The most effective method to train your GPT-powered chatbot involves creating a knowledge base portal and publishing help articles about your product, features, FAQs, and instructions. The bot automatically trains on these articles and updates whenever you add, edit, or delete them. Please note that only published articles are used for training; the bot does not use draft articles.
If needed, you can enrich your bot's knowledge using the sources page:
Web content: Import entire websites, a list of URLs, or fetch URLs from sitemaps. If your knowledge base is managed by Answered.so, there's no need to import it here. As mentioned earlier, your bot automatically synchronizes with all published articles in the knowledge base.
Text snippets: Enhance your bot's knowledge by creating text content. This is particularly useful if you wish to cover specific topics without adding this content into knowledge base articles.
Files: Knowledge can also be incorporated from PDF files.
What are training characters?
The number of training characters on each pricing plan defines how much data you can import to train your chatbots. An average A4 page typically contains around 2,000 to 3,000 characters. Training characters are defined at the account level and shared among all workspaces. They do not renew each month.
As a rough estimate, 1 million characters are roughly equivalent to 400 A4 pages of textual content.
Can I delete some training files/chatbots in order to regain the used training characters?
Yes! You can delete sources and recover the used training characters.