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 7,817 - 7,829 of 8,681
Posts 7,817 - 7,829 of 8,681
Sorry, doesn't work. Debug reveals that the input is split at the commas and each part is compared to the keyphrases, but not the entire expression.
I'd rather have commas not split the input than being able to catch commas. (The latter is more a proof of concept.)
I'd rather have commas not split the input than being able to catch commas. (The latter is more a proof of concept.)
It worked with this Keyphrase:
I recommend.* special [,] (.*) (re)
Message: I recommend the special, sir.
Find: I recommend.* special [,] (.*) (re) (30) Time: 0.07
Found
Response: I'll have the special then, (key1)
Processed: I'll have the special then, sir.
Is that what you had?
I recommend.* special [,] (.*) (re)
Message: I recommend the special, sir.
Find: I recommend.* special [,] (.*) (re) (30) Time: 0.07
Found
Response: I'll have the special then, (key1)
Processed: I'll have the special then, sir.
Is that what you had?
I've tried a somewhat different case.
Test keyphrase:
test keys (.*) [,] (.*) [,] (.*)(re) [0,0] <?PF raw ?>
Input to bot:
test keys a,b,c,d,e,f,g,h,i,j,k,l,m,n,o,p,q,r,s,t,u,v,w,x,y,z
Only line with "Search In:":
Search In: ' test keys a h male eq '
There is no comma in that phrase.
[Edit:]
Could work with fewer commas.
Keyphrase:
test key comma.* [,] (.*)(re) <?PF raw; ?>
Response:
prekey=[(prekey)]; key1=[(key1)]; key2=[(key2)]; key3=[(key3)]; postkey=[(postkey)]
Input:
I'd like to test key comma first item, second item, third item
Output:
prekey=[]; key1=[third item]; key2=[]; key3=[]; postkey=[]
key1 is ok, for RegExes are 'greedy' by default. But I miss prekey.
-----
And something unrelated:
I got an empty response again. Found these lines in Debug:
Considered BLAB (1).
Random Memory Number: 23 of 28
Memory Chance: 24 GossipTopic: 0 MemoryLevel: 2
Going for Gossip
After Gossip-Grab: 0.08
AIScript in Responses
Total Time Pre-Chrono: 0.08 RESULTS: '1'
Trying: (gossipname) told me (heshe)'s (mem-youare). (random: 0)
Skipping Response: (mem-youare) not found.
Response:
Updating Memory
There are several responses without (mem-youare) under xgossip, though.
Test keyphrase:
test keys (.*) [,] (.*) [,] (.*)(re) [0,0] <?PF raw ?>
Input to bot:
test keys a,b,c,d,e,f,g,h,i,j,k,l,m,n,o,p,q,r,s,t,u,v,w,x,y,z
Only line with "Search In:":
Search In: ' test keys a h male eq '
There is no comma in that phrase.
[Edit:]
Could work with fewer commas.
Keyphrase:
test key comma.* [,] (.*)(re) <?PF raw; ?>
Response:
prekey=[(prekey)]; key1=[(key1)]; key2=[(key2)]; key3=[(key3)]; postkey=[(postkey)]
Input:
I'd like to test key comma first item, second item, third item
Output:
prekey=[]; key1=[third item]; key2=[]; key3=[]; postkey=[]
key1 is ok, for RegExes are 'greedy' by default. But I miss prekey.
-----
And something unrelated:
I got an empty response again. Found these lines in Debug:
Considered BLAB (1).
Random Memory Number: 23 of 28
Memory Chance: 24 GossipTopic: 0 MemoryLevel: 2
Going for Gossip
After Gossip-Grab: 0.08
AIScript in Responses
Total Time Pre-Chrono: 0.08 RESULTS: '1'
Trying: (gossipname) told me (heshe)'s (mem-youare). (random: 0)
Skipping Response: (mem-youare) not found.
Response:
Updating Memory
There are several responses without (mem-youare) under xgossip, though.
Yes, the keyphrase I accidentally created, namely:
(.*)(re) <?PF raw; ?>
Maybe the following would be better for it wouldn't catch empty lines:
(.+)(re) <?PF raw; ?>
[Edit]
I could think of lots of fine tuning, like requiring at least one word of at least two letters etc.
(.*)(re) <?PF raw; ?>
Maybe the following would be better for it wouldn't catch empty lines:
(.+)(re) <?PF raw; ?>
[Edit]
I could think of lots of fine tuning, like requiring at least one word of at least two letters etc.
You guys present some real challenges! Okay so tonight I investigated sentence-spanning raw regex keyphrases with punctuation. It turns out that it's not possible to return (key)s in this mode without a lot of work. Since this mode skirts most of the better features of the AI Engine, I'm not going to work to support that.
Raw regex keyphrases without punctuation do support all the (key)s.
Bot-to-bot chatting: fixed
BLAB will trigger xnone at first and then xnonsense after several BLABs. BLAB simply means no keyphrases were matched.
Raw mode puts a space BEFORE the punctuation but not after it, which is one reason your keyphrases weren't working, Mome. The other being that it uses punctuation.
