Ollama Delphi API Client

Ollama makes it easy to run large language models locally on your own hardware — no cloud dependency, no API costs, and full data privacy. For Delphi developers looking to integrate local AI capabilities into their applications, sgcWebSockets provides TsgcHTTP_API_Ollama — a native component that wraps the entire Ollama API with clean, type-safe Delphi code.

Whether you need to keep sensitive data on-premises, build offline-capable AI features, manage your own model library, or generate embeddings for local vector search, this component gives you direct access to every Ollama feature. No cloud accounts. No recurring API fees. Just drop the component, point it at your Ollama instance, and start building.

Complete API Coverage

Every major feature of the Ollama API is supported out of the box.

Getting Started

Integrate Ollama into your Delphi project in under a minute. Drop the component, configure the host, and send your first message.

// Create the component and configure the host
var
  Ollama: TsgcHTTP_API_Ollama;
  vResponse: string;
begin
  Ollama := TsgcHTTP_API_Ollama.Create(nil);
  Try
    Ollama.OllamaOptions.Host := 'http://localhost:11434';

    // Send a simple message to a local model
    vResponse := Ollama._CreateMessage(
      'llama3', 'Hello, Ollama!');

    ShowMessage(vResponse);
  Finally
    Ollama.Free;
  End;
end;

Chat Completions & Streaming

The Chat Completions API works with any model you have pulled into your Ollama instance. Send text with optional system prompts, and receive responses synchronously or streamed in real-time.

System Prompts

Control model behavior by providing a system prompt that sets the context, personality, or constraints for the conversation.

vResponse := Ollama._CreateMessageWithSystem(
  'llama3',
  'You are a helpful assistant that responds in Spanish.',
  'What is the capital of France?');
// Returns: "La capital de Francia es París."

Real-Time Streaming

For responsive user interfaces, stream the model's response token-by-token using Server-Sent Events.

// Enable streaming via SSE
Ollama.OnHTTPAPISSE := OnSSEEvent;
Ollama._CreateMessageStream('llama3',
  'Write a short poem about Delphi programming.');

procedure TForm1.OnSSEEvent(Sender: TObject;
  const aEvent, aData: string; var Cancel: Boolean);
begin
  // aData: JSON payload with generated content
  Memo1.Lines.Add(aData);
end;

Advanced Typed API

For full control over request parameters — temperature, top-p, stop sequences, max tokens — use the typed request and response classes.

var
  oRequest: TsgcOllamaClass_Request_ChatCompletion;
  oMessage: TsgcOllamaClass_Request_Message;
  oResponse: TsgcOllamaClass_Response_ChatCompletion;
begin
  oRequest := TsgcOllamaClass_Request_ChatCompletion.Create;
  Try
    oRequest.Model := 'llama3';
    oRequest.MaxTokens := 2048;
    oRequest.Temperature := 0.7;

    oMessage := TsgcOllamaClass_Request_Message.Create;
    oMessage.Role := 'user';
    oMessage.Content := 'Explain quantum computing in simple terms.';
    oRequest.Messages.Add(oMessage);

    oResponse := Ollama.CreateMessage(oRequest);
    Try
      if Length(oResponse.Choices) > 0 then
        ShowMessage(oResponse.Choices[0].MessageContent);
    Finally
      oResponse.Free;
    End;
  Finally
    oRequest.Free;
  End;
end;

Model Management

Manage your entire local model library from Delphi code. Pull new models, inspect their details, list available tags, and delete models you no longer need — all programmatically.

Pull a Model

// Download a model from the Ollama registry
Ollama._PullModel('llama3');

Show Model Details

// Get detailed information about a model
vDetails := Ollama._ShowModel('llama3');
ShowMessage(vDetails);

List Models and Tags

// List all models via OpenAI-compatible endpoint
vModels := Ollama._GetModels;

// List model tags with detailed metadata (name, size, digest)
vTags := Ollama._GetTags;

// Typed API: access tag properties directly
var
  oTags: TsgcOllamaClass_Response_Tags;
  i: Integer;
begin
  oTags := Ollama.GetTags;
  Try
    for i := 0 to Length(oTags.Models) - 1 do
      Memo1.Lines.Add(Format('%s (%d bytes)',
        [oTags.Models[i].Name, oTags.Models[i].Size]));
  Finally
    oTags.Free;
  End;
end;

Delete a Model

// Remove a model from the local system
Ollama._DeleteModel('old-model:latest');

Embeddings

Generate vector embeddings locally using any embedding-capable model. Embeddings power semantic search, document clustering, and classification — all without sending data to external servers.

// Generate embeddings locally
var
  vEmbedding: string;
begin
  vEmbedding := Ollama._CreateEmbeddings(
    'nomic-embed-text',
    'Delphi is a powerful programming language.');
  ShowMessage(vEmbedding);
end;

For full control, use the typed API to access the raw embedding values.

var
  oResponse: TsgcOllamaClass_Response_Embeddings;
  i: Integer;
begin
  oResponse := Ollama.CreateEmbeddings(
    'nomic-embed-text',
    'Delphi is a powerful programming language.');
  Try
    for i := 0 to oResponse.EmbeddingCount - 1 do
      Memo1.Lines.Add(FloatToStr(oResponse.GetEmbeddingValue(i)));
  Finally
    oResponse.Free;
  End;
end;

Configuration & Options

Fine-tune the component behavior with comprehensive configuration options.

Supported Models

Ollama supports hundreds of open-source models. Here are some popular choices: