Microsoft.Extensions.AI abstractions — including Semantic Kernel — can run embeddings entirely in-process.

dotnet add package Kjarni.Extensions.AI

using Kjarni.Extensions.AI;
using Microsoft.Extensions.AI;

using IEmbeddingGenerator<string, Embedding<float>> generator =
    new KjarniEmbeddingGenerator("minilm-l6-v2");

var embeddings = await generator.GenerateAsync(["semantic search, locally"]);
Console.WriteLine(embeddings[0].Vector.Length);   // 384

The model downloads on first use and is cached on disk. After that, nothing touches the network.

Why

Every other IEmbeddingGenerator provider calls a hosted service. This one doesn't call anything — inference runs inside your process against a native library. That makes it the option that works when data cannot leave the building: air-gapped deployments, regulated environments, on-premise installs, or a laptop on a plane.

Dependency injection

builder.Services.AddKjarniEmbeddingGenerator("minilm-l6-v2");

Registered as a singleton — weights load once, and the generator serializes concurrent calls internally, so one instance is safe to share across requests.

Then inject it anywhere:

public class SearchService(IEmbeddingGenerator<string, Embedding<float>> embeddings)
{
    public async Task<ReadOnlyMemory<float>> VectorFor(string text) =>
        (await embeddings.GenerateAsync([text]))[0].Vector;
}

Wrapping an existing Embedder

If you already hold a Kjarni Embedder — because you also use it for similarity or search — wrap it rather than loading the model twice:

using Kjarni;
using Microsoft.Extensions.AI;

using var embedder = new Embedder("minilm-l6-v2");

var generator = embedder.AsEmbeddingGenerator("minilm-l6-v2");

AsEmbeddingGenerator and AddKjarniEmbeddingGenerator live in the Microsoft.Extensions.AI namespace, so they light up alongside the abstractions themselves — no extra using to discover.

By default the generator borrows the embedder: disposing the generator leaves the embedder alive, since you created it. Pass ownsEmbedder: true to transfer ownership.

With Semantic Kernel

Semantic Kernel builds on the Microsoft.Extensions.AI abstractions, so registration is the same:

var builder = Kernel.CreateBuilder();
builder.Services.AddKjarniEmbeddingGenerator("minilm-l6-v2");

var kernel = builder.Build();

Any SK component that resolves IEmbeddingGenerator<string, Embedding<float>> — memory stores, vector connectors, RAG pipelines — now runs against local inference.

Batching

Pass the whole collection in one call. Kjarni batches natively, and it is substantially faster than looping:

string[] documents = File.ReadAllLines("corpus.txt");

var embeddings = await generator.GenerateAsync(documents);
// embeddings[i] corresponds to documents[i]

Models

Any Kjarni embedding model works. minilm-l6-v2 is the default — smallest and fastest.

var generator = new KjarniEmbeddingGenerator("mpnet-base-v2");

GPU

var generator = new KjarniEmbeddingGenerator("minilm-l6-v2", device: "gpu");

GPU inference uses WebGPU — Vulkan on Linux, DX12 or Vulkan on Windows, Metal on macOS. CUDA is not required.

Behavior notes

• Generation is synchronous under the hood. Inference is compute-bound, not I/O-bound, so the work runs on the calling thread and returns a completed task. There is no hidden thread-pool hop.
• One model per instance. Kjarni loads a single model per embedder. Passing an EmbeddingGenerationOptions.ModelId that differs from the instance's model throws NotSupportedException rather than silently ignoring it. Construct a second generator instead.
• Fixed dimensions. These models don't support Matryoshka truncation, so EmbeddingGenerationOptions.Dimensions throws if it doesn't match the model's native size.
• Metadata is available through GetService(typeof(EmbeddingGeneratorMetadata)), reporting provider name kjarni, the model id, and its dimension count.

License

MIT.