Source types
Single page
Index one URL, right now. It must be a web page or text document. Upload PDFs as files instead.
Entire site
Enter the site’s URL. Thea uses its
sitemap.xml when it exists, including gzipped sitemaps and sitemap indexes. Otherwise it follows same-site links up to your plan’s crawl depth. On re-sync, only changed pages are re-embedded.Status page
A status page that re-syncs on a schedule. Pair it with Settings → General → Status page alerts to post when an incident opens or resolves.
Upload file
PDF, Markdown, or plain text. Size limits: 2 MB (Free), 20 MB (Pro), 50 MB (Business).
FAQ
Structured question-and-answer pairs. They get a retrieval boost because they look like how members ask.
Notion
Sync the pages you share with a Notion integration. Pro and Business. Counts as one source for the whole connection.
Zendesk
Articles from a public Zendesk help center. Pro and Business.
Discord thread
Index a resolved thread so Thea can reuse an answer your team already wrote.
Help center articles
Published articles from your Thea help center are indexed automatically.
Content API
Pages you push with an API key. Pro and Business. See Content API.
Crawl depth
When you add an entire site without a sitemap, Thea follows links up to:
Site crawls skip links to PDFs, images, archives, media, and fonts. To index a PDF, upload it as a file.
Each page or sitemap Thea fetches has the same size limit as file uploads: 2 MB (Free), 20 MB (Pro), or 50 MB (Business).
Add a source
1
Open Knowledge
In the server dashboard, click Knowledge.
2
Click Add source
Choose the type.
3
Connect it
Paste a URL, upload a file, or choose Connect platform for Notion or Zendesk.
4
Wait for Ready
Status moves through Queued, Building, Ready, or Error. Stale means a scheduled sync failed or the source changed out from under the index.
Sync
Choose Re-sync from a source’s menu to fetch it now. This works for single pages, sites, status pages, and Notion or Zendesk connections. Status-page alerts post only from scheduled syncs (see Settings → General), so if a manual re-sync picks up a change first, no alert is posted for that change.
A re-sync removes pages only after it reaches the whole site and confirms they are gone. If part of the site times out, is rate limited, or returns a server error, Thea keeps the pages it could not check and notes that on the source. A page that moved to a new address keeps its indexed copy.
Credentials in your docs
Thea scans everything it indexes for values that look like credentials, such as API keys, tokens, private keys, database passwords, and lines likeDB_PASSWORD=. It also catches a label and its value that sit in different table cells, paragraphs, or a heading. Thea replaces those values with a redaction marker before indexing, so answers never quote them.
A source with redactions shows a credentials badge on Knowledge. This applies to every source type, including Notion, Zendesk, help-center articles, and Content API pages. Treat the badge as a published credential: rotate it, remove it from the page, then re-sync. The badge clears once no passage from that source still contains a redaction.