Advanced Usage

4. Using Retry Policies

   To handle API rate limits and transient errors, use Polly to define retry policies:

   AsyncPolicy httpTimeoutAndRetryPolicy = Policy .Handle<Exception>(ex => ex.IsTransientError()) .WaitAndRetryAsync( retryCount: 6, sleepDurationProvider: retryAttempt => TimeSpan.FromSeconds(Math.Pow(2, retryAttempt)) + TimeSpan.FromMilliseconds(new Random().Next(0, 3000)), onRetry: (exception, timespan, retryCount, context) => { logger.LogError($"Retry {retryCount} after {timespan.TotalSeconds} seconds due to: {exception.Message}"); });

5. Adding New AI Providers

   To add support for a new AI provider, follow these steps:

   • Create a Configuration Class: Define a new configuration class inheriting from AIServiceProviderConfiguration.

   • Implement a Kernel Pool Class: Create a new kernel pool class inheriting from AIServicePool<T>.

   • Register the New Provider: Add the registration method in the ServiceExtension class to register your new provider with the DI container.

   For example, to add a new "CustomAI" provider:

   public record CustomAIConfiguration : AIServiceProviderConfiguration { public required string ModelId { get; init; } public required string ApiKey { get; init; } // Additional settings... } class CustomAIKernelPool( CustomAIConfiguration config, ILoggerFactory loggerFactory) : AIServicePool<CustomAIConfiguration>(config) { protected override void RegisterChatCompletionService(IKernelBuilder kernelBuilder, CustomAIConfiguration config, HttpClient? httpClient) { // Register service logic... } protected override ILogger Logger { get; } = loggerFactory.CreateLogger<CustomAIKernelPool>(); } public static class ServiceExtension { public static void UseCustomAIKernelPool(this IServiceProvider serviceProvider) { var registrar = serviceProvider.GetRequiredService<IKernelPoolFactoryRegistrar>(); registrar.RegisterKernelPoolFactory( AIServiceProviderType.CustomAI, (aiServiceProviderConfiguration, loggerFactory) => new CustomAIKernelPool((CustomAIConfiguration)aiServiceProviderConfiguration, loggerFactory)); } }

Supported Providers

• OpenAI: Use OpenAIConfiguration and OpenAIKernelPool to interact with OpenAI services.
• Azure OpenAI: Use AzureOpenAIConfiguration and AzureOpenAIKernelPool for Azure OpenAI.
• HuggingFace: Use HuggingFaceConfiguration and HuggingFaceKernelPool to integrate with HuggingFace models.
• Google AI: Use GoogleConfiguration and GoogleKernelPool for Google AI services.
• Mistral AI: Use MistralAIConfiguration and MistralAIKernelPool to leverage Mistral AI services.
• Custom Providers: Easily extend to support other providers by following the extensibility guidelines.

Contributing

Contributions are welcome! Please fork the repository, make your changes, and submit a pull request. Ensure your code adheres to the project's coding standards and includes appropriate tests.

License

This project is licensed under the MIT License. See the LICENSE file for more details.

Acknowledgments

• Special thanks to the contributors of Microsoft Semantic Kernel and all integrated AI service providers.
• Inspired by the need for efficient AI resource management in .NET applications.