ollama basic consuming

Since you already have bge-m3:latest downloaded in Ollama, you can consume it via the REST API or official client libraries. Since your file path (~/Code/DotNest-Core-SDK/...) suggests you are working in a .NET/C# environment, I will provide examples for both CLI/Shell and C#.

1. Test via CLI (cURL)

The quickest way to verify the model is working is by sending a POST request to Ollama's local endpoint.

curl http://localhost:11434/api/embed -d '{
  "model": "bge-m3",
  "input": "Llamas are members of the camelid family"
}'

• Response: You will receive a JSON object containing a large array of numbers (the vector).
• Note: bge-m3 is a high-performance model that supports a sequence length of up to 8192 tokens, making it great for long documents.

2. Consume in .NET (C#)

Since you are working on the DotNest-Core-SDK, you'll likely want to integrate this programmatically. The most popular library for this is OllamaSharp.

Install the package:

dotnet add package OllamaSharp

Basic Usage:

using OllamaSharp;

// Initialize the client
var uri = new Uri("http://localhost:11434");
var ollama = new OllamaApiClient(uri);

// Set the model you just listed
ollama.SelectedModel = "bge-m3";

// Generate an embedding for a string
var request = new EmbedRequest 
{ 
    Model = "bge-m3", 
    Input = "How do I use embeddings in my SDK?" 
};

var response = await ollama.Embed(request);

// Access the vector
float[] vector = response.Embeddings[0]; 
Console.WriteLine($"Generated vector with {vector.Length} dimensions.");

3. Usage via Python (Alternative)

If you are doing data science or preprocessing scripts outside of your SDK:

Install the library:

pip install ollama

Script:

import ollama

response = ollama.embed(
    model='bge-m3',
    input='The sky is blue because of Rayleigh scattering.',
)

print(response['embeddings'])

Key Technical Details for bge-m3

• Dimensions: This model typically outputs vectors with 1024 dimensions.
• Multilingual: bge-m3 stands for Multi-functionality, Multi-linguality, and Multi-granularity. It works exceptionally well for non-English languages.
• Context Window: Unlike standard models (like all-MiniLM-L6-v2 which caps at 256 or 512 tokens), bge-m3 handles up to 8k tokens, allowing you to embed entire chapters or long technical docs at once.

Are you planning to store these embeddings in a specific vector database (like ChromaDB, Qdrant, or pgvector), or are you performing simple in-memory similarity comparisons?