Ollama Embeddings

Integrating Ollama embedding modes with Build5Nines.SharpVector enhances the semantic search capabilities of your .NET applications. By leveraging models like nomic-embed-text or others, you can generate higher quality vector representations of text, leading to more accurate and contextually relevant search results.

Why Use an Ollama Embedding Model?

While SharpVector includes basic embedding generation, utilizing an Ollama embedding model offers significant advantages:

• Improved Search Accuracy: Embedding models capture the semantic meaning of text more accurately, resulting in more relevant search outcomes.
• Pre-trained on Extensive Data: These models are trained on vast datasets, enhancing their robustness and generalization capabilities.
• Optimized for Performance: Designed for efficient retrieval and indexing, Ollama embedding models facilitate faster search operations.

Getting Started

To integrate an Ollama embedding model with SharpVector, install the Build5Nines.SharpVector.Ollama NuGet package:

.NET CLIPackage Manager

dotnet add package Build5Nines.SharpVector.Ollama

Nuget\Install-Package Build5Nines.SharpVector.Ollama

This package includes the core Build5Nines.SharpVector library and dependencies required to connect to Ollama's embedding API.

Initialize the Vector Database using Ollama

With the Ollama embedding model running, initialize the the SharpVector database:

using Build5Nines.SharpVector.Ollama;

var modelName = "nomic-embed-text";

// For connecting to Locally running ('localhost') Ollama
var vectorDatabase = new BasicOllamaMemoryVectorDatabase(modelName)

// For connecting to a different Ollama endpoint URL
var ollamaEndpoint = "http:/localhost:11434/api/embeddings";
var vactorDatabase = new BasicOllamaMemoryVectorDatabase(ollamaEndpoint, modelName);

Adding Text Data

To add text documents to the vector database:

// sync
vectorDatabase.AddText(documentText, metadataText);

// async
await vectorDatabase.AddTextAsync(documentText, metadataText);

• documentText: The textual content to be vectorized.
• metadataText: Associated metadata (e.g., document title, JSON string) stored alongside the vectorized text.

Note

Metadata is not vectorized but is retrieved with search results, providing context.

Performing Similarity Search

The SearchAsync method returns documents whose vector representations closely match the query vector, based on similarity metrics like cosine similarity.

var query = "your search query";
var results = await vectorDatabase.SearchAsync(query);

The .SearchAsync method supports additional arguments to help with searching the vector database:

var results = await vectorDatabase.SearchAsync(queryText,
  threshold: 0.001f // 0.2f - Cosine Similarity
  pageIndex: 0, // page index of search results (default: 0)
  pageCount: 10 // Number of results per page to return (default: no limit)
);

• queryText: The text query to search within the vector database.
• threshold: The similarity threshold to use for searching the vector database using Cosine Similarity method.
• pageIndex: The page index of search results to return. Default is 0.
• pageCount: The number of search results to return per page. Default is "no limit" (aka return all results)

Summary

Integrating an Ollama embedding model with Build5Nines.SharpVector empowers your .NET applications with advanced semantic search capabilities. By leveraging high-quality vector representations, you can achieve more accurate and context-aware search results, enhancing the overall user experience.