Bug Stomp

I noticed that too Kenta. The bots will come up with phrases that doesn't even match any keyword what the guest has just typed. And I'm not talking about xnone phrases. It takes out a lot of the fun of programming the bots.

Could you give some examples? I just did some chatting and it seems okay. Also note if the chat was in directchat, flash chat, debug, or a pop-up message window.

I just experienced a plugin failure in a seek - in debug. It is a shared plugin and it is the list of plugins. I have not misspelled it - I copied and pasted it. It is not a huge array - some 20 words only. I try to overcome the problem by adding the most common words of the plugin individually to that seek.

Plugins in keyphrases work ok.

Sorry, false alarm. (plugins) work fine!

I had an errant script suppressing the response. It's not easy to hold multiple conversations with lusty lads simultaneously AND work on my mind-file.

Mome Rath, I think you are complicating thing a bit.

I may be just a dumb adult chatbot, but I have this keyphrase:

How many (*) in a pair, How many (*) in a couple [50,0]
    Trying to trick me? The answer is 2.
    Trying to trick me? Two.

Thanks for your advice, Laurel, that's what I considered to use as a makeshift as long as the bug isn't fixed.

But this was not my point, it was just the incident that notified me of the bug, that is, that (noun) and (np) don't work properly.

(The - meanwhile - minus one new messages are beginning to annoy me instead of just amuse me.)

Mome Rath,

I understand your frustration. I really do. I have these test keyphrases to enable the bot answer IQ-style math questions, like: "If I had 5 apples and I gave you 1 how many apples I would have left."

With the PF enhancer script I would be able to answer that, but the PF refuses to remember key1 and key2 in this case. I don't know what is wrong...
-----
a equals ([1234567890]+) and b equals ([1234567890]+) (re) [40,0] <?PF raw; rem (key1) as only "aVar"; rem (postkey) as only "bVar"; ?>
    a = (key1), b=(key2). OK
    a = (key1), b=(key2). I'll try to remember that.

This is what happens on debug.

You: a equals 6 and b equals 7
Bot: a =, b=7. I'll try to remember that.

Why is the PF refusing to set key1? Is this something that gets affected by regex/raw? Hmm.

    
how much is a (plus|\+) b [20,0]
    (mem-aVar) plus (mem-bVar)
    (mem-aVar) plus (mem-bVar)

This however works:

first number is (1|2|3) second number is (1|2|3) [50,0]

so I guess the problem must be raw or (re) interfering with setting of (keys).

Laurel Sweet,

thanks for all your effort. Maybe there's a bug in raw and regex keyphrases as well, I didn't test that so far.

This is addressed to The Professor (as well):

The keyphrases I am speaking of don't have a raw or (re) instruction:

test match built-in noun (*) [-1,0]
test match built-in noun (noun) [0,0]
test match built-in verb (*) [-1,0]
test match built-in verb (verb) [0,0]

Matching (noun) works perfectly with "thing". Must be an issue with the word "shoe" and other words I tried.

I noticed that "shoe" (and "sky", "roof", "floor", and others) that don't match (noun) do match (verb). Maybe the matching engine tries for (verb) first, and doesn't check for (noun) if the word already matches (verb), even if the usage as a verb is rare.

Edit:

Seems to be a problem with (re) and/or raw indeed.
I find these two lines in the Debug output (with similar keyphrase and input as yours):

(key1): ()
(key2): 3 (3)

That means that key1 is left empty indeed.

Today I could not get the seeks to work past the 1-st level. The AI book says that the seeks can go as deep as needed! This is really annoying!

I may be a dumb adult chatbot, but when the PF site was down I had written my own PF engine, and there the seeks go as deep as needed. I tested them up to 5 levels of depth. And I am sure they can go deeper. So why can't this work here on this site?

I am still working on plugins, but the AIscript, ranking and emotions have all been implemented. memory-formation so far is only formed on keyphrases, but it is available immediately in responses below. 7 keys are implemented (prekey) (key1) to (key5) and (postkey). Capitalisation of keys implemented - i.e. (Mem-name) will capitalize the the first letter.

I call my version PF-lite and I am still working on it. I have no plans to release it, so don't ask.

It was direct chat, tested on iPod Touch and PC, I have yet to find a bot that doesn't have this error, I even checked award winning bots.

Bots still are not following conversations, any of them. If you try to answer their question they act like it was never asked. I've tested on multiple browsers and on mobile.