Blog
Notes from the team.
What we're building, what we've learned, and the things we wish someone had told us about running scrapers in production.
-
JSON schema vs CSS selectors: two ways to scrape, compared
Both get data off a page. One couples you to the layout, the other to the meaning. Here's an honest comparison of when each one wins.
-
Pulling structured data out of documentation sites
Docs pages are dense, consistent, and server-rendered, which makes them ideal for extraction. Here's how to turn a reference page into a queryable dataset.
-
How we keep an LLM scraper from making things up
Putting a language model in your extraction pipeline is easy. Making it trustworthy is the actual work. Here's what that took.
-
Turning GitHub repositories into clean JSON
GitHub renders most of what you need server-side, which makes it a great first target. Here's a schema for pulling repo metadata into structured data.
-
How to scrape a website without writing a single selector
CSS selectors are the part of scraping that breaks. Here's how to pull structured data from a page by describing what you want instead of where it lives.
-
Why we built a structured web-scraping API
Maintaining selectors is the worst part of any scraping job. Here's the bet we made instead, and what it took to make it production-grade.