Locallab

Fixed
- Fixed HuggingFace token handling and validation in model loading
- Fixed ngrok token environment variable usage to use official `NGROK_AUTHTOKEN` name
- Fixed token storage and retrieval in config and environment variables

Improved
- Improved CLI UX for token input and management
- Removed token masking for better visibility
- Show current token values when available
- Added proper token validation
- Enhanced token handling across the package
- Standardized environment variable names
- Better string handling for token values
- Consistent token validation
- Better error messages for token-related issues
- Improved networking setup with proper token handling

Changed
- Updated environment variable names to use official standards
- `NGROK_AUTHTOKEN` for ngrok token
- `HUGGINGFACE_TOKEN` for HuggingFace token
- Standardized token management functions in config.py

Fixed
- Fixed critical error with ngrok URL handling in Google Colab
- Fixed NgrokTunnel type error during server initialization
- Improved error messages for ngrok connection issues
- Updated footer design for better visibility
- Clarified URL usage in documentation (localhost vs ngrok)

Changed
- Simplified footer design in server output
- Enhanced ngrok tunnel setup process with better error handling
- Updated documentation to clearly distinguish between local and ngrok URLs

Added
- Added support for HuggingFace token through CLI and environment variables
- Interactive prompt for HuggingFace token when required
- Secure token handling in configuration
- Improved error messages for model loading issues

Changed
- Made HuggingFace token optional but with interactive prompt when needed
- Enhanced model loading process with better token handling
- Updated documentation with HuggingFace token configuration details

Fixed
- Fixed critical issue with BERT model loading by removing device_map for BERT models
- Added proper BERT model configuration for text generation
- Improved model loading process with better architecture detection
- Enhanced error handling for different model architectures
- Fixed memory management for CPU-only environments
- Added automatic model type detection and configuration
- Improved compatibility with various model architectures
- Enhanced error messages for better debugging

Added
- Added support for BERT models in text generation mode
- Implemented automatic model architecture detection
- Added proper model-specific configurations
- Enhanced memory optimization for different model types

Fixed
- Fixed critical issue with server not terminating properly when Ctrl+C is pressed
- Improved process termination by using os._exit() instead of sys.exit() for clean shutdown
- Added CPU compatibility by disabling quantization when CUDA is not available
- Fixed bitsandbytes error for CPU-only systems with clear warning messages
- Enhanced user experience with better error handling for non-GPU environments

Added
- Added beautiful footer section with author information and social media links
- Included GitHub, Twitter, and Instagram links in the footer
- Added project repository link with star request
- Enhanced server startup display with comprehensive information

Fixed
- Fixed critical issue with server not shutting down properly when Ctrl+C is pressed
- Improved signal handling in ServerWithCallback class to ensure clean shutdown
- Enhanced main_loop method to respond faster to shutdown signals
- Implemented more robust server shutdown process with proper resource cleanup
- Added additional logging during shutdown to help diagnose issues
- Increased shutdown timeout to allow proper cleanup of all resources
- Fixed multiple shutdown attempts when Ctrl+C is pressed repeatedly
- Ensured all server components are properly closed during shutdown