A: Learning resources can include files, such as Word documents, PDFs, spreadsheets, and CSV files, or web resources accessible via a link, like a blog post or even an entire website. Our system can read both single web pages and full websites; however, only text content (excluding images) is processed. When uploading learning content, this consumes context pages—each tier allows a certain amount of data to be uploaded.

To illustrate, imagine a bot learning from a book. On average, one page is equivalent to 1,800 characters. So, if a blog post contains 5,000 characters, it would use about three context pages.

When it comes to adding information, the most effective approach is to include data in as much detail as possible. Providing comprehensive, precise content helps the system understand and process the information accurately, enhancing the overall quality and relevance of responses. The more detailed the input, the better equipped the bot is to deliver clear, informed answers, so it’s beneficial to cover specifics and nuances whenever possible.

To help improve your bot’s accuracy, we recommend keeping an eye on conversation history to spot any messages where the bot may have deviated from the intended answer. This typically occurs when there’s a lack of detail in the knowledge base for specific questions. In such cases, you can click the “Provide a better answer” button in the conversation history to provide a more detailed answer that the bot can use next time its faced with a similar question. This simple step helps strengthen the bot’s overall knowledge base, ensuring more accurate responses in future interactions.

Thank you for your suggestions - there are more videos and tutorials in our roadmap. If you need any help setting up your assistants don't hesitate contacting us.