LLM Kit Azure

Azure OpenAI provider for LLM Kit - Complete integration with Azure OpenAI Service for chat, completions, embeddings, and image generation.

Note: This provider uses the standardized builder pattern. See the Quick Start section for the recommended usage.

Features

• Text Generation: Generate text using GPT-4, GPT-3.5-turbo, and other Azure OpenAI chat models
• Streaming: Stream responses in real-time for interactive applications
• Tool Calling: Support for function calling with Azure OpenAI models
• Text Embedding: Generate embeddings with text-embedding-ada-002 and other embedding models
• Image Generation: Create images using DALL-E 3 and other image generation models
• Multi-modal: Support for text and images in conversations
• Completion Models: Access to GPT-3.5-turbo-instruct and other completion models
• Azure-specific Authentication: Uses api-key header for Azure authentication
• Flexible URL Formats: Supports both v1 API and deployment-based URLs
• Multiple API Versions: Configure API version for different Azure OpenAI endpoints

Installation

Add this to your Cargo.toml:

[dependencies]
llm-kit-azure = "0.1"
llm-kit-core = "0.1"
llm-kit-provider = "0.1"
tokio = { version = "1", features = ["full"] }

Quick Start

Using the Client Builder (Recommended)

use llm_kit_azure::AzureClient;
use llm_kit_provider::LanguageModel;

#[tokio::main]
async fn main() -> Result<(), Box<dyn std::error::Error>> {
    // Create provider using the client builder
    let provider = AzureClient::new()
        .resource_name("my-azure-resource")
        .api_key("your-api-key")  // Or use AZURE_API_KEY env var
        .build();
    
    // Get a chat model using your deployment name
    let model = provider.chat_model("gpt-4-deployment");
    
    println!("Model: {}", model.model_id());
    println!("Provider: {}", model.provider());
    Ok(())
}

Using Settings Directly (Alternative)

use llm_kit_azure::{AzureOpenAIProvider, AzureOpenAIProviderSettings};
use llm_kit_provider::LanguageModel;

#[tokio::main]
async fn main() -> Result<(), Box<dyn std::error::Error>> {
    // Create provider with settings
    let provider = AzureOpenAIProvider::new(
        AzureOpenAIProviderSettings::new()
            .with_resource_name("my-azure-resource")
            .with_api_key("your-api-key")
    );
    
    let model = provider.chat_model("gpt-4-deployment");
    
    println!("Model: {}", model.model_id());
    Ok(())
}

Configuration

Environment Variables

Set your Azure OpenAI credentials as environment variables:

export AZURE_API_KEY=your-api-key
export AZURE_RESOURCE_NAME=my-azure-resource  # Or use AZURE_BASE_URL

Using the Client Builder

use llm_kit_azure::AzureClient;

// With resource name (most common)
let provider = AzureClient::new()
    .resource_name("my-azure-resource")
    .api_key("your-api-key")
    .api_version("2024-02-15-preview")
    .build();

// With custom base URL
let provider = AzureClient::new()
    .base_url("https://my-resource.openai.azure.com/openai")
    .api_key("your-api-key")
    .build();

// With custom headers
let provider = AzureClient::new()
    .resource_name("my-resource")
    .api_key("your-api-key")
    .header("X-Custom-Header", "value")
    .build();

// With deployment-based URLs (legacy format)
let provider = AzureClient::new()
    .resource_name("my-resource")
    .api_key("your-api-key")
    .use_deployment_based_urls(true)
    .build();

Using Settings Directly

use llm_kit_azure::{AzureOpenAIProvider, AzureOpenAIProviderSettings};

let settings = AzureOpenAIProviderSettings::new()
    .with_resource_name("my-azure-resource")
    .with_api_key("your-api-key")
    .with_api_version("2024-02-15-preview")
    .with_header("X-Custom-Header", "value");

let provider = AzureOpenAIProvider::new(settings);

Builder Methods

The AzureClient builder supports:

• .resource_name(name) - Set Azure OpenAI resource name
• .base_url(url) - Set custom base URL
• .api_key(key) - Set the API key
• .api_version(version) - Set API version (default: "v1")
• .header(key, value) - Add a single custom header
• .headers(map) - Add multiple custom headers
• .use_deployment_based_urls(bool) - Use legacy deployment-based URL format
• .build() - Build the provider

Azure-Specific Features

URL Formats

Azure OpenAI supports two URL formats:

V1 API Format (Default)

https://{resource}.openai.azure.com/openai/v1{path}?api-version={version}

This is the recommended format.

