Darkspyre.Inference.Tensor.Decision.Laya
11.20.0
Prefix Reserved
dotnet add package Darkspyre.Inference.Tensor.Decision.Laya --version 11.20.0
NuGet\Install-Package Darkspyre.Inference.Tensor.Decision.Laya -Version 11.20.0
<PackageReference Include="Darkspyre.Inference.Tensor.Decision.Laya" Version="11.20.0" />
<PackageVersion Include="Darkspyre.Inference.Tensor.Decision.Laya" Version="11.20.0" />
<PackageReference Include="Darkspyre.Inference.Tensor.Decision.Laya" />
paket add Darkspyre.Inference.Tensor.Decision.Laya --version 11.20.0
#r "nuget: Darkspyre.Inference.Tensor.Decision.Laya, 11.20.0"
#:package Darkspyre.Inference.Tensor.Decision.Laya@11.20.0
#addin nuget:?package=Darkspyre.Inference.Tensor.Decision.Laya&version=11.20.0
#tool nuget:?package=Darkspyre.Inference.Tensor.Decision.Laya&version=11.20.0
Darkspyre.Inference.Tensor.Decision.Laya
Typed decisions with Laya-family models inside the application's process. The ModernBERT encoder and the typed
decision heads run through ONNX Runtime: Core ML on macOS and the CPU provider elsewhere. Tokenization, sequence
framing and calibration run in C#. The engine implements Common's IDecisionInference and
IDecisionInputMeasurement.
Install
dotnet add package Darkspyre.Inference.Tensor.Decision.Laya --prerelease
Register
Construct LayaDecisionInference over a TensorModelPackManager, or call AddTensorEngines() from
Darkspyre.Inference.Tensor.Registry, which registers LayaEngineProvider. The package has no built-in pack: the
application registers the Laya packs it offers, built with TensorDecisionModels.LayaRuntimeArchitecture.
Example
var catalog = new TensorModelCatalog([customPack]);
using var packs = new TensorModelPackManager(catalog, packDirectory);
await packs.InstallAsync(customPack.Key, cancellationToken: cancellationToken);
await using var laya = new LayaDecisionInference(packs);
await laya.StartAsync(cancellationToken);
await laya.LoadAsync(customPack.Key, cancellationToken); // Accepts loading, not completion.
Subscribe to EngineStatusChanged before loading. After it reports Loaded, make requests; handle Failed instead of
sending a request after a failed load. Awaiting LoadAsync alone does not wait for Loaded.
DecisionInferenceResult result = await laya.DecideAsync(
request with { ModelId = customPack.Key.ModelId }, cancellationToken);
MeasureAsync reports how many tokens a request's state encodes to and how many each question keeps, without running
the model, so the application can tell when a question truncates the state.
Settings
LayaSetup.CreateSettingsManager opens the engine's settings over the application's model list: the model with its
install state, "Runs on", the fields that import a model from a file, folder or URL, and the download, check, load and
backend-check actions.
Limits
- A pack must provide the
modelONNX graph,tokenizerandagent-configartifacts, and keep the Laya architecture, typed heads and tokenizer conventions. A pack that declares CUDA is refused. - On Core ML the graph compiles on every load, so loading a large checkpoint takes tens of seconds. Load in the background and keep the instance loaded.
- A request with more options than the graph's option slots fails with
InvalidRequest. - A state longer than a question's room is truncated without an error; measure first to detect it.
The Darkspyre.Inference packages
This package is one library of the Darkspyre.Inference family. The family runs AI inference from .NET applications through typed, provider-neutral contracts.
Darkspyre.Inference.Commonholds the contracts every other package shares: requests, results, capabilities, failures and the provider setup model.Darkspyre.Inference.Connect.Commonholds the conversation and embedding services. Each Connect provider package adds one provider, andAddLlmConnect()discovers every provider package the application references.- The Tensor packages load and run models inside the application's own process.
- The UI.Settings packages provide the setup editors a host places in its own window or page.
Every package of the family carries the same version. Reference all of them at the same version: each package depends on the exact version of the family packages it uses.
The packages are licensed under the MIT license. LICENSE.md in each package holds the license text.
| Package | Holds |
|---|---|
| Darkspyre.Inference.Common | Shared contracts for Connect and Tensor |
| Darkspyre.Inference.Connect | Image, speech and audio connectors |
| Darkspyre.Inference.Connect.Common | Conversation and embedding services and provider discovery |
| Darkspyre.Inference.Connect.Codex | Conversations on a ChatGPT subscription |
| Darkspyre.Inference.Connect.Claude | Conversations on a Claude subscription through the Claude CLI |
| Darkspyre.Inference.Connect.HostTools | The host-tools local service and its client |
| Darkspyre.Inference.Connect.Antigravity | Conversations on a Google subscription through the Antigravity CLI |
| Darkspyre.Inference.Connect.Grok | Conversations on a Grok subscription |
| Darkspyre.Inference.Connect.Anthropic | Conversations through the Anthropic API with an API key, with hosted web search |
| Darkspyre.Inference.Connect.OpenAiApi | Conversations and embeddings through the OpenAI API |
| Darkspyre.Inference.Connect.SelfHost | Conversations and embeddings on engines the application runs |
| Darkspyre.Inference.Tensor.Common | Model packs, the model lifecycle and the Tensor capability interfaces |
| Darkspyre.Inference.Tensor.Decision.Laya | Laya typed decisions through ONNX Runtime |
| Darkspyre.Inference.Tensor.Embedding | Text embeddings from GGUF sentence encoders |
| Darkspyre.Inference.Tensor.Embedding.Gemma | EmbeddingGemma text and image embeddings |
| Darkspyre.Inference.Tensor.Ggml | The process's GGML and MLX native runtime |
| Darkspyre.Inference.Tensor.Image.Supra | Supra2-IMG image generation |
| Darkspyre.Inference.Tensor.LLM.Qwen | Qwen GGUF text generation |
| Darkspyre.Inference.Tensor.Registry | Registration and discovery of the Tensor engines |
| Darkspyre.Inference.Tensor.STT.Whisper | Whisper speech transcription |
| Darkspyre.Inference.UI.Settings.Core | Editor state and host contracts both setup editors share |
| Darkspyre.Inference.UI.Settings.Avalonia | The Avalonia setup editor |
| Darkspyre.Inference.UI.Settings.Blazor | The Blazor setup editor |
| Product | Versions Compatible and additional computed target framework versions. |
|---|---|
| .NET | net11.0 is compatible. |
-
net11.0
- Darkspyre.Inference.Tensor.Common (= 11.20.0)
- Darkspyre.ML.Tokenizers (>= 2.0.1.1)
- Microsoft.ML.OnnxRuntime (>= 1.30.0)
- TensorSharp.Runtime (>= 2026.10.11)
NuGet packages (1)
Showing the top 1 NuGet packages that depend on Darkspyre.Inference.Tensor.Decision.Laya:
| Package | Downloads |
|---|---|
|
Darkspyre.Inference.Tensor.Registry
Registration, validation and ordered discovery of the Darkspyre Tensor engine implementations. |
GitHub repositories
This package is not used by any popular GitHub repositories.
| Version | Downloads | Last Updated |
|---|---|---|
| 11.20.0 | 104 | 11/15/2026 |
| 11.19.4-rc.3 | 83 | 11/6/2026 |
| 11.19.3-rc.3 | 66 | 11/5/2026 |
| 11.19.3-rc.2 | 80 | 10/31/2026 |
| 11.17.3-rc.2 | 90 | 10/26/2026 |