Knowledge Base Integration
The AI bot uses your Knowledge Base via RAG (Retrieval Augmented Generation) — it searches for relevant articles and snippets and uses them as context for its answers.
How it works
- Visitor asks a question
- The question is converted into an embedding vector
- The most similar KB chunks (article sections and snippets) are searched
- The found chunks are passed as context to the AI model
- The model generates an answer based on the context
Configuration
Settings → AI Chatbot:
| Setting | Description | Default |
|---|---|---|
| Use Knowledge Base | AI bot should use KB as context | Yes |
| Confidence Threshold | Minimum similarity (0.0–1.0) for a chunk to be used | 0.75 |
| Max Articles | Maximum number of KB chunks in context | 3 |
| Snippet Priority | Snippets are preferred over articles | Yes |
| Source Attribution | Bot names the source of its answer | Yes |
| Gap Tracking | Save unanswered questions as knowledge gaps | Yes |
Confidence Threshold
The Confidence Threshold determines from which similarity a KB entry is considered relevant:
- 0.9+: Very strict — only almost exact matches. Fewer answers, but higher quality.
- 0.75 (Default): Balanced. Good compromise.
- 0.5–0.6: Loose — more hits, but also more irrelevant results.
Too low? The bot answers with irrelevant information. Too high? The bot cannot answer many questions and triggers handoffs.
Source Attribution
If enabled, the bot adds source references at the end of its answer, for example:
"Source: How do I set up my widget?"
Visitors can click on the link to read the full article.
Gap Tracking
If the bot does not find any relevant KB entries (or the confidence is too low), the visitor's question is saved as a gap. This helps you identify missing KB content.
Evaluate gaps: Knowledge Base → Gaps