A highly optimized, thread-safe, zero-latency streaming text interceptor and sensitive words filter for LLM outputs. Supports Python dicts, raw text, and OpenAI SSE streams out-of-the-box. 专为大模型（LLM）流式输出设计的高性能、线程安全、零延迟文本拦截器与敏感词过滤清洗工具。支持原始文本、Python字典、以及 OpenAI SSE 流。
