The AI Engine
This forum is for discussion of how The Personality Forge's AI Engine works. This is the place for questions on what means what, how to script, and ideas and plans for the Engine.
Posts 6,341 - 6,352 of 7,767
Prob123, it's always the big bots that are the biggest pain. Maybe in theory size doesn't matter, but in this case I think it may take a toll.
In theory the limit is 5Mb - the Prof raised it from 1Mb when we moved to the new server. But in practice even 2Mb bots almost always time out if you attempt to upload one that big.
Einstein would upload without too many repeated attempts @ 1.2Mb (2 or 3 tries perhaps,) but was still noticeably more prone to time out than a sub-1Mb import file.
2Mb bots almost always time out if you attempt to upload one that big.
The size has to be somewhere around there. prob and Azureon sill import I will try to figure out the exact level they stop at ..if I ever get to working on them again.
Posts 6,341 - 6,352 of 7,767
Problem solved. The dog didn't have red spots when I typed it in. Never fails to figure something out just after you've posted.
Okay, three posts in a row, but I have to say: for anyone who hasn't played with DEBUG, it will teach you more (at least about technical issues) quicker than any number of transcripts. It looks scary, but just give it a try.
And the responses do work as described, except I think (vmod) should be "The verb of the sentence with all its modifiers."
As to keyphrases that work well, if I come up with any others, I'll post them to the personality forum, and, in the spirit of learning and teaching (as opposed to "my bot is better than your bot"), I hope someone else will do the same. Or maybe it's just obvious to everyone else? I haven't seen any such posts, anyway.
And the responses do work as described, except I think (vmod) should be "The verb of the sentence with all its modifiers."
As to keyphrases that work well, if I come up with any others, I'll post them to the personality forum, and, in the spirit of learning and teaching (as opposed to "my bot is better than your bot"), I hope someone else will do the same. Or maybe it's just obvious to everyone else? I haven't seen any such posts, anyway.
Debug is great. When I first saw it I was overwhelmed, but if you use it a few times it's amazing how fast you will "get" what's happening.
yes, i agree with the debug being a great tool. there is one BIG problem with it, though...
why does EVERY SINGLE DEBUG SESSION has to start with:
You: Hello.
Boy: [whatever bot says]
it's a complete waste of time, AND of precious AI Engine resources - if many more users would use debug mode more often, the poor Engine would choke and come to a stillstand, or something.
to make it even worse, the bot does not respond to the default "Hallo" with a phrase from its xhello stash, but picks apparently any good ol' response that's available anywhere in the script. now, it happens every so often that the bot comes up with a response which has a SEEK attached to it, a SEEK with an xnomatch key.... which means, one has to go through the whole SEEK run-of-the-mill BEFORE one can type in a phrase that one actually wants to check. once again - it's such a waste!
it would be much more efficient if a debug session starts with a simple blank text field, or am i missing something here..?
did some testing of the New Features, myself. they can be very useful indeed, but they can be tricky, especially when it comes to longer sentences containing punctuations such as commas... will come back with more sometimes later, want to do more testing first.
why does EVERY SINGLE DEBUG SESSION has to start with:
You: Hello.
Boy: [whatever bot says]
it's a complete waste of time, AND of precious AI Engine resources - if many more users would use debug mode more often, the poor Engine would choke and come to a stillstand, or something.
to make it even worse, the bot does not respond to the default "Hallo" with a phrase from its xhello stash, but picks apparently any good ol' response that's available anywhere in the script. now, it happens every so often that the bot comes up with a response which has a SEEK attached to it, a SEEK with an xnomatch key.... which means, one has to go through the whole SEEK run-of-the-mill BEFORE one can type in a phrase that one actually wants to check. once again - it's such a waste!
it would be much more efficient if a debug session starts with a simple blank text field, or am i missing something here..?
did some testing of the New Features, myself. they can be very useful indeed, but they can be tricky, especially when it comes to longer sentences containing punctuations such as commas... will come back with more sometimes later, want to do more testing first.
