OllamaClientLibrary 1.3.0

OllamaClientLibrary

OllamaClientLibrary is a .NET Standard 2.1 library designed to interact seamlessly with the Ollama API. It offers robust functionality for generating text and chat completions using a variety of models. This library simplifies the process of configuring and utilizing the Ollama API, making it easier for developers to integrate advanced text generation capabilities into their applications.

Features

• Predefined configuration for the local Ollama setup, such as host, model, temperature, timeout, prompt size, etc.
• Auto-install a model if it is not available on your local machine.
• Get JSON completion with an automatically recognized JSON schema of the response DTO, so you no longer need to specify it in the prompt.
• Get chat completion with streaming. The library provides access to the conversation history, allowing you to store it in the database if needed.
• Get text completion. Provide a prompt and get a simple text response.
• Get completion with tools. Based on the Ollama response, the library can dynamically call local methods and provide the necessary parameters.
• Get embeddings completions for a given text. The library provides functionality to convert text into numerical vectors (embeddings) that can be used for various machine learning tasks such as similarity search, clustering, and classification. This is useful for applications like semantic search, recommendation systems, and natural language understanding.
• List available local and remote models with filtering options. Now you have access to see all models installed on your local machine, as well as all models available on Ollama's library
• Pull a model from Ollama's library to your local machine.
• Delete a model from the local machine

Prerequisites:

Setting Up Ollama Server

Download and install Ollama from https://ollama.com/download

Installing a Model

1. Execute ollama run qwen2.5:1.5b in your terminal to start the Ollama server and install the necessary models. You can find a list of available models on Ollama's library.
2. Confirm the installation by running ollama list to see the models installed on your local machine.

Installing a Model (Alternative)

using var client = new OllamaClient(new OllamaOptions()
{
    AutoInstallModel = true, // Default is false
});

Installation

You can install the package via NuGet:

Install-Package OllamaClientLibrary

Usage

Generate JSON Completion sample

using OllamaClientLibrary;
using OllamaClientLibrary.Abstractions;
using OllamaClientLibrary.Constants;

using System.ComponentModel;

// Setup OllamaClient
using IOllamaClient client = new OllamaClient(new OllamaOptions() // If no options are provided, OllamaOptions will be used with the default settings
{
    Host = "http://localhost:11434", // Default host is http://localhost:11434
    Model = "qwen2.5:1.5b", // Default model is "qwen2.5:1.5b"
    Temperature = Temperature.DataCleaningOrAnalysis, // Default temperature is Temperature.GeneralConversationOrTranslation
    AutoInstallModel = true, // Default is false. The library will automatically install the model if it is not available on your local machine
    Timeout = TimeSpan.FromSeconds(30), // Default is 60 seconds.
    MaxPromptTokenSize = 4096, // Default is 4096 tokens. Increase this value if you want to send larger prompts
    AssistantBehavior = "You are a professional .NET developer.", // Optional. Default is "You are a world class AI Assistant"
    ApiKey = "your-api-key", // Optional. It is not required by default for the local setup
    Tools = null // Optional. You can use the ToolFactory to create tools, e.g. ToolFactory.Create<WeatherService>() where WeatherService is your class
});

// Call Ollama API
var response = await client.GetJsonCompletionAsync<Response>(
    "Return a list of all available .NET Core versions from the past five years."
);

// Display results
if (response?.Data != null)
{
    foreach (var item in response.Data.OrderBy(s => s.ReleaseDate))
    {
        Console.WriteLine($"Version: {item.Version}, Release Date: {item.ReleaseDate}, End of Life: {item.EndOfLife}");
    }
}

// Configure the response DTO
class Response
{
    public List<DotNetCore>? Data { get; set; } // A wrapper class is required for the response if the data is an array.
}

class DotNetCore
{
    [Description("Version number in the format of Major.Minor.Patch")]
    public string? Version { get; set; }

    [Description("Release date of the version")]
    public DateTime? ReleaseDate { get; set; }

    [Description("End of life date of the version")]
    public DateTime? EndOfLife { get; set; }
}

More samples

• Get Chat Completion
• Get JSON Completion
• Get JSON Completion with Tools
• Get Text Completion
• Get Text Completion with Tools
• Get Embedding Completion
• List Local Models
• List Remote Models
• Pull Model
• Delete Model

Changelog

• v1.3.0: Introduced MaxPromptTokenSize and AssistantBehavior properties to OllamaOptions, and added the DeleteModelAsync method. Removed KeepChatHistory. Fixed handling of multiple ToolCalls. Updated to call AI after receiving a response from the tools. Modified ToolFactory to retrieve all public methods automatically instead of specifying them manually. Added support for async methods in ToolCalls.
• v1.2.0: Renamed methods to be more descriptive, added AutoInstallModel and Timeout properties to OllamaOptions, started using the chat completion API instead of the generate API added the pull model API and added IOllamaClient interface.
• v1.1.0: Changed the default model to qwen2.5:1.5b, fixed parsing of ModifiedAt for the Models list endpoint, added support for Tools, added Chat History, added integration tests, and configured CI.
• v1.0.1: Allowed setting the ApiKey in OllamaOptions.
• v1.0.0: Initial release with basic functionality for text and chat completions.

License

This project is licensed under the MIT License.

Repository

For more information, visit the GitHub repository.

Author

Oleksandr Kushnir