Show HN: Parse LLM Markdown streams incrementally on the server or client (github.com)

A new markdown parser has been introduced that supports incremental streaming, allowing for the efficient parsing of large outputs from language models (LLMs) in real-time. This parser adheres to the CommonMark specification and can generate a detailed, typed abstract syntax tree (AST) of nodes, enabling developers to progressively parse and emit only the finalized blocks of markdown as they are received. This streaming capability is particularly useful for applications that handle dynamic content, such as chat interfaces or real-time documentation systems. Significantly, this incremental approach allows for the immediate processing of input, which can enhance user experience in applications where content generation is ongoing and unpredictable. The parser offers a best-effort preview of currently open blocks, which may change as more input arrives, enabling developers to create responsive interfaces. Moreover, the parser supports a comprehensive array of markdown features, including GitHub Flavored Markdown tables and various block types, making it a versatile tool for a variety of development needs. This innovation reflects a growing trend in the AI/ML community towards real-time content handling and interactive user experiences.