Pidog AI voice assistant HELP

I have Pidog v2 w pi5 4g.
I have completed calibration OK.
All 13 “fun projects” work great. For “see, hear, respond” steps 14 thru 19 work great.
Step 20 is my problem, When i run the script"sudo python3 20_voice_active_dog_gpt.py" Pidog says "hi there"then its listening for hey buddy.
When i say “hay buddy” LED go pink, when i ask Anything the LEds go briefly yellow and then NOTHING - program goes back to “listening” mode, I can again say hay buddy and it will wake, but any voice input resets to “listen mode”
I have updated only the 20_voice file with the API and Gemini info. Its the same info. as in file "sudo python3 18.online_llm_test.py. where it is used in step 18 and that worked great.

I have reloaded the OS and Pidog stuff 3 times with same results.
A screen shot if it helps.

thanks

The robot is not responding to the wake word.

Check if the microphone is functioning properly.
Ensure WAKE_ENABLE = True.
Adjust the wake word to match your pronunciation.

You can try switching to other LLMs or TTS services to see if that resolves the issue.

Hello

thanks for reply.

My robot works 100%, microphone, speaker everything works OK until I get to step 20.

The wake up word works great, no problem

The robot wakes up to wake word, accepts input (leds go briefly to yellow), then nothing- robot resets to listening. The wake word will wake up robot then it goes right back to listening. all other function s work. I can repeat this wake reset cycle.

I have reloaded the OS and pidog 3 times with same results.

Any Ideas how to fix this???

thanks

How’s your internet upload speeds? If low you could try to disable the camera just for debug purposes and see what happens

Change

WITH_IMAGE = True

To

WITH_IMAGE = False

OK that Image false flag helped. pidog now responds as expected(except no images). Thanks

How do i do that multimode LLM thing so the camera will work?? I am using Gemini 2.5 flash

thanks

IMHO you probably cannot. If that experiment worked, then you are likely lacking performance, possibly lack of RAM, but my best guess would be your Internet upload speed or stability is blocking the camera upload thread. The local vosk stuff will still work fine.

I’ve neither run nor looked at the code myself, I just took a best guess from your summary, which I could be totally wrong, and it may be something entirely different. More investigation would really be needed by those with more skill than I. E.g Try somewhere with faster Internet, offload some work to an external computer etc

I am getting a 16gb pi5 and we will see if that alone is the issue, one step at a time.

thanks for input

Usually the llm software reports a lack of memory anyway, which is why I don’t think yours is a memory problem, but I could be wrong!

You could try increasing swap and also use zram tools as a trial to see if more “RAM” would help.? Not as fast as real memory but may be functional. You could also try an ethernet cable to see if that helps as a trial.

Good luck!

If you get time: Please post feed back here if it works. It will help others with a similar problem.

Continuing the discussion from Pidog AI voice assistant HELP:

I connected via cat cable and same issues.

my internet speed is 122 mbps down ad 17 mbps up.

I do not know if that is good or bad?

ALSO I get the below warnings, could this be related this issue? how to fix??

thanks

That’s plenty of speed!

I’m jealous :wink: I have 5M down and 0.5M up, of course, these things don’t work for me very well at all, but should be more than enough for you for this project. So its not that then

Gemini appears to have stricter requirements for image formats. It only allows JPEG and does not allow JPG (even though they are technically the same format). The solution is to modify the file extension of the image from .jpg to .jpeg.

To fix the issue, run the following command to update the dependency to version 1.0.1:

sudo pip install --break git+https://github.com/sunfounder/sunfounder-voice-assistant.git

1 Like

That appears to have corrected my issue. Pidog can now tell the difference between a box and a can. lol. Thanks