Building Your Knowledge Base
Train your AI chatbot using website content, uploaded documents, and custom Q&A pairs.
Building Your Knowledge Base
Your chatbot's answer quality depends entirely on the quality of your knowledge base. Suppabot supports four content sources: crawled website pages, uploaded documents, free-form text blocks, and explicit Q&A pairs.
Website crawling
When you add a website URL, Suppabot automatically starts a crawl. The crawler:
- Visits every page reachable from your root URL.
- Extracts the visible text content (navigation, footers, and boilerplate are filtered out).
- Chunks the content into segments and builds a vector index.
The AI searches this index to find relevant content for each visitor question.
Re-crawling manually
Your website content changes over time. To re-train from the latest content:
- Go to Dashboard → your website → Settings.
- Click Re-train.
- A new crawl starts. The existing index remains active until the new one is ready.
Automatic re-crawl (Growth and Scale plans)
Growth and Scale plan subscribers can set a crawl schedule under Settings → Re-train Schedule:
- Monthly
- Weekly
- Daily
This keeps your chatbot's knowledge in sync without manual intervention.
Page quota
Each plan includes a maximum number of indexed pages. Your current usage is shown in Dashboard → Usage. If your site exceeds the quota, the crawler indexes pages in order of discovery (breadth-first from the root URL). Upgrade your plan to increase the page limit.
Uploading documents
Upload PDFs and DOCX files to supplement your crawled content — useful for product manuals, pricing sheets, internal policies, or any content not published on your website.
- Go to Dashboard → Knowledge Base → Upload.
- Select one or more files.
- Click Upload. Processing starts immediately.
Supported formats: PDF, DOCX. Max file size: 20 MB per file.
Once processed, document content is chunked and indexed alongside your crawled pages. The AI treats all sources equally when searching for answers.
Text sources
Add free-form text blocks for content that is not a file and not on your website — company policies, return procedures, hours of operation, contact details, or anything else you want the AI to know.
- Go to Dashboard → Knowledge Base → Add Text.
- Give the block a descriptive title (e.g. "Shipping Policy").
- Paste or write your content.
- Click Save. The block is indexed immediately.
You can edit or delete text blocks at any time. Changes take effect within seconds.
Chat Q&A pairs
Q&A pairs let you teach the AI specific answers for high-priority questions. These take priority over crawled content and documents — useful for ensuring consistent, precise answers on sensitive or frequently asked topics.
- Go to Dashboard → Knowledge Base → Q&A.
- Click Add Q&A.
- Enter the question as a visitor would ask it (e.g. "Do you offer refunds?").
- Enter the exact answer you want the AI to give.
- Click Save.
Tips:
- Add multiple phrasings of the same question as separate entries for better matching.
- Use Q&A pairs for legal disclaimers, pricing confirmations, or any answer where exact wording matters.
- Keep answers concise. The AI may expand or rephrase them slightly to fit conversational context.
Checking coverage
Use the Test Chat button (available on any knowledge base page) to ask questions and verify that the AI answers correctly.
Each response in Test Chat shows source citations — the specific pages, documents, or Q&A pairs the AI used to generate the answer. If an answer is wrong or missing a source, you know exactly what content to add or update.
Common issues:
| Symptom | Likely cause | Fix |
|---|---|---|
| AI says "I don't know" | Content not crawled or uploaded | Add missing content or re-crawl |
| AI gives outdated information | Website changed after last crawl | Trigger a manual re-crawl |
| AI gives wrong answer on a key topic | Conflicting content in index | Add a Q&A pair to override |
| AI is too verbose | System prompt not constraining length | Edit AI instructions (see Customizing Your AI) |
