Docs/Using Chervil

Your Web

Chervil can keep a private, local index of the pages you read β€” so you can ask what you read last month, pull together everything you've read on a subject, and have Sprig answer from your own reading. Local, off by default, and yours to delete.

Every other browser keeps a list of where you went and throws away what was there. Chervil can keep the text too β€” in a private index on your own machine β€” so your reading becomes something you can actually ask questions of.

Everything here is local. The index is a file on this computer. It is never synced and never uploaded, and no AI sees a page just because you read it: the text only reaches your AI provider if you ask a question it answers, exactly like attaching a file.

Turning it on

Composed pages are indexed automatically β€” that's what makes the Library searchable. Keeping the text of websites you visit is separate, and off until you turn it on:

Settings β†’ Security β†’ Your Web β†’ "Keep the text of websites I read"

Once it's on, Chervil quietly files the readable text of pages you visit as you browse.

What is never kept:

  • anything in a private tab
  • anything behind a login or paywall
  • any site you list under Never keep these sites

Asking about your own reading

With an index built up, you can ask things the web can't answer:

  • "What was that laptop review I read last month?"
  • "Find the article about the Dayton office move."

Sprig searches your index, answers from the pages themselves, and tells you which page and when you read it, linking the original so you can go back to it. If it can't find the thing, it says so plainly rather than quietly substituting a web answer β€” a guess here is worse than a miss, because you'd believe you'd read something you never did.

Pull a topic together

Ask Sprig to pull together everything you've read on a subject and it synthesizes across your own pages rather than summarizing each in turn: the through-line, where your sources agree, where they contradict each other, what's still open β€” and, usefully, what you haven't looked at yet.

Answering from your reading first

By default your reading is only used when you ask for it. You can also let Chervil check it before composing anything:

Settings β†’ Security β†’ Your Web β†’ "Answer from my own reading first"

With this on, every ordinary question quietly checks your index first, and Sprig uses what you've already read where it genuinely helps β€” marking those parts so you can get back to the original. Anything it drew on is listed under From your own reading in the Sources panel.

This one is off by default on purpose. It means pages from your reading travel with questions you never framed as being about your reading, which is a real change to what leaves your machine β€” so Chervil asks first rather than assuming.

It's also deliberately conservative about what counts as a match. When you ask about your reading, a loose match beats a miss. Here you didn't ask, so a loose match would drag an unrelated page into an unrelated answer β€” it requires every meaningful word of your question to appear in the page, and otherwise stays quiet.

The toggle needs "Keep the text of websites I read" switched on first; there's nothing to answer from otherwise.

Deleting what's kept

All under Settings β†’ Security β†’ Your Web:

  • Forget the last hour / Forget today β€” quick cleanups after a browsing session
  • Forget a site… β€” remove everything from one domain
  • Delete everything β€” empty the index entirely

These really delete. The rows are removed and the file is compacted, so the text isn't left sitting in free space β€” the same standard the rest of Chervil's privacy controls hold to.

Deleted composed pages go to Trash first so you can get them back, and are removed after 30 days by default (Settings β†’ Your Web β†’ Trash).