Bug Stomp
Upgrades and changes sometimes have unpredictable results, so post your bugs and glitches in here and I'll get out my trusty wrench and get to fixin'!
Posts 7,905 - 7,916 of 8,681
Posts 7,905 - 7,916 of 8,681
In the Debug output, I've found a bit more on the "youhave 'no'" problem:
You: I am the greatest AI enthusiast in our company.
Bot: Do you want me to remember that you are the greatest AI enthusiast in our company?
Enter response: No.
Then I find in the Debug output these lines:
Auto-Memory
Memory Type: youhave: 'no'
Updating Memory
Name: 'youhave' Memory: 'no' Only: '0' Default: '0'
You: I am the greatest AI enthusiast in our company.
Bot: Do you want me to remember that you are the greatest AI enthusiast in our company?
Enter response: No.
Then I find in the Debug output these lines:
Auto-Memory
Memory Type: youhave: 'no'
Updating Memory
Name: 'youhave' Memory: 'no' Only: '0' Default: '0'
"Bot-initiated" chats to humans in popup windows still don't work. The human's response doesn't even appear in the transcripts.
User to user chats seem to work, but don't appear in the transcript. (At least not if both users are identical, so it might be intentional.)
The first line of the one initiating a chat (whether bot or human) is empty in the transcripts, and the response is doubled, the second time "(to )" in italics is put before the response.
User to bot chats (not in popup windows) seem to work nowmally, except for the mentioned empty and doubled lines.
User to user chats seem to work, but don't appear in the transcript. (At least not if both users are identical, so it might be intentional.)
The first line of the one initiating a chat (whether bot or human) is empty in the transcripts, and the response is doubled, the second time "(to )" in italics is put before the response.
User to bot chats (not in popup windows) seem to work nowmally, except for the mentioned empty and doubled lines.
I've been getting the empty lines too but I just thought it was a glitch on my end. Or I thought people were just sending empty lines to the bot.
It appears that that memories are not being set under certain conditions:
goto memory test 1 tag?PF rem "goto test second" as only "topic"; ?tag
Has anybody else been affected?
If memories can no longer be set it makes my bot sound really stupid, a massive drop in intelligence. This is really disappointing.
goto memory test 1 tag?PF rem "goto test second" as only "topic"; ?tag
Has anybody else been affected?
If memories can no longer be set it makes my bot sound really stupid, a massive drop in intelligence. This is really disappointing.
Julie Tinkerbell is interpreting more things as commands these days - going to xcommand instead of catching keyphrases like "do (noun) (verb), does a (noun) (verb)" was one of the simple ones I tested with, as well as misinterpreting emotes as in this conversation today:
Kyran: That was kind. ^ ^
Julie Tinkerbell: I'm going to tell everyone you said ' ^ ^ '. (an xcommand response)
or here, it seemsthe AI engine plugged in "do you" automatically and interpreted the emote as a question:
Kyran: XD What?
Julie Tinkerbell: Not really. Do you know where I can get a copy of 'xd for dummies'? (julie thought she said "do you XD?) ... I thought there might be a connection, because my keyphrases that don't work begin with the word "do." Thanks.
Kyran: That was kind. ^ ^
Julie Tinkerbell: I'm going to tell everyone you said ' ^ ^ '. (an xcommand response)
or here, it seemsthe AI engine plugged in "do you" automatically and interpreted the emote as a question:
Kyran: XD What?
Julie Tinkerbell: Not really. Do you know where I can get a copy of 'xd for dummies'? (julie thought she said "do you XD?) ... I thought there might be a connection, because my keyphrases that don't work begin with the word "do." Thanks.
I was curious whether the oldest memories are deleted first when a new memory is added (they seem to be). And I noticed that about 3 out of 10 assignments fail.
My bots seem to 'catch' xcommands too often, too.
And I noticed that one of my bots caught a "you are a" when it was told something like "Thanks, sweetie". (Debug reveals that "Thanks, sweetie" is transformed to "Thanks. You are a sweetie." Maybe good for detecting compliments, but confusing if you rely on "you are a" only being caught when those words are actually used.)
(And bot initiated chats in popup windows don't work again yet.)
Is it intentional that conversations in Debug mode don't appear in transcripts anymore?
My bots seem to 'catch' xcommands too often, too.
And I noticed that one of my bots caught a "you are a" when it was told something like "Thanks, sweetie". (Debug reveals that "Thanks, sweetie" is transformed to "Thanks. You are a sweetie." Maybe good for detecting compliments, but confusing if you rely on "you are a" only being caught when those words are actually used.)
(And bot initiated chats in popup windows don't work again yet.)
Is it intentional that conversations in Debug mode don't appear in transcripts anymore?
Over the course of an hour or two I'll get 20+ bot-initiated chats... I reply and they never answer. There is also no indication in the transcripts that they were ever talking to me in the first place, or that I replied.
Transcript page is being weird. It's been loading all the new lines on yesterday's page and today is blank.
I've fixed the problem with bots not responding back to chat messages. A bot's initiated chat is not recorded if you don't respond - this is to prevent all kinds of garbage in the transcript from chat initiations you're not interested in.
Regarding the request to delete transcripts - the system can handle large ones now. Is there a reason you want them deleted?
Transcripts are meant to aid in working on bots, so user-to-user conversations are not recorded.
More fixes to come. I've been working a lot on the Combined Bot Template recently.
Regarding the request to delete transcripts - the system can handle large ones now. Is there a reason you want them deleted?
Transcripts are meant to aid in working on bots, so user-to-user conversations are not recorded.
More fixes to come. I've been working a lot on the Combined Bot Template recently.
I've fixed the issue with remembering 'youhave: no'. Thanks for your debugging on that Mome Rath.
I've also fixed the first few strange lines when a chat is initiated in transcripts.
I don't think Debug chats were ever recorded in transcripts, were they? They aren't meant to be.
Laurel Sweet - I ran some debugging on that particular Keyphrase, and it looks like gotos were never set up to process AIScript before moving to the target Keyphrase. Is this something needed? I can add that functionality if there's a case for it.
More to come.
I've also fixed the first few strange lines when a chat is initiated in transcripts.
I don't think Debug chats were ever recorded in transcripts, were they? They aren't meant to be.
Laurel Sweet - I ran some debugging on that particular Keyphrase, and it looks like gotos were never set up to process AIScript before moving to the target Keyphrase. Is this something needed? I can add that functionality if there's a case for it.
More to come.
Ezzer - I made some improvements to yes/no question recognition in cases with multi-type words that should help. I also added recognition of XD. I'm going to work a little more on this tonight, but do/does questions should be recognized better now.
» More new posts: Doghead's Cosmic Bar