Table of Contents

Class Chat

Namespace
OllamaSharp
Assembly
OllamaSharp.dll

A chat helper that handles the chat logic internally and automatically extends the message history.

A simple interactive chat can be implemented in just a handful of lines: 
var ollama = new OllamaApiClient("http://localhost:11434", "llama3.2-vision:latest");
var chat = new Chat(ollama);
// ...
while (true)
{
	Console.Write("You: ");
	var message = Console.ReadLine()!;
	Console.Write("Ollama: ");
	await foreach (var answerToken in chat.SendAsync(message))
		Console.Write(answerToken);
	// ...
	Console.WriteLine();
}
// ...
// Output:
// You: Write a haiku about AI models
// Ollama: Code whispers secrets
//   Intelligent designs unfold
//   Minds beyond our own
 
public class Chat
Inheritance
Chat
Inherited Members

Constructors

Chat(IOllamaApiClient)

Initializes a new instance of the Chat class. This basic constructor sets up the chat without a predefined system prompt.

public Chat(IOllamaApiClient client)

Parameters

client IOllamaApiClient

An implementation of the IOllamaApiClient interface, used for managing communication with the chat backend.

Examples

Setting up a chat instance without a system prompt:

var client = new OllamaApiClient("http://localhost:11434", "llama3.2-vision:latest");
var chat = new Chat(client);

// Sending a message to the chat
chat.SendMessage("Hello, how are you?");

Exceptions

ArgumentNullException

Thrown when the client parameter is null.

Chat(IOllamaApiClient, string)

Initializes a new instance of the Chat class with a custom system prompt. This constructor allows you to define the assistant's initial behavior or personality using a system prompt.

public Chat(IOllamaApiClient client, string systemPrompt)

Parameters

client IOllamaApiClient

An implementation of the IOllamaApiClient interface, used for managing communication with the chat backend.

systemPrompt string

A string representing the system prompt that defines the behavior and context for the chat assistant. For example, you can set the assistant to be helpful, humorous, or focused on a specific domain.

Examples

Creating a chat instance with a custom system prompt:

var client = new OllamaApiClient("http://localhost:11434", "llama3.2-vision:latest");
var systemPrompt = "You are an expert assistant specializing in data science.";
var chat = new Chat(client, systemPrompt);

// Sending a message to the chat
chat.SendMessage("Can you explain neural networks?");

Exceptions

ArgumentNullException

Thrown when the client parameter is null.

Properties

AllowRecursiveToolCalls

Allow recursive tool calls when a model decides to call different tools after each other.

public bool AllowRecursiveToolCalls { get; set; }

Property Value

bool

Client

Gets the Ollama API client

public IOllamaApiClient Client { get; }

Property Value

IOllamaApiClient

Messages

Gets or sets the messages of the chat history

public List<Message> Messages { get; set; }

Property Value

List<Message>

Model

Gets or sets the AI model to chat with

public string Model { get; set; }

Property Value

string

Options

Gets or sets the RequestOptions to chat with

public RequestOptions? Options { get; set; }

Property Value

RequestOptions

Think

Gets or sets a value to enable or disable thinking. Use reasoning models like openthinker, qwen3, deepseek-r1, phi4-reasoning that support thinking when activating this option. This might cause errors with non-reasoning models, see https://github.com/awaescher/OllamaSharp/releases/tag/5.2.0 More information: https://github.com/ollama/ollama/releases/tag/v0.9.0

public ThinkValue? Think { get; set; }

Property Value

ThinkValue?

ToolInvoker

Gets or sets the class instance that invokes provided tools requested by the AI model

public IToolInvoker ToolInvoker { get; set; }

Property Value

IToolInvoker

Methods

SendAsAsync(ChatRole, string, IEnumerable<IEnumerable<byte>>?, CancellationToken)

Sends a message in a given role to the currently selected model and streams its response asynchronously.

public IAsyncEnumerable<string> SendAsAsync(ChatRole role, string message, IEnumerable<IEnumerable<byte>>? imagesAsBytes, CancellationToken cancellationToken = default)

Parameters

role ChatRole

The role in which the message should be sent. Refer to ChatRole for supported roles.

message string

The message to send to the model.

imagesAsBytes IEnumerable<IEnumerable<byte>>

Optional images represented as byte arrays to include in the request. This parameter can be null.

cancellationToken CancellationToken

A cancellation token to observe while waiting for the response. By default, this parameter is set to None.

Returns

IAsyncEnumerable<string>

An IAsyncEnumerable<T> of strings representing the streamed response generated by the model.

