Dragwyb Click to Chat can use a knowledge base to retrieve relevant information from selected WordPress content and other sources while preparing AI chatbot replies. Configure the source material, indexing behavior, embedding provider, and vector storage in the Knowledge Base settings, then test the results with questions whose answers are present in your content.
Where to find the settings
Open Click to Chat → AI Assistant → Knowledge Base. The screen contains two tabs:
- Knowledge Sources: Add direct text such as FAQs or company information, web page URLs, and supported training documents.
- Database & Indexing: Select public website content types, choose an embedding provider and vector database, manage automatic synchronization, and start indexing.
How knowledge retrieval works
- Choose which supported public post types to include and add any additional knowledge sources.
- Select an embedding provider. The current interface offers OpenAI
text-embedding-3-smalland Google Geminigemini-embedding-001; the matching provider API key must be configured. - Choose vector storage: SQLite (local storage) or Pinecone (cloud storage, requiring its connection details).
- Save the settings and select Index Content Now to process the selected sources. Use the displayed indexing status to follow progress where available.
- Ask test questions based on known source content and review whether the retrieved answer is relevant.
Retrieval controls and source links
The settings include a minimum evidence confidence threshold: Strict (0.75), Balanced (0.65), or Permissive (0.50). Retrieved chunks below the selected similarity threshold are filtered out. You can also require indexed evidence before the assistant answers and configure the fallback message used when suitable context is unavailable. These controls help limit weak matches; they do not guarantee that every AI response will be correct.
When source display is enabled, supported AI replies can include up to three retrieved source links. Check that each link opens the expected page and supports the answer. Automatic synchronization can update selected website content when supported posts are published or edited and remove it when content becomes non-public, provided auto-sync is enabled.
Guides in this section
- Build the Website Knowledge Base — choose content sources and create an index.
- Configure SQLite Vector Storage — use the local vector-storage option.
- Connect Pinecone Vector Storage — configure a Pinecone index and connection.
- Improve Knowledge Retrieval Quality — review source coverage and test question relevance.
- Refresh the Website Knowledge Base — refresh indexed content after site changes.
- Review AI Answer Source Links — inspect links attached to supported AI replies.
After changing the selected post types or other indexing settings, re-index the content where needed and test again. The exact results depend on the configured providers, available sources, and successful indexing; this guide does not imply that a live indexing run or AI provider request has been tested.