HiroSet/kwiksap-backend

fix: improve speaker identification response

Improve speaker identification matching

security: add Firebase token authentication

Increase API rate limit to 60 requests per minute

fix: revert the changes but adjust the similarity thereshold to 45%

fix: improve voice identification stability and similarity validation

fix(speaker): correct similarity threshold logic and stabilize identification

feat(status): improve API monitoring and Supabase metrics

fix(voice): improve speaker recognition and embedding matching

feat(status): add an endpoint status for return the api status

feat: update the loaded_user from the Supabase

feat: add Supabase persistence for voice embeddings to fix cache loss

fix: reduce the similarity_threshold for balance recognition

fix(backend): resolve init exit code 6 and update Supabase config

fix: add supabase to requirements for deployment

feat(backend): connect root endpoint to Supabase for live user statistics

feat: A new feature (e.g., adding the delete endpoint to the API)

fix:audio handling with tempfile and librosa

feat: optimize speaker ID with session caching and robust security middleware

feat: add audio error handling and align API parameters

feat: add rate limiting and normalization for safety

fix: adding the webrtcvad for the resemblyzer

feat: initialize FastAPI backend with Docker configuration

initial commit