Universal.LMStudio.Client 2.0.0

Universal.LMStudio.Client

.NET client for LM Studio's native and compatibility v1 APIs.

Start the LM Studio server and load a model first. All clients default to http://localhost:1234.

Native API

Use the native API for stateful chats, LM Studio MCP integrations, and model lifecycle operations.

using Universal.LMStudio.Client.Api.V1;
using Universal.LMStudio.Client.Api.V1.Chat;

using var client = new LMStudioClient();

ChatResponse response = await client.Chat.CreateAsync(new ChatRequest
{
    Model = "qwen3.5-2b",
    Input = "Reply with one short sentence.",
    Reasoning = "off",
    Store = false,
});

foreach (ChatOutputItem item in response.Output)
    Console.WriteLine(item.Content);

Continue a stored native chat with its response ID:

ChatResponse first = await client.Chat.CreateAsync(new ChatRequest
{
    Model = "qwen3.5-2b",
    Input = "Remember the word cobalt.",
    Store = true,
});

ChatResponse next = await client.Chat.CreateAsync(new ChatRequest
{
    Model = "qwen3.5-2b",
    Input = "What word did I ask you to remember?",
    PreviousResponseId = first.ResponseId,
    Store = true,
});

OpenAI-compatible chat

The compatibility types mirror Universal.OpenAI.Client naming where LM Studio supports the corresponding endpoint.

using Universal.LMStudio.Client.V1.Chat;

using var chat = new ChatClient("lm-studio");

CreateCompletionResponse response = await chat.CreateCompletionAsync(
    new CreateCompletionRequest
    {
        Model = "qwen3.5-2b",
        Messages = [new UserMessage("Hello from LM Studio")],
        Temperature = 0.2f,
    });

Console.WriteLine(response.Choices[0].Message.Content);

Function tools

LM Studio returns the requested tool call; your application executes it and sends back the result.

using Universal.Common.Json;
using Universal.LMStudio.Client.V1.Chat;

var weather = new Tool
{
    Type = ToolTypes.Function,
    Function = new Function
    {
        Name = "get_weather",
        Description = "Get the weather for a city.",
        Parameters = new JsonSchema
        {
            Type = "object",
            Properties = new Dictionary<string, JsonSchema>
            {
                ["city"] = new JsonSchema { Type = "string" },
            },
            Required = ["city"],
        },
    },
};

using var chat = new ChatClient("lm-studio");
CreateCompletionResponse response = await chat.CreateCompletionAsync(
    new CreateCompletionRequest
    {
        Model = "qwen3.5-2b",
        Messages = [new UserMessage("What is the weather in Sydney?")],
        Tools = [weather],
        ToolChoice = "required",
    });

AssistantMessage assistant = (AssistantMessage)response.Choices[0].Message;
ToolCall call = assistant.ToolCalls[0];
Console.WriteLine($"Call {call.Function.Name}: {call.Function.Arguments}");

// Execute the function, then continue the conversation.
CreateCompletionResponse final = await chat.CreateCompletionAsync(
    new CreateCompletionRequest
    {
        Model = "qwen3.5-2b",
        Messages =
        [
            new UserMessage("What is the weather in Sydney?"),
            assistant,
            new ToolMessage("{\"temperature_c\":22}", call.Id),
        ],
        Tools = [weather],
    });

Responses API

using Universal.LMStudio.Client.V1.Responses;

using var responses = new ResponsesClient("lm-studio");

CreateResponseResponse response = await responses.CreateResponseAsync(
    new CreateResponseRequest
    {
        Model = "qwen3.5-2b",
        Input = "Give me a prime number below 20.",
        Reasoning = new ReasoningOptions { Effort = "low" },
    });

Console.WriteLine(response.OutputText);

Continue a stateful Responses conversation with PreviousResponseId:

CreateResponseResponse next = await responses.CreateResponseAsync(
    new CreateResponseRequest
    {
        Model = "qwen3.5-2b",
        Input = "Multiply it by two.",
        PreviousResponseId = response.Id,
    });

Anthropic-compatible Messages

using Universal.LMStudio.Client.V1.Messages;

using var messages = new AnthropicClient("lm-studio");

MessageResponse response = await messages.CreateMessageAsync(new MessageRequest
{
    Model = "qwen3.5-2b",
    MaxTokens = 256,
    System = "Answer concisely.",
    Messages = [new Message(Roles.User, "Say hello from LM Studio.")],
});

foreach (TextContentBlock block in response.Content.OfType<TextContentBlock>())
    Console.WriteLine(block.Text);

Streaming

Chat Completions, Responses, Messages, and the native chat API support streaming.

using Universal.LMStudio.Client.V1.Chat;

using var chat = new ChatClient("lm-studio");
var request = new CreateCompletionRequest
{
    Model = "qwen3.5-2b",
    Messages = [new UserMessage("Write one sentence.")],
    Stream = true,
};

await foreach (CompletionChunk chunk in chat.CreateCompletionStreamingAsync(request))
{
    foreach (CompletionChunkChoice choice in chunk.Choices)
        Console.Write(choice.Delta?.Content);
}

Embeddings and models

using Universal.LMStudio.Client.V1;

using var client = new V1Client("lm-studio");

var models = await client.Models.ListModelsAsync();
foreach (var model in models.Data)
    Console.WriteLine(model.Id);

var embeddings = await client.Embeddings.CreateEmbeddingsAsync(new()
{
    Model = "text-embedding-nomic-embed-text-v1.5",
    Input = "Embed this text.",
});

float[] vector = embeddings.Data[0].Embedding;

V1Client aggregates every /v1/* endpoint through Chat, Responses, Embeddings, Completions, Models, and Messages.

Custom endpoint and authentication

Pass LMStudioClientOptions when LM Studio is remote, mounted below a path, requires authentication, or needs a custom timeout.

using Universal.LMStudio.Client.Api.V1;
using Universal.LMStudio.Client.V1;

var options = new LMStudioClientOptions
{
    Endpoint = new Uri("http://lmstudio.example:1234/"),
    ApiToken = Environment.GetEnvironmentVariable("LM_STUDIO_API_TOKEN"),
    Timeout = TimeSpan.FromMinutes(5),
};

using var client = new V1Client(options);

Namespace map

• /api/v1/* — Universal.LMStudio.Client.Api.V1.*
• /v1/chat/completions — Universal.LMStudio.Client.V1.Chat
• /v1/responses — Universal.LMStudio.Client.V1.Responses
• /v1/embeddings — Universal.LMStudio.Client.V1.Embeddings
• /v1/completions — Universal.LMStudio.Client.V1.Completions
• /v1/models — Universal.LMStudio.Client.V1.Models
• /v1/messages — Universal.LMStudio.Client.V1.Messages