Jina AI Reader exposes an API to convert URLs to LLM-friendly input and to perform web searches via dedicated endpoints. It supports reading PDFs, MS Office docs, images, and HTML with configurable output through headers.
Collecting history — the radar snapshots this repo daily. The trend line appears after 3 days of data (1 so far).
What it is
Reader converts URLs to LLM-friendly input using https://r.jina.ai/https://your.url and can perform web searches via https://s.jina.ai/. It also documents usage patterns, request headers for output control, and local/self-hosted deployment via Docker.
How it works
Reader reads web pages, PDFs, MS Office documents, and images, processing them into text/markdown formats. It offers a range of request headers to control output format, engine, caching, timeouts, content retention, and markdown options. The README describes behind-the-scenes behavior for search (top 5 results, fetch, apply r.jina.ai), and the various header controls (x-respond-with, x-engine, x-cache-tolerance, x-no-cache, x-target-selector, x-wait-for-selector, x-timeout, x-max-tokens, x-token-budget, x-respond-timing, x-with-generated-alt, x-retain-images, x-retain-links, x-retain-media, x-with-links-summary, x-markdown-chunking, x-preset, x-detach-invisibles, x-set-cookie, x-md-*).
Getting started
- Usage examples show prepending https://r.jina.ai/ to a URL to fetch a read-friendly version. Example: https://r.jina.ai/https://en.wikipedia.org/wiki/Artificial_intelligence
- The repository supports a self-hosted Docker path with prebuilt images (ghcr.io/jina-ai/reader:oss) and two exposed ports (8080 h2c and 8081 HTTP/1.1) per the Run section.
- Local development and cookbooks.md are referenced for various presets and use cases.
Recent releases
- The latest release entry shows: - none
Traction
- Stars: 11801
Behind the repo
- Open-source branch of code behind r.jina.ai and s.jina.ai; SaaS MongoDB storage layer is not included here. Docker-based local development is described.
Caveats
- License: Apache-2.0
- Created: 2024-04-10
- Last push: 2026-05-22
- Language: TypeScript
- Repository notes that the OSS branch runs stateless by default with optional MinIO/S3-compatible bucket caching via docker compose; the SaaS storage layer is stripped in this branch.
- README mentions PDFs and MS Office uploads, image captioning, and various headers; specific internal implementations or API rate limits are described via linked docs and cookbooks rather than code excerpts here.