Show HN: Runo – open-source web scraping that returns typed JSON (github.com)

Runo, an open-source web scraping tool, has launched, allowing users to extract structured, typed JSON from any URL through a custom schema. Initially developed as a closed-source service, the creator has shifted to open-source, enabling broader access and collaboration. Users can input field names, types, and example values, after which Runo utilizes a large language model (LLM) to extract data while rendering JavaScript on demand. This approach eliminates the need for traditional web scraping techniques like selectors and XPath, making the tool more resilient to website redesigns—fields that can't be found return as null rather than disappearing completely. Significant for the AI/ML community, Runo leverages LLM capabilities for semantic extraction, enhancing data retrieval accuracy without being compromised by changes in the Document Object Model (DOM). It employs a smart rendering strategy that first tries plain HTTP requests for static pages, engaging a headless browser only when necessary. The tool supports three modes of operation: extracting from a single URL, processing multiple URLs with a batch schema, and crawling web links from a seed URL. Additionally, Runo is designed for flexibility, offering interfaces for command-line usage, local servers, and a Python library—requiring only a Google Gemini API key for functionality. This innovation promises to streamline workflows for data scientists and developers alike, facilitating efficient data extraction processes.