Nemu Reader vs. the alternatives
Where deterministic, model-free reading wins, and where a JS-rendering scraper is the better tool.
What Reader optimizes for
Reader does one thing: fetch a URL and return clean Markdown, deterministically. HTML, PDF and DOCX come back as the same structured text every time, with no model in the loop. That makes it fast, cheap and, crucially for pipelines, reproducible.
Where the alternatives win
A headless-browser scraper is the right tool when the content only exists after JavaScript runs, or when you need to interact with the page before extracting. An LLM-extraction service is the right tool when the target is a structured record buried in messy prose. Reader will not pretend to be either.
For the common case of turning an article, doc or PDF into text a model can use, determinism beats cleverness. You know what you stored, you can diff it, and it costs €0.003 instead of a model call.