jina-ai/reader

Extracts clean, structured markdown from any webpage or PDF while intelligently removing boilerplate content via readability filtering, and includes AI-generated image captions for multimodal LLM reasoning. Complements single-URL reading with a web search endpoint (`s.jina.ai`) that automatically fetches and processes top-5 results end-to-end, eliminating the need for downstream LLMs to handle browser rendering or JavaScript execution. Built on Puppeteer for headless Chrome rendering, supports fine-grained control via request headers (custom selectors, proxy routing, cache management), and handles complex single-page applications through timeout tuning and hash-based routing detection.

10,191 stars. No commits in the last 6 months.