Player not loading? Watch on YouTube
This tutorial connects Obscura's web scraping output to a local LLM through Ollama. The presenter installs a prebuilt binary on Ubuntu, checks that it runs, and saves a live website as Markdown in page.md. He describes Obscura as an open source headless browser written in Rust that executes JavaScript through V8 without requiring Chrome or a Rust installation.
The demonstration uses Qwen 3.6 with 27 billion parameters on an Nvidia RTX A6000 with 48 GB of VRAM. That is the presenter's hardware, rather than a stated minimum requirement. He says another model can be used if it supports tool use. A shell pipeline sends a scraped blog post to the model for a five-bullet summary; a reusable script then accepts a URL and sends the Markdown through the Ollama API.
The presenter claims roughly 30 MB of browser memory use and page loads around 85 milliseconds. His website scrape takes about 15 seconds by his own estimate and produces 267 lines. He also notes that headers and footers may need cleanup before production use.
The architecture explanation covers eight Rust crates and a shared, single-threaded V8 isolate that serializes JavaScript work. The final example calls Obscura's MCP server over standard input and output to navigate to a blog post and return Markdown. The workflow fetches live web pages while keeping model inference on the demonstrated machine.
