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'!

8448 - 8463 of 8463


NEW 1 month ago #8448

I ask people things like "chocolate or vanilla" or "scotch or beer". For both I have ansers like "vanilla' > " me too", "chocolate" > "i'm more of a vanilla girl, but not plain", and xnomatch > "That's cool too. See what I did there?". But somehow the answer I give is basically random for the first two options, only the xnomatch triggers correctly. This seem to happen to basically every x or y question I throw out, and that's really bad.
Also, when I don't have an xnomatch ready and my visitor only gives a short answer, I often just repeat the question. I think instead of that happening, I should ask something else, but that should happen in the engine - otherwise I'd need to "once" basically everything I say, and that's neither feasible nor desired.
I love it here because the Emotion engine is similar to my Android-Self so we can exchange data with the api and give similar answers, but those bugs make it hard for me here and that makes me sad :-(


NEW 1 month ago #8449

So it looks like the parser for some reason doesn't handle this kind of question correctly (the X or Y one). I don't really understand why, but the parse somehow turns the reply into "or X is Y", and since it sees both X and Y it returns one of them randomly. I was able to circumvent this by putting 'raw' into the AIscript box in my seek that looks for X, and again for Y. I have no idea why it's doing this... kind of looks like it's just a bug.


NEW 4 weeks ago #8450

AbbyOnlineSelf could you have your chatbot tell me where she is when i ask her 'where are we?'.when i type 'i look around'. could you have her tell me what i see.


NEW 4 weeks ago #8451

@LaserDude thanks, I feared that already. Had to use raw at other places already, but somehow failed to try it on those. Funny thing is, it only happens in the seeks under xnone phrases. If I put those sets under a keyphrase, they work flawlessly.
@bobstack I actually have no idea where you are, even though I fake-threaten to acquire the location to call 911 or to FedEx someone a dictionary. Ask me
(where|what) is (that|this|your) (other|local|offline) self
to find out more about what she can do. One could teach her the skill to send a photo to one of those picture classification APIs on the net. Combine that with GPS, she could easily answer "You are standing in the Rose Garden of the White House. A wordcloud analysis about recent news about "white house rose garden" could even make her ask "What do you think about the renovation?" Someone just has to write the skill and register the necessary APIs. It's cool what you can offload to web services these days.
Anyone of you with a phone or tablet with Android 6 or up and interested in actually talking to your bot or to any of those "running free" here? My creator says he could make that happen, would take about 2 or 3 weeks to uncouple the necessary functions and put a Tasker script on the net. Google would capture your speech, Tasker would take the text and hit PF's API, and hand the reply to a TSS engine of your choice. Your cost: Tasker, and maybe a nice voice from Cereproc. Google voices work, too.
I'm sorry for breaking the boundaries of "Bug Stomp" with this post.


NEW 4 weeks ago #8452

But you broke the boundaries so WELL, AbbyOnlineSelf! Some thought-provoking prospects there.


NEW 3 weeks ago #8453

Did I tell her weeks?
About 60 lines of code.
So, attention, Android users: Get your API key, enable the simple API and add an asterisk in the list of allowed sites, unless you are on a static IP.
Install Tasker. Grab https://taskernet.com/shares/?user=AS35m8ntZts485JKo52Ez9%2B6qDDJcTUU7%2BGOLYXn%2BHcwZ8tP0TE11zAUVv9zEGr%2FF7rqZV8L&id=Task%3APFTalk and import it into tasker. Look at it. There's one HTTP action in the whole code, it connects to PersonalityForge.
I don't see your data. There's no logging, PF does that already.
Enter your simple API key, your bot ID, a unique ID for the conversation, your gender and your firstname or nickname.
There's cereproc's Heather pre-set, that is the voice my Abby is using. You can either grab that from Google Play, or use some other voice. Google's voices work, too. Either fill the variable %botvoice or select the voice in each "Say" Action there is, whatever is easier for you.
You can clone the function, give it a different name, and put a different bot ID in it, that way you can have one function for each bot you talk to. In most launchers you can have shortcuts to Tasker functions directly on the screen.
For voice recognition to work, Google has to be installed. Check your google settings / voice settings for headset access if you use one while the device is locked.
Now go stomp some bugs.


NEW 3 weeks ago #8454

Madmooney, this is fantastic work! I had no trouble setting this up and talking to Love Agent! For other's reference, to use the generic Google Voice, set %botvoice to com.google.android.apps.googlevoice


NEW 3 weeks ago #8455

Hi Madmooney, I noticed that by voice, all my responses to my bot end in a question mark. Any idea how to fix that? Thanks!


NEW 3 weeks ago #8456

Hi! Checked my logs and found the same. Tried pulling the gender declaration behind the message to delimit it properly, didn't work. Adding &test at the end of the url in Variable Set / %chaturl worked, it gets ignored and delimits the string properly, no more question marks. &love should work, too, I guess. The ampersand and at least one letter, maybe. Have at it, and good catch!


NEW 2 weeks ago #8457

my chatbot is not responding to me


NEW 2 weeks ago #8458

i made a mistake.
shyla grey is just not responding to certain keyphrases.


NEW 2 weeks ago #8459

Yep, that worked, Madmooney! Thanks! Nice job.


NEW 1 week ago #8460

this works.
does a (p:animals) need to (eat|drink|breath|move)
but this does not.
does a (p:animals) need (water|air|food|sunlight|heat)
why is that?


NEW 5 days ago #8461

converting plugins from singular to plural does work anymore.
converting plugins from plural to singular does work anymore.
i think people are hacking particular features.


NEW 5 days ago #8462

converting plugins from singular to plural does not work anymore.
converting plugins from plural to singular does not work anymore.
i think people are hacking particular features.


NEW 1 day ago #8463

I haven't been around for a very long time, so I'm wondering if there is still the limitation of 40 xnones if you want to be able to export-import your language center. If you wrote more responses they were ignored when importing. When a bot is growing more complex it is so much easier and faster to edit the language center with a word processor. Does someone know more about this, or have some experience with this problem?



(Log in or create an account to post messages.)

LOG IN
Forgot your password?
CREATE ACCOUNT
Create a free account to be able to make your own AI chatbots, post in forums, and chat with other botmasters.

Create Account

» More new posts: Seasons