# Importing your docs

> Import a web page, a whole documentation site, a list of pages or pasted text into the AI knowledge base, then keep it fresh with re-sync, the daily auto-update and the update history.

Source: https://joely.io/documentation/ai/importing-docs
Last updated: 2026-11-10

| Feature | Plan | Permission |
|---------|------|------------|
| Import a web page, paste text, delete pages | All plans | Add/delete knowledge entries |
| Import a sitemap or a list of pages | Team and higher | Add/delete knowledge entries |
| Re-sync a website, folder or page | Team and higher | Add/delete knowledge entries |
| Daily auto-update | Plus and higher | Add/delete knowledge entries |
| Browse imported pages, view indexed content, update history | All plans | View AI settings and usage |

Pages are indexed as written, with no review step, and serve AI replies right away. When a page and a curated fact disagree, [the fact wins](https://joely.io/documentation/ai/knowledge.md#how-the-ai-picks-its-context).

## Web import

Choose **Add Knowledge → A web page**.

![Add Knowledge dialog: a full documentation, a list of pages, a web page or plain text](https://joely.io/images/docs/ai/add-knowledge-chooser.webp)
*Add Knowledge offers four ways to import.*

1. Submit one public HTTP or HTTPS address. Private and internal hosts are blocked.
2. Pick a [scope](https://joely.io/documentation/ai/knowledge.md#scoping-a-fact), or leave **Auto-detect package** on to let Joely pick it from the content.
3. Optionally, add CSS selectors to target or remove parts of the page, up to 10 of each.
4. Joely fetches the page (up to 100,000 characters), strips navigation and boilerplate, and indexes it in sections.
5. The modal ends on the page and its scope. Fix a wrong scope there, with no re-import.

When the site publishes each page's Markdown (GitBook, Mintlify, or an `llms.txt` plugin), Joely reads it instead, so titles and steps stay as written. CSS selectors turn this off.

A failed import shows a retry button on its row. A page with nothing left after cleanup gets a **No content** badge and keeps any content indexed before. Importing the same address again refreshes it (see [Refresh semantics](#refresh-semantics)).

### Images

- Joely describes each useful image from the image and its surrounding text, and skips logos, banners, ads and icons. This uses none of your AI responses.
- It describes at most 30 new images per page and import. Images above 20 MB are linked without a description.
- The AI adds an image to a reply when it shows what the customer asks about, up to **3 per reply**, always with a text answer. Images are linked to your site, not copied: keep them online.

## Pasting text

Choose **Add Knowledge → Plain text** and enter 10 to 100,000 characters. The text is indexed exactly as written. Edit or delete it in the **From text** tab.

On the **Team** plan or higher, **Fix with AI** (grammar, spelling, tone, in the text's language) and **Translate** sit next to the character counter. Both act on your selection, or on the whole text, up to 10,000 characters per request. The toast offers **Undo**, and nothing is indexed until you save.

## Sitemap import and refresh

Choose **Add Knowledge → A full documentation** (Team and higher).

1. Paste a `sitemap.xml`, an `llms.txt`, or any page of the site. Joely looks in `robots.txt`, `/sitemap.xml` and the page's parent folders, then for an `llms.txt`. Private hosts are blocked.
2. Filter by path, such as `/docs/`, and select up to **100 pages** per import. If you pasted a docs page, only it and the pages under it start selected. Pages already imported never start selected.
3. Set the scope and CSS selectors once, for every page. With auto-detection, pages under the same parent take the package most of them agree on.
4. Pages import one at a time. The progress panel marks each one indexed, unchanged, without readable content, or failed. Closing it does not stop the import.

### Import a list of pages

For a site with no sitemap and no `llms.txt`, choose **Add Knowledge → A list of pages** (Team and higher). Paste up to **100** addresses, one per line. Empty lines, duplicates and list markers (`- `, `1. `) are ignored. Lines that are not `http` or `https` addresses are skipped and listed with their line number. Each site gets its own source, and scope, selectors and progress work as for a sitemap.

