I'm gonna tell you something that you won't want to hear and will totally ignore: you're wasting your time on this, by which I mean the time of the AI/s you're using to vibe code and guess the numbers, and you will eventually realize that yourself and abandon it.
I tried using AIs before, out of curiosity, with just a 3b local LLM, simply asking 'how much is domain.com worth'.
Using sales from Namebio and DNJournal it was amazingly accurate, picking narrow ranges of about $5k-10k which matched the actual selling price. Ie. $15-25k for names that sold for about $17k.
I thought this is easy as pie, I'll throw up the best appraisal site evarrr.
Then I got one that was way off. The name supposedly sold for $25k, but the AI would insist it was just about $5k (which I agreed with).
I could not get it to go above $12k, and changing stuff to make it go higher just messed up other previously 'accurate' appraisals.
Then I tried different price ranges of sold names, and they were way off too, meaning not as accurate as the lower end.
And it never rated a name as worthless until I told it to (for unregged/unsold names), then it did that too often.
Then I had to tell it that no LLL.com sells for less than 5-figures, and nobody buys hyphens except Germans, and don't rate names at 8+ figures because that almost never happens, etc etc etc, until there are so many rules you're just writing a book.
That's exactly what you're doing now. Trying to plug holes and causing other leaks to appear.
And what for? If you don't know the value of your domains, or at least what you want for them, how is a computerized parrot that operates on guesswork ever going to know?
There's no science or possible way to calculate a 'correct' domain name price, which is perhaps why you have at least 5 different values that cover literally everything: low, medium, high, "likely range", mid-to-high range and a plus sign. It's meaningless.
Good luck!