Examples

Sending a user message with optional images:

var client = new OllamaApiClient("http://localhost:11434", "llama3.2-vision:latest");
var chat = new Chat(client);
var role = new ChatRole("user");
var message = "What's the weather like today?";
var images = new List<IEnumerable<byte>> { File.ReadAllBytes("exampleImage.jpg") };
await foreach (var response in chat.SendAsAsync(role, message, images, CancellationToken.None))
{
  Console.WriteLine(response);
}

SendAsAsync(ChatRole, string, IEnumerable<object>?, IEnumerable<string>?, object?, CancellationToken)

Sends a message as a specified role to the current model and streams back its response as an asynchronous enumerable.

public IAsyncEnumerable<string> SendAsAsync(ChatRole role, string message, IEnumerable<object>? tools, IEnumerable<string>? imagesAsBase64 = null, object? format = null, CancellationToken cancellationToken = default)

Parameters

role ChatRole

The role in which the message should be sent. This determines the context or perspective of the message.

message string

The message that needs to be sent to the chat model.

tools IEnumerable<object>

A collection of tools available for the model to utilize. Tools can alter the behavior of the model, such as turning off response streaming automatically when used.

imagesAsBase64 IEnumerable<string>

An optional collection of images encoded in Base64 format, which are sent along with the message to the model.

format object

Defines the response format. Acceptable values include "json" or a schema object created with JsonSerializerOptions.Default.GetJsonSchemaAsNode.

cancellationToken CancellationToken

A token to cancel the ongoing operation if required.

Returns

IAsyncEnumerable<string>

An asynchronous enumerable of response strings streamed from the model.

Examples

Using the SendAsAsync(ChatRole, string, IEnumerable<object>?, IEnumerable<string>?, object?, CancellationToken) method to send a message and stream the model's response:

var chat = new Chat(client);
var role = new ChatRole("assistant");
var tools = new List<Tool>();
var images = new List<string> { "base64EncodedImageData" };
await foreach (var response in chat.SendAsAsync(role, "Generate a summary for the attached image", tools, images))
{
  Console.WriteLine($"Received response: {response}");
}

Exceptions

NotSupportedException

Thrown if the format argument is of type CancellationToken by mistake, or if any unsupported types are passed.

SendAsAsync(ChatRole, string, IEnumerable<string>?, CancellationToken)

Sends a message with a specified role to the current model and streams the response as an asynchronous sequence of strings.

public IAsyncEnumerable<string> SendAsAsync(ChatRole role, string message, IEnumerable<string>? imagesAsBase64, CancellationToken cancellationToken = default)

Parameters

role ChatRole

The role from which the message originates, such as "User" or "Assistant".

message string

The message to send to the model.

imagesAsBase64 IEnumerable<string>

Optional collection of images, encoded in Base64 format, to include with the message.

cancellationToken CancellationToken

A token that can be used to cancel the operation.

Returns

IAsyncEnumerable<string>

An asynchronous sequence of strings representing the streamed response from the model.

Examples

var client = new OllamaApiClient("http://localhost:11434", "llama3.2-vision:latest");
var chat = new Chat(client)
{
  Model = "llama3.2-vision:latest"
};
// Sending a message as a user role and processing the response
await foreach (var response in chat.SendAsAsync(ChatRole.User, "Describe the image", null))
{
  Console.WriteLine(response);
}

SendAsAsync(ChatRole, string, CancellationToken)

Sends a message in a given role to the currently selected model and streams its response.

public IAsyncEnumerable<string> SendAsAsync(ChatRole role, string message, CancellationToken cancellationToken = default)

Parameters

role ChatRole

The role in which the message should be sent, represented by a ChatRole.

message string

The message to be sent as a string.

cancellationToken CancellationToken

An optional CancellationToken to observe while waiting for the response.

Returns

IAsyncEnumerable<string>

An IAsyncEnumerable<T> of strings representing the streamed response from the server.

Examples

Example usage of the SendAsAsync(ChatRole, string, CancellationToken) method:

var client = new OllamaApiClient("http://localhost:11434", "llama3.2-vision:latest");
var chat = new Chat(client);
var role = new ChatRole("assistant");
var responseStream = chat.SendAsAsync(role, "How can I assist you today?");
await foreach (var response in responseStream)
{
	Console.WriteLine(response); // Streams and prints the response from the server
}

SendAsync(string, IEnumerable<IEnumerable<byte>>?, CancellationToken)

Sends a message to the currently selected model and streams its response

public IAsyncEnumerable<string> SendAsync(string message, IEnumerable<IEnumerable<byte>>? imagesAsBytes, CancellationToken cancellationToken = default)

Parameters

message string

The message to send

imagesAsBytes IEnumerable<IEnumerable<byte>>

Images in byte representation to send to the model

cancellationToken CancellationToken

The token to cancel the operation with

Returns

IAsyncEnumerable<string>

An IAsyncEnumerable<T> that streams the response.

