Hello there.  I’m still working on the old Pandorabots platform, mainly because I use some javascript and rely on the custom html to post-process some responses before serving to the user (and because I found the new interface a bit confusing to be honest). 

My question is this - when testing new categories, I often make tweaks to the aiml and ask the same question a few times.  I’ve noticed on two occaisons now that after asking the same question (or just saying ‘hi’) about 10 times in a row, Pandorabots ignores that template and refuses to answer.  Is this a known spam filter of sorts?  If so, does it get reset after an elapse time? 

I emailed Pandorabots support who just answered back that they don’t support the free service and wouldn’t comment - it would have taken less words to actually answer my question!

Thanks, Alan

There are no spam filters in the free server and I couldn’t reproduce the error. Is this in the training interface or the web interface after you have published your bot?

Hi Steve - we last communicated a few years ago on knytetrypper’s board.  I have found this error both on published web interface and the training interface.  In the first instance, I was testing my custom html and repeatedly typed ‘hi’ each time as it was the quickest thing to type and eventually this pattern was just ignored and I got a reply back from the pickup file.  I even made up a new category and srai’d this to ‘hi’ and still it wouldn’t work.  Eventually, I downloaded all my aiml files, created a new bot and then uploaded them and I’ve been working with this new one since.

Now when testing I am cautious not to repeat this same activity but I got caught out again yesterday, this time in the training interface where I was testing a category with a ‘that’ condition so I kept telling the bot ‘say xyz’ to get the bot to reply ‘xyz’ which it was doing fine until at some point it stopped recognising the pattern and I just started getting replies back from the pickup file.  However, I tried ‘say xyz’ again this morning and its back working again.

This is why I thought there may be some sort of mechansim to stop Pandorabots being spammed by another automated bot and it interpreted my constant repeat of the pattern to be such an attack.  I think I remember reading a thread somewhere about bot to bot conversations being stopped by Pandorabots to prevent exessive consumption of processing resources.

I wonder if it’s something strange in your custom HTML? Are you using your own AIML sets or the default ALICE ones? Perhaps you coded some anti-spam method in your own files?

Standard Alice ones, updated and amended to personalise but nothing strange.  Nope, nothing anti-spam in my files.  My custom html just catches instances of where a url opened by a category includes <star> such as in a Google search which ordinarily wouldn’t work in the aiml.  The only thing I could think of was some anti-spam facility at Pandorabots.  As soon as I started another bot with all the same aiml files, everything worked fine again.

Now as an experiment just now, I opened up my old bot, said ‘hi’ and it didn’t comprehend and showed me a default response.  It is using the same aiml files and custom html as my new one so it still seems to be blocked.  Did not Pandorabots announce some time back that they were putting in measures to stop bot to bot conversations? 

Anyway, I found a workaround which was to start a new bot with the same aiml files and same custom html but it was quite annoying because obviously I thought I had made an error somewhere in my aiml and was driving myself nuts trying to find it until I suspected it was a Pandorabots thing.

Steve - saw this snippet from Richard Wallace (Oct 11, 2010) on thread ‘http://pandorabots-general.shapado.com/questions/is-there-a-problem-with-the-new-free-hosting-server’:

If you see a bot with an unusual name reach the top of the list, or a bot has an unusually large number of interactions, it could mean that someone is spamming the server and we can take steps to block them.

This is why I thought my bot (or at least a particular pattern) might have been blocked because it repeatedly asked the same question.

This poser has been solved as well.  I figured that during testing I must have typed an answer in the ‘say instead’ field instead of the ‘ask’ field by mistake and not realised it.  So went to check the update file and sure enough there was an entry there for category ‘Hi’ which was why my bot was refusing to say hello back.  Slaps head for not realising sooner.