ollama_easy 0.2.0

A Flutter-friendly Ollama client with one-line prompts, typed chat, streaming, embeddings, and model helpers.

ollama_easy

A small, Flutter-friendly client for Ollama. Use local AI models in your apps with one-line prompts, typed chat, streaming, and built-in support for Android emulators.

final ollama = Ollama.local(defaultModel: 'qwen2.5:1.5b');

// Simple one-line prompt
final answer = await ollama.ask('Explain Flutter isolates in one sentence.');

// Real-time token streaming
await for (final token in ollama.askStream('Write a tiny haiku.')) {
  stdout.write(token);
}

ollama.close();

Features

• Multi-Platform: Support for Windows, Web, Android, iOS, macOS, and Linux.
• Easy Streaming: Built-in support for Stream<String> tokens.
• Android Emulator Support: Easily switch to 10.0.2.2 for emulator testing.
• Typed Chat: Structured messages with system, user, and assistant roles.
• JSON Mode: Get structured output easily with askJson().
• Model Helpers: Pull, check, and list models with progress streams.
• CLI Tool: Run prompts directly from your terminal.

Install

dependencies:
  ollama_easy: ^0.2.0

Getting Started

Ensure the Ollama server is running on your machine before executing your code.

Local Ollama

final ollama = Ollama.local(defaultModel: 'qwen2.5:1.5b');

if (!await ollama.isRunning()) {
  throw StateError('Start Ollama first.');
}

Android Emulator

When running in an Android emulator, localhost refers to the emulator itself. To reach the host machine where Ollama is running, use:

final ollama = Ollama(baseUrl: Uri.parse('http://10.0.2.2:11434'));

Usage

Chat

final reply = await ollama.chatText([
  const OllamaMessage.system('You answer like a helpful Flutter mentor.'),
  const OllamaMessage.user('How should I stream text into a Text widget?'),
]);

Structured JSON

final json = await ollama.askJson(
  'Return a JSON object with "title" and "tags" for a note about Dart streams.',
);

print(json['title']);

Embeddings

final result = await ollama.embed(
  'Flutter makes beautiful apps.',
  model: 'qwen2.5:1.5b',
);

final vector = result.first;

CLI Tool

You can use ollama_easy directly from your terminal to test prompts:

dart run bin/ollama_easy.dart "Tell me a joke"

Flutter Web

Flutter Web requires Ollama to have CORS configured. Start Ollama with the following environment variable:

OLLAMA_ORIGINS="*" ollama serve

API Summary

final ollama = Ollama();

await ollama.ask('...');
ollama.askStream('...');

await ollama.generate('...');
ollama.generateStream('...');

await ollama.chatText([...]);
await ollama.chat([...]);
ollama.chatTextStream([...]);
ollama.chatStream([...]);

await ollama.askJson('...');
await ollama.embed('...');
await ollama.models();
await ollama.hasModel('qwen2.5:1.5b');
await ollama.ensureModel('qwen2.5:1.5b');
await ollama.pull('qwen2.5:1.5b');
ollama.pullStream('qwen2.5:1.5b');
await ollama.version();
await ollama.isRunning();

ollama.close();