Real-time streaming support

Currently, when interacting with Claude/GPT-4 in Cursor, responses appear only after complete generation. Adding streaming support would:

  • Allow users to begin reading/processing responses sooner
  • Create a more interactive and engaging experience

This feature would be particularly valuable when receiving longer responses or code explanations, as users wouldn’t need to wait for the complete response before beginning to process the information.