Knowledge Base Overview 📚
When a visitor asks Konvoq a question, the AI doesn't guess — it searches through everything you've taught it and picks the most relevant content before forming an answer.
Here's exactly how that works.
How Konvoq finds the right answer
Think of it like a really good librarian. When someone walks in and asks a question, the librarian:
- Listens carefully to understand what the visitor really wants
- Checks the FAQ shelf — if there's an exact match, they hand that over immediately
- Searches the stacks — if not, they look through all your docs and web pages for the most relevant passages
- Gives a clear answer — using the retrieved content as their source, not their imagination
Visitor message
↓
Intent classification
↓
FAQ exact match? ──── Yes ──→ Return FAQ answer
↓ No
Semantic search (all content)
↓
OpenAI generates final answer using found context
Content types and when to use each
| Source | Best for | Priority |
|---|---|---|
| FAQs | Single definitive answers | 🥇 Checked first |
| Documents | Long-form guides, policies, manuals | 🥈 Searched after FAQs |
| Scraped pages | Website content, product pages, blogs | 🥈 Same as documents |
| Text entries | Short facts — hours, address, pricing | 🥈 Same as documents |
What is "semantic search"?
Normal search looks for exact words. Semantic search understands meaning.
So if your docs say "We offer a 30-day money-back guarantee" and a visitor asks "Can I get a refund?", Konvoq connects those two — even though the words are different.
Keeping your knowledge fresh
The AI only knows what's currently indexed. If you update your website or docs:
- Re-scrape the updated pages in Knowledge → Scraped Pages → Re-scrape
- Re-upload updated documents — delete the old version first
- Update or delete outdated FAQs
Stale content = stale answers. A quick re-scrape after any major update keeps Konvoq accurate.
Vector storage
Every piece of content you add gets converted into a vector — a mathematical representation that makes semantic search possible. Your plan includes a vector storage limit.
Go to Dashboard → Usage to see how much of your quota you've used. If you're running low, delete outdated scraped pages to free up space.