Rag.NET.Embeddings.Onnx
1.0.0
dotnet add package Rag.NET.Embeddings.Onnx --version 1.0.0
NuGet\Install-Package Rag.NET.Embeddings.Onnx -Version 1.0.0
<PackageReference Include="Rag.NET.Embeddings.Onnx" Version="1.0.0" />
<PackageVersion Include="Rag.NET.Embeddings.Onnx" Version="1.0.0" />
<PackageReference Include="Rag.NET.Embeddings.Onnx" />
paket add Rag.NET.Embeddings.Onnx --version 1.0.0
#r "nuget: Rag.NET.Embeddings.Onnx, 1.0.0"
#:package Rag.NET.Embeddings.Onnx@1.0.0
#addin nuget:?package=Rag.NET.Embeddings.Onnx&version=1.0.0
#tool nuget:?package=Rag.NET.Embeddings.Onnx&version=1.0.0
Rag.NET.Embeddings.Onnx
Local ONNX Runtime embeddings for Rag.NET: sentence embeddings, token-level embeddings (the input late chunking needs), and SPLADE sparse encoding — all on your own hardware, no embedding API calls.
Install
dotnet add package Rag.NET.Embeddings.Onnx
Install alongside the core pipeline package (dotnet add package Rag.NET), which supplies
the AddRagNet(...) builder the generators register into.
Setup
Inside your AddRagNet(...) builder callback, point at an exported ONNX model and its
tokenizer vocabulary:
using Rag.NET.Embeddings.Onnx;
rag.UseOnnxEmbeddings(o =>
{
o.ModelPath = "models/all-MiniLM-L6-v2.onnx";
o.TokenizerVocabPath = "models/vocab.txt";
});
Example
Token-level embeddings feed UseLateChunking from Rag.NET.Chunking, and the SPLADE
encoder produces learned sparse vectors for hybrid search in stores that accept them:
using Rag.NET.Embeddings.Onnx;
rag.UseOnnxTokenEmbeddings(o =>
{
o.ModelPath = "models/jina-embeddings-v2-base-en.onnx";
o.TokenizerVocabPath = "models/vocab.txt";
o.MaxTokens = 8192; // default: long-context token embedding
});
rag.UseSpladeEncoder(o =>
{
o.ModelPath = "models/splade-v3.onnx";
o.TokenizerVocabPath = "models/vocab.txt";
o.TopTerms = 256; // default: sparse terms kept per text
});
Full guide
| Product | Versions Compatible and additional computed target framework versions. |
|---|---|
| .NET | net10.0 is compatible. net10.0-android was computed. net10.0-browser was computed. net10.0-ios was computed. net10.0-maccatalyst was computed. net10.0-macos was computed. net10.0-tvos was computed. net10.0-windows was computed. |
-
net10.0
- Microsoft.Extensions.Logging.Abstractions (>= 10.0.12)
- Microsoft.ML.OnnxRuntime (>= 1.30.0)
- Microsoft.ML.Tokenizers (>= 1.0.3)
- Rag.NET.Abstractions (>= 1.0.0)
NuGet packages
This package is not used by any NuGet packages.
GitHub repositories
This package is not used by any popular GitHub repositories.