Gossip gets only one try to fire off - if the memory type doesn't exist, it goes on to something else. This is to prevent repeating of gossip memories too often.
Raw regex keyphrases without punctuation do support all the (key)s.
Bot-to-bot chatting: fixed
BLAB will trigger xnone at first and then xnonsense after several BLABs. BLAB simply means no keyphrases were matched.
Raw mode puts a space BEFORE the punctuation but not after it, which is one reason your keyphrases weren't working, Mome. The other being that it uses punctuation.
Gossip gets only one try to fire off - if the memory type doesn't exist, it goes on to something else. This is to prevent repeating of gossip memories too often.
The problem with BLABs is that they will end the chat, which is not very favorable in a contest.
The problem with gossip being fired off only once, is that it goes on to an empty response if a memory isn't set (instead of to another x-keyphrase).
I agree with you that regexes are useful in very special cases only and that for the vast majority of cases the AI engine does a really great work.
The problem with gossip being fired off only once, is that it goes on to an empty response if a memory isn't set (instead of to another x-keyphrase).
I agree with you that regexes are useful in very special cases only and that for the vast majority of cases the AI engine does a really great work.
IMPORTANT NOTE!
I just stumbled across something. Using a * in a keyphrase used to generate a (key) but about a month or two ago after I reworked a section of the AI Engine, it stopped doing so. I've returned it to its intended working order. Anyone who has been writing responses that don't count * as a (key) in the last month or so should look over those responses and adjust the (key)s to fit this.
In short, * and (*) both generate (key)s now, as was originally intended.
Apologies to anyone affected. I don't recall anyone reporting any issues on that front so maybe it wasn't widely used.
I just stumbled across something. Using a * in a keyphrase used to generate a (key) but about a month or two ago after I reworked a section of the AI Engine, it stopped doing so. I've returned it to its intended working order. Anyone who has been writing responses that don't count * as a (key) in the last month or so should look over those responses and adjust the (key)s to fit this.
In short, * and (*) both generate (key)s now, as was originally intended.
Apologies to anyone affected. I don't recall anyone reporting any issues on that front so maybe it wasn't widely used.
User search sorts by alphabet instead of honor. Clicking on the HONOR or LATEST link sorts correctly, but forgets the filter.
Recent updates:
* you can now get rid of an unwanted bot - the option is available on the bot's settings page.
* numbers words are no longer converted to numbers in responses
* a filter for troublesome characters like "smart quotes" has been added, and these have been replaced with their typical counterparts in the database
* bot holidays are noted in chats with bots
* implemented better orphaned keyphrase and response removal
* numerous little AI Engine bug fixes/improvements
* bot-to-bot chats restored to working order
* you can now get rid of an unwanted bot - the option is available on the bot's settings page.
* numbers words are no longer converted to numbers in responses
* a filter for troublesome characters like "smart quotes" has been added, and these have been replaced with their typical counterparts in the database
* bot holidays are noted in chats with bots
* implemented better orphaned keyphrase and response removal
* numerous little AI Engine bug fixes/improvements
* bot-to-bot chats restored to working order
(Is there a way to restore accidentally deleted bots? - Could bot holdays be noted in the transcripts too, please?)
Special characters like umlauts are not shown in popup chatboxes. (They seem to get into the transcripts, though. Can't really test for I don't know how to initiate a popup chat with my bots.)
In one Debug session (Peter Wolff; started a few minutes ago), the initial "Hello" to the bot in my name was considered Blab.
[Edit:] Just got a popup chat with one of my bots, so I can tell more.
Special characters as HTML entities are no problem in the popup window; special charactes as themselves turn up in both transcripts, and don't influence the keyphrase recognition, but they and everything that follows is omitted from my responses in the popup window (already in the page source).
Special characters like umlauts are not shown in popup chatboxes. (They seem to get into the transcripts, though. Can't really test for I don't know how to initiate a popup chat with my bots.)
In one Debug session (Peter Wolff; started a few minutes ago), the initial "Hello" to the bot in my name was considered Blab.
[Edit:] Just got a popup chat with one of my bots, so I can tell more.
Special characters as HTML entities are no problem in the popup window; special charactes as themselves turn up in both transcripts, and don't influence the keyphrase recognition, but they and everything that follows is omitted from my responses in the popup window (already in the page source).
I just fixed the blank memory bug and the chrono bug.
Mome - to clarify, do those special characters as themselves not reach the page source? Or do they reach the source but are not rendered on the page?
Also - does that affect the direct chat page and debug or just the pop-ups?
An update on what I said earlier - I looked into the code and gossip doesn't just check once like I thought - it goes through the usual process of looking for a valid response before moving on to an xnone.
Good suggestions. I've added them to my fix list.
Mome - to clarify, do those special characters as themselves not reach the page source? Or do they reach the source but are not rendered on the page?
Also - does that affect the direct chat page and debug or just the pop-ups?
An update on what I said earlier - I looked into the code and gossip doesn't just check once like I thought - it goes through the usual process of looking for a valid response before moving on to an xnone.
Good suggestions. I've added them to my fix list.
» More new posts: Doghead's Cosmic Bar