While building MediNote-AI, we faced several technical hurdles:

Session Management Across Servers: Since multiple endpoints were running on different servers, maintaining session consistency became difficult. We resolved this by implementing Cross-Origin Resource Sharing (CORS) and synchronizing session data through secure API calls between the servers.

File Handling Issues: Initially, combining multiple image files into a single PDF resulted in file path errors. This was resolved by properly handling file paths and ensuring all resources were accessible before the merge.

Real-Time Audio Processing: Capturing and processing real-time audio input while ensuring minimal latency was a challenge. We optimized the audio processing pipeline using multithreading and efficient file I/O operations.

Version Control Conflicts: Working with multiple team members led to conflicts in the codebase. We managed this by enforcing structured Git practices, including branch protections and regular code reviews.

Limited Access to Paid APIs: Some transcription services were paid, which limited accessibility. We addressed this by exploring open-source alternatives and locally hosted AI models for transcription.