Importing a website doesn't always finish on the first try — the open web is messy, and a page can time out or a site can be momentarily out of reach. What used to make that worse was the message you got: a confusing, internal-sounding error that didn't tell you what happened or what to do. Now an import that stumbles explains itself in plain language and gives you a one-click way to pick up exactly where it left off.
A status you can actually read
When an import doesn't complete, the knowledge base now shows a plainly worded status instead of a raw technical error. If the crawler was briefly too busy to finish reading a site, it says exactly that — "could not finish reading this website, please retry in a few minutes" — so you know it's a temporary hiccup, not a problem with your account or your setup.
See exactly where it stopped
Every import moves through four stages — crawling the website, parsing the content, storing it, and enriching the knowledge graph — and the knowledge base now shows that pipeline as a row of steps. If something goes wrong, the exact step that failed is marked, with the plain-language reason right beside it. You no longer have to guess whether a base is half-built or simply still working; the pipeline tells you at a glance.
Retry in one click, from where it failed
Next to the failed step is a Retry button, and it resumes the import from the step that stopped — not from scratch. The work already done isn't thrown away, and pages that already came through aren't crawled again. One click, and the import continues from the exact point it left off.
What you'll see
- A readable reason — a short, human explanation of what happened, in your language, instead of an internal error code.
- A visible pipeline — crawl, parse, store, enrich, with the failed step clearly marked so you know precisely where things stand.
- One-click retry — resume from the failed step without re-adding the source or starting over.
- No leaked internals — error messages no longer expose implementation details that were never meant for you.
How to use it
- If a knowledge base shows an error, open it and read the status — it explains what happened.
- Check the processing pipeline to see which step stopped.
- Hit Retry to resume from that step. Most temporary failures clear on a second attempt a few minutes later.
Why it matters
Hitting a wall is frustrating; hitting a wall with no idea why is worse. A vague error makes you wonder whether you did something wrong, whether your account is broken, or whether the feature works at all — when the real answer is often just that a page was momentarily out of reach. Turning that dead end into a clear status and a one-click retry means a stalled import becomes a few seconds of work instead of a support ticket, and your knowledge base ends up with the sources you actually wanted.
It's also a deliberate stance on error messages in general: the person reading one is trying to get something done, not debug our system, so a message should always answer "what happened and what do I do now" in plain language. We've scrubbed the internal details that used to leak through here and replaced them with a status you can act on — paired with resilient import behind the scenes that retries transient failures on its own before they ever reach you.