Sources

Sources feed the Knowledge Base. Each source becomes a document with processing state.

Manual

• Click New document.
• Write or paste the content.
• Give a clear name to make citations easier.

PDF

• Drag or select a PDF file.
• A progress bar tracks the upload.
• The backend extracts text and splits it into chunks.
• Supported formats: PDFs with selectable text. Scanned PDFs may fail.

URL

• Provide the URL and a name.
• The API scrapes the visible content.
• Ideal for FAQs or product pages.

Site crawl

• Provide the name, the starting URL and the maximum number of pages (1–10, subject to a server cap).
• The platform walks the domain and creates one document per page.

Via API

Every source can also be created programmatically with the API Channel (knowledge:write scope): create bases, manual documents, import URLs and upload PDFs — ideal for bulk onboarding from your CMS or CRM.

Sync

Periodic syncing of crawl sources is configurable via API (the source's crawl_enabled and sync_interval_minutes fields); each source's state includes last_synced_at, next_sync_at, last_status and last_error.

Reindex

If you change a manual source, click Reindex to regenerate embeddings.