Azure Dart Package

The azure Dart package is a monorepo consisting of multiple sub-packages, starting with the open_ai and ai packages. The open_ai package provides an easy-to-use interface to interact with Azure OpenAI services, including features such as completions, embeddings, chat completions, image generation, transcriptions, and translations. The ai package is a placeholder for future Azure AI services.

Table of Contents

• Features
• Installation
• Usage
  • Initialization
  • Completions
  • Embeddings
  • Chat Completions
  • Image Generation
  • Audio Transcriptions
  • Audio Translations
• Project Structure
• Future-proofing

Features

• Completions: Generate text based on provided prompts.
• Embeddings: Get vector representations of given inputs for use in machine learning models.
• Chat Completions: Create conversational AI responses.
• Image Generation: Generate images based on text descriptions using DALL-E.
• Audio Transcriptions: Transcribe audio into text.
• Audio Translations: Translate audio into English.

Installation

Add the package to your pubspec.yaml:

dependencies:
  azure: ^1.0.0

Then, run:

dart pub get

Usage

Initialization

To start using the package, you need to initialize the Azure class by providing your Azure endpoint and API key.

import 'package:azure/azure.dart';

void main() {
  final azure = Azure(
    endpoint: 'https://YOUR_RESOURCE_NAME.azure.com',
    apiKey: 'YOUR_API_KEY',
  );
}

Completions

Generate completions based on a prompt.

final completions = await azure.openAI.completions.createCompletion(
  deploymentId: 'YOUR_DEPLOYMENT_ID',
  prompt: 'Tell me a joke about cats.',
  maxTokens: 50,
  temperature: 0.8,
);
print(completions);

Embeddings

Create embeddings for given inputs.

final embeddings = await azure.openAI.embeddings.createEmbedding(
  deploymentId: 'YOUR_DEPLOYMENT_ID',
  inputs: ['This is a test sentence.'],
);
print(embeddings);

Chat Completions

Generate chat completions for a conversation.

final chatResponse = await azure.openAI.chat.createChatCompletion(
  deploymentId: 'YOUR_DEPLOYMENT_ID',
  messages: [
    {'role': 'user', 'content': 'What is the weather like today?'}
  ],
);
print(chatResponse);

Image Generation

Generate images based on a prompt using DALL-E.

final images = await azure.openAI.images.generateImage(
  deploymentId: 'YOUR_DEPLOYMENT_ID',
  prompt: 'A futuristic city skyline at sunset.',
);
print(images);

Audio Transcriptions

Transcribe an audio file.

final transcription = await azure.openAI.transcriptions.transcribe(
  deploymentId: 'YOUR_DEPLOYMENT_ID',
  filePath: 'path/to/audio/file.wav',
);
print(transcription);

Audio Translations

Translate audio to English.

final translation = await azure.openAI.translations.translate(
  deploymentId: 'YOUR_DEPLOYMENT_ID',
  filePath: 'path/to/audio/file.wav',
);
print(translation);

Project Structure

The package is organized into separate files for handling different functionalities:

lib/
│   azure.dart                // Main entry point
│   api_client.dart           // Core HTTP client for making requests
│
├── open_ai/                  // OpenAI sub-package for Azure services
│   ├── endpoints/            // API endpoints for different Azure OpenAI services
│   │       completions.dart       // API for Completions
│   │       embeddings.dart        // API for Embeddings
│   │       chat_completions.dart  // API for Chat Completions
│   │       images.dart            // API for Image Generation
│   │       audio_transcriptions.dart // API for Transcriptions
│   │       audio_translations.dart   // API for Translations
│   └── models/               // Models for request/response data
│           completions_model.dart
│           embeddings_model.dart
│           chat_model.dart
│           images_model.dart
│           audio_model.dart
│
├── ai/                       // Placeholder for future Azure AI services

Future-proofing

The azure package is designed to support future changes in Azure REST API versions. The core HTTP client (api_client.dart) allows specifying the API version, making it easy to adapt to future updates by modifying the api-version parameter.

If Azure releases a new API version, you can update the apiVersion parameter accordingly without changing the core logic of your code.

Contributing

Contributions are welcome! Please submit issues and pull requests to help improve the package.

License

This package is available under the MIT License.