For dynamic content, I’d avoid manual retraining and use an automated ingestion/sync pipeline. Keep the knowledge source updated from your CMS, database, or API, trigger updates when content changes, and use webhooks or scheduled syncing where appropriate. For high-traffic pages, I’d also load the Chatbase script asynchronously or on user interaction to minimize performance impact.The key is keeping content freshness automated while separating it from the chatbot’s core configuration.

Call me what you like but I enjoy manual retraining very much.

Automating the sync definitely makes more sense once the content starts changing regularly, manual retraining might work for smaller setups, but with a lot of updates it could get pretty time consuming, keeping the knowledge source connected to the chatbot seems like a much easier way to keep everything up to date.

For static content like FAQs, product guides, and documentation, normal ingestion works well. But for information that changes frequently, order status, account data, inventory, pricing, or subscription details, I would avoid relying on repeated retraining and instead connect the agent to a live source or action.

Automating the sync is definitely the way to go for dynamic content, manual retraining gets old fast when information changes regularly.

For dynamic content, I’d prioritize automated data syncing through APIs or scheduled ingestion rather than manual retraining. Keeping the knowledge source updated at the data layer makes the chatbot much easier to maintain as products, pricing, or business information changes. I’d also monitor response accuracy after updates to catch stale or conflicting information early.

For dynamic content, I’d lean toward keeping frequently changing information in connected data sources or APIs rather than relying on repeated manual retraining. That keeps the agent closer to the current source of truth and reduces maintenance as content changes. I’d also separate stable knowledge from real-time data, so policies and documentation can stay in the knowledge base while things like inventory, pricing, or order status are pulled dynamically when needed.

Automating that sync saves a lot of maintenance and keeps responses accurate without manual retraining every time something changes.

far too many marketing and CX teams load widget embed synchronously tanking the page’s interaction to next paint(INP) and largest contentful paint(LCP). delaying the script initialization until a user physically clicks a custom chat with us button or setting up a strict async/defer wrapper is non negotiable practice