Currently, we require the user to provide a document.
This allows us to ground the generated content in a resource, resulting in less hallucinations.
The issue with this is that one must start with a document, and if you don't have a document, it's frustrating.
We should allow people to simply state the topic they want to learn about. After this, we can find resources for them, and let them select the most relevant.
There is also the opportunity to pull existing documents from the user's connected accounts -- such as Google Drive or Dropbox -- but this will of course require pre-authentication
Currently, we require the user to provide a document.
This allows us to ground the generated content in a resource, resulting in less hallucinations.
The issue with this is that one must start with a document, and if you don't have a document, it's frustrating.
We should allow people to simply state the topic they want to learn about. After this, we can find resources for them, and let them select the most relevant.
There is also the opportunity to pull existing documents from the user's connected accounts -- such as Google Drive or Dropbox -- but this will of course require pre-authentication