What knowledge sources are
Vex AI is only as good as what you teach it. A knowledge source is any body of content Vex reads and searches when composing a reply. You can connect multiple source types, and Vex searches all of them every time it responds.
Where to manage sources
Go to Automation → AI, open your Vex chatbot, and switch to the General tab. The widget table shows each connected widget and the sources attached to it. Click the edit icon on a widget row to add, remove, or rescan sources.
Knowledge base
Your Yaplet knowledge base is the highest-value source. It auto-syncs every time you publish or update an article — no manual refresh ever needed. Vex can also cite KB articles inline in its replies so visitors can read the full article for more detail.
To connect: choose Knowledge Base in the source picker and select from your available bases. One chatbot can read multiple knowledge bases.
Documentation
If you have a Yaplet documentation site, link it the same way. Documentation is best for longer technical content — API references, integration guides, developer docs — that end-users might also look up.
Public URLs
Paste a single URL or a sitemap URL and Vex scrapes the content. Good for:
- Static help pages or FAQs hosted outside Yaplet
- Policy pages (terms of service, privacy policy, refund policy)
- Your main website's product or feature pages
Vex rescans public URLs automatically every two weeks. You can trigger a manual rescrape at any time from the source row — the button has a 5-minute cooldown between rescans.
Avoid dynamic pages that change often (blog index pages, live pricing with real-time updates, news feeds) — the scrape can go stale between automatic refreshes.
Documents
Upload PDF, DOCX, or TXT files directly. Vex chunks the content automatically for retrieval. Good for:
- Internal policies or employee handbooks you can't publish publicly
- Product manuals or spec sheets
- Contracts, SOWs, or any document a visitor might ask about
To update a document, delete it and re-upload the new version. Status shows as "Processing" until the embedding is complete (usually under a minute for typical documents).
Product feeds
Upload a CSV product feed so Vex can recommend products and answer product-specific questions. Vex rescans the feed automatically every 24 hours, or you can trigger a manual rescan (same 5-minute cooldown). Each product can be individually enabled or disabled from the feed table.
API tools (context tools)
For real-time data that changes too fast for any scraped source — account status, order history, subscription tier — you can wire Vex to a live HTTP endpoint. Vex calls the API mid-conversation, receives the response, and uses the data in its reply. This is an advanced setup; full details are in the developer documentation.
Processing status
After adding any source, Vex processes it in the background (chunking and embedding the content). Check the status indicator in the source row:
- Processing: Content is being indexed — Vex can't use it yet.
- Active: Ready — Vex searches this source on every reply.
- Warning: Partial failure — some pages or chunks couldn't be processed. Check the URL or file format.
- Inactive: Disabled — Vex ignores this source until you re-enable it.
Next step
After adding sources, preview a few questions to check quality: Test your chatbots before going live. After real conversations happen, find what Vex couldn't answer: Find knowledge gaps Vex couldn't answer.