June 23rd: first steps for audio!!

8:46 PM:
got audio to work yippie!
the audio

9:07 PM:
YESSIR, got STT to work finally

time spent: ~30m

June 24th: WAKE WORD DETECTION & PROMPT PROCESSING

4:30-5:11 PM:
ITS NOT WORKINGGGGGGGGGGG
so basically im trying to use vosk to listen constantly (don't worry I wont log it so you don't get spied on) and trying to get partial phrases to detect the wake word but idk whats happening
for some reason its not working the second time the wake word is said
normal function:

second time saying it:


5:31 PM:
i fixed it. all i needed to do was reset last_detected, questioning my own intelligence
wish me luck on working on the AI and voice recognition

7:20 PM:
I GOT IT TO WORK!
i have gotten the script to record a single phrase after the wake word is said, identify it with vosk, and send the file & phrase over to the main Python file!!!!
next step: add voice recognition and assistant response


8:50 PM:
just got voice recognition to work! i used resemblyzer and volunteered my sister to train the recognizer as well. it works for both of us and recognizes our separate voices. i obviously censored the names, but you can see the separate names.


10:11 PM:
after a short break, i uploaded all of the voice recognition code and made simple code to easily add voices.
time spent: 2hr
# June 25th: AI & INTERNET 7:40 PM:
ive worked on the AI integration for the entire afternoon, and i got it mostly to work. i let the AI search the web for results, but i think the AI is too stupid to understand my instructions, and is doing the opposite of WHAT I TELL IT. im thinking of switching from llama to zephyr.


8:12 PM:
there may be an issue. i was testing to see if my raspberry pi 5 was capable of running the AI (it was 4.1 gb), but when i tried downloading it weird colors started appearing on screen so i immediatley stopped it.
i think it was about to crash but i thought maybe enabling adding more swap memory would help, so i overrode the limit and gave myself 2 gb over the max (4gb) of swap memory. hope i dont break the rpi. wish me luck. might have cooked the pi

8:25 PM:
there may be another issue.
the pi wont run ai anymore even small models. i might have to find another solution but this might force me to find another computer or run it on an old home computer...
ill hopefully find a solution soon

10:06 PM:

time spent: ~ 2hr 57m

June 26th: big fixes with AI

7:53 PM:
YEAAAAAAH IT WORKED!!!!! i fixed the system prompt and made it have the capability to search the internet.

and yes that was accurate


8:01 PM:
okay, maybe the AI is still not smart...

time spent: 1hr 52m

June 27th: TTS & small updates

3:42 PM:
updated the code with changes i made last night, minor changes like clearing up system prompt and fixing a problem where if the wake word is said accidentally, it would keep listening for a question forever. planning on adding TTS today

8:29:
after a few bugs and a WHILE of looking for a decent voice (which ill have to redo for when I transfer it to my rpi) i settled on a voice called Daniel which was the only decent voice.
apparently for every OS there is a different set of voices, so if anyone wants to use this, they have to choose a different voice
i cant really put an image for audio...
time spent: ~1h
(btw I forgot to add time spent on all of the things up to today, but luckily every time I put a journal I posted a devlog on SOM, so I just got the times from here)