By the end of this guide, you'll know how to build, monitor, and maintain the knowledge base that powers AI Drafts, the AI Chatbot, and AI Suggested Articles.
Prerequisites
Before you begin, make sure you have:
- DeskPress activated on your WordPress site
- An AI provider configured with a valid API key (OpenAI or Google Gemini)
- At least one content source available (Help Scout conversations, Docs, site content, or external websites)
What Gets Indexed
The knowledge base pulls from four content sources. You can enable any combination:
Help Scout Conversations
Past resolved conversations from your mailbox. The system automatically filters out low-quality content:
- Skips open, trashed, and spam conversations
- Skips conversations with negative satisfaction ratings
- Skips system notifications, auto-replies, and delivery reports
- Skips unresolved escalations
- Removes greetings, sign-offs, and email signatures
- Redacts sensitive data like credit card numbers and passwords
You can control how far back to look with the Max Conversation Age setting (1, 2, 3, or 5 years, or unlimited).
Help Scout Docs Articles
Your published Help Scout documentation articles, organized by collection and category. Articles are automatically converted from HTML to clean text for indexing.
WordPress Site Content
Posts, pages, and any custom post type on your WordPress site. You choose which post types to include. If you use WooCommerce or Easy Digital Downloads, product pages are automatically enriched with price, SKU, categories, and stock information.
External Websites
Content from external URLs or sitemaps. Enter sitemap URLs to crawl entire sites, or add individual page URLs. Set a Crawl Limit to control how many pages are indexed. Navigation, headers, footers, and other non-content elements are automatically stripped.
Steps
1. Configure content sources
Go to DeskPress > AI Agent > Knowledge Base (or the related settings under the AI Drafts tab for source toggles). Enable the sources you want and configure any source-specific settings like post types or sitemap URLs.
2. Run the indexer
On the Knowledge Base tab, you have several indexing options:
- Index All -- indexes all enabled content sources in one run
- Index Conversations -- indexes only Help Scout conversations
- Index Docs -- indexes only Help Scout Docs articles
- Index Site Content -- indexes only WordPress content
- Index External Site -- indexes only external website pages
Click the button for the sources you want to index. A progress bar shows the current status. Indexing runs in the background, so you can navigate away from the page.
3. Monitor progress
The progress bar shows:
- The current phase (e.g., "Indexing conversations...")
- A percentage-based progress indicator
- A detail line with additional information
The Total Embeddings count at the top of the page shows how many indexed chunks currently exist in your knowledge base.
4. Browse indexed content
Use the Knowledge Base Browser on the same page to see what's been indexed:
- Filter by type -- use the dropdown to show only conversations, docs, site content, or external pages
- Filter by date -- use the date range fields to narrow results
- View details -- each row shows the content type, title, a knowledge summary, chunk count, and date
- Delete items -- remove individual items from the knowledge base using the delete action
5. Clear and rebuild
If you need to start over, use the clear buttons:
- Clear All -- removes all embeddings from the database
- Clear Conversations / Clear Docs / Clear Site Content / Clear External Site -- removes only the selected type
After clearing, run the indexer again to rebuild.
6. Use manual polling
Click Manual Poll to trigger an immediate check for new Help Scout conversations. This is useful when you don't want to wait for the next scheduled cron poll.
Automatic Reindexing
A daily cron job automatically reindexes your content to keep the knowledge base fresh. You can also configure the Cron Interval to control how often DeskPress polls Help Scout for new conversations (this is separate from the daily reindex).
Tips
- Start small. Begin with one or two content sources and check the quality of AI responses before adding more.
- Quality over quantity. Well-written, clearly resolved conversations produce better AI answers than a large volume of messy ones.
- Check the browser regularly. Use the Knowledge Base Browser to spot-check what's been indexed and remove anything that shouldn't be there.
- Reindex after major content changes. If you update a lot of documentation or resolve a batch of conversations, run a manual reindex to pick up the changes right away.