A complete, self-contained .NET client library for the OpenAI API. It provides interfaces, request/response models, and manager classes for all major OpenAI endpoints including chat completions, embeddings, images, assistants, fine-tuning, batches, and vector stores. The SDK uses a consistent HTTP client pattern and supports both Azure and standard OpenAI endpoints. It can be extracted and used in any .NET project with minimal modification.

A clean abstraction over Redis operations supporting string and hash field manipulations with three modes: overwrite, append, and numeric add. It encapsulates connection management and provides a simple interface (IRedisService) for common caching patterns. The implementation is self-contained and can be reused in any .NET application requiring Redis caching with atomic operations.

A thread-safe manager for handling cancellation of streaming AI chat responses. Uses a ConcurrentDictionary mapping user IDs to a tuple of SemaphoreSlim and CancellationTokenSource. This pattern allows multiple concurrent chat sessions with proper cancellation support and is reusable in any real-time streaming scenario where per-user cancellation is needed.

A generic workflow engine that chains AI calls and plugin executions based on a node-based configuration. It processes WorkFlowNodeData to execute sequential or parallel steps, supporting branching and conditional logic. The engine is decoupled from specific AI providers and can be extracted to build custom automation pipelines in other projects.