Help fine-tuning onset detection

[Emmanuel Alabi]

Hello this is kind of very specific, but ive been trying to use librosa to write a script to detect beats and generate an edl file that i can use to import markers in davinci resolve. It mostly works, but the problem i have is with the actual onset detection. When i use onset detection with a song, say montagem danza, it literally detects every little change in sound. How can i adjust it to detect only the beat drops, like what i would want if im creating a tiktok edit? Heres the code i have so far, i used a lot of help from chatgpt. https://github.com/emjjkk/beat-detection