Examples

Getting a response from the model with an image:

var client = new HttpClient();
var cat = await client.GetByteArrayAsync("https://cataas.com/cat");
var ollama = new OllamaApiClient("http://localhost:11434", "llama3.2-vision:latest");
var chat = new Chat(ollama);
var response = chat.SendAsync("What do you see?", [cat]);
await foreach (var answerToken in response) Console.Write(answerToken);

// Output: The image shows a white kitten with black markings on its
//         head and tail, sitting next to an orange tabby cat. The kitten
//         is looking at the camera while the tabby cat appears to be
//         sleeping or resting with its eyes closed. The two cats are
//         lying in a blanket that has been rumpled up.

SendAsync(string, IEnumerable<object>?, IEnumerable<string>?, object?, CancellationToken)

Sends a message to the currently selected model and streams its response. Allows for optional tools, images, or response formatting to customize the interaction.

public IAsyncEnumerable<string> SendAsync(string message, IEnumerable<object>? tools, IEnumerable<string>? imagesAsBase64 = null, object? format = null, CancellationToken cancellationToken = default)

Parameters

message string

The message to send to the chat model as a string.

tools IEnumerable<object>

A collection of Tool instances that the model can utilize. Enabling tools automatically disables response streaming. For more information, see the tools documentation: Tool Support.

imagesAsBase64 IEnumerable<string>

An optional collection of images encoded as Base64 strings to pass into the model.

format object

Specifies the response format. Can be set to "json" or an object created with JsonSerializerOptions.Default.GetJsonSchemaAsNode.

cancellationToken CancellationToken

A CancellationToken to observe while waiting for the operation to complete.

Returns

IAsyncEnumerable<string>

An asynchronous enumerable stream of string responses from the model.

Examples

Example usage of SendAsync(string, IEnumerable<object>?, IEnumerable<string>?, object?, CancellationToken):

var client = new OllamaApiClient("http://localhost:11434", "llama3.2-vision:latest");
var chat = new Chat(client);
var tools = new List<Tool> { new Tool() }; // Example tools
var images = new List<string> { ConvertImageToBase64("path-to-image.jpg") };
await foreach (var response in chat.SendAsync(
  "Tell me about recent advancements in AI.",
  tools: tools,
  imagesAsBase64: images,
  format: "json",
  cancellationToken: CancellationToken.None))
{
  Console.WriteLine(response);
}

SendAsync(string, IEnumerable<string>?, CancellationToken)

Sends a message to the currently selected model and streams its response

public IAsyncEnumerable<string> SendAsync(string message, IEnumerable<string>? imagesAsBase64, CancellationToken cancellationToken = default)

Parameters

message string

The message to send

imagesAsBase64 IEnumerable<string>

Base64 encoded images to send to the model

cancellationToken CancellationToken

The token to cancel the operation with

Returns

IAsyncEnumerable<string>

An IAsyncEnumerable<T> that streams the response.

Examples

Getting a response from the model with an image:

var client = new HttpClient();
var cat = await client.GetByteArrayAsync("https://cataas.com/cat");
var base64Cat = Convert.ToBase64String(cat);
var ollama = new OllamaApiClient("http://localhost:11434", "llama3.2-vision:latest");
var chat = new Chat(ollama);
var response = chat.SendAsync("What do you see?", [base64Cat]);
await foreach (var answerToken in response) Console.Write(answerToken);

// Output:
// The image shows a cat lying on the floor next to an iPad. The cat is looking
// at the screen, which displays a game with fish and other sea creatures. The
// cat's paw is touching the screen, as if it is playing the game. The background
// of the image is a wooden floor.

SendAsync(string, CancellationToken)

Sends a message to the currently selected model and streams its response

public IAsyncEnumerable<string> SendAsync(string message, CancellationToken cancellationToken = default)

Parameters

message string

The message to send

cancellationToken CancellationToken

The token to cancel the operation with

Returns

IAsyncEnumerable<string>

An IAsyncEnumerable<T> that streams the response.

Examples

Getting a response from the model:

var response = await chat.SendAsync("Write a haiku about AI models");
await foreach (var answerToken in response)
	 Console.WriteLine(answerToken);

Events

OnThink

Event that gets fired for each token that the AI model is thinking. This will just work for models that support thinking according to their Ollama manifest and if Think is set to true. If Think is set to null, think tokens will be written to the default model output. If Think is false, think tokens will not be emitted.

public event EventHandler<string>? OnThink

Event Type

EventHandler<string>

OnToolCall

Gets fired when the AI model wants to invoke a tool.

public event EventHandler<Message.ToolCall>? OnToolCall

Event Type

EventHandler<Message.ToolCall>

OnToolResult

Gets fired after a tool was invoked and the result is available.

public event EventHandler<ToolResult>? OnToolResult

Event Type

EventHandler<ToolResult>
Edit this page