Contoso.AI.TextEmbedder.MiniLML6 0.1.6-beta

Contoso.AI.TextEmbedder.MiniLML6

AI-powered text embedding library using the all-MiniLM-L6-v2 ONNX model from sentence-transformers. Converts text into 384-dimensional embeddings for semantic search, similarity tasks, clustering, and more.

✨ Features

• 384-Dimensional Embeddings - Rich semantic representations for text
• Automatic Model Download - Model downloads automatically at build time
• GPU Acceleration - Leverages DirectML for faster inference when available
• Easy-to-Use API - Simple async factory pattern with CreateAsync()
• Cosine Similarity - Built-in similarity calculation between embeddings
• Semantic Search - Find semantically similar texts
• NuGet Package - Easy integration into any .NET project

📋 Requirements

• Windows 10 SDK 19041 or later
• .NET 8.0 or later
• GPU (optional) - DirectML-compatible GPU for hardware acceleration
• Falls back to CPU execution if GPU is not available

🚀 Quick Start

Installation

dotnet add package Contoso.AI.TextEmbedder.MiniLML6

Or via Package Manager Console:

Install-Package Contoso.AI.TextEmbedder.MiniLML6

Basic Usage

using Contoso.AI;

// Check if the feature is ready
var readyState = TextEmbedderMiniLML6.GetReadyState();

if (readyState != AIFeatureReadyState.Ready)
{
    // Prepare the feature (downloads model if needed)
    var readyResult = await TextEmbedderMiniLML6.EnsureReadyAsync();
    if (readyResult.Status != AIFeatureReadyResultState.Success)
    {
        Console.WriteLine($"Failed to initialize: {readyResult.ExtendedError?.Message}");
        return;
    }
}

// Create embedder instance (returns ITextEmbedder interface)
using ITextEmbedder embedder = await TextEmbedderMiniLML6.CreateAsync();

// Generate embeddings
var texts = new[]
{
    "The quick brown fox jumps over the lazy dog",
    "Machine learning is transforming technology"
};

var embeddings = embedder.GenerateEmbeddings(texts);

// Calculate similarity
var similarity = embeddings[0].CosineSimilarity(embeddings[1]);
Console.WriteLine($"Similarity: {similarity:F3}");

Semantic Search Example

using Contoso.AI;

using ITextEmbedder embedder = await TextEmbedderMiniLML6.CreateAsync();

// Index some documents
var documents = new[]
{
    "Python is a programming language",
    "Cats make great pets",
    "Machine learning uses neural networks",
    "Dogs are loyal companions"
};

var docEmbeddings = embedder.GenerateEmbeddings(documents);

// Search with a query
var query = "Domestic animals";
var queryEmbedding = embedder.GenerateEmbeddings(query)[0];

// Find most similar documents
var results = docEmbeddings
    .Select((emb, idx) => new { 
        Document = documents[idx], 
        Score = queryEmbedding.CosineSimilarity(emb) 
    })
    .OrderByDescending(x => x.Score)
    .ToList();

foreach (var result in results)
{
    Console.WriteLine($"{result.Score:F3} - {result.Document}");
}

📖 API Reference

TextEmbedderMiniLML6 Class

ITextEmbedder Interface

Embedding Class

🏗️ Development

Building from Source

# Clone the repository
git clone https://github.com/contoso/Contoso.AI.git
cd Contoso.AI/TextEmbedder

# Restore and build (model downloads automatically)
dotnet restore
dotnet build

# Run the console test
dotnet run --project Contoso.AI.TextEmbedder.MiniLML6.ConsoleTest

Model Information

This project uses the all-MiniLM-L6-v2 model from sentence-transformers:

• Source: sentence-transformers/all-MiniLM-L6-v2 on Hugging Face
• Commit: d83dd3760b5bfe921f2fe125446b17bf0b7eda8c
• Format: ONNX
• Output Dimensions: 384
• License: Apache 2.0

The model is automatically downloaded during the first build and cached in the obj/Models directory.

🧪 Use Cases

• Semantic Search - Find documents similar to a query
• Text Clustering - Group similar texts together
• Duplicate Detection - Identify similar or duplicate content
• Recommendation Systems - Recommend similar items based on text descriptions
• Question Answering - Find answers similar to questions
• Content Classification - Classify text by similarity to categories

🤝 Contributing

Contributions are welcome! Please feel free to submit a Pull Request.

📄 License

This project is licensed under the MIT License - see the LICENSE file for details.

🙏 Acknowledgments

• sentence-transformers for the all-MiniLM-L6-v2 model
• ONNX Runtime for the inference engine
• Microsoft AI Dev Gallery for embedding generation code samples

Made with ❤️ by Contoso