A list source cannot discover new pages on re-sync: paste them in a new list.

## Refresh semantics

| Action | Finds new pages? | Deletes missing pages? | Cost if unchanged | Plan |
|--------|------------------|------------------------|-------------------|------|
| Re-import a URL | No | No | Nothing | All plans |
| Rerun a sitemap import | Yes, you select them | No | Nothing | Team and higher |
| Re-sync a source | Yes, from its sitemap (not a list source) | No, flags them | Nothing | Team and higher |
| Re-sync a page | No | No | Nothing | Team and higher |
| Daily auto-update | No | No | Nothing beyond the fetch | Plus and higher |

A changed page has its sections replaced by the new capture, and only the changed sections cost anything. Pages missing from the site are **never deleted automatically**: delete them yourself.

## Knowledge sources and re-sync

The **From web** tab lists every imported page, grouped by website, with its sections, freshness and import state. A website gets a warning badge after **30 days** without a sync. The eye icon shows a page as indexed for the AI, images included.

![From web tab listing the pages of an imported documentation site](https://joely.io/images/docs/ai/sources-tab.webp)
*Pages grouped by website, with their scope and actions.*
1. Re-syncs the whole website
2. Shows the page as indexed for the AI

![Ember Garage page as indexed for the AI: configuration options and troubleshooting sections](https://joely.io/images/docs/ai/indexed-page-sections.webp)
*The eye icon: what the AI reads from a page, cleaned and filtered.*

To re-sync a website or a folder:

1. A free check finds the sitemap again and marks each page unchanged, changed, new, missing or failed.
2. Changed and new pages start checked. Untick any, or use **Pick only imported**. With nothing to update, the re-sync ends on its own.
3. Checked pages are re-indexed, and a summary counts the results.
4. Missing pages are flagged, never deleted. A 404 or 410 page is struck through, other failures get an error badge. The content stays until you delete it.

A re-sync handles at most **300 pages**. Above that, it is refused: delete unused pages or tick fewer pages.

To re-sync **one page**, use its row. Joely tells you if it is unchanged or offline. If it changed, the eye button shows the changes line by line before you re-import.

### Automatic daily update

On the **Plus** plan or higher, the **Auto-update** switch re-checks each website 24 hours after its last import or re-sync.

- Only imported pages are checked: the update adds no page and deletes no page.
- A changed page is re-indexed with its CSS selectors and **keeps its package, category and shop scope**.
- Each run covers up to **300 pages** per website, the ones refreshed longest ago.
- While it runs, the website shows **Auto-updating…** and you can stop it.
- You are notified only when pages were updated or became unreachable, once per failing page.
- Updates pause while the AI assistant is off or the plan lacks the feature, then resume on their own.

### Update history

The history button, next to **Auto-update**, lists what happened to your imported pages.

| Event | When |
|-------|------|
| Added | A page is imported, found by a re-sync, or imported again after deletion |
| Updated | New content was re-indexed. **See changes** shows the lines removed and added, up to 1,000 |
| No change | A check found identical content. Shown with **Show unchanged checks** |
| Failed | The page could not be read. It keeps serving its previous content |
| Removed | The page was deleted, alone or with its folder or website |

Each entry shows its cause and who started it, and you can filter by website and event. Pasted text is not listed. History is kept for **90 days**.

## Deleting pages

Delete a page, a folder or a whole website from the **From web** tab, after a preview of what goes. The content leaves the AI immediately. Re-importing the same address brings the page back, for free if it has not changed.

## Package documentation in the ticket sidebar

When a ticket's packages have imported pages, the ticket details show a **Documentation** row listing them, on the dashboard and on the portal. Staff can also insert a **Documentation** button in a reply from the composer's `/` menu. Pasted text never appears there.

![Documentation dialog listing the Ember Garage page](https://joely.io/images/docs/ai/sidebar-documentation.webp)
*The Documentation row of the ticket details opens the linked pages.*

## Older imports

A page imported before image support or the update history catches up at its next re-import or re-sync.
