Inspiration
Code reviews take time, and most of that time is spent writing down what has already been said and keeping track of what needs to be done
What it does
The extension records microphone audio and automatically adds PR comments to the code and to github based on what has been said and what needs to be changed
How we built it
Google cloud API for transcription, pyanote for speaker diarization, Google Gemini used for classifying, splitting, transforming, and selecting a location for comments. Github api for duplicating comments onto the Github PR.
Challenges we ran into
Speaker diarization model was difficult to implement, and is at times unreliable, and various parameters had to be changed to best ensure consistent capture of audio. Duplicating the local comments to Github was challenging and ensuring proper file pathing regardless of what view the user is using. Placing the comments in the right spot was also difficult, ensuring Gemini had enough context to make good comments and place correctly.
Accomplishments that we're proud of
Gemini API integration to create useful comments and removing unrelated conversation from being added as comments, as well as finding a good spot for the comments. Implementation of a speaker diarization model to allow more context for the model to work with.
What we learned
Gemini API integration, Github API integration, voice based software and working with the microphone.
What's next for PR Buddy
Quicker response times, more use cases, chrome extension version for online code reviews.
AI Use
We vibe coded parts of this submission
Log in or sign up for Devpost to join the conversation.