Hope the following is taken in the spirit intended, as I don’t want to appear callous about this. Based on your comments in this, and other threads, I do not believe your current approach to using DEVONthink is going to serve you well. In fact, over the years I’ve seen others take similar approaches and sooner or later they abandon DEVONthink, stating something along the lines of ‘DEVONthink just didn’t work for me’ or ‘the learning curve is just too steep’.
I say this because you appear to want to jump in with the more advanced usages of the app, while making many assumptions about what DEVONthink will, or will not, do without first becoming familiar with DEVONthink by actually putting it to work with the data you are working on right now. It also complicates things somewhat in that you want to take on DEVONagent with the same approach, similar assumptions (hint: DEVONagent’s searching is MUCH more powerful than using Google alone) at the same time that you are still just beginning to learn DEVONthink.
For now, forget about data from past computers that is stored in layer after layer in your Documents folder. If you haven’t needed that data in years, odds are that you never will. With the ephemeral state of knowledge today, I suspect that most of that data is obsolete now anyway.
Assuming that you do have volumes of unorganized data that has value, using DEVONthink as a Finder substitute is not going to be beneficial. You also cannot just ‘move folders into DTPO, do the work with classify, then export again’. The classify function of DEVONthink requires a level of group organization to begin with to effectively classify newly-added data to the database.
Forget about the preoccupation to work with scripting (beyond the scripts that are included with DEVONthink), utilities to organize window position, means to shortcut opening data with external apps. Fiddling with this now just adds complexity when the time would perhaps be better spent learning the basics of what DEVONthink can do.
Forget about all the rebuilding/repairing/verifying of your databases. If you are getting error messages indicating that you need to do this on a regular basis, something is very wrong.
But to get to your specific question, I would forget about indexing documents unless I could state a very explicit reason why your needs cannot be met by importing documents into discrete databases. There are multiple reasons why I say this, but perhaps the most significant reason is that replicants are one of the more powerful features of DEVONthink, yet when indexing they will totally frustrate an inexperienced user that expects the filesystem (Finder folders) to match document by document the organization of indexed groups in the database.
In summary, my recommendation would be to create a database, or databases, that relate to what you are working on now as well as current interests for your development. Import/capture data into DEVONthink that has value. If you know you have captured data in the past that adds value, search with Spotlight, review the results, import the data into DEVONthink that has value, and delete or archive the rest. Over time you will become familiar with how DEVONthink works and you will have built an organizational system in your databases that will allow See Also/Classify to work effectively. In reality there is more to it than this, but I’ve already spent more time drafting this than I could afford today. Hopefully some of this will be helpful to you.