Crate llm_kit_azure 

Azure OpenAI Provider for LLM Kit

This crate provides an Azure OpenAI provider implementation for the LLM Kit. It allows you to use Azure OpenAI models for text generation, embeddings, and image generation.

Features

• Chat Models: GPT-4, GPT-3.5-turbo, and other chat models
• Completion Models: GPT-3.5-turbo-instruct and other completion models
• Embedding Models: text-embedding-ada-002 and other embedding models
• Image Models: DALL-E 3 and other image generation models
• Azure-specific Authentication: Uses api-key header
• Flexible URL Formats: Supports both v1 API and deployment-based URLs

Quick Start (Recommended: Builder Pattern)

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

#[tokio::main]
async fn main() -> Result<(), Box<dyn std::error::Error>> {
    // Create Azure OpenAI provider using the builder
    let provider = AzureClient::new()
        .resource_name("my-azure-resource")
        .api_key("your-api-key")
        .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(())
}

Alternative: Direct Instantiation

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 Options

Using Resource Name

use llm_kit_azure::AzureClient;

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

Using Custom Base URL

use llm_kit_azure::AzureClient;

let provider = AzureClient::new()
    .base_url("https://my-resource.openai.azure.com/openai")
    .api_key("key")
    .build();

With Custom API Version

use llm_kit_azure::AzureClient;

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

With Custom Headers

use llm_kit_azure::AzureClient;

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

With Deployment-Based URLs (Legacy Format)

use llm_kit_azure::AzureClient;

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

URL Formats

Azure OpenAI supports two URL formats:

V1 API Format (Default)

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

Deployment-Based Format (Legacy)

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

Use .with_use_deployment_based_urls(true) to enable the legacy format.

Environment Variables

The provider will read from these environment variables if not explicitly configured:

• AZURE_API_KEY - API key for authentication
• AZURE_RESOURCE_NAME - Azure OpenAI resource name

Model Types

Chat Models

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

let model = provider.chat_model("gpt-4-deployment");

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");

Structs

AzureClient

Builder for creating an Azure OpenAI provider.

AzureOpenAIProvider

Azure OpenAI provider implementation.

AzureOpenAIProviderSettings

Configuration options for creating an Azure OpenAI provider.