Learn how to add reference documents, write custom instructions, sync URLs, and build Q&A pairs to train your AI sales and support agent.
The Knowledge Base is your AI agent's brain. Feeding it accurate, well-structured resources ensures it responds correctly to customer questions about your products, shipping policies, and refund guidelines.
1. Knowledge Base Directory
The dashboard lets you search and filter existing training resources by type or status. For each resource, the system displays:
- File Title: The custom title or name of the knowledge source (e.g., About Our Online Store).
- Source: The format type of the resource (Text, File, Website URLs, or Q&A).
- Status: The active state of the content, shown as 'Published' (green) or 'Draft' (gray).
- Author: Who created the source (e.g., AI or Admin).
- Created At: The timestamp when the resource was added.
Click the three-dot vertical action menu on any row to manage the source:
- View: Open a details panel to read the content text or download the source document.
- Edit: Modify the title or body of the knowledge base item.
- Unpublish: Demote a published item back to draft status to hide it from the AI agent's active lookup.
- Delete: Permanently delete the item from the system.
Important Note: Newly created or edited knowledge base items start as drafts. They are not visible to your Bloopy AI agent until you publish them. Once published, avoid deleting files, as this removes the knowledge from the AI. To maintain the knowledge base without losing information, use the 'Unpublish' function or archive the item instead.
2. Choosing a Source Type
Click '+ Add New' on the top right to select the type of knowledge base content you want to add:
- File: Upload documents (PDF, DOCX, or text files) containing policies, catalogs, or FAQ guides.
- Text: Write custom text instructions, guidelines, or business context manually inside the dashboard editor.
- Website URLs: Sync public links or web pages so the AI can scan them for information.
- Q&A: Set up direct question-and-answer pairs for specific support cases.