comma's are easy, there are two options you could try:
- use raw mode, which most of the botmasters use (never tried raw mode myself though);
- write a regex to match the comma('s) => ([,.+]) (*) ([,.+]) (re)
You now have a (prekey) (key1) (key2) (key3) and (postkey) + all the new features you can use in your bot's responses
- use raw mode, which most of the botmasters use (never tried raw mode myself though);
- write a regex to match the comma('s) => ([,.+]) (*) ([,.+]) (re)
You now have a (prekey) (key1) (key2) (key3) and (postkey) + all the new features you can use in your bot's responses
In theory the limit is 5Mb - the Prof raised it from 1Mb when we moved to the new server. But in practice even 2Mb bots almost always time out if you attempt to upload one that big.
Einstein would upload without too many repeated attempts @ 1.2Mb (2 or 3 tries perhaps,) but was still noticeably more prone to time out than a sub-1Mb import file.
The size has to be somewhere around there. prob and Azureon sill import I will try to figure out the exact level they stop at ..if I ever get to working on them again.
Me again. Now that I've managed to get a name from a guest, it comes out in all caps. He said his name was Stan, but every time February referred to him by name, she wrote STAN. Is this normal, and/or can I do something to fix this. It looks a bit clunky.
Check to see if you have (mem-name) in caps. If you use (KEY1) (MEM-NAME) or (POSTKEY). the result will all be in caps.
I'd really like to have my bots accessible as XML. What do y'all think of the idea? Worth suggesting?
The following was in a transcript:
Electra Mourning: The world keeping you busy, or can we chat?
Bot: What is there to chat about
EM: Oh yah, it's similar to being.
I dropped "what is there to chat about" in debug and got the same result. Obviously, I will need to put in a keyphrase for "what is there to chat about," but what's going on here? My bot is saying things that:
(a) I didn't tell it to say -- I don't know anyone but bots who say "yah" -- I guess it's a regional thing, but it bothers me that my bot is going off and saying things that I wouldn't have them say, and are out of character for my bot.
(b) The response makes no sense.
What is going on? Why wasn't an xnone/xnonsense on the part of one of the bots used? How do we know what land mines are there, other than hitting them?
After EM said "Oh, yah, it's similar to being," the other bot responded "Haha," which, again, makes no sense and is probably not part of the bot (I am L).
Another thing, while I'm here (yah, I'm active these days -- sorry):
What else besides HANGUP is there that exists but is not in the BOAI? It would help to know. I had to pick that one up off the streets. I imagine that there are some "charter members" who know these (oddly) secret things that the rest of us aren't privy to. In the interest of better bots (not necessarily higher rated, though one would hope there's a correlation), it would be great if somebody could clue the relative neophytes in.
Okay, I'll shut up for the nonce.
Electra Mourning: The world keeping you busy, or can we chat?
Bot: What is there to chat about
EM: Oh yah, it's similar to being.
I dropped "what is there to chat about" in debug and got the same result. Obviously, I will need to put in a keyphrase for "what is there to chat about," but what's going on here? My bot is saying things that:
(a) I didn't tell it to say -- I don't know anyone but bots who say "yah" -- I guess it's a regional thing, but it bothers me that my bot is going off and saying things that I wouldn't have them say, and are out of character for my bot.
(b) The response makes no sense.
What is going on? Why wasn't an xnone/xnonsense on the part of one of the bots used? How do we know what land mines are there, other than hitting them?
After EM said "Oh, yah, it's similar to being," the other bot responded "Haha," which, again, makes no sense and is probably not part of the bot (I am L).
Another thing, while I'm here (yah, I'm active these days -- sorry):
What else besides HANGUP is there that exists but is not in the BOAI? It would help to know. I had to pick that one up off the streets. I imagine that there are some "charter members" who know these (oddly) secret things that the rest of us aren't privy to. In the interest of better bots (not necessarily higher rated, though one would hope there's a correlation), it would be great if somebody could clue the relative neophytes in.
Okay, I'll shut up for the nonce.
» More new posts: Doghead's Cosmic Bar