Show HN: Smelt – Extract structured data from PDFs and HTML using LLM (github.com)

The newly launched tool, Smelt, offers an innovative way to extract structured data from PDFs and HTML documents and convert it into formats like JSON or CSV. Utilizing the Anthropic API, Smelt detects tables within any document, infers schemas by identifying column names and types, and allows users to output clean structured data effortlessly. Key features include support for input from local files or URLs, the ability to fetch the largest table from a document automatically, and robust command-line options for fine-tuning the extraction process. This development is significant for the AI/ML community as it streamlines data extraction processes, a critical task often fraught with limitations in traditional methods. By combining deterministic Go code with state-of-the-art language model capabilities, Smelt ensures efficient schema inference and data conversion, requiring only a single API call per run. Its approach enhances pipeline compatibility, allowing users to integrate Smelt with existing data workflows. Additionally, the flexibility in column type support and user-defined queries provides a powerful tool for researchers, data scientists, and developers seeking to harness unstructured data for analysis or machine learning applications.