Release Notes

The prompt limit for Google AI Mode results has doubled, from 4,000 to 8,000 characters. This gives you room to submit longer, more detailed prompts, so complex queries return richer, more precise results in a single request.

You can now control page interactions with custom browser instructions when scraping LLM sources, on top of the batch processing already available. Together, these give you more control over how large-scale LLM data collection runs, letting you handle bigger workloads with the precision your use case needs.

You can now collect clean and structured product data from Home Depot product pages without building or maintaining a parser yourself. This adds another major retailer to our dedicated parser lineup – getting you ready-to-use fields faster and with less custom development on your end.

Your requests to CAPTCHA-protected sites are now handled with improved logic, resulting in higher success rates – particularly for jobs that previously retry rates or came back empty – should run cleaner, with fewer gaps in your data.

The change applies automatically across existing configurations, so no updates are needed on your side.

We've improved how pricing fields are extracted for Walmart e-commerce sources. Prices now appear more consistently in parsed responses across product pages and search result types. More reliable values will appear automatically in new requests, with no manual work from your end.

Headless Browser is now refined to build and check its fingerprints, so emulated sessions look more like real devices. We've made its browser fingerprints more realistic and consistent, so your sessions get interrupted less often and more of your requests get through – all automatically, with nothing to set up.