Right now, when a user clicks on a conversation starter, it automatically sends the conversation starter text to ChatGPT for processing. What would be nice is for there to be three types of conversation starters: basic, open-ended, and guided:
- Basic conversation starters would behave like the current conversation starter does now. A user clicks on it and it and it sends it off.
- Open-ended conversation starters would behave similarly, except when a user clicks on it, it populates the input field but doesn’t send it. This is useful in situations where the GPT wants to help the user start their question but knows it’s an incomplete question (for example, a video summarizing tool could use this to let users say: “Summarize this YouTube video…”. The user clicks it, manually enters the URL they had in mind, and then clicks the submit button).
- Guided conversation starters would be slightly more advanced. It would support fill-in-the-blanks with a limited answer set. So imagine a GPT whose job it is to convert something from one format to another but it only supports certain conversions. A user could click a guided conversation starter that reads “Convert from ___ to ___”. When clicked, a simple modal appears below the input window that says “Convert from ___ to ___” except the blanks are replaced with a dropdown list of options previously set up by the GPT author. When selected, the main input window gets populated. (Note that there may be a better way to design it than this, but you get the idea.)