Deployment-Based Format (Legacy)

https://{resource}.openai.azure.com/openai/deployments/{deployment}{path}?api-version={version}

Enable this format with .use_deployment_based_urls(true).

API Versions

Azure OpenAI supports multiple API versions. Common versions include:

• "v1" - Default, recommended version
• "2023-05-15" - Stable release
• "2024-02-15-preview" - Preview features
• "2024-08-01-preview" - Latest preview

Set the API version using the builder:

let provider = AzureClient::new()
    .resource_name("my-resource")
    .api_key("your-api-key")
    .api_version("2024-02-15-preview")
    .build();

Prerequisites

To use this provider, you need:

1. An Azure subscription
2. An Azure OpenAI resource
3. Deployed models in your Azure OpenAI resource
4. API key from Azure portal

Supported Models

Azure OpenAI supports various model types through deployments. You must first deploy models in your Azure OpenAI resource before using them.

Chat Models

GPT-4 and GPT-3.5 models for conversational AI:

• gpt-4 - Most capable model
• gpt-4-turbo - Faster GPT-4 variant
• gpt-35-turbo - Fast and efficient
• And other chat models available in your Azure deployment

Embedding Models

Text embedding models:

• text-embedding-ada-002 - Standard embedding model
• text-embedding-3-small - Smaller embedding model
• text-embedding-3-large - Larger embedding model

Image Models

Image generation models:

• dall-e-3 - Latest DALL-E model
• dall-e-2 - Previous generation

Completion Models

Text completion models:

• gpt-35-turbo-instruct - Instruction-tuned completion model

Note: Model availability depends on your Azure OpenAI resource region and deployment. Use your deployment name (not the base model name) when creating models.

Usage Examples

Model Types

Chat Models

Use .chat_model() or .model() for conversational AI:

use llm_kit_azure::AzureClient;
use llm_kit_core::{GenerateText, Prompt};

let provider = AzureClient::new()
    .resource_name("my-resource")
    .api_key("your-api-key")
    .build();

// Use your deployment name
let model = provider.chat_model("gpt-4-deployment");

let result = GenerateText::new(model, Prompt::text("Hello!"))
    .execute()
    .await?;

Completion Models

Use .completion_model() for text completion:

let model = provider.completion_model("gpt-35-turbo-instruct");

Embedding Models

Use .text_embedding_model() for embeddings:

let model = provider.text_embedding_model("text-embedding-ada-002");

Image Models

Use .image_model() for image generation:

let model = provider.image_model("dall-e-3");

Streaming

Stream responses for real-time output. See examples/stream.rs for a complete example.

Tool Calling

Azure OpenAI supports function calling for tool integration. See examples/chat_tool_calling.rs for a complete example.

Examples

See the examples/ directory for complete examples:

• chat.rs - Basic chat completion
• stream.rs - Streaming responses
• chat_tool_calling.rs - Tool calling with chat models
• stream_tool_calling.rs - Streaming with tool calling
• text_embedding.rs - Text embeddings
• image_generation.rs - Image generation with DALL-E

Run examples with:

cargo run --example chat
cargo run --example stream
cargo run --example chat_tool_calling
cargo run --example text_embedding
cargo run --example image_generation

Documentation

• API Documentation
• LLM Kit Documentation
• Azure OpenAI Documentation

License

Licensed under Apache License, Version 2.0 (LICENSE)

Contributing

Contributions are welcome! Please see the Contributing Guide for more details.